chore: bump version to 0.14.2 (#6032)

* chore: only retry when retry-able in flow (#5987)

* chore: only retry when retry-able

* chore: revert dbg change

* refactor: per review

* fix: check for available frontend first

* docs: more explain&longer timeout&feat: more retry at every level&try send select 1

* fix: use `sql` method for "SELECT 1"

* fix: also put recover flows in spawned task and a dead loop

* test: update transient error in flow rebuild test

* chore: sleep after sqlness sleep

* chore: add a warning

* chore: wait even more time after reboot

* fix: sanitize_connection_string (#6012)

* fix: disable recursion limit in prost (#6010)

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* ci: fix the bugs of release-dev-builder-images and add update-dev-builder-image-tag (#6009)

* fix: the dev-builder release job is not triggered by merged event

* ci: add update-dev-builder-image-tag

* fix: always create mito engine (#6018)

* fix: force streaming mode for instant source table (#6031)

* fix: force streaming mode for instant source table

* tests: sqlness test&refactor: get table

* refactor: per review

* chore: bump version to 0.14.2

---------

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: jeremyhi <jiachun_feng@proton.me>
Co-authored-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: zyy17 <zyylsxm@gmail.com>
Co-authored-by: Lei, HUANG <6406592+v0y4g3r@users.noreply.github.com>

.coderabbit.yaml

@@ -1,15 +0,0 @@
-# yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json
-language: "en-US"
-early_access: false
-reviews:
-  profile: "chill"
-  request_changes_workflow: false
-  high_level_summary: true
-  poem: true
-  review_status: true
-  collapse_walkthrough: false
-  auto_review:
-    enabled: false
-    drafts: false
-chat:
-  auto_reply: true

.github/actions/setup-greptimedb-cluster/with-remote-wal.yaml

@@ -7,7 +7,8 @@ meta:
     provider = "kafka"
     broker_endpoints = ["kafka.kafka-cluster.svc.cluster.local:9092"]
     num_topics = 3
-    auto_prune_topic_records = true
+    auto_prune_interval = "30s"
+    trigger_flush_threshold = 100
 
     [datanode]
     [datanode.client]
@@ -22,6 +23,7 @@ datanode:
     provider = "kafka"
     broker_endpoints = ["kafka.kafka-cluster.svc.cluster.local:9092"]
     linger = "2ms"
+    overwrite_entry_start_id = true
 frontend:
   configData: |-
     [runtime]

.github/scripts/update-dev-builder-version.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+DEV_BUILDER_IMAGE_TAG=$1
+
+update_dev_builder_version() {
+  if [ -z "$DEV_BUILDER_IMAGE_TAG" ]; then
+    echo "Error: Should specify the dev-builder image tag" 
+    exit 1
+  fi
+
+  # Configure Git configs.
+  git config --global user.email greptimedb-ci@greptime.com
+  git config --global user.name greptimedb-ci
+
+  # Checkout a new branch.
+  BRANCH_NAME="ci/update-dev-builder-$(date +%Y%m%d%H%M%S)"
+  git checkout -b $BRANCH_NAME
+
+  # Update the dev-builder image tag in the Makefile.
+  gsed -i "s/DEV_BUILDER_IMAGE_TAG ?=.*/DEV_BUILDER_IMAGE_TAG ?= ${DEV_BUILDER_IMAGE_TAG}/g" Makefile
+
+  # Commit the changes.
+  git add Makefile
+  git commit -m "ci: update dev-builder image tag"
+  git push origin $BRANCH_NAME
+
+  # Create a Pull Request.
+  gh pr create \
+      --title "ci: update dev-builder image tag" \
+      --body "This PR updates the dev-builder image tag" \
+      --base main \
+      --head $BRANCH_NAME \
+      --reviewer zyy17 \
+      --reviewer daviderli614
+}
+
+update_dev_builder_version

.github/workflows/grafana.yml

@@ -21,32 +21,6 @@ jobs:
         run: sudo apt-get install -y jq
 
       # Make the check.sh script executable
-      - name: Make check.sh executable
-        run: chmod +x grafana/check.sh
-
-      # Run the check.sh script
-      - name: Run check.sh
-        run: ./grafana/check.sh
-
-      # Only run summary.sh for pull_request events (not for merge queues or final pushes)
-      - name: Check if this is a pull request
-        id: check-pr
+      - name: Check grafana dashboards
         run: |
-          if [[ "${{ github.event_name }}" == "pull_request" ]]; then
-            echo "is_pull_request=true" >> $GITHUB_OUTPUT
-          else
-            echo "is_pull_request=false" >> $GITHUB_OUTPUT
-          fi
-
-      # Make the summary.sh script executable
-      - name: Make summary.sh executable
-        if: steps.check-pr.outputs.is_pull_request == 'true'
-        run: chmod +x grafana/summary.sh
-
-      # Run the summary.sh script and add its output to the GitHub Job Summary
-      - name: Run summary.sh and add to Job Summary
-        if: steps.check-pr.outputs.is_pull_request == 'true'
-        run: |
-          SUMMARY=$(./grafana/summary.sh)
-          echo "### Summary of Grafana Panels" >> $GITHUB_STEP_SUMMARY
-          echo "$SUMMARY" >> $GITHUB_STEP_SUMMARY
+          make check-dashboards

.github/workflows/release-dev-builder-images.yaml

@@ -24,11 +24,19 @@ on:
         description: Release dev-builder-android image
         required: false
         default: false
+      update_dev_builder_image_tag:
+        type: boolean
+        description: Update the DEV_BUILDER_IMAGE_TAG in Makefile and create a PR
+        required: false
+        default: false
 
 jobs:
   release-dev-builder-images:
     name: Release dev builder images
-    if: ${{ inputs.release_dev_builder_ubuntu_image || inputs.release_dev_builder_centos_image || inputs.release_dev_builder_android_image }} # Only manually trigger this job.
+    # The jobs are triggered by the following events:
+    # 1. Manually triggered workflow_dispatch event
+    # 2. Push event when the PR that modifies the `rust-toolchain.toml` or `docker/dev-builder/**` is merged to main
+    if: ${{ github.event_name == 'push' || inputs.release_dev_builder_ubuntu_image || inputs.release_dev_builder_centos_image || inputs.release_dev_builder_android_image }}
     runs-on: ubuntu-latest
     outputs:
       version: ${{ steps.set-version.outputs.version }}
@@ -57,9 +65,9 @@ jobs:
           version: ${{ env.VERSION }}
           dockerhub-image-registry-username: ${{ secrets.DOCKERHUB_USERNAME }}
           dockerhub-image-registry-token: ${{ secrets.DOCKERHUB_TOKEN }}
-          build-dev-builder-ubuntu: ${{ inputs.release_dev_builder_ubuntu_image }}
-          build-dev-builder-centos: ${{ inputs.release_dev_builder_centos_image }}
-          build-dev-builder-android: ${{ inputs.release_dev_builder_android_image }}
+          build-dev-builder-ubuntu: ${{ inputs.release_dev_builder_ubuntu_image || github.event_name == 'push' }}
+          build-dev-builder-centos: ${{ inputs.release_dev_builder_centos_image || github.event_name == 'push' }}
+          build-dev-builder-android: ${{ inputs.release_dev_builder_android_image || github.event_name == 'push' }}
 
   release-dev-builder-images-ecr:
     name: Release dev builder images to AWS ECR
@@ -85,7 +93,7 @@ jobs:
 
       - name: Push dev-builder-ubuntu image
         shell: bash
-        if: ${{ inputs.release_dev_builder_ubuntu_image }}
+        if: ${{ inputs.release_dev_builder_ubuntu_image || github.event_name == 'push' }}
         env:
           IMAGE_VERSION: ${{ needs.release-dev-builder-images.outputs.version }}
           IMAGE_NAMESPACE: ${{ vars.IMAGE_NAMESPACE }}
@@ -106,7 +114,7 @@ jobs:
 
       - name: Push dev-builder-centos image
         shell: bash
-        if: ${{ inputs.release_dev_builder_centos_image }}
+        if: ${{ inputs.release_dev_builder_centos_image || github.event_name == 'push' }}
         env:
           IMAGE_VERSION: ${{ needs.release-dev-builder-images.outputs.version }}
           IMAGE_NAMESPACE: ${{ vars.IMAGE_NAMESPACE }}
@@ -127,7 +135,7 @@ jobs:
 
       - name: Push dev-builder-android image
         shell: bash
-        if: ${{ inputs.release_dev_builder_android_image }}
+        if: ${{ inputs.release_dev_builder_android_image || github.event_name == 'push' }}
         env:
           IMAGE_VERSION: ${{ needs.release-dev-builder-images.outputs.version }}
           IMAGE_NAMESPACE: ${{ vars.IMAGE_NAMESPACE }}
@@ -162,7 +170,7 @@ jobs:
 
       - name: Push dev-builder-ubuntu image
         shell: bash
-        if: ${{ inputs.release_dev_builder_ubuntu_image }}
+        if: ${{ inputs.release_dev_builder_ubuntu_image || github.event_name == 'push' }}
         env:
           IMAGE_VERSION: ${{ needs.release-dev-builder-images.outputs.version }}
           IMAGE_NAMESPACE: ${{ vars.IMAGE_NAMESPACE }}
@@ -176,7 +184,7 @@ jobs:
 
       - name: Push dev-builder-centos image
         shell: bash
-        if: ${{ inputs.release_dev_builder_centos_image }}
+        if: ${{ inputs.release_dev_builder_centos_image || github.event_name == 'push' }}
         env:
           IMAGE_VERSION: ${{ needs.release-dev-builder-images.outputs.version }}
           IMAGE_NAMESPACE: ${{ vars.IMAGE_NAMESPACE }}
@@ -190,7 +198,7 @@ jobs:
 
       - name: Push dev-builder-android image
         shell: bash
-        if: ${{ inputs.release_dev_builder_android_image }}
+        if: ${{ inputs.release_dev_builder_android_image || github.event_name == 'push' }}
         env:
           IMAGE_VERSION: ${{ needs.release-dev-builder-images.outputs.version }}
           IMAGE_NAMESPACE: ${{ vars.IMAGE_NAMESPACE }}
@@ -201,3 +209,24 @@ jobs:
             quay.io/skopeo/stable:latest \
             copy -a docker://docker.io/$IMAGE_NAMESPACE/dev-builder-android:$IMAGE_VERSION \
             docker://$ACR_IMAGE_REGISTRY/$IMAGE_NAMESPACE/dev-builder-android:$IMAGE_VERSION
+  
+  update-dev-builder-image-tag:
+    name: Update dev-builder image tag
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+      pull-requests: write
+    if: ${{ github.event_name == 'push' || inputs.update_dev_builder_image_tag }}
+    needs: [
+      release-dev-builder-images
+    ]
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Update dev-builder image tag
+        shell: bash
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          ./.github/scripts/update-dev-builder-version.sh ${{ needs.release-dev-builder-images.outputs.version }}

Cargo.toml

@@ -68,7 +68,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "0.14.0"
+version = "0.14.2"
 edition = "2021"
 license = "Apache-2.0"
 
@@ -112,15 +112,15 @@ clap = { version = "4.4", features = ["derive"] }
 config = "0.13.0"
 crossbeam-utils = "0.8"
 dashmap = "6.1"
-datafusion = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "5bbedc6704162afb03478f56ffb629405a4e1220" }
-datafusion-common = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "5bbedc6704162afb03478f56ffb629405a4e1220" }
-datafusion-expr = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "5bbedc6704162afb03478f56ffb629405a4e1220" }
-datafusion-functions = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "5bbedc6704162afb03478f56ffb629405a4e1220" }
-datafusion-optimizer = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "5bbedc6704162afb03478f56ffb629405a4e1220" }
-datafusion-physical-expr = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "5bbedc6704162afb03478f56ffb629405a4e1220" }
-datafusion-physical-plan = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "5bbedc6704162afb03478f56ffb629405a4e1220" }
-datafusion-sql = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "5bbedc6704162afb03478f56ffb629405a4e1220" }
-datafusion-substrait = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "5bbedc6704162afb03478f56ffb629405a4e1220" }
+datafusion = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "e104c7cf62b11dd5fe41461b82514978234326b4" }
+datafusion-common = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "e104c7cf62b11dd5fe41461b82514978234326b4" }
+datafusion-expr = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "e104c7cf62b11dd5fe41461b82514978234326b4" }
+datafusion-functions = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "e104c7cf62b11dd5fe41461b82514978234326b4" }
+datafusion-optimizer = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "e104c7cf62b11dd5fe41461b82514978234326b4" }
+datafusion-physical-expr = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "e104c7cf62b11dd5fe41461b82514978234326b4" }
+datafusion-physical-plan = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "e104c7cf62b11dd5fe41461b82514978234326b4" }
+datafusion-sql = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "e104c7cf62b11dd5fe41461b82514978234326b4" }
+datafusion-substrait = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "e104c7cf62b11dd5fe41461b82514978234326b4" }
 deadpool = "0.12"
 deadpool-postgres = "0.14"
 derive_builder = "0.20"
@@ -129,7 +129,7 @@ etcd-client = "0.14"
 fst = "0.4.7"
 futures = "0.3"
 futures-util = "0.3"
-greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "b6d9cffd43c4e6358805a798f17e03e232994b82" }
+greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "e82b0158cd38d4021edb4e4c0ae77f999051e62f" }
 hex = "0.4"
 http = "1"
 humantime = "2.1"
@@ -162,7 +162,7 @@ paste = "1.0"
 pin-project = "1.0"
 prometheus = { version = "0.13.3", features = ["process"] }
 promql-parser = { version = "0.5.1", features = ["ser"] }
-prost = "0.13"
+prost = { version = "0.13", features = ["no-recursion-limit"] }
 raft-engine = { version = "0.4.1", default-features = false }
 rand = "0.9"
 ratelimit = "0.10"
@@ -191,7 +191,7 @@ simd-json = "0.15"
 similar-asserts = "1.6.0"
 smallvec = { version = "1", features = ["serde"] }
 snafu = "0.8"
-sqlparser = { git = "https://github.com/GreptimeTeam/sqlparser-rs.git", rev = "e98e6b322426a9d397a71efef17075966223c089", features = [
+sqlparser = { git = "https://github.com/GreptimeTeam/sqlparser-rs.git", rev = "0cf6c04490d59435ee965edd2078e8855bd8471e", features = [
     "visitor",
     "serde",
 ] } # branch = "v0.54.x"
@@ -269,6 +269,9 @@ metric-engine = { path = "src/metric-engine" }
 mito2 = { path = "src/mito2" }
 object-store = { path = "src/object-store" }
 operator = { path = "src/operator" }
+otel-arrow-rust = { git = "https://github.com/open-telemetry/otel-arrow", rev = "5d551412d2a12e689cde4d84c14ef29e36784e51", features = [
+    "server",
+] }
 partition = { path = "src/partition" }
 pipeline = { path = "src/pipeline" }
 plugins = { path = "src/plugins" }

Makefile

@@ -222,6 +222,16 @@ start-cluster: ## Start the greptimedb cluster with etcd by using docker compose
 stop-cluster: ## Stop the greptimedb cluster that created by docker compose.
 	docker compose -f ./docker/docker-compose/cluster-with-etcd.yaml stop
 
+##@ Grafana
+
+.PHONY: check-dashboards
+check-dashboards: ## Check the Grafana dashboards.
+	@./grafana/scripts/check.sh
+
+.PHONY: dashboards
+dashboards: ## Generate the Grafana dashboards for standalone mode and intermediate dashboards.
+	@./grafana/scripts/gen-dashboards.sh
+
 ##@ Docs
 config-docs: ## Generate configuration documentation from toml files.
 	docker run --rm \

README.md

@@ -6,7 +6,7 @@
   </picture>
 </p>
 
-<h2 align="center">Unified & Cost-Effective Observability  Database for Metrics, Logs, and Events</h2>
+<h2 align="center">Real-Time & Cloud-Native Observability  Database<br/>for metrics, logs, and traces</h2>
 
 <div align="center">
 <h3 align="center">

config/config.md

@@ -319,6 +319,7 @@
 | `selector` | String | `round_robin` | Datanode selector type.<br/>- `round_robin` (default value)<br/>- `lease_based`<br/>- `load_based`<br/>For details, please see "https://docs.greptime.com/developer-guide/metasrv/selector". |
 | `use_memory_store` | Bool | `false` | Store data in memory. |
 | `enable_region_failover` | Bool | `false` | Whether to enable region failover.<br/>This feature is only available on GreptimeDB running on cluster mode and<br/>- Using Remote WAL<br/>- Using shared storage (e.g., s3). |
+| `allow_region_failover_on_local_wal` | Bool | `false` | Whether to allow region failover on local WAL.<br/>**This option is not recommended to be set to true, because it may lead to data loss during failover.** |
 | `node_max_idle_time` | String | `24hours` | Max allowed idle time before removing node info from metasrv memory. |
 | `enable_telemetry` | Bool | `true` | Whether to enable greptimedb telemetry. Enabled by default. |
 | `runtime` | -- | -- | The runtime options. |

config/metasrv.example.toml

@@ -50,6 +50,10 @@ use_memory_store = false
 ## - Using shared storage (e.g., s3).
 enable_region_failover = false
 
+## Whether to allow region failover on local WAL.
+## **This option is not recommended to be set to true, because it may lead to data loss during failover.**
+allow_region_failover_on_local_wal = false
+
 ## Max allowed idle time before removing node info from metasrv memory.
 node_max_idle_time = "24hours"
 

grafana/README.md

@@ -1,61 +1,89 @@
-Grafana dashboard for GreptimeDB
---------------------------------
+# Grafana dashboards for GreptimeDB
 
-GreptimeDB's official Grafana dashboard.
+## Overview
 
-Status notify: we are still working on this config. It's expected to change frequently in the recent days. Please feel free to submit your feedback and/or contribution to this dashboard 🤗
+This repository maintains the Grafana dashboards for GreptimeDB. It has two types of dashboards:
 
-If you use Helm [chart](https://github.com/GreptimeTeam/helm-charts) to deploy GreptimeDB cluster, you can enable self-monitoring by setting the following values in your Helm chart:
+- `cluster/dashboard.json`: The Grafana dashboard for the GreptimeDB cluster. Read the [dashboard.md](./dashboards/cluster/dashboard.md) for more details.
+- `standalone/dashboard.json`: The Grafana dashboard for the standalone GreptimeDB instance. **It's generated from the `cluster/dashboard.json` by removing the instance filter through the `make dashboards` command**. Read the [dashboard.md](./dashboards/standalone/dashboard.md) for more details.
+
+As the rapid development of GreptimeDB, the metrics may be changed, and please feel free to submit your feedback and/or contribution to this dashboard 🤗
+
+**NOTE**: 
+
+- The Grafana version should be greater than 9.0.
+
+- If you want to modify the dashboards, you only need to modify the `cluster/dashboard.json` and run the `make dashboards` command to generate the `standalone/dashboard.json` and other related files.
+
+To maintain the dashboards easily, we use the [`dac`](https://github.com/zyy17/dac) tool to generate the intermediate dashboards and markdown documents:
+
+- `cluster/dashboard.yaml`: The intermediate dashboard for the GreptimeDB cluster.
+- `standalone/dashboard.yaml`: The intermediate dashboard for the standalone GreptimeDB instance.
+
+## Data Sources
+
+There are two data sources for the dashboards to fetch the metrics:
+
+- **Prometheus**: Expose the metrics of GreptimeDB.
+- **Information Schema**: It is the MySQL port of the current monitored instance. The `overview` dashboard will use this datasource to show the information schema of the current instance.
+
+## Instance Filters
+
+To deploy the dashboards for multiple scenarios (K8s, bare metal, etc.), we prefer to use the `instance` label when filtering instances.
+
+Additionally, we recommend including the `pod` label in the legend to make it easier to identify each instance, even though this field will be empty in bare metal scenarios.
+
+For example, the following query is recommended:
+
+```promql
+sum(process_resident_memory_bytes{instance=~"$datanode"}) by (instance, pod)
+```
+
+And the legend will be like: `[{{instance}}]-[{{ pod }}]`.
+
+## Deployment
+
+### Helm
+
+If you use the Helm [chart](https://github.com/GreptimeTeam/helm-charts) to deploy a GreptimeDB cluster, you can enable self-monitoring by setting the following values in your Helm chart:
 
 - `monitoring.enabled=true`: Deploys a standalone GreptimeDB instance dedicated to monitoring the cluster;
 - `grafana.enabled=true`: Deploys Grafana and automatically imports the monitoring dashboard;
 
-The standalone GreptimeDB instance will collect metrics from your cluster and the dashboard will be available in the Grafana UI. For detailed deployment instructions, please refer to our [Kubernetes deployment guide](https://docs.greptime.com/nightly/user-guide/deployments/deploy-on-kubernetes/getting-started).
+The standalone GreptimeDB instance will collect metrics from your cluster, and the dashboard will be available in the Grafana UI. For detailed deployment instructions, please refer to our [Kubernetes deployment guide](https://docs.greptime.com/nightly/user-guide/deployments/deploy-on-kubernetes/getting-started).
 
-# How to use
+### Self-host Prometheus and import dashboards manually
 
-## `greptimedb.json`
+1. **Configure Prometheus to scrape the cluster**
 
-Open Grafana Dashboard page, choose `New` -> `Import`. And upload `greptimedb.json` file.
+   The following is an example configuration(**Please modify it according to your actual situation**):
 
-## `greptimedb-cluster.json`
+    ```yml
+    # example config
+    # only to indicate how to assign labels to each target
+    # modify yours accordingly
+    scrape_configs:
+      - job_name: metasrv
+        static_configs:
+        - targets: ['<metasrv-ip>:<port>']
 
-This cluster dashboard provides a comprehensive view of incoming requests, response statuses, and internal activities such as flush and compaction, with a layered structure from frontend to datanode. Designed with a focus on alert functionality, its primary aim is to highlight any anomalies in metrics, allowing users to quickly pinpoint the cause of errors.
+      - job_name: datanode
+        static_configs:
+        - targets: ['<datanode0-ip>:<port>', '<datanode1-ip>:<port>', '<datanode2-ip>:<port>']
 
-We use Prometheus to scrape off metrics from nodes in GreptimeDB cluster, Grafana to visualize the diagram. Any compatible stack should work too.
+      - job_name: frontend
+        static_configs:
+        - targets: ['<frontend-ip>:<port>']
+    ```
 
-__Note__: This dashboard is still in an early stage of development. Any issue or advice on improvement is welcomed.
+2. **Configure the data sources in Grafana**
 
-### Configuration
+   You need to add two data sources in Grafana:
 
-Please ensure the following configuration before importing the dashboard into Grafana.
+   - Prometheus: It is the Prometheus instance that scrapes the GreptimeDB metrics.
+   - Information Schema: It is the MySQL port of the current monitored instance. The dashboard will use this datasource to show the information schema of the current instance.
 
-__1. Prometheus scrape config__
+3. **Import the dashboards based on your deployment scenario**
 
-Configure Prometheus to scrape the cluster.
-
-```yml
-# example config
-# only to indicate how to assign labels to each target
-# modify yours accordingly
-scrape_configs:
-  - job_name: metasrv
-    static_configs:
-    - targets: ['<metasrv-ip>:<port>']
-
-  - job_name: datanode
-    static_configs:
-    - targets: ['<datanode0-ip>:<port>', '<datanode1-ip>:<port>', '<datanode2-ip>:<port>']
-
-  - job_name: frontend
-    static_configs:
-    - targets: ['<frontend-ip>:<port>']
-```
-
-__2. Grafana config__
-
-Create a Prometheus data source in Grafana before using this dashboard. We use `datasource` as a variable in Grafana dashboard so that multiple environments are supported.
-
-### Usage
-
-Use `datasource` or `instance` on the upper-left corner to filter data from certain node.
+   - **Cluster**: Import the `cluster/dashboard.json` dashboard.
+   - **Standalone**: Import the `standalone/dashboard.json` dashboard.

grafana/check.sh

@@ -1,19 +0,0 @@
-#!/usr/bin/env bash
-
-BASEDIR=$(dirname "$0")
-
-# Use jq to check for panels with empty or missing descriptions
-invalid_panels=$(cat $BASEDIR/greptimedb-cluster.json | jq -r '
-  .panels[]
-  | select((.type == "stats" or .type == "timeseries") and (.description == "" or .description == null))
-')
-
-# Check if any invalid panels were found
-if [[ -n "$invalid_panels" ]]; then
-  echo "Error: The following panels have empty or missing descriptions:"
-  echo "$invalid_panels"
-  exit 1
-else
-  echo "All panels with type 'stats' or 'timeseries' have valid descriptions."
-  exit 0
-fi

grafana/dashboards/cluster/dashboard.md

@@ -0,0 +1,97 @@
+# Overview
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Uptime | `time() - process_start_time_seconds` | `stat` | The start time of GreptimeDB. | `prometheus` | `s` | `__auto` |
+| Version | `SELECT pkg_version FROM information_schema.build_info` | `stat` | GreptimeDB version. | `mysql` | -- | -- |
+| Total Ingestion Rate | `sum(rate(greptime_table_operator_ingest_rows[$__rate_interval]))` | `stat` | Total ingestion rate. | `prometheus` | `rowsps` | `__auto` |
+| Total Storage Size | `select SUM(disk_size) from information_schema.region_statistics;` | `stat` | Total number of data file size. | `mysql` | `decbytes` | -- |
+| Total Rows | `select SUM(region_rows) from information_schema.region_statistics;` | `stat` | Total number of data rows in the cluster. Calculated by sum of rows from each region. | `mysql` | `sishort` | -- |
+| Deployment | `SELECT count(*) as datanode FROM information_schema.cluster_info WHERE peer_type = 'DATANODE';`<br/>`SELECT count(*) as frontend FROM information_schema.cluster_info WHERE peer_type = 'FRONTEND';`<br/>`SELECT count(*) as metasrv FROM information_schema.cluster_info WHERE peer_type = 'METASRV';`<br/>`SELECT count(*) as flownode FROM information_schema.cluster_info WHERE peer_type = 'FLOWNODE';` | `stat` | The deployment topology of GreptimeDB. | `mysql` | -- | -- |
+| Database Resources | `SELECT COUNT(*) as databases FROM information_schema.schemata WHERE schema_name NOT IN ('greptime_private', 'information_schema')`<br/>`SELECT COUNT(*) as tables FROM information_schema.tables WHERE table_schema != 'information_schema'`<br/>`SELECT COUNT(region_id) as regions FROM information_schema.region_peers`<br/>`SELECT COUNT(*) as flows FROM information_schema.flows` | `stat` | The number of the key resources in GreptimeDB. | `mysql` | -- | -- |
+| Data Size | `SELECT SUM(memtable_size) * 0.42825 as WAL FROM information_schema.region_statistics;`<br/>`SELECT SUM(index_size) as index FROM information_schema.region_statistics;`<br/>`SELECT SUM(manifest_size) as manifest FROM information_schema.region_statistics;` | `stat` | The data size of wal/index/manifest in the GreptimeDB. | `mysql` | `decbytes` | -- |
+# Ingestion
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Total Ingestion Rate | `sum(rate(greptime_table_operator_ingest_rows{instance=~"$frontend"}[$__rate_interval]))` | `timeseries` | Total ingestion rate.<br/><br/>Here we listed 3 primary protocols:<br/><br/>- Prometheus remote write<br/>- Greptime's gRPC API (when using our ingest SDK)<br/>- Log ingestion http API<br/> | `prometheus` | `rowsps` | `ingestion` |
+| Ingestion Rate by Type | `sum(rate(greptime_servers_http_logs_ingestion_counter[$__rate_interval]))`<br/>`sum(rate(greptime_servers_prometheus_remote_write_samples[$__rate_interval]))` | `timeseries` | Total ingestion rate.<br/><br/>Here we listed 3 primary protocols:<br/><br/>- Prometheus remote write<br/>- Greptime's gRPC API (when using our ingest SDK)<br/>- Log ingestion http API<br/> | `prometheus` | `rowsps` | `http-logs` |
+# Queries
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Total Query Rate | `sum (rate(greptime_servers_mysql_query_elapsed_count{instance=~"$frontend"}[$__rate_interval]))`<br/>`sum (rate(greptime_servers_postgres_query_elapsed_count{instance=~"$frontend"}[$__rate_interval]))`<br/>`sum (rate(greptime_servers_http_promql_elapsed_counte{instance=~"$frontend"}[$__rate_interval]))` | `timeseries` | Total rate of query API calls by protocol. This metric is collected from frontends.<br/><br/>Here we listed 3 main protocols:<br/>- MySQL<br/>- Postgres<br/>- Prometheus API<br/><br/>Note that there are some other minor query APIs like /sql are not included | `prometheus` | `reqps` | `mysql` |
+# Resources
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Datanode Memory per Instance | `sum(process_resident_memory_bytes{instance=~"$datanode"}) by (instance, pod)` | `timeseries` | Current memory usage by instance | `prometheus` | `decbytes` | `[{{instance}}]-[{{ pod }}]` |
+| Datanode CPU Usage per Instance | `sum(rate(process_cpu_seconds_total{instance=~"$datanode"}[$__rate_interval]) * 1000) by (instance, pod)` | `timeseries` | Current cpu usage by instance | `prometheus` | `none` | `[{{ instance }}]-[{{ pod }}]` |
+| Frontend Memory per Instance | `sum(process_resident_memory_bytes{instance=~"$frontend"}) by (instance, pod)` | `timeseries` | Current memory usage by instance | `prometheus` | `decbytes` | `[{{ instance }}]-[{{ pod }}]` |
+| Frontend CPU Usage per Instance | `sum(rate(process_cpu_seconds_total{instance=~"$frontend"}[$__rate_interval]) * 1000) by (instance, pod)` | `timeseries` | Current cpu usage by instance | `prometheus` | `none` | `[{{ instance }}]-[{{ pod }}]-cpu` |
+| Metasrv Memory per Instance | `sum(process_resident_memory_bytes{instance=~"$metasrv"}) by (instance, pod)` | `timeseries` | Current memory usage by instance | `prometheus` | `decbytes` | `[{{ instance }}]-[{{ pod }}]-resident` |
+| Metasrv CPU Usage per Instance | `sum(rate(process_cpu_seconds_total{instance=~"$metasrv"}[$__rate_interval]) * 1000) by (instance, pod)` | `timeseries` | Current cpu usage by instance | `prometheus` | `none` | `[{{ instance }}]-[{{ pod }}]` |
+| Flownode Memory per Instance | `sum(process_resident_memory_bytes{instance=~"$flownode"}) by (instance, pod)` | `timeseries` | Current memory usage by instance | `prometheus` | `decbytes` | `[{{ instance }}]-[{{ pod }}]` |
+| Flownode CPU Usage per Instance | `sum(rate(process_cpu_seconds_total{instance=~"$flownode"}[$__rate_interval]) * 1000) by (instance, pod)` | `timeseries` | Current cpu usage by instance | `prometheus` | `none` | `[{{ instance }}]-[{{ pod }}]` |
+# Frontend Requests
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| HTTP QPS per Instance | `sum by(instance, pod, path, method, code) (rate(greptime_servers_http_requests_elapsed_count{instance=~"$frontend",path!~"/health\|/metrics"}[$__rate_interval]))` | `timeseries` | HTTP QPS per Instance. | `prometheus` | `reqps` | `[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]` |
+| HTTP P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, path, method, code) (rate(greptime_servers_http_requests_elapsed_bucket{instance=~"$frontend",path!~"/health\|/metrics"}[$__rate_interval])))` | `timeseries` | HTTP P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]-p99` |
+| gRPC QPS per Instance | `sum by(instance, pod, path, code) (rate(greptime_servers_grpc_requests_elapsed_count{instance=~"$frontend"}[$__rate_interval]))` | `timeseries` | gRPC QPS per Instance. | `prometheus` | `reqps` | `[{{instance}}]-[{{pod}}]-[{{path}}]-[{{code}}]` |
+| gRPC P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, path, code) (rate(greptime_servers_grpc_requests_elapsed_bucket{instance=~"$frontend"}[$__rate_interval])))` | `timeseries` | gRPC P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]-p99` |
+| MySQL QPS per Instance | `sum by(pod, instance)(rate(greptime_servers_mysql_query_elapsed_count{instance=~"$frontend"}[$__rate_interval]))` | `timeseries` | MySQL QPS per Instance. | `prometheus` | `reqps` | `[{{instance}}]-[{{pod}}]` |
+| MySQL P99 per Instance | `histogram_quantile(0.99, sum by(pod, instance, le) (rate(greptime_servers_mysql_query_elapsed_bucket{instance=~"$frontend"}[$__rate_interval])))` | `timeseries` | MySQL P99 per Instance. | `prometheus` | `s` | `[{{ instance }}]-[{{ pod }}]-p99` |
+| PostgreSQL QPS per Instance | `sum by(pod, instance)(rate(greptime_servers_postgres_query_elapsed_count{instance=~"$frontend"}[$__rate_interval]))` | `timeseries` | PostgreSQL QPS per Instance. | `prometheus` | `reqps` | `[{{instance}}]-[{{pod}}]` |
+| PostgreSQL P99 per Instance | `histogram_quantile(0.99, sum by(pod,instance,le) (rate(greptime_servers_postgres_query_elapsed_bucket{instance=~"$frontend"}[$__rate_interval])))` | `timeseries` | PostgreSQL P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-p99` |
+# Frontend to Datanode
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Ingest Rows per Instance | `sum by(instance, pod)(rate(greptime_table_operator_ingest_rows{instance=~"$frontend"}[$__rate_interval]))` | `timeseries` | Ingestion rate by row as in each frontend | `prometheus` | `rowsps` | `[{{instance}}]-[{{pod}}]` |
+| Region Call QPS per Instance | `sum by(instance, pod, request_type) (rate(greptime_grpc_region_request_count{instance=~"$frontend"}[$__rate_interval]))` | `timeseries` | Region Call QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{request_type}}]` |
+| Region Call P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, request_type) (rate(greptime_grpc_region_request_bucket{instance=~"$frontend"}[$__rate_interval])))` | `timeseries` | Region Call P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{request_type}}]` |
+# Mito Engine
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Request OPS per Instance | `sum by(instance, pod, type) (rate(greptime_mito_handle_request_elapsed_count{instance=~"$datanode"}[$__rate_interval]))` | `timeseries` | Request QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{type}}]` |
+| Request P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, type) (rate(greptime_mito_handle_request_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))` | `timeseries` | Request P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{type}}]` |
+| Write Buffer per Instance | `greptime_mito_write_buffer_bytes{instance=~"$datanode"}` | `timeseries` | Write Buffer per Instance. | `prometheus` | `decbytes` | `[{{instance}}]-[{{pod}}]` |
+| Write Rows per Instance | `sum by (instance, pod) (rate(greptime_mito_write_rows_total{instance=~"$datanode"}[$__rate_interval]))` | `timeseries` | Ingestion size by row counts. | `prometheus` | `rowsps` | `[{{instance}}]-[{{pod}}]` |
+| Flush OPS per Instance | `sum by(instance, pod, reason) (rate(greptime_mito_flush_requests_total{instance=~"$datanode"}[$__rate_interval]))` | `timeseries` | Flush QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{reason}}]` |
+| Write Stall per Instance | `sum by(instance, pod) (greptime_mito_write_stall_total{instance=~"$datanode"})` | `timeseries` | Write Stall per Instance. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]` |
+| Read Stage OPS per Instance | `sum by(instance, pod) (rate(greptime_mito_read_stage_elapsed_count{instance=~"$datanode", stage="total"}[$__rate_interval]))` | `timeseries` | Read Stage OPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]` |
+| Read Stage P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_read_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))` | `timeseries` | Read Stage P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]` |
+| Write Stage P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_write_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))` | `timeseries` | Write Stage P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]` |
+| Compaction OPS per Instance | `sum by(instance, pod) (rate(greptime_mito_compaction_total_elapsed_count{instance=~"$datanode"}[$__rate_interval]))` | `timeseries` | Compaction OPS per Instance. | `prometheus` | `ops` | `[{{ instance }}]-[{{pod}}]` |
+| Compaction P99 per Instance by Stage | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))` | `timeseries` | Compaction latency by stage | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-p99` |
+| Compaction P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le,stage) (rate(greptime_mito_compaction_total_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))` | `timeseries` | Compaction P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-compaction` |
+| WAL write size | `histogram_quantile(0.95, sum by(le,instance, pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))`<br/>`histogram_quantile(0.99, sum by(le,instance,pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))`<br/>`sum by (instance, pod)(rate(raft_engine_write_size_sum[$__rate_interval]))` | `timeseries` | Write-ahead logs write size as bytes. This chart includes stats of p95 and p99 size by instance, total WAL write rate. | `prometheus` | `bytes` | `[{{instance}}]-[{{pod}}]-req-size-p95` |
+| Cached Bytes per Instance | `greptime_mito_cache_bytes{instance=~"$datanode"}` | `timeseries` | Cached Bytes per Instance. | `prometheus` | `decbytes` | `[{{instance}}]-[{{pod}}]-[{{type}}]` |
+| Inflight Compaction | `greptime_mito_inflight_compaction_count` | `timeseries` | Ongoing compaction task count | `prometheus` | `none` | `[{{instance}}]-[{{pod}}]` |
+| WAL sync duration seconds | `histogram_quantile(0.99, sum by(le, type, node, instance, pod) (rate(raft_engine_sync_log_duration_seconds_bucket[$__rate_interval])))` | `timeseries` | Raft engine (local disk) log store sync latency, p99 | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-p99` |
+| Log Store op duration seconds | `histogram_quantile(0.99, sum by(le,logstore,optype,instance, pod) (rate(greptime_logstore_op_elapsed_bucket[$__rate_interval])))` | `timeseries` | Write-ahead log operations latency at p99 | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{logstore}}]-[{{optype}}]-p99` |
+| Inflight Flush | `greptime_mito_inflight_flush_count` | `timeseries` | Ongoing flush task count | `prometheus` | `none` | `[{{instance}}]-[{{pod}}]` |
+# OpenDAL
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| QPS per Instance | `sum by(instance, pod, scheme, operation) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode"}[$__rate_interval]))` | `timeseries` | QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]` |
+| Read QPS per Instance | `sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode", operation="read"}[$__rate_interval]))` | `timeseries` | Read QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]` |
+| Read P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{instance=~"$datanode",operation="read"}[$__rate_interval])))` | `timeseries` | Read P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-{{scheme}}` |
+| Write QPS per Instance | `sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode", operation="write"}[$__rate_interval]))` | `timeseries` | Write QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-{{scheme}}` |
+| Write P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{instance=~"$datanode", operation="write"}[$__rate_interval])))` | `timeseries` | Write P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]` |
+| List QPS per Instance | `sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode", operation="list"}[$__rate_interval]))` | `timeseries` | List QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]` |
+| List P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{instance=~"$datanode", operation="list"}[$__rate_interval])))` | `timeseries` | List P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]` |
+| Other Requests per Instance | `sum by(instance, pod, scheme, operation) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode",operation!~"read\|write\|list\|stat"}[$__rate_interval]))` | `timeseries` | Other Requests per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]` |
+| Other Request P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, scheme, operation) (rate(opendal_operation_duration_seconds_bucket{instance=~"$datanode", operation!~"read\|write\|list"}[$__rate_interval])))` | `timeseries` | Other Request P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]` |
+| Opendal traffic | `sum by(instance, pod, scheme, operation) (rate(opendal_operation_bytes_sum{instance=~"$datanode"}[$__rate_interval]))` | `timeseries` | Total traffic as in bytes by instance and operation | `prometheus` | `decbytes` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]` |
+| OpenDAL errors per Instance | `sum by(instance, pod, scheme, operation, error) (rate(opendal_operation_errors_total{instance=~"$datanode", error!="NotFound"}[$__rate_interval]))` | `timeseries` | OpenDAL error counts per Instance. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]-[{{error}}]` |
+# Metasrv
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Region migration datanode | `greptime_meta_region_migration_stat{datanode_type="src"}`<br/>`greptime_meta_region_migration_stat{datanode_type="desc"}` | `state-timeline` | Counter of region migration by source and destination | `prometheus` | `none` | `from-datanode-{{datanode_id}}` |
+| Region migration error | `greptime_meta_region_migration_error` | `timeseries` | Counter of region migration error | `prometheus` | `none` | `__auto` |
+| Datanode load | `greptime_datanode_load` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `none` | `__auto` |
+# Flownode
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Flow Ingest / Output Rate | `sum by(instance, pod, direction) (rate(greptime_flow_processed_rows[$__rate_interval]))` | `timeseries` | Flow Ingest / Output Rate. | `prometheus` | -- | `[{{pod}}]-[{{instance}}]-[{{direction}}]` |
+| Flow Ingest Latency | `histogram_quantile(0.95, sum(rate(greptime_flow_insert_elapsed_bucket[$__rate_interval])) by (le, instance, pod))`<br/>`histogram_quantile(0.99, sum(rate(greptime_flow_insert_elapsed_bucket[$__rate_interval])) by (le, instance, pod))` | `timeseries` | Flow Ingest Latency. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]-p95` |
+| Flow Operation Latency | `histogram_quantile(0.95, sum(rate(greptime_flow_processing_time_bucket[$__rate_interval])) by (le,instance,pod,type))`<br/>`histogram_quantile(0.99, sum(rate(greptime_flow_processing_time_bucket[$__rate_interval])) by (le,instance,pod,type))` | `timeseries` | Flow Operation Latency. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]-[{{type}}]-p95` |
+| Flow Buffer Size per Instance | `greptime_flow_input_buf_size` | `timeseries` | Flow Buffer Size per Instance. | `prometheus` | -- | `[{{instance}}]-[{{pod}]` |
+| Flow Processing Error per Instance | `sum by(instance,pod,code) (rate(greptime_flow_errors[$__rate_interval]))` | `timeseries` | Flow Processing Error per Instance. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]-[{{code}}]` |

grafana/dashboards/cluster/dashboard.yaml

@@ -0,0 +1,769 @@
+groups:
+    - title: Overview
+      panels:
+        - title: Uptime
+          type: stat
+          description: The start time of GreptimeDB.
+          unit: s
+          queries:
+            - expr: time() - process_start_time_seconds
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: __auto
+        - title: Version
+          type: stat
+          description: GreptimeDB version.
+          queries:
+            - expr: SELECT pkg_version FROM information_schema.build_info
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Total Ingestion Rate
+          type: stat
+          description: Total ingestion rate.
+          unit: rowsps
+          queries:
+            - expr: sum(rate(greptime_table_operator_ingest_rows[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: __auto
+        - title: Total Storage Size
+          type: stat
+          description: Total number of data file size.
+          unit: decbytes
+          queries:
+            - expr: select SUM(disk_size) from information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Total Rows
+          type: stat
+          description: Total number of data rows in the cluster. Calculated by sum of rows from each region.
+          unit: sishort
+          queries:
+            - expr: select SUM(region_rows) from information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Deployment
+          type: stat
+          description: The deployment topology of GreptimeDB.
+          queries:
+            - expr: SELECT count(*) as datanode FROM information_schema.cluster_info WHERE peer_type = 'DATANODE';
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT count(*) as frontend FROM information_schema.cluster_info WHERE peer_type = 'FRONTEND';
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT count(*) as metasrv FROM information_schema.cluster_info WHERE peer_type = 'METASRV';
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT count(*) as flownode FROM information_schema.cluster_info WHERE peer_type = 'FLOWNODE';
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Database Resources
+          type: stat
+          description: The number of the key resources in GreptimeDB.
+          queries:
+            - expr: SELECT COUNT(*) as databases FROM information_schema.schemata WHERE schema_name NOT IN ('greptime_private', 'information_schema')
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT COUNT(*) as tables FROM information_schema.tables WHERE table_schema != 'information_schema'
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT COUNT(region_id) as regions FROM information_schema.region_peers
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT COUNT(*) as flows FROM information_schema.flows
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Data Size
+          type: stat
+          description: The data size of wal/index/manifest in the GreptimeDB.
+          unit: decbytes
+          queries:
+            - expr: SELECT SUM(memtable_size) * 0.42825 as WAL FROM information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT SUM(index_size) as index FROM information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT SUM(manifest_size) as manifest FROM information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+    - title: Ingestion
+      panels:
+        - title: Total Ingestion Rate
+          type: timeseries
+          description: |
+            Total ingestion rate.
+
+            Here we listed 3 primary protocols:
+
+            - Prometheus remote write
+            - Greptime's gRPC API (when using our ingest SDK)
+            - Log ingestion http API
+          unit: rowsps
+          queries:
+            - expr: sum(rate(greptime_table_operator_ingest_rows{instance=~"$frontend"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: ingestion
+        - title: Ingestion Rate by Type
+          type: timeseries
+          description: |
+            Total ingestion rate.
+
+            Here we listed 3 primary protocols:
+
+            - Prometheus remote write
+            - Greptime's gRPC API (when using our ingest SDK)
+            - Log ingestion http API
+          unit: rowsps
+          queries:
+            - expr: sum(rate(greptime_servers_http_logs_ingestion_counter[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: http-logs
+            - expr: sum(rate(greptime_servers_prometheus_remote_write_samples[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: prometheus-remote-write
+    - title: Queries
+      panels:
+        - title: Total Query Rate
+          type: timeseries
+          description: |-
+            Total rate of query API calls by protocol. This metric is collected from frontends.
+
+            Here we listed 3 main protocols:
+            - MySQL
+            - Postgres
+            - Prometheus API
+
+            Note that there are some other minor query APIs like /sql are not included
+          unit: reqps
+          queries:
+            - expr: sum (rate(greptime_servers_mysql_query_elapsed_count{instance=~"$frontend"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: mysql
+            - expr: sum (rate(greptime_servers_postgres_query_elapsed_count{instance=~"$frontend"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: pg
+            - expr: sum (rate(greptime_servers_http_promql_elapsed_counte{instance=~"$frontend"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: promql
+    - title: Resources
+      panels:
+        - title: Datanode Memory per Instance
+          type: timeseries
+          description: Current memory usage by instance
+          unit: decbytes
+          queries:
+            - expr: sum(process_resident_memory_bytes{instance=~"$datanode"}) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{ pod }}]'
+        - title: Datanode CPU Usage per Instance
+          type: timeseries
+          description: Current cpu usage by instance
+          unit: none
+          queries:
+            - expr: sum(rate(process_cpu_seconds_total{instance=~"$datanode"}[$__rate_interval]) * 1000) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+        - title: Frontend Memory per Instance
+          type: timeseries
+          description: Current memory usage by instance
+          unit: decbytes
+          queries:
+            - expr: sum(process_resident_memory_bytes{instance=~"$frontend"}) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+        - title: Frontend CPU Usage per Instance
+          type: timeseries
+          description: Current cpu usage by instance
+          unit: none
+          queries:
+            - expr: sum(rate(process_cpu_seconds_total{instance=~"$frontend"}[$__rate_interval]) * 1000) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]-cpu'
+        - title: Metasrv Memory per Instance
+          type: timeseries
+          description: Current memory usage by instance
+          unit: decbytes
+          queries:
+            - expr: sum(process_resident_memory_bytes{instance=~"$metasrv"}) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]-resident'
+        - title: Metasrv CPU Usage per Instance
+          type: timeseries
+          description: Current cpu usage by instance
+          unit: none
+          queries:
+            - expr: sum(rate(process_cpu_seconds_total{instance=~"$metasrv"}[$__rate_interval]) * 1000) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+        - title: Flownode Memory per Instance
+          type: timeseries
+          description: Current memory usage by instance
+          unit: decbytes
+          queries:
+            - expr: sum(process_resident_memory_bytes{instance=~"$flownode"}) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+        - title: Flownode CPU Usage per Instance
+          type: timeseries
+          description: Current cpu usage by instance
+          unit: none
+          queries:
+            - expr: sum(rate(process_cpu_seconds_total{instance=~"$flownode"}[$__rate_interval]) * 1000) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+    - title: Frontend Requests
+      panels:
+        - title: HTTP QPS per Instance
+          type: timeseries
+          description: HTTP QPS per Instance.
+          unit: reqps
+          queries:
+            - expr: sum by(instance, pod, path, method, code) (rate(greptime_servers_http_requests_elapsed_count{instance=~"$frontend",path!~"/health|/metrics"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]'
+        - title: HTTP P99 per Instance
+          type: timeseries
+          description: HTTP P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, path, method, code) (rate(greptime_servers_http_requests_elapsed_bucket{instance=~"$frontend",path!~"/health|/metrics"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]-p99'
+        - title: gRPC QPS per Instance
+          type: timeseries
+          description: gRPC QPS per Instance.
+          unit: reqps
+          queries:
+            - expr: sum by(instance, pod, path, code) (rate(greptime_servers_grpc_requests_elapsed_count{instance=~"$frontend"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{path}}]-[{{code}}]'
+        - title: gRPC P99 per Instance
+          type: timeseries
+          description: gRPC P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, path, code) (rate(greptime_servers_grpc_requests_elapsed_bucket{instance=~"$frontend"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]-p99'
+        - title: MySQL QPS per Instance
+          type: timeseries
+          description: MySQL QPS per Instance.
+          unit: reqps
+          queries:
+            - expr: sum by(pod, instance)(rate(greptime_servers_mysql_query_elapsed_count{instance=~"$frontend"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: MySQL P99 per Instance
+          type: timeseries
+          description: MySQL P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(pod, instance, le) (rate(greptime_servers_mysql_query_elapsed_bucket{instance=~"$frontend"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]-p99'
+        - title: PostgreSQL QPS per Instance
+          type: timeseries
+          description: PostgreSQL QPS per Instance.
+          unit: reqps
+          queries:
+            - expr: sum by(pod, instance)(rate(greptime_servers_postgres_query_elapsed_count{instance=~"$frontend"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: PostgreSQL P99 per Instance
+          type: timeseries
+          description: PostgreSQL P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(pod,instance,le) (rate(greptime_servers_postgres_query_elapsed_bucket{instance=~"$frontend"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-p99'
+    - title: Frontend to Datanode
+      panels:
+        - title: Ingest Rows per Instance
+          type: timeseries
+          description: Ingestion rate by row as in each frontend
+          unit: rowsps
+          queries:
+            - expr: sum by(instance, pod)(rate(greptime_table_operator_ingest_rows{instance=~"$frontend"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Region Call QPS per Instance
+          type: timeseries
+          description: Region Call QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, request_type) (rate(greptime_grpc_region_request_count{instance=~"$frontend"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{request_type}}]'
+        - title: Region Call P99 per Instance
+          type: timeseries
+          description: Region Call P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, request_type) (rate(greptime_grpc_region_request_bucket{instance=~"$frontend"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{request_type}}]'
+    - title: Mito Engine
+      panels:
+        - title: Request OPS per Instance
+          type: timeseries
+          description: Request QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, type) (rate(greptime_mito_handle_request_elapsed_count{instance=~"$datanode"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]'
+        - title: Request P99 per Instance
+          type: timeseries
+          description: Request P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, type) (rate(greptime_mito_handle_request_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]'
+        - title: Write Buffer per Instance
+          type: timeseries
+          description: Write Buffer per Instance.
+          unit: decbytes
+          queries:
+            - expr: greptime_mito_write_buffer_bytes{instance=~"$datanode"}
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Write Rows per Instance
+          type: timeseries
+          description: Ingestion size by row counts.
+          unit: rowsps
+          queries:
+            - expr: sum by (instance, pod) (rate(greptime_mito_write_rows_total{instance=~"$datanode"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Flush OPS per Instance
+          type: timeseries
+          description: Flush QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, reason) (rate(greptime_mito_flush_requests_total{instance=~"$datanode"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{reason}}]'
+        - title: Write Stall per Instance
+          type: timeseries
+          description: Write Stall per Instance.
+          queries:
+            - expr: sum by(instance, pod) (greptime_mito_write_stall_total{instance=~"$datanode"})
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Read Stage OPS per Instance
+          type: timeseries
+          description: Read Stage OPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod) (rate(greptime_mito_read_stage_elapsed_count{instance=~"$datanode", stage="total"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Read Stage P99 per Instance
+          type: timeseries
+          description: Read Stage P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_read_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]'
+        - title: Write Stage P99 per Instance
+          type: timeseries
+          description: Write Stage P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_write_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]'
+        - title: Compaction OPS per Instance
+          type: timeseries
+          description: Compaction OPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod) (rate(greptime_mito_compaction_total_elapsed_count{instance=~"$datanode"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{pod}}]'
+        - title: Compaction P99 per Instance by Stage
+          type: timeseries
+          description: Compaction latency by stage
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]-p99'
+        - title: Compaction P99 per Instance
+          type: timeseries
+          description: Compaction P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le,stage) (rate(greptime_mito_compaction_total_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]-compaction'
+        - title: WAL write size
+          type: timeseries
+          description: Write-ahead logs write size as bytes. This chart includes stats of p95 and p99 size by instance, total WAL write rate.
+          unit: bytes
+          queries:
+            - expr: histogram_quantile(0.95, sum by(le,instance, pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-req-size-p95'
+            - expr: histogram_quantile(0.99, sum by(le,instance,pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-req-size-p99'
+            - expr: sum by (instance, pod)(rate(raft_engine_write_size_sum[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-throughput'
+        - title: Cached Bytes per Instance
+          type: timeseries
+          description: Cached Bytes per Instance.
+          unit: decbytes
+          queries:
+            - expr: greptime_mito_cache_bytes{instance=~"$datanode"}
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]'
+        - title: Inflight Compaction
+          type: timeseries
+          description: Ongoing compaction task count
+          unit: none
+          queries:
+            - expr: greptime_mito_inflight_compaction_count
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: WAL sync duration seconds
+          type: timeseries
+          description: Raft engine (local disk) log store sync latency, p99
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(le, type, node, instance, pod) (rate(raft_engine_sync_log_duration_seconds_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-p99'
+        - title: Log Store op duration seconds
+          type: timeseries
+          description: Write-ahead log operations latency at p99
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(le,logstore,optype,instance, pod) (rate(greptime_logstore_op_elapsed_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{logstore}}]-[{{optype}}]-p99'
+        - title: Inflight Flush
+          type: timeseries
+          description: Ongoing flush task count
+          unit: none
+          queries:
+            - expr: greptime_mito_inflight_flush_count
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+    - title: OpenDAL
+      panels:
+        - title: QPS per Instance
+          type: timeseries
+          description: QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme, operation) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]'
+        - title: Read QPS per Instance
+          type: timeseries
+          description: Read QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode", operation="read"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]'
+        - title: Read P99 per Instance
+          type: timeseries
+          description: Read P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{instance=~"$datanode",operation="read"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-{{scheme}}'
+        - title: Write QPS per Instance
+          type: timeseries
+          description: Write QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode", operation="write"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-{{scheme}}'
+        - title: Write P99 per Instance
+          type: timeseries
+          description: Write P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{instance=~"$datanode", operation="write"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]'
+        - title: List QPS per Instance
+          type: timeseries
+          description: List QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode", operation="list"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]'
+        - title: List P99 per Instance
+          type: timeseries
+          description: List P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{instance=~"$datanode", operation="list"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]'
+        - title: Other Requests per Instance
+          type: timeseries
+          description: Other Requests per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme, operation) (rate(opendal_operation_duration_seconds_count{instance=~"$datanode",operation!~"read|write|list|stat"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]'
+        - title: Other Request P99 per Instance
+          type: timeseries
+          description: Other Request P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, scheme, operation) (rate(opendal_operation_duration_seconds_bucket{instance=~"$datanode", operation!~"read|write|list"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]'
+        - title: Opendal traffic
+          type: timeseries
+          description: Total traffic as in bytes by instance and operation
+          unit: decbytes
+          queries:
+            - expr: sum by(instance, pod, scheme, operation) (rate(opendal_operation_bytes_sum{instance=~"$datanode"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]'
+        - title: OpenDAL errors per Instance
+          type: timeseries
+          description: OpenDAL error counts per Instance.
+          queries:
+            - expr: sum by(instance, pod, scheme, operation, error) (rate(opendal_operation_errors_total{instance=~"$datanode", error!="NotFound"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]-[{{error}}]'
+    - title: Metasrv
+      panels:
+        - title: Region migration datanode
+          type: state-timeline
+          description: Counter of region migration by source and destination
+          unit: none
+          queries:
+            - expr: greptime_meta_region_migration_stat{datanode_type="src"}
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: from-datanode-{{datanode_id}}
+            - expr: greptime_meta_region_migration_stat{datanode_type="desc"}
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: to-datanode-{{datanode_id}}
+        - title: Region migration error
+          type: timeseries
+          description: Counter of region migration error
+          unit: none
+          queries:
+            - expr: greptime_meta_region_migration_error
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: __auto
+        - title: Datanode load
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: none
+          queries:
+            - expr: greptime_datanode_load
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: __auto
+    - title: Flownode
+      panels:
+        - title: Flow Ingest / Output Rate
+          type: timeseries
+          description: Flow Ingest / Output Rate.
+          queries:
+            - expr: sum by(instance, pod, direction) (rate(greptime_flow_processed_rows[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{pod}}]-[{{instance}}]-[{{direction}}]'
+        - title: Flow Ingest Latency
+          type: timeseries
+          description: Flow Ingest Latency.
+          queries:
+            - expr: histogram_quantile(0.95, sum(rate(greptime_flow_insert_elapsed_bucket[$__rate_interval])) by (le, instance, pod))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-p95'
+            - expr: histogram_quantile(0.99, sum(rate(greptime_flow_insert_elapsed_bucket[$__rate_interval])) by (le, instance, pod))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-p99'
+        - title: Flow Operation Latency
+          type: timeseries
+          description: Flow Operation Latency.
+          queries:
+            - expr: histogram_quantile(0.95, sum(rate(greptime_flow_processing_time_bucket[$__rate_interval])) by (le,instance,pod,type))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]-p95'
+            - expr: histogram_quantile(0.99, sum(rate(greptime_flow_processing_time_bucket[$__rate_interval])) by (le,instance,pod,type))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]-p99'
+        - title: Flow Buffer Size per Instance
+          type: timeseries
+          description: Flow Buffer Size per Instance.
+          queries:
+            - expr: greptime_flow_input_buf_size
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}]'
+        - title: Flow Processing Error per Instance
+          type: timeseries
+          description: Flow Processing Error per Instance.
+          queries:
+            - expr: sum by(instance,pod,code) (rate(greptime_flow_errors[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{code}}]'

grafana/dashboards/standalone/dashboard.md

@@ -0,0 +1,97 @@
+# Overview
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Uptime | `time() - process_start_time_seconds` | `stat` | The start time of GreptimeDB. | `prometheus` | `s` | `__auto` |
+| Version | `SELECT pkg_version FROM information_schema.build_info` | `stat` | GreptimeDB version. | `mysql` | -- | -- |
+| Total Ingestion Rate | `sum(rate(greptime_table_operator_ingest_rows[$__rate_interval]))` | `stat` | Total ingestion rate. | `prometheus` | `rowsps` | `__auto` |
+| Total Storage Size | `select SUM(disk_size) from information_schema.region_statistics;` | `stat` | Total number of data file size. | `mysql` | `decbytes` | -- |
+| Total Rows | `select SUM(region_rows) from information_schema.region_statistics;` | `stat` | Total number of data rows in the cluster. Calculated by sum of rows from each region. | `mysql` | `sishort` | -- |
+| Deployment | `SELECT count(*) as datanode FROM information_schema.cluster_info WHERE peer_type = 'DATANODE';`<br/>`SELECT count(*) as frontend FROM information_schema.cluster_info WHERE peer_type = 'FRONTEND';`<br/>`SELECT count(*) as metasrv FROM information_schema.cluster_info WHERE peer_type = 'METASRV';`<br/>`SELECT count(*) as flownode FROM information_schema.cluster_info WHERE peer_type = 'FLOWNODE';` | `stat` | The deployment topology of GreptimeDB. | `mysql` | -- | -- |
+| Database Resources | `SELECT COUNT(*) as databases FROM information_schema.schemata WHERE schema_name NOT IN ('greptime_private', 'information_schema')`<br/>`SELECT COUNT(*) as tables FROM information_schema.tables WHERE table_schema != 'information_schema'`<br/>`SELECT COUNT(region_id) as regions FROM information_schema.region_peers`<br/>`SELECT COUNT(*) as flows FROM information_schema.flows` | `stat` | The number of the key resources in GreptimeDB. | `mysql` | -- | -- |
+| Data Size | `SELECT SUM(memtable_size) * 0.42825 as WAL FROM information_schema.region_statistics;`<br/>`SELECT SUM(index_size) as index FROM information_schema.region_statistics;`<br/>`SELECT SUM(manifest_size) as manifest FROM information_schema.region_statistics;` | `stat` | The data size of wal/index/manifest in the GreptimeDB. | `mysql` | `decbytes` | -- |
+# Ingestion
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Total Ingestion Rate | `sum(rate(greptime_table_operator_ingest_rows{}[$__rate_interval]))` | `timeseries` | Total ingestion rate.<br/><br/>Here we listed 3 primary protocols:<br/><br/>- Prometheus remote write<br/>- Greptime's gRPC API (when using our ingest SDK)<br/>- Log ingestion http API<br/> | `prometheus` | `rowsps` | `ingestion` |
+| Ingestion Rate by Type | `sum(rate(greptime_servers_http_logs_ingestion_counter[$__rate_interval]))`<br/>`sum(rate(greptime_servers_prometheus_remote_write_samples[$__rate_interval]))` | `timeseries` | Total ingestion rate.<br/><br/>Here we listed 3 primary protocols:<br/><br/>- Prometheus remote write<br/>- Greptime's gRPC API (when using our ingest SDK)<br/>- Log ingestion http API<br/> | `prometheus` | `rowsps` | `http-logs` |
+# Queries
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Total Query Rate | `sum (rate(greptime_servers_mysql_query_elapsed_count{}[$__rate_interval]))`<br/>`sum (rate(greptime_servers_postgres_query_elapsed_count{}[$__rate_interval]))`<br/>`sum (rate(greptime_servers_http_promql_elapsed_counte{}[$__rate_interval]))` | `timeseries` | Total rate of query API calls by protocol. This metric is collected from frontends.<br/><br/>Here we listed 3 main protocols:<br/>- MySQL<br/>- Postgres<br/>- Prometheus API<br/><br/>Note that there are some other minor query APIs like /sql are not included | `prometheus` | `reqps` | `mysql` |
+# Resources
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Datanode Memory per Instance | `sum(process_resident_memory_bytes{}) by (instance, pod)` | `timeseries` | Current memory usage by instance | `prometheus` | `decbytes` | `[{{instance}}]-[{{ pod }}]` |
+| Datanode CPU Usage per Instance | `sum(rate(process_cpu_seconds_total{}[$__rate_interval]) * 1000) by (instance, pod)` | `timeseries` | Current cpu usage by instance | `prometheus` | `none` | `[{{ instance }}]-[{{ pod }}]` |
+| Frontend Memory per Instance | `sum(process_resident_memory_bytes{}) by (instance, pod)` | `timeseries` | Current memory usage by instance | `prometheus` | `decbytes` | `[{{ instance }}]-[{{ pod }}]` |
+| Frontend CPU Usage per Instance | `sum(rate(process_cpu_seconds_total{}[$__rate_interval]) * 1000) by (instance, pod)` | `timeseries` | Current cpu usage by instance | `prometheus` | `none` | `[{{ instance }}]-[{{ pod }}]-cpu` |
+| Metasrv Memory per Instance | `sum(process_resident_memory_bytes{}) by (instance, pod)` | `timeseries` | Current memory usage by instance | `prometheus` | `decbytes` | `[{{ instance }}]-[{{ pod }}]-resident` |
+| Metasrv CPU Usage per Instance | `sum(rate(process_cpu_seconds_total{}[$__rate_interval]) * 1000) by (instance, pod)` | `timeseries` | Current cpu usage by instance | `prometheus` | `none` | `[{{ instance }}]-[{{ pod }}]` |
+| Flownode Memory per Instance | `sum(process_resident_memory_bytes{}) by (instance, pod)` | `timeseries` | Current memory usage by instance | `prometheus` | `decbytes` | `[{{ instance }}]-[{{ pod }}]` |
+| Flownode CPU Usage per Instance | `sum(rate(process_cpu_seconds_total{}[$__rate_interval]) * 1000) by (instance, pod)` | `timeseries` | Current cpu usage by instance | `prometheus` | `none` | `[{{ instance }}]-[{{ pod }}]` |
+# Frontend Requests
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| HTTP QPS per Instance | `sum by(instance, pod, path, method, code) (rate(greptime_servers_http_requests_elapsed_count{path!~"/health\|/metrics"}[$__rate_interval]))` | `timeseries` | HTTP QPS per Instance. | `prometheus` | `reqps` | `[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]` |
+| HTTP P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, path, method, code) (rate(greptime_servers_http_requests_elapsed_bucket{path!~"/health\|/metrics"}[$__rate_interval])))` | `timeseries` | HTTP P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]-p99` |
+| gRPC QPS per Instance | `sum by(instance, pod, path, code) (rate(greptime_servers_grpc_requests_elapsed_count{}[$__rate_interval]))` | `timeseries` | gRPC QPS per Instance. | `prometheus` | `reqps` | `[{{instance}}]-[{{pod}}]-[{{path}}]-[{{code}}]` |
+| gRPC P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, path, code) (rate(greptime_servers_grpc_requests_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | gRPC P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]-p99` |
+| MySQL QPS per Instance | `sum by(pod, instance)(rate(greptime_servers_mysql_query_elapsed_count{}[$__rate_interval]))` | `timeseries` | MySQL QPS per Instance. | `prometheus` | `reqps` | `[{{instance}}]-[{{pod}}]` |
+| MySQL P99 per Instance | `histogram_quantile(0.99, sum by(pod, instance, le) (rate(greptime_servers_mysql_query_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | MySQL P99 per Instance. | `prometheus` | `s` | `[{{ instance }}]-[{{ pod }}]-p99` |
+| PostgreSQL QPS per Instance | `sum by(pod, instance)(rate(greptime_servers_postgres_query_elapsed_count{}[$__rate_interval]))` | `timeseries` | PostgreSQL QPS per Instance. | `prometheus` | `reqps` | `[{{instance}}]-[{{pod}}]` |
+| PostgreSQL P99 per Instance | `histogram_quantile(0.99, sum by(pod,instance,le) (rate(greptime_servers_postgres_query_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | PostgreSQL P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-p99` |
+# Frontend to Datanode
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Ingest Rows per Instance | `sum by(instance, pod)(rate(greptime_table_operator_ingest_rows{}[$__rate_interval]))` | `timeseries` | Ingestion rate by row as in each frontend | `prometheus` | `rowsps` | `[{{instance}}]-[{{pod}}]` |
+| Region Call QPS per Instance | `sum by(instance, pod, request_type) (rate(greptime_grpc_region_request_count{}[$__rate_interval]))` | `timeseries` | Region Call QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{request_type}}]` |
+| Region Call P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, request_type) (rate(greptime_grpc_region_request_bucket{}[$__rate_interval])))` | `timeseries` | Region Call P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{request_type}}]` |
+# Mito Engine
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Request OPS per Instance | `sum by(instance, pod, type) (rate(greptime_mito_handle_request_elapsed_count{}[$__rate_interval]))` | `timeseries` | Request QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{type}}]` |
+| Request P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, type) (rate(greptime_mito_handle_request_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | Request P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{type}}]` |
+| Write Buffer per Instance | `greptime_mito_write_buffer_bytes{}` | `timeseries` | Write Buffer per Instance. | `prometheus` | `decbytes` | `[{{instance}}]-[{{pod}}]` |
+| Write Rows per Instance | `sum by (instance, pod) (rate(greptime_mito_write_rows_total{}[$__rate_interval]))` | `timeseries` | Ingestion size by row counts. | `prometheus` | `rowsps` | `[{{instance}}]-[{{pod}}]` |
+| Flush OPS per Instance | `sum by(instance, pod, reason) (rate(greptime_mito_flush_requests_total{}[$__rate_interval]))` | `timeseries` | Flush QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{reason}}]` |
+| Write Stall per Instance | `sum by(instance, pod) (greptime_mito_write_stall_total{})` | `timeseries` | Write Stall per Instance. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]` |
+| Read Stage OPS per Instance | `sum by(instance, pod) (rate(greptime_mito_read_stage_elapsed_count{ stage="total"}[$__rate_interval]))` | `timeseries` | Read Stage OPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]` |
+| Read Stage P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_read_stage_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | Read Stage P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]` |
+| Write Stage P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_write_stage_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | Write Stage P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]` |
+| Compaction OPS per Instance | `sum by(instance, pod) (rate(greptime_mito_compaction_total_elapsed_count{}[$__rate_interval]))` | `timeseries` | Compaction OPS per Instance. | `prometheus` | `ops` | `[{{ instance }}]-[{{pod}}]` |
+| Compaction P99 per Instance by Stage | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | Compaction latency by stage | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-p99` |
+| Compaction P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le,stage) (rate(greptime_mito_compaction_total_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | Compaction P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-compaction` |
+| WAL write size | `histogram_quantile(0.95, sum by(le,instance, pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))`<br/>`histogram_quantile(0.99, sum by(le,instance,pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))`<br/>`sum by (instance, pod)(rate(raft_engine_write_size_sum[$__rate_interval]))` | `timeseries` | Write-ahead logs write size as bytes. This chart includes stats of p95 and p99 size by instance, total WAL write rate. | `prometheus` | `bytes` | `[{{instance}}]-[{{pod}}]-req-size-p95` |
+| Cached Bytes per Instance | `greptime_mito_cache_bytes{}` | `timeseries` | Cached Bytes per Instance. | `prometheus` | `decbytes` | `[{{instance}}]-[{{pod}}]-[{{type}}]` |
+| Inflight Compaction | `greptime_mito_inflight_compaction_count` | `timeseries` | Ongoing compaction task count | `prometheus` | `none` | `[{{instance}}]-[{{pod}}]` |
+| WAL sync duration seconds | `histogram_quantile(0.99, sum by(le, type, node, instance, pod) (rate(raft_engine_sync_log_duration_seconds_bucket[$__rate_interval])))` | `timeseries` | Raft engine (local disk) log store sync latency, p99 | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-p99` |
+| Log Store op duration seconds | `histogram_quantile(0.99, sum by(le,logstore,optype,instance, pod) (rate(greptime_logstore_op_elapsed_bucket[$__rate_interval])))` | `timeseries` | Write-ahead log operations latency at p99 | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{logstore}}]-[{{optype}}]-p99` |
+| Inflight Flush | `greptime_mito_inflight_flush_count` | `timeseries` | Ongoing flush task count | `prometheus` | `none` | `[{{instance}}]-[{{pod}}]` |
+# OpenDAL
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| QPS per Instance | `sum by(instance, pod, scheme, operation) (rate(opendal_operation_duration_seconds_count{}[$__rate_interval]))` | `timeseries` | QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]` |
+| Read QPS per Instance | `sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{ operation="read"}[$__rate_interval]))` | `timeseries` | Read QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]` |
+| Read P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{operation="read"}[$__rate_interval])))` | `timeseries` | Read P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-{{scheme}}` |
+| Write QPS per Instance | `sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{ operation="write"}[$__rate_interval]))` | `timeseries` | Write QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-{{scheme}}` |
+| Write P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{ operation="write"}[$__rate_interval])))` | `timeseries` | Write P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]` |
+| List QPS per Instance | `sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{ operation="list"}[$__rate_interval]))` | `timeseries` | List QPS per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]` |
+| List P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{ operation="list"}[$__rate_interval])))` | `timeseries` | List P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]` |
+| Other Requests per Instance | `sum by(instance, pod, scheme, operation) (rate(opendal_operation_duration_seconds_count{operation!~"read\|write\|list\|stat"}[$__rate_interval]))` | `timeseries` | Other Requests per Instance. | `prometheus` | `ops` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]` |
+| Other Request P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, scheme, operation) (rate(opendal_operation_duration_seconds_bucket{ operation!~"read\|write\|list"}[$__rate_interval])))` | `timeseries` | Other Request P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]` |
+| Opendal traffic | `sum by(instance, pod, scheme, operation) (rate(opendal_operation_bytes_sum{}[$__rate_interval]))` | `timeseries` | Total traffic as in bytes by instance and operation | `prometheus` | `decbytes` | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]` |
+| OpenDAL errors per Instance | `sum by(instance, pod, scheme, operation, error) (rate(opendal_operation_errors_total{ error!="NotFound"}[$__rate_interval]))` | `timeseries` | OpenDAL error counts per Instance. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]-[{{error}}]` |
+# Metasrv
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Region migration datanode | `greptime_meta_region_migration_stat{datanode_type="src"}`<br/>`greptime_meta_region_migration_stat{datanode_type="desc"}` | `state-timeline` | Counter of region migration by source and destination | `prometheus` | `none` | `from-datanode-{{datanode_id}}` |
+| Region migration error | `greptime_meta_region_migration_error` | `timeseries` | Counter of region migration error | `prometheus` | `none` | `__auto` |
+| Datanode load | `greptime_datanode_load` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `none` | `__auto` |
+# Flownode
+| Title | Query | Type | Description | Datasource | Unit | Legend Format |
+| --- | --- | --- | --- | --- | --- | --- |
+| Flow Ingest / Output Rate | `sum by(instance, pod, direction) (rate(greptime_flow_processed_rows[$__rate_interval]))` | `timeseries` | Flow Ingest / Output Rate. | `prometheus` | -- | `[{{pod}}]-[{{instance}}]-[{{direction}}]` |
+| Flow Ingest Latency | `histogram_quantile(0.95, sum(rate(greptime_flow_insert_elapsed_bucket[$__rate_interval])) by (le, instance, pod))`<br/>`histogram_quantile(0.99, sum(rate(greptime_flow_insert_elapsed_bucket[$__rate_interval])) by (le, instance, pod))` | `timeseries` | Flow Ingest Latency. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]-p95` |
+| Flow Operation Latency | `histogram_quantile(0.95, sum(rate(greptime_flow_processing_time_bucket[$__rate_interval])) by (le,instance,pod,type))`<br/>`histogram_quantile(0.99, sum(rate(greptime_flow_processing_time_bucket[$__rate_interval])) by (le,instance,pod,type))` | `timeseries` | Flow Operation Latency. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]-[{{type}}]-p95` |
+| Flow Buffer Size per Instance | `greptime_flow_input_buf_size` | `timeseries` | Flow Buffer Size per Instance. | `prometheus` | -- | `[{{instance}}]-[{{pod}]` |
+| Flow Processing Error per Instance | `sum by(instance,pod,code) (rate(greptime_flow_errors[$__rate_interval]))` | `timeseries` | Flow Processing Error per Instance. | `prometheus` | -- | `[{{instance}}]-[{{pod}}]-[{{code}}]` |

grafana/dashboards/standalone/dashboard.yaml

@@ -0,0 +1,769 @@
+groups:
+    - title: Overview
+      panels:
+        - title: Uptime
+          type: stat
+          description: The start time of GreptimeDB.
+          unit: s
+          queries:
+            - expr: time() - process_start_time_seconds
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: __auto
+        - title: Version
+          type: stat
+          description: GreptimeDB version.
+          queries:
+            - expr: SELECT pkg_version FROM information_schema.build_info
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Total Ingestion Rate
+          type: stat
+          description: Total ingestion rate.
+          unit: rowsps
+          queries:
+            - expr: sum(rate(greptime_table_operator_ingest_rows[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: __auto
+        - title: Total Storage Size
+          type: stat
+          description: Total number of data file size.
+          unit: decbytes
+          queries:
+            - expr: select SUM(disk_size) from information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Total Rows
+          type: stat
+          description: Total number of data rows in the cluster. Calculated by sum of rows from each region.
+          unit: sishort
+          queries:
+            - expr: select SUM(region_rows) from information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Deployment
+          type: stat
+          description: The deployment topology of GreptimeDB.
+          queries:
+            - expr: SELECT count(*) as datanode FROM information_schema.cluster_info WHERE peer_type = 'DATANODE';
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT count(*) as frontend FROM information_schema.cluster_info WHERE peer_type = 'FRONTEND';
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT count(*) as metasrv FROM information_schema.cluster_info WHERE peer_type = 'METASRV';
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT count(*) as flownode FROM information_schema.cluster_info WHERE peer_type = 'FLOWNODE';
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Database Resources
+          type: stat
+          description: The number of the key resources in GreptimeDB.
+          queries:
+            - expr: SELECT COUNT(*) as databases FROM information_schema.schemata WHERE schema_name NOT IN ('greptime_private', 'information_schema')
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT COUNT(*) as tables FROM information_schema.tables WHERE table_schema != 'information_schema'
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT COUNT(region_id) as regions FROM information_schema.region_peers
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT COUNT(*) as flows FROM information_schema.flows
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+        - title: Data Size
+          type: stat
+          description: The data size of wal/index/manifest in the GreptimeDB.
+          unit: decbytes
+          queries:
+            - expr: SELECT SUM(memtable_size) * 0.42825 as WAL FROM information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT SUM(index_size) as index FROM information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+            - expr: SELECT SUM(manifest_size) as manifest FROM information_schema.region_statistics;
+              datasource:
+                type: mysql
+                uid: ${information_schema}
+    - title: Ingestion
+      panels:
+        - title: Total Ingestion Rate
+          type: timeseries
+          description: |
+            Total ingestion rate.
+
+            Here we listed 3 primary protocols:
+
+            - Prometheus remote write
+            - Greptime's gRPC API (when using our ingest SDK)
+            - Log ingestion http API
+          unit: rowsps
+          queries:
+            - expr: sum(rate(greptime_table_operator_ingest_rows{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: ingestion
+        - title: Ingestion Rate by Type
+          type: timeseries
+          description: |
+            Total ingestion rate.
+
+            Here we listed 3 primary protocols:
+
+            - Prometheus remote write
+            - Greptime's gRPC API (when using our ingest SDK)
+            - Log ingestion http API
+          unit: rowsps
+          queries:
+            - expr: sum(rate(greptime_servers_http_logs_ingestion_counter[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: http-logs
+            - expr: sum(rate(greptime_servers_prometheus_remote_write_samples[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: prometheus-remote-write
+    - title: Queries
+      panels:
+        - title: Total Query Rate
+          type: timeseries
+          description: |-
+            Total rate of query API calls by protocol. This metric is collected from frontends.
+
+            Here we listed 3 main protocols:
+            - MySQL
+            - Postgres
+            - Prometheus API
+
+            Note that there are some other minor query APIs like /sql are not included
+          unit: reqps
+          queries:
+            - expr: sum (rate(greptime_servers_mysql_query_elapsed_count{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: mysql
+            - expr: sum (rate(greptime_servers_postgres_query_elapsed_count{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: pg
+            - expr: sum (rate(greptime_servers_http_promql_elapsed_counte{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: promql
+    - title: Resources
+      panels:
+        - title: Datanode Memory per Instance
+          type: timeseries
+          description: Current memory usage by instance
+          unit: decbytes
+          queries:
+            - expr: sum(process_resident_memory_bytes{}) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{ pod }}]'
+        - title: Datanode CPU Usage per Instance
+          type: timeseries
+          description: Current cpu usage by instance
+          unit: none
+          queries:
+            - expr: sum(rate(process_cpu_seconds_total{}[$__rate_interval]) * 1000) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+        - title: Frontend Memory per Instance
+          type: timeseries
+          description: Current memory usage by instance
+          unit: decbytes
+          queries:
+            - expr: sum(process_resident_memory_bytes{}) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+        - title: Frontend CPU Usage per Instance
+          type: timeseries
+          description: Current cpu usage by instance
+          unit: none
+          queries:
+            - expr: sum(rate(process_cpu_seconds_total{}[$__rate_interval]) * 1000) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]-cpu'
+        - title: Metasrv Memory per Instance
+          type: timeseries
+          description: Current memory usage by instance
+          unit: decbytes
+          queries:
+            - expr: sum(process_resident_memory_bytes{}) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]-resident'
+        - title: Metasrv CPU Usage per Instance
+          type: timeseries
+          description: Current cpu usage by instance
+          unit: none
+          queries:
+            - expr: sum(rate(process_cpu_seconds_total{}[$__rate_interval]) * 1000) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+        - title: Flownode Memory per Instance
+          type: timeseries
+          description: Current memory usage by instance
+          unit: decbytes
+          queries:
+            - expr: sum(process_resident_memory_bytes{}) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+        - title: Flownode CPU Usage per Instance
+          type: timeseries
+          description: Current cpu usage by instance
+          unit: none
+          queries:
+            - expr: sum(rate(process_cpu_seconds_total{}[$__rate_interval]) * 1000) by (instance, pod)
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]'
+    - title: Frontend Requests
+      panels:
+        - title: HTTP QPS per Instance
+          type: timeseries
+          description: HTTP QPS per Instance.
+          unit: reqps
+          queries:
+            - expr: sum by(instance, pod, path, method, code) (rate(greptime_servers_http_requests_elapsed_count{path!~"/health|/metrics"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]'
+        - title: HTTP P99 per Instance
+          type: timeseries
+          description: HTTP P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, path, method, code) (rate(greptime_servers_http_requests_elapsed_bucket{path!~"/health|/metrics"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]-p99'
+        - title: gRPC QPS per Instance
+          type: timeseries
+          description: gRPC QPS per Instance.
+          unit: reqps
+          queries:
+            - expr: sum by(instance, pod, path, code) (rate(greptime_servers_grpc_requests_elapsed_count{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{path}}]-[{{code}}]'
+        - title: gRPC P99 per Instance
+          type: timeseries
+          description: gRPC P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, path, code) (rate(greptime_servers_grpc_requests_elapsed_bucket{}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{path}}]-[{{method}}]-[{{code}}]-p99'
+        - title: MySQL QPS per Instance
+          type: timeseries
+          description: MySQL QPS per Instance.
+          unit: reqps
+          queries:
+            - expr: sum by(pod, instance)(rate(greptime_servers_mysql_query_elapsed_count{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: MySQL P99 per Instance
+          type: timeseries
+          description: MySQL P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(pod, instance, le) (rate(greptime_servers_mysql_query_elapsed_bucket{}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{ pod }}]-p99'
+        - title: PostgreSQL QPS per Instance
+          type: timeseries
+          description: PostgreSQL QPS per Instance.
+          unit: reqps
+          queries:
+            - expr: sum by(pod, instance)(rate(greptime_servers_postgres_query_elapsed_count{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: PostgreSQL P99 per Instance
+          type: timeseries
+          description: PostgreSQL P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(pod,instance,le) (rate(greptime_servers_postgres_query_elapsed_bucket{}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-p99'
+    - title: Frontend to Datanode
+      panels:
+        - title: Ingest Rows per Instance
+          type: timeseries
+          description: Ingestion rate by row as in each frontend
+          unit: rowsps
+          queries:
+            - expr: sum by(instance, pod)(rate(greptime_table_operator_ingest_rows{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Region Call QPS per Instance
+          type: timeseries
+          description: Region Call QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, request_type) (rate(greptime_grpc_region_request_count{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{request_type}}]'
+        - title: Region Call P99 per Instance
+          type: timeseries
+          description: Region Call P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, request_type) (rate(greptime_grpc_region_request_bucket{}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{request_type}}]'
+    - title: Mito Engine
+      panels:
+        - title: Request OPS per Instance
+          type: timeseries
+          description: Request QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, type) (rate(greptime_mito_handle_request_elapsed_count{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]'
+        - title: Request P99 per Instance
+          type: timeseries
+          description: Request P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, type) (rate(greptime_mito_handle_request_elapsed_bucket{}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]'
+        - title: Write Buffer per Instance
+          type: timeseries
+          description: Write Buffer per Instance.
+          unit: decbytes
+          queries:
+            - expr: greptime_mito_write_buffer_bytes{}
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Write Rows per Instance
+          type: timeseries
+          description: Ingestion size by row counts.
+          unit: rowsps
+          queries:
+            - expr: sum by (instance, pod) (rate(greptime_mito_write_rows_total{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Flush OPS per Instance
+          type: timeseries
+          description: Flush QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, reason) (rate(greptime_mito_flush_requests_total{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{reason}}]'
+        - title: Write Stall per Instance
+          type: timeseries
+          description: Write Stall per Instance.
+          queries:
+            - expr: sum by(instance, pod) (greptime_mito_write_stall_total{})
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Read Stage OPS per Instance
+          type: timeseries
+          description: Read Stage OPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod) (rate(greptime_mito_read_stage_elapsed_count{ stage="total"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: Read Stage P99 per Instance
+          type: timeseries
+          description: Read Stage P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_read_stage_elapsed_bucket{}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]'
+        - title: Write Stage P99 per Instance
+          type: timeseries
+          description: Write Stage P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_write_stage_elapsed_bucket{}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]'
+        - title: Compaction OPS per Instance
+          type: timeseries
+          description: Compaction OPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod) (rate(greptime_mito_compaction_total_elapsed_count{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{ instance }}]-[{{pod}}]'
+        - title: Compaction P99 per Instance by Stage
+          type: timeseries
+          description: Compaction latency by stage
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]-p99'
+        - title: Compaction P99 per Instance
+          type: timeseries
+          description: Compaction P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le,stage) (rate(greptime_mito_compaction_total_elapsed_bucket{}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]-compaction'
+        - title: WAL write size
+          type: timeseries
+          description: Write-ahead logs write size as bytes. This chart includes stats of p95 and p99 size by instance, total WAL write rate.
+          unit: bytes
+          queries:
+            - expr: histogram_quantile(0.95, sum by(le,instance, pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-req-size-p95'
+            - expr: histogram_quantile(0.99, sum by(le,instance,pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-req-size-p99'
+            - expr: sum by (instance, pod)(rate(raft_engine_write_size_sum[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-throughput'
+        - title: Cached Bytes per Instance
+          type: timeseries
+          description: Cached Bytes per Instance.
+          unit: decbytes
+          queries:
+            - expr: greptime_mito_cache_bytes{}
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]'
+        - title: Inflight Compaction
+          type: timeseries
+          description: Ongoing compaction task count
+          unit: none
+          queries:
+            - expr: greptime_mito_inflight_compaction_count
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+        - title: WAL sync duration seconds
+          type: timeseries
+          description: Raft engine (local disk) log store sync latency, p99
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(le, type, node, instance, pod) (rate(raft_engine_sync_log_duration_seconds_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-p99'
+        - title: Log Store op duration seconds
+          type: timeseries
+          description: Write-ahead log operations latency at p99
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(le,logstore,optype,instance, pod) (rate(greptime_logstore_op_elapsed_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{logstore}}]-[{{optype}}]-p99'
+        - title: Inflight Flush
+          type: timeseries
+          description: Ongoing flush task count
+          unit: none
+          queries:
+            - expr: greptime_mito_inflight_flush_count
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]'
+    - title: OpenDAL
+      panels:
+        - title: QPS per Instance
+          type: timeseries
+          description: QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme, operation) (rate(opendal_operation_duration_seconds_count{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]'
+        - title: Read QPS per Instance
+          type: timeseries
+          description: Read QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{ operation="read"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]'
+        - title: Read P99 per Instance
+          type: timeseries
+          description: Read P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{operation="read"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-{{scheme}}'
+        - title: Write QPS per Instance
+          type: timeseries
+          description: Write QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{ operation="write"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-{{scheme}}'
+        - title: Write P99 per Instance
+          type: timeseries
+          description: Write P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{ operation="write"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]'
+        - title: List QPS per Instance
+          type: timeseries
+          description: List QPS per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme) (rate(opendal_operation_duration_seconds_count{ operation="list"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]'
+        - title: List P99 per Instance
+          type: timeseries
+          description: List P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, scheme) (rate(opendal_operation_duration_seconds_bucket{ operation="list"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]'
+        - title: Other Requests per Instance
+          type: timeseries
+          description: Other Requests per Instance.
+          unit: ops
+          queries:
+            - expr: sum by(instance, pod, scheme, operation) (rate(opendal_operation_duration_seconds_count{operation!~"read|write|list|stat"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]'
+        - title: Other Request P99 per Instance
+          type: timeseries
+          description: Other Request P99 per Instance.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(instance, pod, le, scheme, operation) (rate(opendal_operation_duration_seconds_bucket{ operation!~"read|write|list"}[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]'
+        - title: Opendal traffic
+          type: timeseries
+          description: Total traffic as in bytes by instance and operation
+          unit: decbytes
+          queries:
+            - expr: sum by(instance, pod, scheme, operation) (rate(opendal_operation_bytes_sum{}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]'
+        - title: OpenDAL errors per Instance
+          type: timeseries
+          description: OpenDAL error counts per Instance.
+          queries:
+            - expr: sum by(instance, pod, scheme, operation, error) (rate(opendal_operation_errors_total{ error!="NotFound"}[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{scheme}}]-[{{operation}}]-[{{error}}]'
+    - title: Metasrv
+      panels:
+        - title: Region migration datanode
+          type: state-timeline
+          description: Counter of region migration by source and destination
+          unit: none
+          queries:
+            - expr: greptime_meta_region_migration_stat{datanode_type="src"}
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: from-datanode-{{datanode_id}}
+            - expr: greptime_meta_region_migration_stat{datanode_type="desc"}
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: to-datanode-{{datanode_id}}
+        - title: Region migration error
+          type: timeseries
+          description: Counter of region migration error
+          unit: none
+          queries:
+            - expr: greptime_meta_region_migration_error
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: __auto
+        - title: Datanode load
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: none
+          queries:
+            - expr: greptime_datanode_load
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: __auto
+    - title: Flownode
+      panels:
+        - title: Flow Ingest / Output Rate
+          type: timeseries
+          description: Flow Ingest / Output Rate.
+          queries:
+            - expr: sum by(instance, pod, direction) (rate(greptime_flow_processed_rows[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{pod}}]-[{{instance}}]-[{{direction}}]'
+        - title: Flow Ingest Latency
+          type: timeseries
+          description: Flow Ingest Latency.
+          queries:
+            - expr: histogram_quantile(0.95, sum(rate(greptime_flow_insert_elapsed_bucket[$__rate_interval])) by (le, instance, pod))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-p95'
+            - expr: histogram_quantile(0.99, sum(rate(greptime_flow_insert_elapsed_bucket[$__rate_interval])) by (le, instance, pod))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-p99'
+        - title: Flow Operation Latency
+          type: timeseries
+          description: Flow Operation Latency.
+          queries:
+            - expr: histogram_quantile(0.95, sum(rate(greptime_flow_processing_time_bucket[$__rate_interval])) by (le,instance,pod,type))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]-p95'
+            - expr: histogram_quantile(0.99, sum(rate(greptime_flow_processing_time_bucket[$__rate_interval])) by (le,instance,pod,type))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{type}}]-p99'
+        - title: Flow Buffer Size per Instance
+          type: timeseries
+          description: Flow Buffer Size per Instance.
+          queries:
+            - expr: greptime_flow_input_buf_size
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}]'
+        - title: Flow Processing Error per Instance
+          type: timeseries
+          description: Flow Processing Error per Instance.
+          queries:
+            - expr: sum by(instance,pod,code) (rate(greptime_flow_errors[$__rate_interval]))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '[{{instance}}]-[{{pod}}]-[{{code}}]'

grafana/scripts/check.sh

@@ -0,0 +1,54 @@
+#!/usr/bin/env bash
+
+DASHBOARD_DIR=${1:-grafana/dashboards}
+
+check_dashboard_description() {
+  for dashboard in $(find $DASHBOARD_DIR -name "*.json"); do
+    echo "Checking $dashboard description"
+
+    # Use jq to check for panels with empty or missing descriptions
+    invalid_panels=$(cat $dashboard | jq -r '
+      .panels[]
+    | select((.type == "stats" or .type == "timeseries") and (.description == "" or .description == null))')
+
+    # Check if any invalid panels were found
+    if [[ -n "$invalid_panels" ]]; then
+      echo "Error: The following panels have empty or missing descriptions:"
+      echo "$invalid_panels"
+      exit 1
+    else
+      echo "All panels with type 'stats' or 'timeseries' have valid descriptions."
+    fi
+  done
+}
+
+check_dashboards_generation() {
+  ./grafana/scripts/gen-dashboards.sh
+
+  if [[ -n "$(git diff --name-only grafana/dashboards)" ]]; then
+    echo "Error: The dashboards are not generated correctly. You should execute the `make dashboards` command."
+    exit 1
+  fi
+}
+
+check_datasource() {
+  for dashboard in $(find $DASHBOARD_DIR -name "*.json"); do
+    echo "Checking $dashboard datasource"
+    jq -r '.panels[] | select(.type != "row") | .targets[] | [.datasource.type, .datasource.uid] | @tsv' $dashboard | while read -r type uid; do
+    # if the datasource is prometheus, check if the uid is ${metrics}
+    if [[ "$type" == "prometheus" && "$uid" != "\${metrics}" ]]; then
+      echo "Error: The datasource uid of $dashboard is not valid. It should be \${metrics}, got $uid"
+      exit 1
+    fi
+    # if the datasource is mysql, check if the uid is ${information_schema}
+    if [[ "$type" == "mysql" && "$uid" != "\${information_schema}" ]]; then
+      echo "Error: The datasource uid of $dashboard is not valid. It should be \${information_schema}, got $uid"
+      exit 1
+    fi
+    done
+  done
+}
+
+check_dashboards_generation
+check_dashboard_description
+check_datasource

grafana/scripts/gen-dashboards.sh

@@ -0,0 +1,25 @@
+#! /usr/bin/env bash
+
+CLUSTER_DASHBOARD_DIR=${1:-grafana/dashboards/cluster}
+STANDALONE_DASHBOARD_DIR=${2:-grafana/dashboards/standalone}
+DAC_IMAGE=ghcr.io/zyy17/dac:20250423-522bd35
+
+remove_instance_filters() {
+  # Remove the instance filters for the standalone dashboards.
+  sed 's/instance=~\\"$datanode\\",//; s/instance=~\\"$datanode\\"//; s/instance=~\\"$frontend\\",//; s/instance=~\\"$frontend\\"//; s/instance=~\\"$metasrv\\",//; s/instance=~\\"$metasrv\\"//; s/instance=~\\"$flownode\\",//; s/instance=~\\"$flownode\\"//;' $CLUSTER_DASHBOARD_DIR/dashboard.json > $STANDALONE_DASHBOARD_DIR/dashboard.json
+}
+
+generate_intermediate_dashboards_and_docs() {
+  docker run -v ${PWD}:/greptimedb --rm ${DAC_IMAGE} \
+    -i /greptimedb/$CLUSTER_DASHBOARD_DIR/dashboard.json \
+    -o /greptimedb/$CLUSTER_DASHBOARD_DIR/dashboard.yaml \
+    -m /greptimedb/$CLUSTER_DASHBOARD_DIR/dashboard.md
+
+  docker run -v ${PWD}:/greptimedb --rm ${DAC_IMAGE} \
+    -i /greptimedb/$STANDALONE_DASHBOARD_DIR/dashboard.json \
+    -o /greptimedb/$STANDALONE_DASHBOARD_DIR/dashboard.yaml \
+    -m /greptimedb/$STANDALONE_DASHBOARD_DIR/dashboard.md
+}
+
+remove_instance_filters
+generate_intermediate_dashboards_and_docs

grafana/summary.sh

@@ -1,11 +0,0 @@
-#!/usr/bin/env bash
-
-BASEDIR=$(dirname "$0")
-echo '| Title | Description | Expressions |
-|---|---|---|'
-
-cat $BASEDIR/greptimedb-cluster.json | jq -r '
-  .panels |
-  map(select(.type == "stat" or .type == "timeseries")) |
-  .[] | "| \(.title) | \(.description | gsub("\n"; "<br>")) | \(.targets | map(.expr // .rawSql | "`\(.|gsub("\n"; "<br>"))`")  | join("<br>")) |"
-'

src/api/src/helper.rs

@@ -514,6 +514,7 @@ fn query_request_type(request: &QueryRequest) -> &'static str {
         Some(Query::Sql(_)) => "query.sql",
         Some(Query::LogicalPlan(_)) => "query.logical_plan",
         Some(Query::PromRangeQuery(_)) => "query.prom_range",
+        Some(Query::InsertIntoPlan(_)) => "query.insert_into_plan",
         None => "query.empty",
     }
 }

src/catalog/src/table_source.rs

@@ -27,7 +27,7 @@ use session::context::QueryContextRef;
 use snafu::{ensure, OptionExt, ResultExt};
 use table::metadata::TableType;
 use table::table::adapter::DfTableProviderAdapter;
-mod dummy_catalog;
+pub mod dummy_catalog;
 use dummy_catalog::DummyCatalogList;
 use table::TableRef;
 

src/client/src/database.rs

@@ -36,8 +36,8 @@ use common_grpc::flight::{FlightDecoder, FlightMessage};
 use common_query::Output;
 use common_recordbatch::error::ExternalSnafu;
 use common_recordbatch::RecordBatchStreamWrapper;
-use common_telemetry::error;
 use common_telemetry::tracing_context::W3cTrace;
+use common_telemetry::{error, warn};
 use futures::future;
 use futures_util::{Stream, StreamExt, TryStreamExt};
 use prost::Message;
@@ -192,6 +192,36 @@ impl Database {
         from_grpc_response(response)
     }
 
+    /// Retry if connection fails, max_retries is the max number of retries, so the total wait time
+    /// is `max_retries * GRPC_CONN_TIMEOUT`
+    pub async fn handle_with_retry(&self, request: Request, max_retries: u32) -> Result<u32> {
+        let mut client = make_database_client(&self.client)?.inner;
+        let mut retries = 0;
+        let request = self.to_rpc_request(request);
+        loop {
+            let raw_response = client.handle(request.clone()).await;
+            match (raw_response, retries < max_retries) {
+                (Ok(resp), _) => return from_grpc_response(resp.into_inner()),
+                (Err(err), true) => {
+                    // determine if the error is retryable
+                    if is_grpc_retryable(&err) {
+                        // retry
+                        retries += 1;
+                        warn!("Retrying {} times with error = {:?}", retries, err);
+                        continue;
+                    }
+                }
+                (Err(err), false) => {
+                    error!(
+                        "Failed to send request to grpc handle after {} retries, error = {:?}",
+                        retries, err
+                    );
+                    return Err(err.into());
+                }
+            }
+        }
+    }
+
     #[inline]
     fn to_rpc_request(&self, request: Request) -> GreptimeRequest {
         GreptimeRequest {
@@ -368,6 +398,11 @@ impl Database {
     }
 }
 
+/// by grpc standard, only `Unavailable` is retryable, see: https://github.com/grpc/grpc/blob/master/doc/statuscodes.md#status-codes-and-their-use-in-grpc
+pub fn is_grpc_retryable(err: &tonic::Status) -> bool {
+    matches!(err.code(), tonic::Code::Unavailable)
+}
+
 #[derive(Default, Debug, Clone)]
 struct FlightContext {
     auth_header: Option<AuthHeader>,

src/cmd/src/error.rs

@@ -78,13 +78,6 @@ pub enum Error {
         source: datanode::error::Error,
     },
 
-    #[snafu(display("Failed to build object storage manager"))]
-    BuildObjectStorageManager {
-        #[snafu(implicit)]
-        location: Location,
-        source: datanode::error::Error,
-    },
-
     #[snafu(display("Failed to shutdown datanode"))]
     ShutdownDatanode {
         #[snafu(implicit)]
@@ -335,8 +328,6 @@ impl ErrorExt for Error {
                 source.status_code()
             }
 
-            Error::BuildObjectStorageManager { source, .. } => source.status_code(),
-
             Error::MissingConfig { .. }
             | Error::LoadLayeredConfig { .. }
             | Error::IllegalConfig { .. }

src/cmd/src/flownode.rs

@@ -345,7 +345,7 @@ impl StartCommand {
         let client = Arc::new(NodeClients::new(channel_config));
 
         let invoker = FrontendInvoker::build_from(
-            flownode.flow_worker_manager().clone(),
+            flownode.flow_engine().streaming_engine(),
             catalog_manager.clone(),
             cached_meta_backend.clone(),
             layered_cache_registry.clone(),
@@ -355,7 +355,9 @@ impl StartCommand {
         .await
         .context(StartFlownodeSnafu)?;
         flownode
-            .flow_worker_manager()
+            .flow_engine()
+            .streaming_engine()
+            // TODO(discord9): refactor and avoid circular reference
             .set_frontend_invoker(invoker)
             .await;
 

src/cmd/src/metasrv.rs

@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::fmt;
 use std::time::Duration;
 
 use async_trait::async_trait;
@@ -131,8 +132,8 @@ impl SubCommand {
     }
 }
 
-#[derive(Debug, Default, Parser)]
-struct StartCommand {
+#[derive(Default, Parser)]
+pub struct StartCommand {
     /// The address to bind the gRPC server.
     #[clap(long, alias = "bind-addr")]
     rpc_bind_addr: Option<String>,
@@ -171,8 +172,29 @@ struct StartCommand {
     backend: Option<BackendImpl>,
 }
 
+impl fmt::Debug for StartCommand {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        f.debug_struct("StartCommand")
+            .field("rpc_bind_addr", &self.rpc_bind_addr)
+            .field("rpc_server_addr", &self.rpc_server_addr)
+            .field("store_addrs", &self.sanitize_store_addrs())
+            .field("config_file", &self.config_file)
+            .field("selector", &self.selector)
+            .field("use_memory_store", &self.use_memory_store)
+            .field("enable_region_failover", &self.enable_region_failover)
+            .field("http_addr", &self.http_addr)
+            .field("http_timeout", &self.http_timeout)
+            .field("env_prefix", &self.env_prefix)
+            .field("data_home", &self.data_home)
+            .field("store_key_prefix", &self.store_key_prefix)
+            .field("max_txn_ops", &self.max_txn_ops)
+            .field("backend", &self.backend)
+            .finish()
+    }
+}
+
 impl StartCommand {
-    fn load_options(&self, global_options: &GlobalOptions) -> Result<MetasrvOptions> {
+    pub fn load_options(&self, global_options: &GlobalOptions) -> Result<MetasrvOptions> {
         let mut opts = MetasrvOptions::load_layered_options(
             self.config_file.as_deref(),
             self.env_prefix.as_ref(),
@@ -184,6 +206,15 @@ impl StartCommand {
         Ok(opts)
     }
 
+    fn sanitize_store_addrs(&self) -> Option<Vec<String>> {
+        self.store_addrs.as_ref().map(|addrs| {
+            addrs
+                .iter()
+                .map(|addr| common_meta::kv_backend::util::sanitize_connection_string(addr))
+                .collect()
+        })
+    }
+
     // The precedence order is: cli > config file > environment variables > default values.
     fn merge_with_cli_options(
         &self,
@@ -261,7 +292,7 @@ impl StartCommand {
         Ok(())
     }
 
-    async fn build(&self, opts: MetasrvOptions) -> Result<Instance> {
+    pub async fn build(&self, opts: MetasrvOptions) -> Result<Instance> {
         common_runtime::init_global_runtimes(&opts.runtime);
 
         let guard = common_telemetry::init_global_logging(

src/cmd/src/standalone.rs

@@ -44,7 +44,6 @@ use common_meta::peer::Peer;
 use common_meta::region_keeper::MemoryRegionKeeper;
 use common_meta::region_registry::LeaderRegionRegistry;
 use common_meta::sequence::SequenceBuilder;
-use common_meta::snapshot::MetadataSnapshotManager;
 use common_meta::wal_options_allocator::{build_wal_options_allocator, WalOptionsAllocatorRef};
 use common_procedure::{ProcedureInfo, ProcedureManagerRef};
 use common_telemetry::info;
@@ -57,8 +56,8 @@ use datanode::datanode::{Datanode, DatanodeBuilder};
 use datanode::region_server::RegionServer;
 use file_engine::config::EngineConfig as FileEngineConfig;
 use flow::{
-    FlowConfig, FlowWorkerManager, FlownodeBuilder, FlownodeInstance, FlownodeOptions,
-    FrontendClient, FrontendInvoker,
+    FlowConfig, FlownodeBuilder, FlownodeInstance, FlownodeOptions, FrontendClient,
+    FrontendInvoker, GrpcQueryHandlerWithBoxedError, StreamingEngine,
 };
 use frontend::frontend::{Frontend, FrontendOptions};
 use frontend::instance::builder::FrontendBuilder;
@@ -498,10 +497,6 @@ impl StartCommand {
             .build(),
         );
 
-        let object_store_manager = DatanodeBuilder::build_object_store_manager(&dn_opts.storage)
-            .await
-            .context(error::BuildObjectStorageManagerSnafu)?;
-
         let datanode = DatanodeBuilder::new(dn_opts, plugins.clone(), Mode::Standalone)
             .with_kv_backend(kv_backend.clone())
             .with_cache_registry(layered_cache_registry.clone())
@@ -529,17 +524,17 @@ impl StartCommand {
             ..Default::default()
         };
 
-        // TODO(discord9): for standalone not use grpc, but just somehow get a handler to frontend grpc client without
+        // for standalone not use grpc, but get a handler to frontend grpc client without
         // actually make a connection
-        let fe_server_addr = fe_opts.grpc.bind_addr.clone();
-        let frontend_client = FrontendClient::from_static_grpc_addr(fe_server_addr);
+        let (frontend_client, frontend_instance_handler) =
+            FrontendClient::from_empty_grpc_handler();
         let flow_builder = FlownodeBuilder::new(
             flownode_options,
             plugins.clone(),
             table_metadata_manager.clone(),
             catalog_manager.clone(),
             flow_metadata_manager.clone(),
-            Arc::new(frontend_client),
+            Arc::new(frontend_client.clone()),
         );
         let flownode = flow_builder
             .build()
@@ -549,15 +544,15 @@ impl StartCommand {
 
         // set the ref to query for the local flow state
         {
-            let flow_worker_manager = flownode.flow_worker_manager();
+            let flow_streaming_engine = flownode.flow_engine().streaming_engine();
             information_extension
-                .set_flow_worker_manager(flow_worker_manager.clone())
+                .set_flow_streaming_engine(flow_streaming_engine)
                 .await;
         }
 
         let node_manager = Arc::new(StandaloneDatanodeManager {
             region_server: datanode.region_server(),
-            flow_server: flownode.flow_worker_manager(),
+            flow_server: flownode.flow_engine(),
         });
 
         let table_id_sequence = Arc::new(
@@ -596,11 +591,6 @@ impl StartCommand {
         )
         .await?;
 
-        let metadata_snapshot_manager = MetadataSnapshotManager::new(
-            kv_backend.clone(),
-            object_store_manager.default_object_store().clone(),
-        );
-
         let fe_instance = FrontendBuilder::new(
             fe_opts.clone(),
             kv_backend.clone(),
@@ -611,16 +601,24 @@ impl StartCommand {
             StatementStatistics::new(opts.logging.slow_query.clone()),
         )
         .with_plugin(plugins.clone())
-        .with_metadata_snapshot_manager(metadata_snapshot_manager)
         .try_build()
         .await
         .context(error::StartFrontendSnafu)?;
         let fe_instance = Arc::new(fe_instance);
 
-        let flow_worker_manager = flownode.flow_worker_manager();
+        // set the frontend client for flownode
+        let grpc_handler = fe_instance.clone() as Arc<dyn GrpcQueryHandlerWithBoxedError>;
+        let weak_grpc_handler = Arc::downgrade(&grpc_handler);
+        frontend_instance_handler
+            .lock()
+            .unwrap()
+            .replace(weak_grpc_handler);
+
+        // set the frontend invoker for flownode
+        let flow_streaming_engine = flownode.flow_engine().streaming_engine();
         // flow server need to be able to use frontend to write insert requests back
         let invoker = FrontendInvoker::build_from(
-            flow_worker_manager.clone(),
+            flow_streaming_engine.clone(),
             catalog_manager.clone(),
             kv_backend.clone(),
             layered_cache_registry.clone(),
@@ -629,7 +627,7 @@ impl StartCommand {
         )
         .await
         .context(error::StartFlownodeSnafu)?;
-        flow_worker_manager.set_frontend_invoker(invoker).await;
+        flow_streaming_engine.set_frontend_invoker(invoker).await;
 
         let export_metrics_task = ExportMetricsTask::try_new(&opts.export_metrics, Some(&plugins))
             .context(error::ServersSnafu)?;
@@ -705,7 +703,7 @@ pub struct StandaloneInformationExtension {
     region_server: RegionServer,
     procedure_manager: ProcedureManagerRef,
     start_time_ms: u64,
-    flow_worker_manager: RwLock<Option<Arc<FlowWorkerManager>>>,
+    flow_streaming_engine: RwLock<Option<Arc<StreamingEngine>>>,
 }
 
 impl StandaloneInformationExtension {
@@ -714,14 +712,14 @@ impl StandaloneInformationExtension {
             region_server,
             procedure_manager,
             start_time_ms: common_time::util::current_time_millis() as u64,
-            flow_worker_manager: RwLock::new(None),
+            flow_streaming_engine: RwLock::new(None),
         }
     }
 
-    /// Set the flow worker manager for the standalone instance.
-    pub async fn set_flow_worker_manager(&self, flow_worker_manager: Arc<FlowWorkerManager>) {
-        let mut guard = self.flow_worker_manager.write().await;
-        *guard = Some(flow_worker_manager);
+    /// Set the flow streaming engine for the standalone instance.
+    pub async fn set_flow_streaming_engine(&self, flow_streaming_engine: Arc<StreamingEngine>) {
+        let mut guard = self.flow_streaming_engine.write().await;
+        *guard = Some(flow_streaming_engine);
     }
 }
 
@@ -800,7 +798,7 @@ impl InformationExtension for StandaloneInformationExtension {
 
     async fn flow_stats(&self) -> std::result::Result<Option<FlowStat>, Self::Error> {
         Ok(Some(
-            self.flow_worker_manager
+            self.flow_streaming_engine
                 .read()
                 .await
                 .as_ref()

src/cmd/tests/load_config_test.rs

@@ -74,6 +74,7 @@ fn test_load_datanode_example_config() {
                 RegionEngineConfig::File(FileEngineConfig {}),
                 RegionEngineConfig::Metric(MetricEngineConfig {
                     experimental_sparse_primary_key_encoding: false,
+                    flush_metadata_region_interval: Duration::from_secs(30),
                 }),
             ],
             logging: LoggingOptions {
@@ -216,6 +217,7 @@ fn test_load_standalone_example_config() {
                 RegionEngineConfig::File(FileEngineConfig {}),
                 RegionEngineConfig::Metric(MetricEngineConfig {
                     experimental_sparse_primary_key_encoding: false,
+                    flush_metadata_region_interval: Duration::from_secs(30),
                 }),
             ],
             storage: StorageConfig {

src/common/base/src/plugins.rs

@@ -31,7 +31,8 @@ impl Plugins {
     }
 
     pub fn insert<T: 'static + Send + Sync>(&self, value: T) {
-        let _ = self.write().insert(value);
+        let last = self.write().insert(value);
+        assert!(last.is_none(), "each type of plugins must be one and only");
     }
 
     pub fn get<T: 'static + Send + Sync + Clone>(&self) -> Option<T> {
@@ -137,4 +138,12 @@ mod tests {
         assert_eq!(plugins.len(), 2);
         assert!(!plugins.is_empty());
     }
+
+    #[test]
+    #[should_panic(expected = "each type of plugins must be one and only")]
+    fn test_plugin_uniqueness() {
+        let plugins = Plugins::new();
+        plugins.insert(1i32);
+        plugins.insert(2i32);
+    }
 }

src/common/function/src/admin.rs

@@ -15,7 +15,6 @@
 mod add_region_follower;
 mod flush_compact_region;
 mod flush_compact_table;
-mod metadata_snaphost;
 mod migrate_region;
 mod remove_region_follower;
 
@@ -24,7 +23,6 @@ use std::sync::Arc;
 use add_region_follower::AddRegionFollowerFunction;
 use flush_compact_region::{CompactRegionFunction, FlushRegionFunction};
 use flush_compact_table::{CompactTableFunction, FlushTableFunction};
-use metadata_snaphost::{DumpMetadataFunction, RestoreMetadataFunction};
 use migrate_region::MigrateRegionFunction;
 use remove_region_follower::RemoveRegionFollowerFunction;
 
@@ -45,7 +43,5 @@ impl AdminFunction {
         registry.register_async(Arc::new(FlushTableFunction));
         registry.register_async(Arc::new(CompactTableFunction));
         registry.register_async(Arc::new(FlushFlowFunction));
-        registry.register_async(Arc::new(DumpMetadataFunction));
-        registry.register_async(Arc::new(RestoreMetadataFunction));
     }
 }

src/common/function/src/admin/metadata_snaphost.rs

@@ -1,56 +0,0 @@
-use common_macro::admin_fn;
-use common_query::error::{MissingMetadataSnapshotHandlerSnafu, Result};
-use common_query::prelude::{Signature, Volatility};
-use datatypes::prelude::*;
-use session::context::QueryContextRef;
-
-use crate::handlers::MetadataSnapshotHandlerRef;
-
-const METADATA_DIR: &str = "/snaphost/";
-const METADATA_FILE_NAME: &str = "dump_metadata";
-const METADATA_FILE_EXTENSION: &str = "metadata.fb";
-
-#[admin_fn(
-    name = DumpMetadataFunction,
-    display_name = dump_metadata,
-    sig_fn = dump_signature,
-    ret = string
-)]
-pub(crate) async fn dump_metadata(
-    metadata_snapshot_handler: &MetadataSnapshotHandlerRef,
-    _query_ctx: &QueryContextRef,
-    _params: &[ValueRef<'_>],
-) -> Result<Value> {
-    let filename = metadata_snapshot_handler
-        .dump(METADATA_DIR, METADATA_FILE_NAME)
-        .await?;
-    Ok(Value::from(filename))
-}
-
-fn dump_signature() -> Signature {
-    Signature::uniform(0, vec![], Volatility::Immutable)
-}
-
-#[admin_fn(
-    name = RestoreMetadataFunction,
-    display_name = restore_metadata,
-    sig_fn = restore_signature,
-    ret = uint64,
-)]
-pub(crate) async fn restore_metadata(
-    metadata_snapshot_handler: &MetadataSnapshotHandlerRef,
-    _query_ctx: &QueryContextRef,
-    _params: &[ValueRef<'_>],
-) -> Result<Value> {
-    let num_keyvalues = metadata_snapshot_handler
-        .restore(
-            METADATA_DIR,
-            &format!("{METADATA_FILE_NAME}.{METADATA_FILE_EXTENSION}"),
-        )
-        .await?;
-    Ok(Value::from(num_keyvalues))
-}
-
-fn restore_signature() -> Signature {
-    Signature::uniform(0, vec![], Volatility::Immutable)
-}

src/common/function/src/handlers.rs

@@ -89,18 +89,8 @@ pub trait FlowServiceHandler: Send + Sync {
     ) -> Result<api::v1::flow::FlowResponse>;
 }
 
-/// This metadata snapshot handler is only use for dump and restore metadata for now.
-#[async_trait]
-pub trait MetadataSnapshotHandler: Send + Sync {
-    async fn dump(&self, path: &str, filename: &str) -> Result<String>;
-
-    async fn restore(&self, path: &str, filename: &str) -> Result<u64>;
-}
-
 pub type TableMutationHandlerRef = Arc<dyn TableMutationHandler>;
 
 pub type ProcedureServiceHandlerRef = Arc<dyn ProcedureServiceHandler>;
 
 pub type FlowServiceHandlerRef = Arc<dyn FlowServiceHandler>;
-
-pub type MetadataSnapshotHandlerRef = Arc<dyn MetadataSnapshotHandler>;

src/common/function/src/scalars/timestamp.rs

@@ -13,10 +13,8 @@
 // limitations under the License.
 
 use std::sync::Arc;
-mod greatest;
 mod to_unixtime;
 
-use greatest::GreatestFunction;
 use to_unixtime::ToUnixtimeFunction;
 
 use crate::function_registry::FunctionRegistry;
@@ -26,6 +24,5 @@ pub(crate) struct TimestampFunction;
 impl TimestampFunction {
     pub fn register(registry: &FunctionRegistry) {
         registry.register(Arc::new(ToUnixtimeFunction));
-        registry.register(Arc::new(GreatestFunction));
     }
 }

src/common/function/src/scalars/timestamp/greatest.rs

@@ -1,328 +0,0 @@
-// Copyright 2023 Greptime Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-use std::fmt::{self};
-
-use common_query::error::{
-    self, ArrowComputeSnafu, InvalidFuncArgsSnafu, Result, UnsupportedInputDataTypeSnafu,
-};
-use common_query::prelude::{Signature, Volatility};
-use datafusion::arrow::compute::kernels::cmp::gt;
-use datatypes::arrow::array::AsArray;
-use datatypes::arrow::compute::cast;
-use datatypes::arrow::compute::kernels::zip;
-use datatypes::arrow::datatypes::{
-    DataType as ArrowDataType, Date32Type, TimeUnit, TimestampMicrosecondType,
-    TimestampMillisecondType, TimestampNanosecondType, TimestampSecondType,
-};
-use datatypes::prelude::ConcreteDataType;
-use datatypes::types::TimestampType;
-use datatypes::vectors::{Helper, VectorRef};
-use snafu::{ensure, ResultExt};
-
-use crate::function::{Function, FunctionContext};
-
-#[derive(Clone, Debug, Default)]
-pub struct GreatestFunction;
-
-const NAME: &str = "greatest";
-
-macro_rules! gt_time_types {
-    ($ty: ident, $columns:expr) => {{
-        let column1 = $columns[0].to_arrow_array();
-        let column2 = $columns[1].to_arrow_array();
-
-        let column1 = column1.as_primitive::<$ty>();
-        let column2 = column2.as_primitive::<$ty>();
-        let boolean_array = gt(&column1, &column2).context(ArrowComputeSnafu)?;
-
-        let result = zip::zip(&boolean_array, &column1, &column2).context(ArrowComputeSnafu)?;
-        Helper::try_into_vector(&result).context(error::FromArrowArraySnafu)
-    }};
-}
-
-impl Function for GreatestFunction {
-    fn name(&self) -> &str {
-        NAME
-    }
-
-    fn return_type(&self, input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
-        ensure!(
-            input_types.len() == 2,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect exactly two, have: {}",
-                    input_types.len()
-                )
-            }
-        );
-
-        match &input_types[0] {
-            ConcreteDataType::String(_) => Ok(ConcreteDataType::timestamp_millisecond_datatype()),
-            ConcreteDataType::Date(_) => Ok(ConcreteDataType::date_datatype()),
-            ConcreteDataType::Timestamp(ts_type) => Ok(ConcreteDataType::Timestamp(*ts_type)),
-            _ => UnsupportedInputDataTypeSnafu {
-                function: NAME,
-                datatypes: input_types,
-            }
-            .fail(),
-        }
-    }
-
-    fn signature(&self) -> Signature {
-        Signature::uniform(
-            2,
-            vec![
-                ConcreteDataType::string_datatype(),
-                ConcreteDataType::date_datatype(),
-                ConcreteDataType::timestamp_nanosecond_datatype(),
-                ConcreteDataType::timestamp_microsecond_datatype(),
-                ConcreteDataType::timestamp_millisecond_datatype(),
-                ConcreteDataType::timestamp_second_datatype(),
-            ],
-            Volatility::Immutable,
-        )
-    }
-
-    fn eval(&self, _func_ctx: &FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 2,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect exactly two, have: {}",
-                    columns.len()
-                ),
-            }
-        );
-        match columns[0].data_type() {
-            ConcreteDataType::String(_) => {
-                let column1 = cast(
-                    &columns[0].to_arrow_array(),
-                    &ArrowDataType::Timestamp(TimeUnit::Millisecond, None),
-                )
-                .context(ArrowComputeSnafu)?;
-                let column1 = column1.as_primitive::<TimestampMillisecondType>();
-                let column2 = cast(
-                    &columns[1].to_arrow_array(),
-                    &ArrowDataType::Timestamp(TimeUnit::Millisecond, None),
-                )
-                .context(ArrowComputeSnafu)?;
-                let column2 = column2.as_primitive::<TimestampMillisecondType>();
-                let boolean_array = gt(&column1, &column2).context(ArrowComputeSnafu)?;
-                let result =
-                    zip::zip(&boolean_array, &column1, &column2).context(ArrowComputeSnafu)?;
-                Ok(Helper::try_into_vector(&result).context(error::FromArrowArraySnafu)?)
-            }
-            ConcreteDataType::Date(_) => gt_time_types!(Date32Type, columns),
-            ConcreteDataType::Timestamp(ts_type) => match ts_type {
-                TimestampType::Second(_) => gt_time_types!(TimestampSecondType, columns),
-                TimestampType::Millisecond(_) => {
-                    gt_time_types!(TimestampMillisecondType, columns)
-                }
-                TimestampType::Microsecond(_) => {
-                    gt_time_types!(TimestampMicrosecondType, columns)
-                }
-                TimestampType::Nanosecond(_) => {
-                    gt_time_types!(TimestampNanosecondType, columns)
-                }
-            },
-            _ => UnsupportedInputDataTypeSnafu {
-                function: NAME,
-                datatypes: columns.iter().map(|c| c.data_type()).collect::<Vec<_>>(),
-            }
-            .fail(),
-        }
-    }
-}
-
-impl fmt::Display for GreatestFunction {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-        write!(f, "GREATEST")
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use std::sync::Arc;
-
-    use common_time::timestamp::TimeUnit;
-    use common_time::{Date, Timestamp};
-    use datatypes::types::{
-        DateType, TimestampMicrosecondType, TimestampMillisecondType, TimestampNanosecondType,
-        TimestampSecondType,
-    };
-    use datatypes::value::Value;
-    use datatypes::vectors::{
-        DateVector, StringVector, TimestampMicrosecondVector, TimestampMillisecondVector,
-        TimestampNanosecondVector, TimestampSecondVector, Vector,
-    };
-    use paste::paste;
-
-    use super::*;
-    #[test]
-    fn test_greatest_takes_string_vector() {
-        let function = GreatestFunction;
-        assert_eq!(
-            function
-                .return_type(&[
-                    ConcreteDataType::string_datatype(),
-                    ConcreteDataType::string_datatype()
-                ])
-                .unwrap(),
-            ConcreteDataType::timestamp_millisecond_datatype()
-        );
-        let columns = vec![
-            Arc::new(StringVector::from(vec![
-                "1970-01-01".to_string(),
-                "2012-12-23".to_string(),
-            ])) as _,
-            Arc::new(StringVector::from(vec![
-                "2001-02-01".to_string(),
-                "1999-01-01".to_string(),
-            ])) as _,
-        ];
-
-        let result = function
-            .eval(&FunctionContext::default(), &columns)
-            .unwrap();
-        let result = result
-            .as_any()
-            .downcast_ref::<TimestampMillisecondVector>()
-            .unwrap();
-        assert_eq!(result.len(), 2);
-        assert_eq!(
-            result.get(0),
-            Value::Timestamp(Timestamp::from_str("2001-02-01 00:00:00", None).unwrap())
-        );
-        assert_eq!(
-            result.get(1),
-            Value::Timestamp(Timestamp::from_str("2012-12-23 00:00:00", None).unwrap())
-        );
-    }
-
-    #[test]
-    fn test_greatest_takes_date_vector() {
-        let function = GreatestFunction;
-        assert_eq!(
-            function
-                .return_type(&[
-                    ConcreteDataType::date_datatype(),
-                    ConcreteDataType::date_datatype()
-                ])
-                .unwrap(),
-            ConcreteDataType::Date(DateType)
-        );
-
-        let columns = vec![
-            Arc::new(DateVector::from_slice(vec![-1, 2])) as _,
-            Arc::new(DateVector::from_slice(vec![0, 1])) as _,
-        ];
-
-        let result = function
-            .eval(&FunctionContext::default(), &columns)
-            .unwrap();
-        let result = result.as_any().downcast_ref::<DateVector>().unwrap();
-        assert_eq!(result.len(), 2);
-        assert_eq!(
-            result.get(0),
-            Value::Date(Date::from_str_utc("1970-01-01").unwrap())
-        );
-        assert_eq!(
-            result.get(1),
-            Value::Date(Date::from_str_utc("1970-01-03").unwrap())
-        );
-    }
-
-    #[test]
-    fn test_greatest_takes_datetime_vector() {
-        let function = GreatestFunction;
-        assert_eq!(
-            function
-                .return_type(&[
-                    ConcreteDataType::timestamp_millisecond_datatype(),
-                    ConcreteDataType::timestamp_millisecond_datatype()
-                ])
-                .unwrap(),
-            ConcreteDataType::timestamp_millisecond_datatype()
-        );
-
-        let columns = vec![
-            Arc::new(TimestampMillisecondVector::from_slice(vec![-1, 2])) as _,
-            Arc::new(TimestampMillisecondVector::from_slice(vec![0, 1])) as _,
-        ];
-
-        let result = function
-            .eval(&FunctionContext::default(), &columns)
-            .unwrap();
-        let result = result
-            .as_any()
-            .downcast_ref::<TimestampMillisecondVector>()
-            .unwrap();
-        assert_eq!(result.len(), 2);
-        assert_eq!(
-            result.get(0),
-            Value::Timestamp(Timestamp::from_str("1970-01-01 00:00:00", None).unwrap())
-        );
-        assert_eq!(
-            result.get(1),
-            Value::Timestamp(Timestamp::from_str("1970-01-01 00:00:00.002", None).unwrap())
-        );
-    }
-
-    macro_rules! test_timestamp {
-        ($type: expr,$unit: ident) => {
-            paste! {
-                #[test]
-                fn [<test_greatest_takes_ $unit:lower _vector>]() {
-                    let function = GreatestFunction;
-                    assert_eq!(
-                        function.return_type(&[$type, $type]).unwrap(),
-                        ConcreteDataType::Timestamp(TimestampType::$unit([<Timestamp $unit Type>]))
-                    );
-
-                    let columns = vec![
-                        Arc::new([<Timestamp $unit Vector>]::from_slice(vec![-1, 2])) as _,
-                        Arc::new([<Timestamp $unit Vector>]::from_slice(vec![0, 1])) as _,
-                    ];
-
-                    let result = function.eval(&FunctionContext::default(), &columns).unwrap();
-                    let result = result.as_any().downcast_ref::<[<Timestamp $unit Vector>]>().unwrap();
-                    assert_eq!(result.len(), 2);
-                    assert_eq!(
-                        result.get(0),
-                        Value::Timestamp(Timestamp::new(0, TimeUnit::$unit))
-                    );
-                    assert_eq!(
-                        result.get(1),
-                        Value::Timestamp(Timestamp::new(2, TimeUnit::$unit))
-                    );
-                }
-            }
-        }
-    }
-
-    test_timestamp!(
-        ConcreteDataType::timestamp_nanosecond_datatype(),
-        Nanosecond
-    );
-    test_timestamp!(
-        ConcreteDataType::timestamp_microsecond_datatype(),
-        Microsecond
-    );
-    test_timestamp!(
-        ConcreteDataType::timestamp_millisecond_datatype(),
-        Millisecond
-    );
-    test_timestamp!(ConcreteDataType::timestamp_second_datatype(), Second);
-}

src/common/function/src/scalars/uddsketch_calc.rs

@@ -115,6 +115,13 @@ impl Function for UddSketchCalcFunction {
                 }
             };
 
+            // Check if the sketch is empty, if so, return null
+            // This is important to avoid panics when calling estimate_quantile on an empty sketch
+            // In practice, this will happen if input is all null
+            if sketch.bucket_iter().count() == 0 {
+                builder.push_null();
+                continue;
+            }
             // Compute the estimated quantile from the sketch
             let result = sketch.estimate_quantile(perc);
             builder.push(Some(result));

src/common/function/src/state.rs

@@ -12,10 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use crate::handlers::{
-    FlowServiceHandlerRef, MetadataSnapshotHandlerRef, ProcedureServiceHandlerRef,
-    TableMutationHandlerRef,
-};
+use crate::handlers::{FlowServiceHandlerRef, ProcedureServiceHandlerRef, TableMutationHandlerRef};
 
 /// Shared state for SQL functions.
 /// The handlers in state may be `None` in cli command-line or test cases.
@@ -27,8 +24,6 @@ pub struct FunctionState {
     pub procedure_service_handler: Option<ProcedureServiceHandlerRef>,
     // The flownode handler
     pub flow_service_handler: Option<FlowServiceHandlerRef>,
-    // The metadata snapshot handler
-    pub metadata_snapshot_handler: Option<MetadataSnapshotHandlerRef>,
 }
 
 impl FunctionState {
@@ -53,14 +48,10 @@ impl FunctionState {
             CompactTableRequest, DeleteRequest, FlushTableRequest, InsertRequest,
         };
 
-        use crate::handlers::{
-            FlowServiceHandler, MetadataSnapshotHandler, ProcedureServiceHandler,
-            TableMutationHandler,
-        };
+        use crate::handlers::{FlowServiceHandler, ProcedureServiceHandler, TableMutationHandler};
         struct MockProcedureServiceHandler;
         struct MockTableMutationHandler;
         struct MockFlowServiceHandler;
-        struct MockMetadataServiceHandler;
         const ROWS: usize = 42;
 
         #[async_trait]
@@ -159,22 +150,10 @@ impl FunctionState {
             }
         }
 
-        #[async_trait]
-        impl MetadataSnapshotHandler for MockMetadataServiceHandler {
-            async fn dump(&self, _path: &str, _filename: &str) -> Result<String> {
-                Ok("test_filename".to_string())
-            }
-
-            async fn restore(&self, _path: &str, _filename: &str) -> Result<u64> {
-                Ok(100)
-            }
-        }
-
         Self {
             table_mutation_handler: Some(Arc::new(MockTableMutationHandler)),
             procedure_service_handler: Some(Arc::new(MockProcedureServiceHandler)),
             flow_service_handler: Some(Arc::new(MockFlowServiceHandler)),
-            metadata_snapshot_handler: Some(Arc::new(MockMetadataServiceHandler)),
         }
     }
 }

src/common/macro/src/admin_fn.rs

@@ -179,10 +179,6 @@ fn build_struct(
             Ident::new("flow_service_handler", handler_type.span()),
             Ident::new("MissingFlowServiceHandlerSnafu", handler_type.span()),
         ),
-        "MetadataSnapshotHandlerRef" => (
-            Ident::new("metadata_snapshot_handler", handler_type.span()),
-            Ident::new("MissingMetadataSnapshotHandlerSnafu", handler_type.span()),
-        ),
         handler => ok!(error!(
             handler_type.span(),
             format!("Unknown handler type: {handler}")

src/common/meta/Cargo.toml

@@ -41,7 +41,6 @@ deadpool = { workspace = true, optional = true }
 deadpool-postgres = { workspace = true, optional = true }
 derive_builder.workspace = true
 etcd-client.workspace = true
-flexbuffers = "25.2"
 futures.workspace = true
 futures-util.workspace = true
 hex.workspace = true
@@ -49,7 +48,6 @@ humantime-serde.workspace = true
 itertools.workspace = true
 lazy_static.workspace = true
 moka.workspace = true
-object-store.workspace = true
 prometheus.workspace = true
 prost.workspace = true
 rand.workspace = true
@@ -72,7 +70,6 @@ typetag.workspace = true
 [dev-dependencies]
 chrono.workspace = true
 common-procedure = { workspace = true, features = ["testing"] }
-common-test-util.workspace = true
 common-wal = { workspace = true, features = ["testing"] }
 datatypes.workspace = true
 hyper = { version = "0.14", features = ["full"] }

src/common/meta/src/cache/flow/table_flownode.rs

@@ -187,6 +187,7 @@ mod tests {
                     },
                     flownode_ids: BTreeMap::from([(0, 1), (1, 2), (2, 3)]),
                     catalog_name: DEFAULT_CATALOG_NAME.to_string(),
+                    query_context: None,
                     flow_name: "my_flow".to_string(),
                     raw_sql: "sql".to_string(),
                     expire_after: Some(300),

src/common/meta/src/ddl/create_flow.rs

@@ -38,7 +38,7 @@ use table::metadata::TableId;
 use crate::cache_invalidator::Context;
 use crate::ddl::utils::{add_peer_context_if_needed, handle_retry_error};
 use crate::ddl::DdlContext;
-use crate::error::{self, Result};
+use crate::error::{self, Result, UnexpectedSnafu};
 use crate::instruction::{CacheIdent, CreateFlow};
 use crate::key::flow::flow_info::FlowInfoValue;
 use crate::key::flow::flow_route::FlowRouteValue;
@@ -171,7 +171,7 @@ impl CreateFlowProcedure {
         }
         self.data.state = CreateFlowState::CreateFlows;
         // determine flow type
-        self.data.flow_type = Some(determine_flow_type(&self.data.task));
+        self.data.flow_type = Some(get_flow_type_from_options(&self.data.task)?);
 
         Ok(Status::executing(true))
     }
@@ -196,8 +196,8 @@ impl CreateFlowProcedure {
             });
         }
         info!(
-            "Creating flow({:?}) on flownodes with peers={:?}",
-            self.data.flow_id, self.data.peers
+            "Creating flow({:?}, type={:?}) on flownodes with peers={:?}",
+            self.data.flow_id, self.data.flow_type, self.data.peers
         );
         join_all(create_flow)
             .await
@@ -306,8 +306,20 @@ impl Procedure for CreateFlowProcedure {
     }
 }
 
-pub fn determine_flow_type(_flow_task: &CreateFlowTask) -> FlowType {
-    FlowType::Batching
+pub fn get_flow_type_from_options(flow_task: &CreateFlowTask) -> Result<FlowType> {
+    let flow_type = flow_task
+        .flow_options
+        .get(FlowType::FLOW_TYPE_KEY)
+        .map(|s| s.as_str());
+    match flow_type {
+        Some(FlowType::BATCHING) => Ok(FlowType::Batching),
+        Some(FlowType::STREAMING) => Ok(FlowType::Streaming),
+        Some(unknown) => UnexpectedSnafu {
+            err_msg: format!("Unknown flow type: {}", unknown),
+        }
+        .fail(),
+        None => Ok(FlowType::Batching),
+    }
 }
 
 /// The state of [CreateFlowProcedure].
@@ -437,6 +449,7 @@ impl From<&CreateFlowData> for (FlowInfoValue, Vec<(FlowPartitionId, FlowRouteVa
             sink_table_name,
             flownode_ids,
             catalog_name,
+            query_context: Some(value.query_context.clone()),
             flow_name,
             raw_sql: sql,
             expire_after,

src/common/meta/src/ddl/tests/create_flow.rs

@@ -46,7 +46,7 @@ pub(crate) fn test_create_flow_task(
         create_if_not_exists,
         expire_after: Some(300),
         comment: "".to_string(),
-        sql: "raw_sql".to_string(),
+        sql: "select 1".to_string(),
         flow_options: Default::default(),
     }
 }

src/common/meta/src/error.rs

@@ -401,6 +401,13 @@ pub enum Error {
         location: Location,
     },
 
+    #[snafu(display("Invalid flow request body: {:?}", body))]
+    InvalidFlowRequestBody {
+        body: Box<Option<api::v1::flow::flow_request::Body>>,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
     #[snafu(display("Failed to get kv cache, err: {}", err_msg))]
     GetKvCache { err_msg: String },
 
@@ -784,74 +791,12 @@ pub enum Error {
         source: common_procedure::error::Error,
     },
 
-    #[snafu(display("Invalid file path: {}", file_path))]
-    InvalidFilePath {
-        #[snafu(implicit)]
-        location: Location,
-        file_path: String,
-    },
-
-    #[snafu(display("Failed to serialize flexbuffers"))]
-    SerializeFlexbuffers {
+    #[snafu(display("Failed to parse timezone"))]
+    InvalidTimeZone {
         #[snafu(implicit)]
         location: Location,
         #[snafu(source)]
-        error: flexbuffers::SerializationError,
-    },
-
-    #[snafu(display("Failed to deserialize flexbuffers"))]
-    DeserializeFlexbuffers {
-        #[snafu(implicit)]
-        location: Location,
-        #[snafu(source)]
-        error: flexbuffers::DeserializationError,
-    },
-
-    #[snafu(display("Failed to read flexbuffers"))]
-    ReadFlexbuffers {
-        #[snafu(implicit)]
-        location: Location,
-        #[snafu(source)]
-        error: flexbuffers::ReaderError,
-    },
-
-    #[snafu(display("Invalid file name: {}", reason))]
-    InvalidFileName {
-        #[snafu(implicit)]
-        location: Location,
-        reason: String,
-    },
-
-    #[snafu(display("Invalid file extension: {}", reason))]
-    InvalidFileExtension {
-        #[snafu(implicit)]
-        location: Location,
-        reason: String,
-    },
-
-    #[snafu(display("Invalid file context: {}", reason))]
-    InvalidFileContext {
-        #[snafu(implicit)]
-        location: Location,
-        reason: String,
-    },
-
-    #[snafu(display("Failed to write object, file path: {}", file_path))]
-    WriteObject {
-        #[snafu(implicit)]
-        location: Location,
-        file_path: String,
-        #[snafu(source)]
-        error: object_store::Error,
-    },
-
-    #[snafu(display("Failed to read object, file path: {}", file_path))]
-    ReadObject {
-        #[snafu(implicit)]
-        location: Location,
-        file_path: String,
-        #[snafu(source)]
-        error: object_store::Error,
+        error: common_time::error::Error,
     },
 }
 
@@ -871,8 +816,6 @@ impl ErrorExt for Error {
             | SerializeToJson { .. }
             | DeserializeFromJson { .. } => StatusCode::Internal,
 
-            WriteObject { .. } | ReadObject { .. } => StatusCode::StorageUnavailable,
-
             NoLeader { .. } => StatusCode::TableUnavailable,
             ValueNotExist { .. } | ProcedurePoisonConflict { .. } => StatusCode::Unexpected,
 
@@ -909,11 +852,7 @@ impl ErrorExt for Error {
             | ProcedureOutput { .. }
             | FromUtf8 { .. }
             | MetadataCorruption { .. }
-            | ParseWalOptions { .. }
-            | ReadFlexbuffers { .. }
-            | SerializeFlexbuffers { .. }
-            | DeserializeFlexbuffers { .. }
-            | InvalidFileContext { .. } => StatusCode::Unexpected,
+            | ParseWalOptions { .. } => StatusCode::Unexpected,
 
             SendMessage { .. } | GetKvCache { .. } | CacheNotGet { .. } => StatusCode::Internal,
 
@@ -930,9 +869,8 @@ impl ErrorExt for Error {
             | InvalidSetDatabaseOption { .. }
             | InvalidUnsetDatabaseOption { .. }
             | InvalidTopicNamePrefix { .. }
-            | InvalidFileExtension { .. }
-            | InvalidFileName { .. }
-            | InvalidFilePath { .. } => StatusCode::InvalidArguments,
+            | InvalidTimeZone { .. } => StatusCode::InvalidArguments,
+            InvalidFlowRequestBody { .. } => StatusCode::InvalidArguments,
 
             FlowNotFound { .. } => StatusCode::FlowNotFound,
             FlowRouteNotFound { .. } => StatusCode::Unexpected,

src/common/meta/src/key/flow.rs

@@ -452,6 +452,7 @@ mod tests {
         };
         FlowInfoValue {
             catalog_name: catalog_name.to_string(),
+            query_context: None,
             flow_name: flow_name.to_string(),
             source_table_ids,
             sink_table_name,
@@ -625,6 +626,7 @@ mod tests {
         };
         let flow_value = FlowInfoValue {
             catalog_name: "greptime".to_string(),
+            query_context: None,
             flow_name: "flow".to_string(),
             source_table_ids: vec![1024, 1025, 1026],
             sink_table_name: another_sink_table_name,
@@ -864,6 +866,7 @@ mod tests {
         };
         let flow_value = FlowInfoValue {
             catalog_name: "greptime".to_string(),
+            query_context: None,
             flow_name: "flow".to_string(),
             source_table_ids: vec![1024, 1025, 1026],
             sink_table_name: another_sink_table_name,

src/common/meta/src/key/flow/flow_info.rs

@@ -121,6 +121,13 @@ pub struct FlowInfoValue {
     pub(crate) flownode_ids: BTreeMap<FlowPartitionId, FlownodeId>,
     /// The catalog name.
     pub(crate) catalog_name: String,
+    /// The query context used when create flow.
+    /// Although flow doesn't belong to any schema, this query_context is needed to remember
+    /// the query context when `create_flow` is executed
+    /// for recovering flow using the same sql&query_context after db restart.
+    /// if none, should use default query context
+    #[serde(default)]
+    pub(crate) query_context: Option<crate::rpc::ddl::QueryContext>,
     /// The flow name.
     pub(crate) flow_name: String,
     /// The raw sql.
@@ -155,6 +162,10 @@ impl FlowInfoValue {
         &self.catalog_name
     }
 
+    pub fn query_context(&self) -> &Option<crate::rpc::ddl::QueryContext> {
+        &self.query_context
+    }
+
     pub fn flow_name(&self) -> &String {
         &self.flow_name
     }

src/common/meta/src/kv_backend.rs

@@ -35,7 +35,7 @@ pub mod memory;
 pub mod rds;
 pub mod test;
 pub mod txn;
-
+pub mod util;
 pub type KvBackendRef<E = Error> = Arc<dyn KvBackend<Error = E> + Send + Sync>;
 
 #[async_trait]

src/common/meta/src/kv_backend/util.rs

@@ -0,0 +1,85 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+/// Removes sensitive information like passwords from connection strings.
+///
+/// This function sanitizes connection strings by removing credentials:
+/// - For URL format (mysql://user:password@host:port/db): Removes everything before '@'
+/// - For parameter format (host=localhost password=secret): Removes the password parameter
+/// - For URL format without credentials (mysql://host:port/db): Removes the protocol prefix
+///
+/// # Arguments
+///
+/// * `conn_str` - The connection string to sanitize
+///
+/// # Returns
+///
+/// A sanitized version of the connection string with sensitive information removed
+pub fn sanitize_connection_string(conn_str: &str) -> String {
+    // Case 1: URL format with credentials (mysql://user:password@host:port/db)
+    // Extract everything after the '@' symbol
+    if let Some(at_pos) = conn_str.find('@') {
+        return conn_str[at_pos + 1..].to_string();
+    }
+
+    // Case 2: Parameter format with password (host=localhost password=secret dbname=mydb)
+    // Filter out any parameter that starts with "password="
+    if conn_str.contains("password=") {
+        return conn_str
+            .split_whitespace()
+            .filter(|param| !param.starts_with("password="))
+            .collect::<Vec<_>>()
+            .join(" ");
+    }
+
+    // Case 3: URL format without credentials (mysql://host:port/db)
+    // Extract everything after the protocol prefix
+    if let Some(host_part) = conn_str.split("://").nth(1) {
+        return host_part.to_string();
+    }
+
+    // Case 4: Already sanitized or unknown format
+    // Return as is
+    conn_str.to_string()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_sanitize_connection_string() {
+        // Test URL format with username/password
+        let conn_str = "mysql://user:password123@localhost:3306/db";
+        assert_eq!(sanitize_connection_string(conn_str), "localhost:3306/db");
+
+        // Test URL format without credentials
+        let conn_str = "mysql://localhost:3306/db";
+        assert_eq!(sanitize_connection_string(conn_str), "localhost:3306/db");
+
+        // Test parameter format with password
+        let conn_str = "host=localhost port=5432 user=postgres password=secret dbname=mydb";
+        assert_eq!(
+            sanitize_connection_string(conn_str),
+            "host=localhost port=5432 user=postgres dbname=mydb"
+        );
+
+        // Test parameter format without password
+        let conn_str = "host=localhost port=5432 user=postgres dbname=mydb";
+        assert_eq!(
+            sanitize_connection_string(conn_str),
+            "host=localhost port=5432 user=postgres dbname=mydb"
+        );
+    }
+}

src/common/meta/src/lib.rs

@@ -43,7 +43,6 @@ pub mod region_keeper;
 pub mod region_registry;
 pub mod rpc;
 pub mod sequence;
-pub mod snapshot;
 pub mod state_store;
 #[cfg(any(test, feature = "testing"))]
 pub mod test_util;

src/common/meta/src/region_registry.rs

@@ -113,8 +113,10 @@ impl LeaderRegionManifestInfo {
     pub fn prunable_entry_id(&self) -> u64 {
         match self {
             LeaderRegionManifestInfo::Mito {
-                flushed_entry_id, ..
-            } => *flushed_entry_id,
+                flushed_entry_id,
+                topic_latest_entry_id,
+                ..
+            } => (*flushed_entry_id).max(*topic_latest_entry_id),
             LeaderRegionManifestInfo::Metric {
                 data_flushed_entry_id,
                 data_topic_latest_entry_id,

src/common/meta/src/rpc/ddl.rs

@@ -35,17 +35,20 @@ use api::v1::{
 };
 use base64::engine::general_purpose;
 use base64::Engine as _;
-use common_time::DatabaseTimeToLive;
+use common_time::{DatabaseTimeToLive, Timezone};
 use prost::Message;
 use serde::{Deserialize, Serialize};
 use serde_with::{serde_as, DefaultOnNull};
-use session::context::QueryContextRef;
+use session::context::{QueryContextBuilder, QueryContextRef};
 use snafu::{OptionExt, ResultExt};
 use table::metadata::{RawTableInfo, TableId};
 use table::table_name::TableName;
 use table::table_reference::TableReference;
 
-use crate::error::{self, InvalidSetDatabaseOptionSnafu, InvalidUnsetDatabaseOptionSnafu, Result};
+use crate::error::{
+    self, InvalidSetDatabaseOptionSnafu, InvalidTimeZoneSnafu, InvalidUnsetDatabaseOptionSnafu,
+    Result,
+};
 use crate::key::FlowId;
 
 /// DDL tasks
@@ -1202,7 +1205,7 @@ impl From<DropFlowTask> for PbDropFlowTask {
     }
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 pub struct QueryContext {
     current_catalog: String,
     current_schema: String,
@@ -1223,6 +1226,19 @@ impl From<QueryContextRef> for QueryContext {
     }
 }
 
+impl TryFrom<QueryContext> for session::context::QueryContext {
+    type Error = error::Error;
+    fn try_from(value: QueryContext) -> std::result::Result<Self, Self::Error> {
+        Ok(QueryContextBuilder::default()
+            .current_catalog(value.current_catalog)
+            .current_schema(value.current_schema)
+            .timezone(Timezone::from_tz_string(&value.timezone).context(InvalidTimeZoneSnafu)?)
+            .extensions(value.extensions)
+            .channel((value.channel as u32).into())
+            .build())
+    }
+}
+
 impl From<QueryContext> for PbQueryContext {
     fn from(
         QueryContext {

src/common/meta/src/snapshot.rs

@@ -1,365 +0,0 @@
-// Copyright 2023 Greptime Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-pub mod file;
-
-use std::fmt::{Display, Formatter};
-use std::time::Instant;
-
-use common_telemetry::info;
-use file::{Metadata, MetadataContent};
-use futures::TryStreamExt;
-use object_store::ObjectStore;
-use snafu::{OptionExt, ResultExt};
-use strum::Display;
-
-use crate::error::{
-    Error, InvalidFileExtensionSnafu, InvalidFileNameSnafu, InvalidFilePathSnafu, ReadObjectSnafu,
-    Result, WriteObjectSnafu,
-};
-use crate::kv_backend::KvBackendRef;
-use crate::range_stream::{PaginationStream, DEFAULT_PAGE_SIZE};
-use crate::rpc::store::{BatchPutRequest, RangeRequest};
-use crate::rpc::KeyValue;
-use crate::snapshot::file::{Document, KeyValue as FileKeyValue};
-
-/// The format of the backup file.
-#[derive(Debug, PartialEq, Eq, Display, Clone, Copy)]
-pub enum FileFormat {
-    #[strum(serialize = "fb")]
-    FlexBuffers,
-}
-
-impl TryFrom<&str> for FileFormat {
-    type Error = String;
-
-    fn try_from(value: &str) -> std::result::Result<Self, Self::Error> {
-        match value.to_lowercase().as_str() {
-            "fb" => Ok(FileFormat::FlexBuffers),
-            _ => Err(format!("Invalid file format: {}", value)),
-        }
-    }
-}
-
-#[derive(Debug, PartialEq, Eq, Display)]
-#[strum(serialize_all = "lowercase")]
-pub enum DataType {
-    Metadata,
-}
-
-impl TryFrom<&str> for DataType {
-    type Error = String;
-
-    fn try_from(value: &str) -> std::result::Result<Self, Self::Error> {
-        match value.to_lowercase().as_str() {
-            "metadata" => Ok(DataType::Metadata),
-            _ => Err(format!("Invalid data type: {}", value)),
-        }
-    }
-}
-
-#[derive(Debug, PartialEq, Eq)]
-pub struct FileExtension {
-    format: FileFormat,
-    data_type: DataType,
-}
-
-impl FileExtension {
-    pub fn new(format: FileFormat, data_type: DataType) -> Self {
-        Self { format, data_type }
-    }
-}
-
-impl Display for FileExtension {
-    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
-        write!(f, "{}.{}", self.data_type, self.format)
-    }
-}
-
-impl TryFrom<&str> for FileExtension {
-    type Error = Error;
-
-    fn try_from(value: &str) -> Result<Self> {
-        let parts = value.split(".").collect::<Vec<&str>>();
-        if parts.len() != 2 {
-            return InvalidFileExtensionSnafu {
-                reason: format!(
-                    "Extension should be in the format of <datatype>.<format>, got: {}",
-                    value
-                ),
-            }
-            .fail();
-        }
-
-        let data_type = DataType::try_from(parts[0])
-            .map_err(|e| InvalidFileExtensionSnafu { reason: e }.build())?;
-        let format = FileFormat::try_from(parts[1])
-            .map_err(|e| InvalidFileExtensionSnafu { reason: e }.build())?;
-        Ok(FileExtension { format, data_type })
-    }
-}
-
-#[derive(Debug, PartialEq, Eq)]
-pub struct FileName {
-    name: String,
-    extension: FileExtension,
-}
-
-impl Display for FileName {
-    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
-        write!(f, "{}.{}", self.name, self.extension)
-    }
-}
-
-impl TryFrom<&str> for FileName {
-    type Error = Error;
-
-    fn try_from(value: &str) -> Result<Self> {
-        let Some((name, extension)) = value.split_once(".") else {
-            return InvalidFileNameSnafu {
-                reason: format!(
-                    "The file name should be in the format of <name>.<extension>, got: {}",
-                    value
-                ),
-            }
-            .fail();
-        };
-        let extension = FileExtension::try_from(extension)?;
-        Ok(Self {
-            name: name.to_string(),
-            extension,
-        })
-    }
-}
-
-impl FileName {
-    fn new(name: String, extension: FileExtension) -> Self {
-        Self { name, extension }
-    }
-}
-
-/// The manager of the metadata snapshot.
-///
-/// It manages the metadata snapshot, including dumping and restoring.
-pub struct MetadataSnapshotManager {
-    kv_backend: KvBackendRef,
-    object_store: ObjectStore,
-}
-
-/// The maximum size of the request to put metadata, use 1MiB by default.
-const MAX_REQUEST_SIZE: usize = 1024 * 1024;
-
-impl MetadataSnapshotManager {
-    pub fn new(kv_backend: KvBackendRef, object_store: ObjectStore) -> Self {
-        Self {
-            kv_backend,
-            object_store,
-        }
-    }
-
-    /// Restores the metadata from the backup file to the metadata store.
-    pub async fn restore(&self, file_path: &str) -> Result<u64> {
-        let filename = FileName::try_from(
-            file_path
-                .rsplit("/")
-                .next()
-                .context(InvalidFilePathSnafu { file_path })?,
-        )?;
-        let data = self
-            .object_store
-            .read(file_path)
-            .await
-            .context(ReadObjectSnafu { file_path })?;
-        let document = Document::from_slice(&filename.extension.format, &data.to_bytes())?;
-        let metadata_content = document.into_metadata_content()?;
-        let mut req = BatchPutRequest::default();
-        let mut total_request_size = 0;
-        let mut count = 0;
-        let now = Instant::now();
-        for FileKeyValue { key, value } in metadata_content.into_iter() {
-            count += 1;
-            let key_size = key.len();
-            let value_size = value.len();
-            if total_request_size + key_size + value_size > MAX_REQUEST_SIZE {
-                self.kv_backend.batch_put(req).await?;
-                req = BatchPutRequest::default();
-                total_request_size = 0;
-            }
-            req.kvs.push(KeyValue { key, value });
-            total_request_size += key_size + value_size;
-        }
-        if !req.kvs.is_empty() {
-            self.kv_backend.batch_put(req).await?;
-        }
-
-        info!(
-            "Restored metadata from {} successfully, total {} key-value pairs, elapsed {:?}",
-            file_path,
-            count,
-            now.elapsed()
-        );
-        Ok(count)
-    }
-
-    /// Dumps the metadata to the backup file.
-    pub async fn dump(&self, path: &str, filename: &str) -> Result<(String, u64)> {
-        let format = FileFormat::FlexBuffers;
-        let filename = FileName::new(
-            filename.to_string(),
-            FileExtension {
-                format,
-                data_type: DataType::Metadata,
-            },
-        );
-        let file_path = format!("{}/{}", path.trim_end_matches('/'), filename);
-        let now = Instant::now();
-        let req = RangeRequest::new().with_range(vec![0], vec![0]);
-        let stream = PaginationStream::new(self.kv_backend.clone(), req, DEFAULT_PAGE_SIZE, |kv| {
-            Ok(FileKeyValue {
-                key: kv.key,
-                value: kv.value,
-            })
-        })
-        .into_stream();
-        let keyvalues = stream.try_collect::<Vec<_>>().await?;
-        let num_keyvalues = keyvalues.len();
-        let document = Document::new(
-            Metadata::new(),
-            file::Content::Metadata(MetadataContent::new(keyvalues)),
-        );
-        let bytes = document.to_bytes(&format)?;
-        let r = self
-            .object_store
-            .write(&file_path, bytes)
-            .await
-            .context(WriteObjectSnafu {
-                file_path: &file_path,
-            })?;
-        info!(
-            "Dumped metadata to {} successfully, total {} key-value pairs, file size {} bytes, elapsed {:?}",
-            file_path,
-            num_keyvalues,
-            r.content_length(),
-            now.elapsed()
-        );
-
-        Ok((filename.to_string(), num_keyvalues as u64))
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use std::assert_matches::assert_matches;
-    use std::sync::Arc;
-
-    use common_test_util::temp_dir::{create_temp_dir, TempDir};
-    use object_store::services::Fs;
-
-    use super::*;
-    use crate::kv_backend::memory::MemoryKvBackend;
-    use crate::kv_backend::KvBackend;
-    use crate::rpc::store::PutRequest;
-
-    #[test]
-    fn test_file_name() {
-        let file_name = FileName::try_from("test.metadata.fb").unwrap();
-        assert_eq!(file_name.name, "test");
-        assert_eq!(file_name.extension.format, FileFormat::FlexBuffers);
-        assert_eq!(file_name.extension.data_type, DataType::Metadata);
-        assert_eq!(file_name.to_string(), "test.metadata.fb");
-
-        let invalid_file_name = FileName::try_from("test.metadata").unwrap_err();
-        assert_eq!(
-            invalid_file_name.to_string(),
-            "Invalid file extension: Extension should be in the format of <datatype>.<format>, got: metadata"
-        );
-
-        let invalid_file_extension = FileName::try_from("test.metadata.hello").unwrap_err();
-        assert_eq!(
-            invalid_file_extension.to_string(),
-            "Invalid file extension: Invalid file format: hello"
-        );
-    }
-
-    fn test_env(
-        prefix: &str,
-    ) -> (
-        TempDir,
-        Arc<MemoryKvBackend<Error>>,
-        MetadataSnapshotManager,
-    ) {
-        let temp_dir = create_temp_dir(prefix);
-        let kv_backend = Arc::new(MemoryKvBackend::default());
-        let temp_path = temp_dir.path();
-        let data_path = temp_path.join("data").as_path().display().to_string();
-        let builder = Fs::default().root(&data_path);
-        let object_store = ObjectStore::new(builder).unwrap().finish();
-        let manager = MetadataSnapshotManager::new(kv_backend.clone(), object_store);
-        (temp_dir, kv_backend, manager)
-    }
-
-    #[tokio::test]
-    async fn test_dump_and_restore() {
-        common_telemetry::init_default_ut_logging();
-        let (temp_dir, kv_backend, manager) = test_env("test_dump_and_restore");
-        let temp_path = temp_dir.path();
-
-        for i in 0..10 {
-            kv_backend
-                .put(
-                    PutRequest::new()
-                        .with_key(format!("test_{}", i).as_bytes().to_vec())
-                        .with_value(format!("value_{}", i).as_bytes().to_vec()),
-                )
-                .await
-                .unwrap();
-        }
-        let dump_path = temp_path.join("snapshot");
-        manager
-            .dump(
-                &dump_path.as_path().display().to_string(),
-                "metadata_snapshot",
-            )
-            .await
-            .unwrap();
-        // Clean up the kv backend
-        kv_backend.clear();
-
-        let restore_path = dump_path
-            .join("metadata_snapshot.metadata.fb")
-            .as_path()
-            .display()
-            .to_string();
-        manager.restore(&restore_path).await.unwrap();
-
-        for i in 0..10 {
-            let key = format!("test_{}", i);
-            let value = kv_backend.get(key.as_bytes()).await.unwrap().unwrap();
-            assert_eq!(value.value, format!("value_{}", i).as_bytes());
-        }
-    }
-
-    #[tokio::test]
-    async fn test_restore_from_nonexistent_file() {
-        let (temp_dir, _kv_backend, manager) = test_env("test_restore_from_nonexistent_file");
-        let restore_path = temp_dir
-            .path()
-            .join("nonexistent.metadata.fb")
-            .as_path()
-            .display()
-            .to_string();
-        let err = manager.restore(&restore_path).await.unwrap_err();
-        assert_matches!(err, Error::ReadObject { .. })
-    }
-}

src/common/meta/src/snapshot/file.rs

@@ -1,145 +0,0 @@
-// Copyright 2023 Greptime Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-use common_time::util::current_time_millis;
-use flexbuffers::{FlexbufferSerializer, Reader};
-use serde::{Deserialize, Serialize};
-use snafu::ResultExt;
-
-use crate::error::{
-    DeserializeFlexbuffersSnafu, ReadFlexbuffersSnafu, Result, SerializeFlexbuffersSnafu,
-};
-use crate::snapshot::FileFormat;
-
-/// The layout of the backup file.
-#[derive(Debug, PartialEq, Serialize, Deserialize)]
-pub(crate) struct Document {
-    metadata: Metadata,
-    content: Content,
-}
-
-impl Document {
-    /// Creates a new document.
-    pub fn new(metadata: Metadata, content: Content) -> Self {
-        Self { metadata, content }
-    }
-
-    fn serialize_to_flexbuffer(&self) -> Result<Vec<u8>> {
-        let mut builder = FlexbufferSerializer::new();
-        self.serialize(&mut builder)
-            .context(SerializeFlexbuffersSnafu)?;
-        Ok(builder.take_buffer())
-    }
-
-    /// Converts the [`Document`] to a bytes.
-    pub(crate) fn to_bytes(&self, format: &FileFormat) -> Result<Vec<u8>> {
-        match format {
-            FileFormat::FlexBuffers => self.serialize_to_flexbuffer(),
-        }
-    }
-
-    fn deserialize_from_flexbuffer(data: &[u8]) -> Result<Self> {
-        let reader = Reader::get_root(data).context(ReadFlexbuffersSnafu)?;
-        Document::deserialize(reader).context(DeserializeFlexbuffersSnafu)
-    }
-
-    /// Deserializes the [`Document`] from a bytes.
-    pub(crate) fn from_slice(format: &FileFormat, data: &[u8]) -> Result<Self> {
-        match format {
-            FileFormat::FlexBuffers => Self::deserialize_from_flexbuffer(data),
-        }
-    }
-
-    /// Converts the [`Document`] to a [`MetadataContent`].
-    pub(crate) fn into_metadata_content(self) -> Result<MetadataContent> {
-        match self.content {
-            Content::Metadata(metadata) => Ok(metadata),
-        }
-    }
-}
-
-/// The metadata of the backup file.
-#[derive(Debug, PartialEq, Serialize, Deserialize)]
-pub(crate) struct Metadata {
-    // UNIX_EPOCH in milliseconds.
-    created_timestamp_mills: i64,
-}
-
-impl Metadata {
-    /// Create a new metadata.
-    ///
-    /// The `created_timestamp_mills` will be the current time in milliseconds.
-    pub fn new() -> Self {
-        Self {
-            created_timestamp_mills: current_time_millis(),
-        }
-    }
-}
-
-/// The content of the backup file.
-#[derive(Debug, PartialEq, Serialize, Deserialize)]
-pub(crate) enum Content {
-    Metadata(MetadataContent),
-}
-
-/// The content of the backup file.
-#[derive(Debug, PartialEq, Serialize, Deserialize)]
-pub(crate) struct MetadataContent {
-    values: Vec<KeyValue>,
-}
-
-impl MetadataContent {
-    /// Create a new metadata content.
-    pub fn new(values: impl IntoIterator<Item = KeyValue>) -> Self {
-        Self {
-            values: values.into_iter().collect(),
-        }
-    }
-
-    /// Returns an iterator over the key-value pairs.
-    pub fn into_iter(self) -> impl Iterator<Item = KeyValue> {
-        self.values.into_iter()
-    }
-}
-
-/// The key-value pair of the backup file.
-#[derive(Debug, PartialEq, Serialize, Deserialize)]
-pub(crate) struct KeyValue {
-    pub key: Vec<u8>,
-    pub value: Vec<u8>,
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn test_document() {
-        let document = Document::new(
-            Metadata::new(),
-            Content::Metadata(MetadataContent::new(vec![KeyValue {
-                key: b"key".to_vec(),
-                value: b"value".to_vec(),
-            }])),
-        );
-
-        let bytes = document.to_bytes(&FileFormat::FlexBuffers).unwrap();
-        let document_deserialized = Document::from_slice(&FileFormat::FlexBuffers, &bytes).unwrap();
-        assert_eq!(
-            document.metadata.created_timestamp_mills,
-            document_deserialized.metadata.created_timestamp_mills
-        );
-        assert_eq!(document.content, document_deserialized.content);
-    }
-}

src/common/query/src/error.rs

@@ -162,13 +162,6 @@ pub enum Error {
         location: Location,
     },
 
-    #[snafu(display("Failed to do metadata snapshot"))]
-    MetadataSnapshot {
-        source: BoxedError,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
     #[snafu(display("Failed to do procedure task"))]
     ProcedureService {
         source: BoxedError,
@@ -194,12 +187,6 @@ pub enum Error {
         location: Location,
     },
 
-    #[snafu(display("Missing MetadataSnapshotHandler, not expected"))]
-    MissingMetadataSnapshotHandler {
-        #[snafu(implicit)]
-        location: Location,
-    },
-
     #[snafu(display("Invalid function args: {}", err_msg))]
     InvalidFuncArgs {
         err_msg: String,
@@ -264,7 +251,6 @@ impl ErrorExt for Error {
             Error::MissingTableMutationHandler { .. }
             | Error::MissingProcedureServiceHandler { .. }
             | Error::MissingFlowServiceHandler { .. }
-            | Error::MissingMetadataSnapshotHandler { .. }
             | Error::RegisterUdf { .. } => StatusCode::Unexpected,
 
             Error::UnsupportedInputDataType { .. }
@@ -276,8 +262,7 @@ impl ErrorExt for Error {
             Error::DecodePlan { source, .. }
             | Error::Execute { source, .. }
             | Error::ProcedureService { source, .. }
-            | Error::TableMutation { source, .. }
-            | Error::MetadataSnapshot { source, .. } => source.status_code(),
+            | Error::TableMutation { source, .. } => source.status_code(),
 
             Error::PermissionDenied { .. } => StatusCode::PermissionDenied,
         }

src/common/query/src/logical_plan.rs

@@ -18,16 +18,19 @@ mod udaf;
 
 use std::sync::Arc;
 
+use api::v1::TableName;
 use datafusion::catalog::CatalogProviderList;
 use datafusion::error::Result as DatafusionResult;
 use datafusion::logical_expr::{LogicalPlan, LogicalPlanBuilder};
-use datafusion_common::Column;
-use datafusion_expr::col;
+use datafusion_common::{Column, TableReference};
+use datafusion_expr::dml::InsertOp;
+use datafusion_expr::{col, DmlStatement, WriteOp};
 pub use expr::{build_filter_from_timestamp, build_same_type_ts_filter};
+use snafu::ResultExt;
 
 pub use self::accumulator::{Accumulator, AggregateFunctionCreator, AggregateFunctionCreatorRef};
 pub use self::udaf::AggregateFunction;
-use crate::error::Result;
+use crate::error::{GeneralDataFusionSnafu, Result};
 use crate::logical_plan::accumulator::*;
 use crate::signature::{Signature, Volatility};
 
@@ -79,6 +82,74 @@ pub fn rename_logical_plan_columns(
     LogicalPlanBuilder::from(plan).project(projection)?.build()
 }
 
+/// Convert a insert into logical plan to an (table_name, logical_plan)
+/// where table_name is the name of the table to insert into.
+/// logical_plan is the plan to be executed.
+///
+/// if input logical plan is not `insert into table_name <input>`, return None
+///
+/// Returned TableName will use provided catalog and schema if not specified in the logical plan,
+/// if table scan in logical plan have full table name, will **NOT** override it.
+pub fn breakup_insert_plan(
+    plan: &LogicalPlan,
+    default_catalog: &str,
+    default_schema: &str,
+) -> Option<(TableName, Arc<LogicalPlan>)> {
+    if let LogicalPlan::Dml(dml) = plan {
+        if dml.op != WriteOp::Insert(InsertOp::Append) {
+            return None;
+        }
+        let table_name = &dml.table_name;
+        let table_name = match table_name {
+            TableReference::Bare { table } => TableName {
+                catalog_name: default_catalog.to_string(),
+                schema_name: default_schema.to_string(),
+                table_name: table.to_string(),
+            },
+            TableReference::Partial { schema, table } => TableName {
+                catalog_name: default_catalog.to_string(),
+                schema_name: schema.to_string(),
+                table_name: table.to_string(),
+            },
+            TableReference::Full {
+                catalog,
+                schema,
+                table,
+            } => TableName {
+                catalog_name: catalog.to_string(),
+                schema_name: schema.to_string(),
+                table_name: table.to_string(),
+            },
+        };
+        let logical_plan = dml.input.clone();
+        Some((table_name, logical_plan))
+    } else {
+        None
+    }
+}
+
+/// create a `insert into table_name <input>` logical plan
+pub fn add_insert_to_logical_plan(
+    table_name: TableName,
+    table_schema: datafusion_common::DFSchemaRef,
+    input: LogicalPlan,
+) -> Result<LogicalPlan> {
+    let table_name = TableReference::Full {
+        catalog: table_name.catalog_name.into(),
+        schema: table_name.schema_name.into(),
+        table: table_name.table_name.into(),
+    };
+
+    let plan = LogicalPlan::Dml(DmlStatement::new(
+        table_name,
+        table_schema,
+        WriteOp::Insert(InsertOp::Append),
+        Arc::new(input),
+    ));
+    let plan = plan.recompute_schema().context(GeneralDataFusionSnafu)?;
+    Ok(plan)
+}
+
 /// The datafusion `[LogicalPlan]` decoder.
 #[async_trait::async_trait]
 pub trait SubstraitPlanDecoder {

src/common/wal/src/config/kafka/common.rs

@@ -30,10 +30,10 @@ pub const DEFAULT_BACKOFF_CONFIG: BackoffConfig = BackoffConfig {
     deadline: Some(Duration::from_secs(120)),
 };
 
-/// Default interval for active WAL pruning.
-pub const DEFAULT_ACTIVE_PRUNE_INTERVAL: Duration = Duration::ZERO;
-/// Default limit for concurrent active pruning tasks.
-pub const DEFAULT_ACTIVE_PRUNE_TASK_LIMIT: usize = 10;
+/// Default interval for auto WAL pruning.
+pub const DEFAULT_AUTO_PRUNE_INTERVAL: Duration = Duration::ZERO;
+/// Default limit for concurrent auto pruning tasks.
+pub const DEFAULT_AUTO_PRUNE_PARALLELISM: usize = 10;
 /// Default interval for sending flush request to regions when pruning remote WAL.
 pub const DEFAULT_TRIGGER_FLUSH_THRESHOLD: u64 = 0;
 

src/common/wal/src/config/kafka/datanode.rs

@@ -18,8 +18,8 @@ use common_base::readable_size::ReadableSize;
 use serde::{Deserialize, Serialize};
 
 use crate::config::kafka::common::{
-    KafkaConnectionConfig, KafkaTopicConfig, DEFAULT_ACTIVE_PRUNE_INTERVAL,
-    DEFAULT_ACTIVE_PRUNE_TASK_LIMIT, DEFAULT_TRIGGER_FLUSH_THRESHOLD,
+    KafkaConnectionConfig, KafkaTopicConfig, DEFAULT_AUTO_PRUNE_INTERVAL,
+    DEFAULT_AUTO_PRUNE_PARALLELISM, DEFAULT_TRIGGER_FLUSH_THRESHOLD,
 };
 
 /// Kafka wal configurations for datanode.
@@ -47,9 +47,8 @@ pub struct DatanodeKafkaConfig {
     pub dump_index_interval: Duration,
     /// Ignore missing entries during read WAL.
     pub overwrite_entry_start_id: bool,
-    // Active WAL pruning.
-    pub auto_prune_topic_records: bool,
     // Interval of WAL pruning.
+    #[serde(with = "humantime_serde")]
     pub auto_prune_interval: Duration,
     // Threshold for sending flush request when pruning remote WAL.
     // `None` stands for never sending flush request.
@@ -70,10 +69,9 @@ impl Default for DatanodeKafkaConfig {
             create_index: true,
             dump_index_interval: Duration::from_secs(60),
             overwrite_entry_start_id: false,
-            auto_prune_topic_records: false,
-            auto_prune_interval: DEFAULT_ACTIVE_PRUNE_INTERVAL,
+            auto_prune_interval: DEFAULT_AUTO_PRUNE_INTERVAL,
             trigger_flush_threshold: DEFAULT_TRIGGER_FLUSH_THRESHOLD,
-            auto_prune_parallelism: DEFAULT_ACTIVE_PRUNE_TASK_LIMIT,
+            auto_prune_parallelism: DEFAULT_AUTO_PRUNE_PARALLELISM,
         }
     }
 }

src/common/wal/src/config/kafka/metasrv.rs

@@ -17,8 +17,8 @@ use std::time::Duration;
 use serde::{Deserialize, Serialize};
 
 use crate::config::kafka::common::{
-    KafkaConnectionConfig, KafkaTopicConfig, DEFAULT_ACTIVE_PRUNE_INTERVAL,
-    DEFAULT_ACTIVE_PRUNE_TASK_LIMIT, DEFAULT_TRIGGER_FLUSH_THRESHOLD,
+    KafkaConnectionConfig, KafkaTopicConfig, DEFAULT_AUTO_PRUNE_INTERVAL,
+    DEFAULT_AUTO_PRUNE_PARALLELISM, DEFAULT_TRIGGER_FLUSH_THRESHOLD,
 };
 
 /// Kafka wal configurations for metasrv.
@@ -34,6 +34,7 @@ pub struct MetasrvKafkaConfig {
     // Automatically create topics for WAL.
     pub auto_create_topics: bool,
     // Interval of WAL pruning.
+    #[serde(with = "humantime_serde")]
     pub auto_prune_interval: Duration,
     // Threshold for sending flush request when pruning remote WAL.
     // `None` stands for never sending flush request.
@@ -48,9 +49,9 @@ impl Default for MetasrvKafkaConfig {
             connection: Default::default(),
             kafka_topic: Default::default(),
             auto_create_topics: true,
-            auto_prune_interval: DEFAULT_ACTIVE_PRUNE_INTERVAL,
+            auto_prune_interval: DEFAULT_AUTO_PRUNE_INTERVAL,
             trigger_flush_threshold: DEFAULT_TRIGGER_FLUSH_THRESHOLD,
-            auto_prune_parallelism: DEFAULT_ACTIVE_PRUNE_TASK_LIMIT,
+            auto_prune_parallelism: DEFAULT_AUTO_PRUNE_PARALLELISM,
         }
     }
 }

src/datanode/src/datanode.rs

@@ -57,9 +57,9 @@ use tokio::sync::Notify;
 
 use crate::config::{DatanodeOptions, RegionEngineConfig, StorageConfig};
 use crate::error::{
-    self, BuildMitoEngineSnafu, CreateDirSnafu, GetMetadataSnafu, MissingCacheSnafu,
-    MissingKvBackendSnafu, MissingNodeIdSnafu, OpenLogStoreSnafu, Result, ShutdownInstanceSnafu,
-    ShutdownServerSnafu, StartServerSnafu,
+    self, BuildMetricEngineSnafu, BuildMitoEngineSnafu, CreateDirSnafu, GetMetadataSnafu,
+    MissingCacheSnafu, MissingKvBackendSnafu, MissingNodeIdSnafu, OpenLogStoreSnafu, Result,
+    ShutdownInstanceSnafu, ShutdownServerSnafu, StartServerSnafu,
 };
 use crate::event_listener::{
     new_region_server_event_channel, NoopRegionServerEventListener, RegionServerEventListenerRef,
@@ -357,7 +357,6 @@ impl DatanodeBuilder {
             None,
             None,
             None,
-            None,
             false,
             self.plugins.clone(),
             opts.query.clone(),
@@ -399,44 +398,46 @@ impl DatanodeBuilder {
         schema_metadata_manager: SchemaMetadataManagerRef,
         plugins: Plugins,
     ) -> Result<Vec<RegionEngineRef>> {
-        let mut engines = vec![];
-        let mut metric_engine_config = opts.region_engine.iter().find_map(|c| match c {
-            RegionEngineConfig::Metric(config) => Some(config.clone()),
-            _ => None,
-        });
+        let mut metric_engine_config = metric_engine::config::EngineConfig::default();
+        let mut mito_engine_config = MitoConfig::default();
+        let mut file_engine_config = file_engine::config::EngineConfig::default();
 
         for engine in &opts.region_engine {
             match engine {
                 RegionEngineConfig::Mito(config) => {
-                    let mito_engine = Self::build_mito_engine(
-                        opts,
-                        object_store_manager.clone(),
-                        config.clone(),
-                        schema_metadata_manager.clone(),
-                        plugins.clone(),
-                    )
-                    .await?;
-
-                    let metric_engine = MetricEngine::new(
-                        mito_engine.clone(),
-                        metric_engine_config.take().unwrap_or_default(),
-                    );
-                    engines.push(Arc::new(mito_engine) as _);
-                    engines.push(Arc::new(metric_engine) as _);
+                    mito_engine_config = config.clone();
                 }
                 RegionEngineConfig::File(config) => {
-                    let engine = FileRegionEngine::new(
-                        config.clone(),
-                        object_store_manager.default_object_store().clone(), // TODO: implement custom storage for file engine
-                    );
-                    engines.push(Arc::new(engine) as _);
+                    file_engine_config = config.clone();
                 }
-                RegionEngineConfig::Metric(_) => {
-                    // Already handled in `build_mito_engine`.
+                RegionEngineConfig::Metric(metric_config) => {
+                    metric_engine_config = metric_config.clone();
                 }
             }
         }
-        Ok(engines)
+
+        let mito_engine = Self::build_mito_engine(
+            opts,
+            object_store_manager.clone(),
+            mito_engine_config,
+            schema_metadata_manager.clone(),
+            plugins.clone(),
+        )
+        .await?;
+
+        let metric_engine = MetricEngine::try_new(mito_engine.clone(), metric_engine_config)
+            .context(BuildMetricEngineSnafu)?;
+
+        let file_engine = FileRegionEngine::new(
+            file_engine_config,
+            object_store_manager.default_object_store().clone(), // TODO: implement custom storage for file engine
+        );
+
+        Ok(vec![
+            Arc::new(mito_engine) as _,
+            Arc::new(metric_engine) as _,
+            Arc::new(file_engine) as _,
+        ])
     }
 
     /// Builds [MitoEngine] according to options.

src/datanode/src/error.rs

@@ -336,6 +336,13 @@ pub enum Error {
         location: Location,
     },
 
+    #[snafu(display("Failed to build metric engine"))]
+    BuildMetricEngine {
+        source: metric_engine::error::Error,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
     #[snafu(display("Failed to serialize options to TOML"))]
     TomlFormat {
         #[snafu(implicit)]
@@ -452,6 +459,7 @@ impl ErrorExt for Error {
 
             FindLogicalRegions { source, .. } => source.status_code(),
             BuildMitoEngine { source, .. } => source.status_code(),
+            BuildMetricEngine { source, .. } => source.status_code(),
             ConcurrentQueryLimiterClosed { .. } | ConcurrentQueryLimiterTimeout { .. } => {
                 StatusCode::RegionBusy
             }

src/flow/src/adapter.rs

@@ -58,7 +58,7 @@ use crate::metrics::{METRIC_FLOW_INSERT_ELAPSED, METRIC_FLOW_ROWS, METRIC_FLOW_R
 use crate::repr::{self, DiffRow, RelationDesc, Row, BATCH_SIZE};
 use crate::{CreateFlowArgs, FlowId, TableName};
 
-mod flownode_impl;
+pub(crate) mod flownode_impl;
 mod parse_expr;
 pub(crate) mod refill;
 mod stat;
@@ -135,12 +135,13 @@ impl Configurable for FlownodeOptions {
 }
 
 /// Arc-ed FlowNodeManager, cheaper to clone
-pub type FlowWorkerManagerRef = Arc<FlowWorkerManager>;
+pub type FlowStreamingEngineRef = Arc<StreamingEngine>;
 
 /// FlowNodeManager manages the state of all tasks in the flow node, which should be run on the same thread
 ///
 /// The choice of timestamp is just using current system timestamp for now
-pub struct FlowWorkerManager {
+///
+pub struct StreamingEngine {
     /// The handler to the worker that will run the dataflow
     /// which is `!Send` so a handle is used
     pub worker_handles: Vec<WorkerHandle>,
@@ -158,7 +159,8 @@ pub struct FlowWorkerManager {
     flow_err_collectors: RwLock<BTreeMap<FlowId, ErrCollector>>,
     src_send_buf_lens: RwLock<BTreeMap<TableId, watch::Receiver<usize>>>,
     tick_manager: FlowTickManager,
-    node_id: Option<u32>,
+    /// This node id is only available in distributed mode, on standalone mode this is guaranteed to be `None`
+    pub node_id: Option<u32>,
     /// Lock for flushing, will be `read` by `handle_inserts` and `write` by `flush_flow`
     ///
     /// So that a series of event like `inserts -> flush` can be handled correctly
@@ -168,7 +170,7 @@ pub struct FlowWorkerManager {
 }
 
 /// Building FlownodeManager
-impl FlowWorkerManager {
+impl StreamingEngine {
     /// set frontend invoker
     pub async fn set_frontend_invoker(&self, frontend: FrontendInvoker) {
         *self.frontend_invoker.write().await = Some(frontend);
@@ -187,7 +189,7 @@ impl FlowWorkerManager {
         let node_context = FlownodeContext::new(Box::new(srv_map.clone()) as _);
         let tick_manager = FlowTickManager::new();
         let worker_handles = Vec::new();
-        FlowWorkerManager {
+        StreamingEngine {
             worker_handles,
             worker_selector: Mutex::new(0),
             query_engine,
@@ -263,7 +265,7 @@ pub fn batches_to_rows_req(batches: Vec<Batch>) -> Result<Vec<DiffRequest>, Erro
 }
 
 /// This impl block contains methods to send writeback requests to frontend
-impl FlowWorkerManager {
+impl StreamingEngine {
     /// Return the number of requests it made
     pub async fn send_writeback_requests(&self) -> Result<usize, Error> {
         let all_reqs = self.generate_writeback_request().await?;
@@ -534,7 +536,7 @@ impl FlowWorkerManager {
 }
 
 /// Flow Runtime related methods
-impl FlowWorkerManager {
+impl StreamingEngine {
     /// Start state report handler, which will receive a sender from HeartbeatTask to send state size report back
     ///
     /// if heartbeat task is shutdown, this future will exit too
@@ -659,7 +661,7 @@ impl FlowWorkerManager {
         }
         // flow is now shutdown, drop frontend_invoker early so a ref cycle(in standalone mode) can be prevent:
         // FlowWorkerManager.frontend_invoker -> FrontendInvoker.inserter
-        // -> Inserter.node_manager -> NodeManager.flownode -> Flownode.flow_worker_manager.frontend_invoker
+        // -> Inserter.node_manager -> NodeManager.flownode -> Flownode.flow_streaming_engine.frontend_invoker
         self.frontend_invoker.write().await.take();
     }
 
@@ -728,7 +730,7 @@ impl FlowWorkerManager {
 }
 
 /// Create&Remove flow
-impl FlowWorkerManager {
+impl StreamingEngine {
     /// remove a flow by it's id
     pub async fn remove_flow_inner(&self, flow_id: FlowId) -> Result<(), Error> {
         for handle in self.worker_handles.iter() {
@@ -746,7 +748,6 @@ impl FlowWorkerManager {
     /// steps to create task:
     /// 1. parse query into typed plan(and optional parse expire_after expr)
     /// 2. render source/sink with output table id and used input table id
-    #[allow(clippy::too_many_arguments)]
     pub async fn create_flow_inner(&self, args: CreateFlowArgs) -> Result<Option<FlowId>, Error> {
         let CreateFlowArgs {
             flow_id,

src/flow/src/adapter/flownode_impl.rs

@@ -20,35 +20,451 @@ use api::v1::flow::{
     flow_request, CreateRequest, DropRequest, FlowRequest, FlowResponse, FlushFlow,
 };
 use api::v1::region::InsertRequests;
+use catalog::CatalogManager;
 use common_error::ext::BoxedError;
 use common_meta::ddl::create_flow::FlowType;
-use common_meta::error::{Result as MetaResult, UnexpectedSnafu};
+use common_meta::error::Result as MetaResult;
+use common_meta::key::flow::FlowMetadataManager;
 use common_runtime::JoinHandle;
-use common_telemetry::{trace, warn};
+use common_telemetry::{error, info, trace, warn};
 use datatypes::value::Value;
+use futures::TryStreamExt;
 use itertools::Itertools;
-use snafu::{IntoError, OptionExt, ResultExt};
+use session::context::QueryContextBuilder;
+use snafu::{ensure, IntoError, OptionExt, ResultExt};
 use store_api::storage::{RegionId, TableId};
+use tokio::sync::{Mutex, RwLock};
 
-use crate::adapter::{CreateFlowArgs, FlowWorkerManager};
+use crate::adapter::{CreateFlowArgs, StreamingEngine};
 use crate::batching_mode::engine::BatchingEngine;
+use crate::batching_mode::{FRONTEND_SCAN_TIMEOUT, MIN_REFRESH_DURATION};
 use crate::engine::FlowEngine;
-use crate::error::{CreateFlowSnafu, FlowNotFoundSnafu, InsertIntoFlowSnafu, InternalSnafu};
+use crate::error::{
+    CreateFlowSnafu, ExternalSnafu, FlowNotFoundSnafu, IllegalCheckTaskStateSnafu,
+    InsertIntoFlowSnafu, InternalSnafu, JoinTaskSnafu, ListFlowsSnafu, NoAvailableFrontendSnafu,
+    SyncCheckTaskSnafu, UnexpectedSnafu,
+};
 use crate::metrics::METRIC_FLOW_TASK_COUNT;
 use crate::repr::{self, DiffRow};
 use crate::{Error, FlowId};
 
+/// Ref to [`FlowDualEngine`]
+pub type FlowDualEngineRef = Arc<FlowDualEngine>;
+
 /// Manage both streaming and batching mode engine
 ///
 /// including create/drop/flush flow
 /// and redirect insert requests to the appropriate engine
 pub struct FlowDualEngine {
-    streaming_engine: Arc<FlowWorkerManager>,
+    streaming_engine: Arc<StreamingEngine>,
     batching_engine: Arc<BatchingEngine>,
     /// helper struct for faster query flow by table id or vice versa
-    src_table2flow: std::sync::RwLock<SrcTableToFlow>,
+    src_table2flow: RwLock<SrcTableToFlow>,
+    flow_metadata_manager: Arc<FlowMetadataManager>,
+    catalog_manager: Arc<dyn CatalogManager>,
+    check_task: tokio::sync::Mutex<Option<ConsistentCheckTask>>,
 }
 
+impl FlowDualEngine {
+    pub fn new(
+        streaming_engine: Arc<StreamingEngine>,
+        batching_engine: Arc<BatchingEngine>,
+        flow_metadata_manager: Arc<FlowMetadataManager>,
+        catalog_manager: Arc<dyn CatalogManager>,
+    ) -> Self {
+        Self {
+            streaming_engine,
+            batching_engine,
+            src_table2flow: RwLock::new(SrcTableToFlow::default()),
+            flow_metadata_manager,
+            catalog_manager,
+            check_task: Mutex::new(None),
+        }
+    }
+
+    /// Determine if the engine is in distributed mode
+    pub fn is_distributed(&self) -> bool {
+        self.streaming_engine.node_id.is_some()
+    }
+
+    pub fn streaming_engine(&self) -> Arc<StreamingEngine> {
+        self.streaming_engine.clone()
+    }
+
+    pub fn batching_engine(&self) -> Arc<BatchingEngine> {
+        self.batching_engine.clone()
+    }
+
+    /// In distributed mode, scan periodically(1s) until available frontend is found, or timeout,
+    /// in standalone mode, return immediately
+    /// notice here if any frontend appear in cluster info this function will return immediately
+    async fn wait_for_available_frontend(&self, timeout: std::time::Duration) -> Result<(), Error> {
+        if !self.is_distributed() {
+            return Ok(());
+        }
+        let frontend_client = self.batching_engine().frontend_client.clone();
+        let sleep_duration = std::time::Duration::from_millis(1_000);
+        let now = std::time::Instant::now();
+        loop {
+            let frontend_list = frontend_client.scan_for_frontend().await?;
+            if !frontend_list.is_empty() {
+                let fe_list = frontend_list
+                    .iter()
+                    .map(|(_, info)| &info.peer.addr)
+                    .collect::<Vec<_>>();
+                info!("Available frontend found: {:?}", fe_list);
+                return Ok(());
+            }
+            let elapsed = now.elapsed();
+            tokio::time::sleep(sleep_duration).await;
+            info!("Waiting for available frontend, elapsed={:?}", elapsed);
+            if elapsed >= timeout {
+                return NoAvailableFrontendSnafu {
+                    timeout,
+                    context: "No available frontend found in cluster info",
+                }
+                .fail();
+            }
+        }
+    }
+
+    /// Try to sync with check task, this is only used in drop flow&flush flow, so a flow id is required
+    ///
+    /// the need to sync is to make sure flush flow actually get called
+    async fn try_sync_with_check_task(
+        &self,
+        flow_id: FlowId,
+        allow_drop: bool,
+    ) -> Result<(), Error> {
+        // this function rarely get called so adding some log is helpful
+        info!("Try to sync with check task for flow {}", flow_id);
+        let mut retry = 0;
+        let max_retry = 10;
+        // keep trying to trigger consistent check
+        while retry < max_retry {
+            if let Some(task) = self.check_task.lock().await.as_ref() {
+                task.trigger(false, allow_drop).await?;
+                break;
+            }
+            retry += 1;
+            tokio::time::sleep(std::time::Duration::from_millis(500)).await;
+        }
+
+        if retry == max_retry {
+            error!(
+                "Can't sync with check task for flow {} with allow_drop={}",
+                flow_id, allow_drop
+            );
+            return SyncCheckTaskSnafu {
+                flow_id,
+                allow_drop,
+            }
+            .fail();
+        }
+        info!("Successfully sync with check task for flow {}", flow_id);
+
+        Ok(())
+    }
+
+    /// Spawn a task to consistently check if all flow tasks in metasrv is created on flownode,
+    /// so on startup, this will create all missing flow tasks, and constantly check at a interval
+    async fn check_flow_consistent(
+        &self,
+        allow_create: bool,
+        allow_drop: bool,
+    ) -> Result<(), Error> {
+        // use nodeid to determine if this is standalone/distributed mode, and retrieve all flows in this node(in distributed mode)/or all flows(in standalone mode)
+        let nodeid = self.streaming_engine.node_id;
+        let should_exists: Vec<_> = if let Some(nodeid) = nodeid {
+            // nodeid is available, so we only need to check flows on this node
+            // which also means we are in distributed mode
+            let to_be_recover = self
+                .flow_metadata_manager
+                .flownode_flow_manager()
+                .flows(nodeid.into())
+                .try_collect::<Vec<_>>()
+                .await
+                .context(ListFlowsSnafu {
+                    id: Some(nodeid.into()),
+                })?;
+            to_be_recover.into_iter().map(|(id, _)| id).collect()
+        } else {
+            // nodeid is not available, so we need to check all flows
+            // which also means we are in standalone mode
+            let all_catalogs = self
+                .catalog_manager
+                .catalog_names()
+                .await
+                .map_err(BoxedError::new)
+                .context(ExternalSnafu)?;
+            let mut all_flow_ids = vec![];
+            for catalog in all_catalogs {
+                let flows = self
+                    .flow_metadata_manager
+                    .flow_name_manager()
+                    .flow_names(&catalog)
+                    .await
+                    .try_collect::<Vec<_>>()
+                    .await
+                    .map_err(BoxedError::new)
+                    .context(ExternalSnafu)?;
+
+                all_flow_ids.extend(flows.into_iter().map(|(_, id)| id.flow_id()));
+            }
+            all_flow_ids
+        };
+        let should_exists = should_exists
+            .into_iter()
+            .map(|i| i as FlowId)
+            .collect::<HashSet<_>>();
+        let actual_exists = self.list_flows().await?.into_iter().collect::<HashSet<_>>();
+        let to_be_created = should_exists
+            .iter()
+            .filter(|id| !actual_exists.contains(id))
+            .collect::<Vec<_>>();
+        let to_be_dropped = actual_exists
+            .iter()
+            .filter(|id| !should_exists.contains(id))
+            .collect::<Vec<_>>();
+
+        if !to_be_created.is_empty() {
+            if allow_create {
+                info!(
+                    "Recovering {} flows: {:?}",
+                    to_be_created.len(),
+                    to_be_created
+                );
+                let mut errors = vec![];
+                for flow_id in to_be_created {
+                    let flow_id = *flow_id;
+                    let info = self
+                        .flow_metadata_manager
+                        .flow_info_manager()
+                        .get(flow_id as u32)
+                        .await
+                        .map_err(BoxedError::new)
+                        .context(ExternalSnafu)?
+                        .context(FlowNotFoundSnafu { id: flow_id })?;
+
+                    let sink_table_name = [
+                        info.sink_table_name().catalog_name.clone(),
+                        info.sink_table_name().schema_name.clone(),
+                        info.sink_table_name().table_name.clone(),
+                    ];
+                    let args = CreateFlowArgs {
+                        flow_id,
+                        sink_table_name,
+                        source_table_ids: info.source_table_ids().to_vec(),
+                        // because recover should only happen on restart the `create_if_not_exists` and `or_replace` can be arbitrary value(since flow doesn't exist)
+                        // but for the sake of consistency and to make sure recover of flow actually happen, we set both to true
+                        // (which is also fine since checks for not allow both to be true is on metasrv and we already pass that)
+                        create_if_not_exists: true,
+                        or_replace: true,
+                        expire_after: info.expire_after(),
+                        comment: Some(info.comment().clone()),
+                        sql: info.raw_sql().clone(),
+                        flow_options: info.options().clone(),
+                        query_ctx: info
+                            .query_context()
+                            .clone()
+                            .map(|ctx| {
+                                ctx.try_into()
+                                    .map_err(BoxedError::new)
+                                    .context(ExternalSnafu)
+                            })
+                            .transpose()?
+                            // or use default QueryContext with catalog_name from info
+                            // to keep compatibility with old version
+                            .or_else(|| {
+                                Some(
+                                    QueryContextBuilder::default()
+                                        .current_catalog(info.catalog_name().to_string())
+                                        .build(),
+                                )
+                            }),
+                    };
+                    if let Err(err) = self
+                        .create_flow(args)
+                        .await
+                        .map_err(BoxedError::new)
+                        .with_context(|_| CreateFlowSnafu {
+                            sql: info.raw_sql().clone(),
+                        })
+                    {
+                        errors.push((flow_id, err));
+                    }
+                }
+                for (flow_id, err) in errors {
+                    warn!("Failed to recreate flow {}, err={:#?}", flow_id, err);
+                }
+            } else {
+                warn!(
+                    "Flownode {:?} found flows not exist in flownode, flow_ids={:?}",
+                    nodeid, to_be_created
+                );
+            }
+        }
+        if !to_be_dropped.is_empty() {
+            if allow_drop {
+                info!("Dropping flows: {:?}", to_be_dropped);
+                let mut errors = vec![];
+                for flow_id in to_be_dropped {
+                    let flow_id = *flow_id;
+                    if let Err(err) = self.remove_flow(flow_id).await {
+                        errors.push((flow_id, err));
+                    }
+                }
+                for (flow_id, err) in errors {
+                    warn!("Failed to drop flow {}, err={:#?}", flow_id, err);
+                }
+            } else {
+                warn!(
+                    "Flownode {:?} found flows not exist in flownode, flow_ids={:?}",
+                    nodeid, to_be_dropped
+                );
+            }
+        }
+        Ok(())
+    }
+
+    // TODO(discord9): consider sync this with heartbeat(might become necessary in the future)
+    pub async fn start_flow_consistent_check_task(self: &Arc<Self>) -> Result<(), Error> {
+        let mut check_task = self.check_task.lock().await;
+        ensure!(
+            check_task.is_none(),
+            IllegalCheckTaskStateSnafu {
+                reason: "Flow consistent check task already exists",
+            }
+        );
+        let task = ConsistentCheckTask::start_check_task(self).await?;
+        *check_task = Some(task);
+        Ok(())
+    }
+
+    pub async fn stop_flow_consistent_check_task(&self) -> Result<(), Error> {
+        info!("Stopping flow consistent check task");
+        let mut check_task = self.check_task.lock().await;
+
+        ensure!(
+            check_task.is_some(),
+            IllegalCheckTaskStateSnafu {
+                reason: "Flow consistent check task does not exist",
+            }
+        );
+
+        check_task.take().unwrap().stop().await?;
+        info!("Stopped flow consistent check task");
+        Ok(())
+    }
+
+    /// TODO(discord9): also add a `exists` api using flow metadata manager's `exists` method
+    async fn flow_exist_in_metadata(&self, flow_id: FlowId) -> Result<bool, Error> {
+        self.flow_metadata_manager
+            .flow_info_manager()
+            .get(flow_id as u32)
+            .await
+            .map_err(BoxedError::new)
+            .context(ExternalSnafu)
+            .map(|info| info.is_some())
+    }
+}
+
+struct ConsistentCheckTask {
+    handle: JoinHandle<()>,
+    shutdown_tx: tokio::sync::mpsc::Sender<()>,
+    trigger_tx: tokio::sync::mpsc::Sender<(bool, bool, tokio::sync::oneshot::Sender<()>)>,
+}
+
+impl ConsistentCheckTask {
+    async fn start_check_task(engine: &Arc<FlowDualEngine>) -> Result<Self, Error> {
+        let engine = engine.clone();
+        let (tx, mut rx) = tokio::sync::mpsc::channel(1);
+        let (trigger_tx, mut trigger_rx) =
+            tokio::sync::mpsc::channel::<(bool, bool, tokio::sync::oneshot::Sender<()>)>(10);
+        let handle = common_runtime::spawn_global(async move {
+            // first check if available frontend is found
+            if let Err(err) = engine
+                .wait_for_available_frontend(FRONTEND_SCAN_TIMEOUT)
+                .await
+            {
+                warn!("No frontend is available yet:\n {err:?}");
+            }
+
+            // then do recover flows, if failed, always retry
+            let mut recover_retry = 0;
+            while let Err(err) = engine.check_flow_consistent(true, false).await {
+                recover_retry += 1;
+                error!(
+                    "Failed to recover flows:\n {err:?}, retry {} in {}s",
+                    recover_retry,
+                    MIN_REFRESH_DURATION.as_secs()
+                );
+                tokio::time::sleep(MIN_REFRESH_DURATION).await;
+            }
+
+            // then do check flows, with configurable allow_create and allow_drop
+            let (mut allow_create, mut allow_drop) = (false, false);
+            let mut ret_signal: Option<tokio::sync::oneshot::Sender<()>> = None;
+            loop {
+                if let Err(err) = engine.check_flow_consistent(allow_create, allow_drop).await {
+                    error!(err; "Failed to check flow consistent");
+                }
+                if let Some(done) = ret_signal.take() {
+                    let _ = done.send(());
+                }
+                tokio::select! {
+                    _ = rx.recv() => break,
+                    incoming = trigger_rx.recv() => if let Some(incoming) = incoming {
+                        (allow_create, allow_drop) = (incoming.0, incoming.1);
+                        ret_signal = Some(incoming.2);
+                    },
+                    _ = tokio::time::sleep(std::time::Duration::from_secs(10)) => {
+                        (allow_create, allow_drop) = (false, false);
+                    },
+                }
+            }
+        });
+        Ok(ConsistentCheckTask {
+            handle,
+            shutdown_tx: tx,
+            trigger_tx,
+        })
+    }
+
+    async fn trigger(&self, allow_create: bool, allow_drop: bool) -> Result<(), Error> {
+        let (tx, rx) = tokio::sync::oneshot::channel();
+        self.trigger_tx
+            .send((allow_create, allow_drop, tx))
+            .await
+            .map_err(|_| {
+                IllegalCheckTaskStateSnafu {
+                    reason: "Failed to send trigger signal",
+                }
+                .build()
+            })?;
+        rx.await.map_err(|_| {
+            IllegalCheckTaskStateSnafu {
+                reason: "Failed to receive trigger signal",
+            }
+            .build()
+        })?;
+        Ok(())
+    }
+
+    async fn stop(self) -> Result<(), Error> {
+        self.shutdown_tx.send(()).await.map_err(|_| {
+            IllegalCheckTaskStateSnafu {
+                reason: "Failed to send shutdown signal",
+            }
+            .build()
+        })?;
+        // abort so no need to wait
+        self.handle.abort();
+        Ok(())
+    }
+}
+
+#[derive(Default)]
 struct SrcTableToFlow {
     /// mapping of table ids to flow ids for streaming mode
     stream: HashMap<TableId, HashSet<FlowId>>,
@@ -138,35 +554,54 @@ impl FlowEngine for FlowDualEngine {
 
         self.src_table2flow
             .write()
-            .unwrap()
+            .await
             .add_flow(flow_id, flow_type, src_table_ids);
 
         Ok(res)
     }
 
     async fn remove_flow(&self, flow_id: FlowId) -> Result<(), Error> {
-        let flow_type = self.src_table2flow.read().unwrap().get_flow_type(flow_id);
+        let flow_type = self.src_table2flow.read().await.get_flow_type(flow_id);
+
         match flow_type {
             Some(FlowType::Batching) => self.batching_engine.remove_flow(flow_id).await,
             Some(FlowType::Streaming) => self.streaming_engine.remove_flow(flow_id).await,
-            None => FlowNotFoundSnafu { id: flow_id }.fail(),
+            None => {
+                // this can happen if flownode just restart, and is stilling creating the flow
+                // since now that this flow should dropped, we need to trigger the consistent check and allow drop
+                // this rely on drop flow ddl delete metadata first, see src/common/meta/src/ddl/drop_flow.rs
+                warn!(
+                    "Flow {} is not exist in the underlying engine, but exist in metadata",
+                    flow_id
+                );
+                self.try_sync_with_check_task(flow_id, true).await?;
+
+                Ok(())
+            }
         }?;
         // remove mapping
-        self.src_table2flow.write().unwrap().remove_flow(flow_id);
+        self.src_table2flow.write().await.remove_flow(flow_id);
         Ok(())
     }
 
     async fn flush_flow(&self, flow_id: FlowId) -> Result<usize, Error> {
-        let flow_type = self.src_table2flow.read().unwrap().get_flow_type(flow_id);
+        // sync with check task
+        self.try_sync_with_check_task(flow_id, false).await?;
+        let flow_type = self.src_table2flow.read().await.get_flow_type(flow_id);
         match flow_type {
             Some(FlowType::Batching) => self.batching_engine.flush_flow(flow_id).await,
             Some(FlowType::Streaming) => self.streaming_engine.flush_flow(flow_id).await,
-            None => FlowNotFoundSnafu { id: flow_id }.fail(),
+            None => {
+                warn!(
+                    "Currently flow={flow_id} doesn't exist in flownode, ignore flush_flow request"
+                );
+                Ok(0)
+            }
         }
     }
 
     async fn flow_exist(&self, flow_id: FlowId) -> Result<bool, Error> {
-        let flow_type = self.src_table2flow.read().unwrap().get_flow_type(flow_id);
+        let flow_type = self.src_table2flow.read().await.get_flow_type(flow_id);
         // not using `flow_type.is_some()` to make sure the flow is actually exist in the underlying engine
         match flow_type {
             Some(FlowType::Batching) => self.batching_engine.flow_exist(flow_id).await,
@@ -175,6 +610,13 @@ impl FlowEngine for FlowDualEngine {
         }
     }
 
+    async fn list_flows(&self) -> Result<impl IntoIterator<Item = FlowId>, Error> {
+        let stream_flows = self.streaming_engine.list_flows().await?;
+        let batch_flows = self.batching_engine.list_flows().await?;
+
+        Ok(stream_flows.into_iter().chain(batch_flows))
+    }
+
     async fn handle_flow_inserts(
         &self,
         request: api::v1::region::InsertRequests,
@@ -184,7 +626,7 @@ impl FlowEngine for FlowDualEngine {
         let mut to_batch_engine = request.requests;
 
         {
-            let src_table2flow = self.src_table2flow.read().unwrap();
+            let src_table2flow = self.src_table2flow.read().await;
             to_batch_engine.retain(|req| {
                 let region_id = RegionId::from(req.region_id);
                 let table_id = region_id.table_id();
@@ -221,12 +663,7 @@ impl FlowEngine for FlowDualEngine {
                 requests: to_batch_engine,
             })
             .await?;
-        stream_handler.await.map_err(|e| {
-            crate::error::UnexpectedSnafu {
-                reason: format!("JoinError when handle inserts for flow stream engine: {e:?}"),
-            }
-            .build()
-        })??;
+        stream_handler.await.context(JoinTaskSnafu)??;
 
         Ok(())
     }
@@ -307,14 +744,7 @@ impl common_meta::node_manager::Flownode for FlowDualEngine {
                     ..Default::default()
                 })
             }
-            None => UnexpectedSnafu {
-                err_msg: "Missing request body",
-            }
-            .fail(),
-            _ => UnexpectedSnafu {
-                err_msg: "Invalid request body.",
-            }
-            .fail(),
+            other => common_meta::error::InvalidFlowRequestBodySnafu { body: other }.fail(),
         }
     }
 
@@ -339,7 +769,7 @@ fn to_meta_err(
 }
 
 #[async_trait::async_trait]
-impl common_meta::node_manager::Flownode for FlowWorkerManager {
+impl common_meta::node_manager::Flownode for StreamingEngine {
     async fn handle(&self, request: FlowRequest) -> MetaResult<FlowResponse> {
         let query_ctx = request
             .header
@@ -413,14 +843,7 @@ impl common_meta::node_manager::Flownode for FlowWorkerManager {
                     ..Default::default()
                 })
             }
-            None => UnexpectedSnafu {
-                err_msg: "Missing request body",
-            }
-            .fail(),
-            _ => UnexpectedSnafu {
-                err_msg: "Invalid request body.",
-            }
-            .fail(),
+            other => common_meta::error::InvalidFlowRequestBodySnafu { body: other }.fail(),
         }
     }
 
@@ -432,7 +855,7 @@ impl common_meta::node_manager::Flownode for FlowWorkerManager {
     }
 }
 
-impl FlowEngine for FlowWorkerManager {
+impl FlowEngine for StreamingEngine {
     async fn create_flow(&self, args: CreateFlowArgs) -> Result<Option<FlowId>, Error> {
         self.create_flow_inner(args).await
     }
@@ -449,6 +872,16 @@ impl FlowEngine for FlowWorkerManager {
         self.flow_exist_inner(flow_id).await
     }
 
+    async fn list_flows(&self) -> Result<impl IntoIterator<Item = FlowId>, Error> {
+        Ok(self
+            .flow_err_collectors
+            .read()
+            .await
+            .keys()
+            .cloned()
+            .collect::<Vec<_>>())
+    }
+
     async fn handle_flow_inserts(
         &self,
         request: api::v1::region::InsertRequests,
@@ -474,7 +907,7 @@ impl FetchFromRow {
     }
 }
 
-impl FlowWorkerManager {
+impl StreamingEngine {
     async fn handle_inserts_inner(
         &self,
         request: InsertRequests,
@@ -552,7 +985,7 @@ impl FlowWorkerManager {
                             .copied()
                             .map(FetchFromRow::Idx)
                             .or_else(|| col_default_val.clone().map(FetchFromRow::Default))
-                            .with_context(|| crate::error::UnexpectedSnafu {
+                            .with_context(|| UnexpectedSnafu {
                                 reason: format!(
                                     "Column not found: {}, default_value: {:?}",
                                     col_name, col_default_val

src/flow/src/adapter/refill.rs

@@ -31,7 +31,7 @@ use snafu::{ensure, OptionExt, ResultExt};
 use table::metadata::TableId;
 
 use crate::adapter::table_source::ManagedTableSource;
-use crate::adapter::{FlowId, FlowWorkerManager, FlowWorkerManagerRef};
+use crate::adapter::{FlowId, FlowStreamingEngineRef, StreamingEngine};
 use crate::error::{FlowNotFoundSnafu, JoinTaskSnafu, UnexpectedSnafu};
 use crate::expr::error::ExternalSnafu;
 use crate::expr::utils::find_plan_time_window_expr_lower_bound;
@@ -39,10 +39,10 @@ use crate::repr::RelationDesc;
 use crate::server::get_all_flow_ids;
 use crate::{Error, FrontendInvoker};
 
-impl FlowWorkerManager {
+impl StreamingEngine {
     /// Create and start refill flow tasks in background
     pub async fn create_and_start_refill_flow_tasks(
-        self: &FlowWorkerManagerRef,
+        self: &FlowStreamingEngineRef,
         flow_metadata_manager: &FlowMetadataManagerRef,
         catalog_manager: &CatalogManagerRef,
     ) -> Result<(), Error> {
@@ -130,7 +130,7 @@ impl FlowWorkerManager {
 
     /// Starting to refill flows, if any error occurs, will rebuild the flow and retry
     pub(crate) async fn starting_refill_flows(
-        self: &FlowWorkerManagerRef,
+        self: &FlowStreamingEngineRef,
         tasks: Vec<RefillTask>,
     ) -> Result<(), Error> {
         // TODO(discord9): add a back pressure mechanism
@@ -266,7 +266,7 @@ impl TaskState<()> {
     fn start_running(
         &mut self,
         task_data: &TaskData,
-        manager: FlowWorkerManagerRef,
+        manager: FlowStreamingEngineRef,
         mut output_stream: SendableRecordBatchStream,
     ) -> Result<(), Error> {
         let data = (*task_data).clone();
@@ -383,7 +383,7 @@ impl RefillTask {
     /// Start running the task in background, non-blocking
     pub async fn start_running(
         &mut self,
-        manager: FlowWorkerManagerRef,
+        manager: FlowStreamingEngineRef,
         invoker: &FrontendInvoker,
     ) -> Result<(), Error> {
         let TaskState::Prepared { sql } = &mut self.state else {

src/flow/src/adapter/stat.rs

@@ -16,9 +16,9 @@ use std::collections::BTreeMap;
 
 use common_meta::key::flow::flow_state::FlowStat;
 
-use crate::FlowWorkerManager;
+use crate::StreamingEngine;
 
-impl FlowWorkerManager {
+impl StreamingEngine {
     pub async fn gen_state_report(&self) -> FlowStat {
         let mut full_report = BTreeMap::new();
         let mut last_exec_time_map = BTreeMap::new();

src/flow/src/adapter/util.rs

@@ -33,8 +33,8 @@ use crate::adapter::table_source::TableDesc;
 use crate::adapter::{TableName, WorkerHandle, AUTO_CREATED_PLACEHOLDER_TS_COL};
 use crate::error::{Error, ExternalSnafu, UnexpectedSnafu};
 use crate::repr::{ColumnType, RelationDesc, RelationType};
-use crate::FlowWorkerManager;
-impl FlowWorkerManager {
+use crate::StreamingEngine;
+impl StreamingEngine {
     /// Get a worker handle for creating flow, using round robin to select a worker
     pub(crate) async fn get_worker_handle_for_create_flow(&self) -> &WorkerHandle {
         let use_idx = {

src/flow/src/batching_mode.rs

@@ -31,4 +31,19 @@ pub const DEFAULT_BATCHING_ENGINE_QUERY_TIMEOUT: Duration = Duration::from_secs(
 pub const SLOW_QUERY_THRESHOLD: Duration = Duration::from_secs(60);
 
 /// The minimum duration between two queries execution by batching mode task
-const MIN_REFRESH_DURATION: Duration = Duration::new(5, 0);
+pub const MIN_REFRESH_DURATION: Duration = Duration::new(5, 0);
+
+/// Grpc connection timeout
+const GRPC_CONN_TIMEOUT: Duration = Duration::from_secs(5);
+
+/// Grpc max retry number
+const GRPC_MAX_RETRIES: u32 = 3;
+
+/// Flow wait for available frontend timeout,
+/// if failed to find available frontend after FRONTEND_SCAN_TIMEOUT elapsed, return error
+/// which should prevent flownode from starting
+pub const FRONTEND_SCAN_TIMEOUT: Duration = Duration::from_secs(30);
+
+/// Frontend activity timeout
+/// if frontend is down(not sending heartbeat) for more than FRONTEND_ACTIVITY_TIMEOUT, it will be removed from the list that flownode use to connect
+pub const FRONTEND_ACTIVITY_TIMEOUT: Duration = Duration::from_secs(60);

src/flow/src/batching_mode/engine.rs

@@ -17,14 +17,16 @@
 use std::collections::{BTreeMap, HashMap};
 use std::sync::Arc;
 
+use catalog::CatalogManagerRef;
 use common_error::ext::BoxedError;
 use common_meta::ddl::create_flow::FlowType;
 use common_meta::key::flow::FlowMetadataManagerRef;
-use common_meta::key::table_info::TableInfoManager;
+use common_meta::key::table_info::{TableInfoManager, TableInfoValue};
 use common_meta::key::TableMetadataManagerRef;
 use common_runtime::JoinHandle;
-use common_telemetry::info;
 use common_telemetry::tracing::warn;
+use common_telemetry::{debug, info};
+use common_time::TimeToLive;
 use query::QueryEngineRef;
 use snafu::{ensure, OptionExt, ResultExt};
 use store_api::storage::RegionId;
@@ -36,7 +38,9 @@ use crate::batching_mode::task::BatchingTask;
 use crate::batching_mode::time_window::{find_time_window_expr, TimeWindowExpr};
 use crate::batching_mode::utils::sql_to_df_plan;
 use crate::engine::FlowEngine;
-use crate::error::{ExternalSnafu, FlowAlreadyExistSnafu, TableNotFoundMetaSnafu, UnexpectedSnafu};
+use crate::error::{
+    ExternalSnafu, FlowAlreadyExistSnafu, TableNotFoundMetaSnafu, UnexpectedSnafu, UnsupportedSnafu,
+};
 use crate::{CreateFlowArgs, Error, FlowId, TableName};
 
 /// Batching mode Engine, responsible for driving all the batching mode tasks
@@ -45,9 +49,11 @@ use crate::{CreateFlowArgs, Error, FlowId, TableName};
 pub struct BatchingEngine {
     tasks: RwLock<BTreeMap<FlowId, BatchingTask>>,
     shutdown_txs: RwLock<BTreeMap<FlowId, oneshot::Sender<()>>>,
-    frontend_client: Arc<FrontendClient>,
+    /// frontend client for insert request
+    pub(crate) frontend_client: Arc<FrontendClient>,
     flow_metadata_manager: FlowMetadataManagerRef,
     table_meta: TableMetadataManagerRef,
+    catalog_manager: CatalogManagerRef,
     query_engine: QueryEngineRef,
 }
 
@@ -57,6 +63,7 @@ impl BatchingEngine {
         query_engine: QueryEngineRef,
         flow_metadata_manager: FlowMetadataManagerRef,
         table_meta: TableMetadataManagerRef,
+        catalog_manager: CatalogManagerRef,
     ) -> Self {
         Self {
             tasks: Default::default(),
@@ -64,6 +71,7 @@ impl BatchingEngine {
             frontend_client,
             flow_metadata_manager,
             table_meta,
+            catalog_manager,
             query_engine,
         }
     }
@@ -179,6 +187,16 @@ async fn get_table_name(
     table_info: &TableInfoManager,
     table_id: &TableId,
 ) -> Result<TableName, Error> {
+    get_table_info(table_info, table_id).await.map(|info| {
+        let name = info.table_name();
+        [name.catalog_name, name.schema_name, name.table_name]
+    })
+}
+
+async fn get_table_info(
+    table_info: &TableInfoManager,
+    table_id: &TableId,
+) -> Result<TableInfoValue, Error> {
     table_info
         .get(*table_id)
         .await
@@ -187,8 +205,7 @@ async fn get_table_name(
         .with_context(|| UnexpectedSnafu {
             reason: format!("Table id = {:?}, couldn't found table name", table_id),
         })
-        .map(|name| name.table_name())
-        .map(|name| [name.catalog_name, name.schema_name, name.table_name])
+        .map(|info| info.into_inner())
 }
 
 impl BatchingEngine {
@@ -248,7 +265,20 @@ impl BatchingEngine {
         let query_ctx = Arc::new(query_ctx);
         let mut source_table_names = Vec::with_capacity(2);
         for src_id in source_table_ids {
+            // also check table option to see if ttl!=instant
             let table_name = get_table_name(self.table_meta.table_info_manager(), &src_id).await?;
+            let table_info = get_table_info(self.table_meta.table_info_manager(), &src_id).await?;
+            ensure!(
+                table_info.table_info.meta.options.ttl != Some(TimeToLive::Instant),
+                UnsupportedSnafu {
+                    reason: format!(
+                        "Source table `{}`(id={}) has instant TTL, Instant TTL is not supported under batching mode. Consider using a TTL longer than flush interval",
+                        table_name.join("."),
+                        src_id
+                    ),
+                }
+            );
+
             source_table_names.push(table_name);
         }
 
@@ -273,7 +303,14 @@ impl BatchingEngine {
             })
             .transpose()?;
 
-        info!("Flow id={}, found time window expr={:?}", flow_id, phy_expr);
+        info!(
+            "Flow id={}, found time window expr={}",
+            flow_id,
+            phy_expr
+                .as_ref()
+                .map(|phy_expr| phy_expr.to_string())
+                .unwrap_or("None".to_string())
+        );
 
         let task = BatchingTask::new(
             flow_id,
@@ -284,7 +321,7 @@ impl BatchingEngine {
             sink_table_name,
             source_table_names,
             query_ctx,
-            self.table_meta.clone(),
+            self.catalog_manager.clone(),
             rx,
         );
 
@@ -295,10 +332,11 @@ impl BatchingEngine {
         // check execute once first to detect any error early
         task.check_execute(&engine, &frontend).await?;
 
-        // TODO(discord9): also save handle & use time wheel or what for better
-        let _handle = common_runtime::spawn_global(async move {
+        // TODO(discord9): use time wheel or what for better
+        let handle = common_runtime::spawn_global(async move {
             task_inner.start_executing_loop(engine, frontend).await;
         });
+        task.state.write().unwrap().task_handle = Some(handle);
 
         // only replace here not earlier because we want the old one intact if something went wrong before this line
         let replaced_old_task_opt = self.tasks.write().await.insert(flow_id, task);
@@ -326,15 +364,23 @@ impl BatchingEngine {
     }
 
     pub async fn flush_flow_inner(&self, flow_id: FlowId) -> Result<usize, Error> {
+        debug!("Try flush flow {flow_id}");
         let task = self.tasks.read().await.get(&flow_id).cloned();
         let task = task.with_context(|| UnexpectedSnafu {
             reason: format!("Can't found task for flow {flow_id}"),
         })?;
 
+        task.mark_all_windows_as_dirty()?;
+
         let res = task
             .gen_exec_once(&self.query_engine, &self.frontend_client)
             .await?;
+
         let affected_rows = res.map(|(r, _)| r).unwrap_or_default() as usize;
+        debug!(
+            "Successfully flush flow {flow_id}, affected rows={}",
+            affected_rows
+        );
         Ok(affected_rows)
     }
 
@@ -357,6 +403,9 @@ impl FlowEngine for BatchingEngine {
     async fn flow_exist(&self, flow_id: FlowId) -> Result<bool, Error> {
         Ok(self.flow_exist_inner(flow_id).await)
     }
+    async fn list_flows(&self) -> Result<impl IntoIterator<Item = FlowId>, Error> {
+        Ok(self.tasks.read().await.keys().cloned().collect::<Vec<_>>())
+    }
     async fn handle_flow_inserts(
         &self,
         request: api::v1::region::InsertRequests,

src/flow/src/batching_mode/frontend_client.rs

@@ -14,44 +14,112 @@
 
 //! Frontend client to run flow as batching task which is time-window-aware normal query triggered every tick set by user
 
-use std::sync::Arc;
+use std::sync::{Arc, Weak};
+use std::time::SystemTime;
 
-use client::{Client, Database, DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
-use common_error::ext::BoxedError;
+use api::v1::greptime_request::Request;
+use api::v1::CreateTableExpr;
+use client::{Client, Database};
+use common_error::ext::{BoxedError, ErrorExt};
 use common_grpc::channel_manager::{ChannelConfig, ChannelManager};
 use common_meta::cluster::{NodeInfo, NodeInfoKey, Role};
 use common_meta::peer::Peer;
 use common_meta::rpc::store::RangeRequest;
+use common_query::Output;
+use common_telemetry::warn;
+use itertools::Itertools;
 use meta_client::client::MetaClient;
-use snafu::ResultExt;
+use servers::query_handler::grpc::GrpcQueryHandler;
+use session::context::{QueryContextBuilder, QueryContextRef};
+use snafu::{OptionExt, ResultExt};
 
-use crate::batching_mode::DEFAULT_BATCHING_ENGINE_QUERY_TIMEOUT;
-use crate::error::{ExternalSnafu, UnexpectedSnafu};
+use crate::batching_mode::{
+    DEFAULT_BATCHING_ENGINE_QUERY_TIMEOUT, FRONTEND_ACTIVITY_TIMEOUT, GRPC_CONN_TIMEOUT,
+    GRPC_MAX_RETRIES,
+};
+use crate::error::{ExternalSnafu, InvalidRequestSnafu, NoAvailableFrontendSnafu, UnexpectedSnafu};
 use crate::Error;
 
-fn default_channel_mgr() -> ChannelManager {
-    let cfg = ChannelConfig::new().timeout(DEFAULT_BATCHING_ENGINE_QUERY_TIMEOUT);
-    ChannelManager::with_config(cfg)
+/// Just like [`GrpcQueryHandler`] but use BoxedError
+///
+/// basically just a specialized `GrpcQueryHandler<Error=BoxedError>`
+///
+/// this is only useful for flownode to
+/// invoke frontend Instance in standalone mode
+#[async_trait::async_trait]
+pub trait GrpcQueryHandlerWithBoxedError: Send + Sync + 'static {
+    async fn do_query(
+        &self,
+        query: Request,
+        ctx: QueryContextRef,
+    ) -> std::result::Result<Output, BoxedError>;
 }
 
-fn client_from_urls(addrs: Vec<String>) -> Client {
-    Client::with_manager_and_urls(default_channel_mgr(), addrs)
+/// auto impl
+#[async_trait::async_trait]
+impl<
+        E: ErrorExt + Send + Sync + 'static,
+        T: GrpcQueryHandler<Error = E> + Send + Sync + 'static,
+    > GrpcQueryHandlerWithBoxedError for T
+{
+    async fn do_query(
+        &self,
+        query: Request,
+        ctx: QueryContextRef,
+    ) -> std::result::Result<Output, BoxedError> {
+        self.do_query(query, ctx).await.map_err(BoxedError::new)
+    }
 }
 
+type HandlerMutable = Arc<std::sync::Mutex<Option<Weak<dyn GrpcQueryHandlerWithBoxedError>>>>;
+
 /// A simple frontend client able to execute sql using grpc protocol
-#[derive(Debug)]
+///
+/// This is for computation-heavy query which need to offload computation to frontend, lifting the load from flownode
+#[derive(Debug, Clone)]
 pub enum FrontendClient {
     Distributed {
         meta_client: Arc<MetaClient>,
+        chnl_mgr: ChannelManager,
     },
     Standalone {
         /// for the sake of simplicity still use grpc even in standalone mode
         /// notice the client here should all be lazy, so that can wait after frontend is booted then make conn
-        /// TODO(discord9): not use grpc under standalone mode
-        database_client: DatabaseWithPeer,
+        database_client: HandlerMutable,
     },
 }
 
+impl FrontendClient {
+    /// Create a new empty frontend client, with a `HandlerMutable` to set the grpc handler later
+    pub fn from_empty_grpc_handler() -> (Self, HandlerMutable) {
+        let handler = Arc::new(std::sync::Mutex::new(None));
+        (
+            Self::Standalone {
+                database_client: handler.clone(),
+            },
+            handler,
+        )
+    }
+
+    pub fn from_meta_client(meta_client: Arc<MetaClient>) -> Self {
+        Self::Distributed {
+            meta_client,
+            chnl_mgr: {
+                let cfg = ChannelConfig::new()
+                    .connect_timeout(GRPC_CONN_TIMEOUT)
+                    .timeout(DEFAULT_BATCHING_ENGINE_QUERY_TIMEOUT);
+                ChannelManager::with_config(cfg)
+            },
+        }
+    }
+
+    pub fn from_grpc_handler(grpc_handler: Weak<dyn GrpcQueryHandlerWithBoxedError>) -> Self {
+        Self::Standalone {
+            database_client: Arc::new(std::sync::Mutex::new(Some(grpc_handler))),
+        }
+    }
+}
+
 #[derive(Debug, Clone)]
 pub struct DatabaseWithPeer {
     pub database: Database,
@@ -62,29 +130,24 @@ impl DatabaseWithPeer {
     fn new(database: Database, peer: Peer) -> Self {
         Self { database, peer }
     }
-}
 
-impl FrontendClient {
-    pub fn from_meta_client(meta_client: Arc<MetaClient>) -> Self {
-        Self::Distributed { meta_client }
-    }
-
-    pub fn from_static_grpc_addr(addr: String) -> Self {
-        let peer = Peer {
-            id: 0,
-            addr: addr.clone(),
-        };
-
-        let client = client_from_urls(vec![addr]);
-        let database = Database::new(DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME, client);
-        Self::Standalone {
-            database_client: DatabaseWithPeer::new(database, peer),
-        }
+    /// Try sending a "SELECT 1" to the database
+    async fn try_select_one(&self) -> Result<(), Error> {
+        // notice here use `sql` for `SELECT 1` return 1 row
+        let _ = self
+            .database
+            .sql("SELECT 1")
+            .await
+            .with_context(|_| InvalidRequestSnafu {
+                context: format!("Failed to handle `SELECT 1` request at {:?}", self.peer),
+            })?;
+        Ok(())
     }
 }
 
 impl FrontendClient {
-    async fn scan_for_frontend(&self) -> Result<Vec<(NodeInfoKey, NodeInfo)>, Error> {
+    /// scan for available frontend from metadata
+    pub(crate) async fn scan_for_frontend(&self) -> Result<Vec<(NodeInfoKey, NodeInfo)>, Error> {
         let Self::Distributed { meta_client, .. } = self else {
             return Ok(vec![]);
         };
@@ -114,35 +177,177 @@ impl FrontendClient {
         Ok(res)
     }
 
-    /// Get the database with max `last_activity_ts`
-    async fn get_last_active_frontend(&self) -> Result<DatabaseWithPeer, Error> {
-        if let Self::Standalone { database_client } = self {
-            return Ok(database_client.clone());
-        }
-
-        let frontends = self.scan_for_frontend().await?;
-        let mut peer = None;
-
-        if let Some((_, val)) = frontends.iter().max_by_key(|(_, val)| val.last_activity_ts) {
-            peer = Some(val.peer.clone());
-        }
-
-        let Some(peer) = peer else {
-            UnexpectedSnafu {
-                reason: format!("No frontend available: {:?}", frontends),
+    /// Get the database with maximum `last_activity_ts`& is able to process query
+    async fn get_latest_active_frontend(
+        &self,
+        catalog: &str,
+        schema: &str,
+    ) -> Result<DatabaseWithPeer, Error> {
+        let Self::Distributed {
+            meta_client: _,
+            chnl_mgr,
+        } = self
+        else {
+            return UnexpectedSnafu {
+                reason: "Expect distributed mode",
             }
-            .fail()?
+            .fail();
         };
-        let client = client_from_urls(vec![peer.addr.clone()]);
-        let database = Database::new(DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME, client);
-        Ok(DatabaseWithPeer::new(database, peer))
+
+        let mut interval = tokio::time::interval(GRPC_CONN_TIMEOUT);
+        interval.tick().await;
+        for retry in 0..GRPC_MAX_RETRIES {
+            let frontends = self.scan_for_frontend().await?;
+            let now_in_ms = SystemTime::now()
+                .duration_since(SystemTime::UNIX_EPOCH)
+                .unwrap()
+                .as_millis() as i64;
+
+            // found node with maximum last_activity_ts
+            for (_, node_info) in frontends
+                .iter()
+                .sorted_by_key(|(_, node_info)| node_info.last_activity_ts)
+                .rev()
+                // filter out frontend that have been down for more than 1 min
+                .filter(|(_, node_info)| {
+                    node_info.last_activity_ts + FRONTEND_ACTIVITY_TIMEOUT.as_millis() as i64
+                        > now_in_ms
+                })
+            {
+                let addr = &node_info.peer.addr;
+                let client = Client::with_manager_and_urls(chnl_mgr.clone(), vec![addr.clone()]);
+                let database = Database::new(catalog, schema, client);
+                let db = DatabaseWithPeer::new(database, node_info.peer.clone());
+                match db.try_select_one().await {
+                    Ok(_) => return Ok(db),
+                    Err(e) => {
+                        warn!(
+                            "Failed to connect to frontend {} on retry={}: \n{e:?}",
+                            addr, retry
+                        );
+                    }
+                }
+            }
+            // no available frontend
+            // sleep and retry
+            interval.tick().await;
+        }
+
+        NoAvailableFrontendSnafu {
+            timeout: GRPC_CONN_TIMEOUT,
+            context: "No available frontend found that is able to process query",
+        }
+        .fail()
     }
 
-    /// Get a database client, and possibly update it before returning.
-    pub async fn get_database_client(&self) -> Result<DatabaseWithPeer, Error> {
+    pub async fn create(
+        &self,
+        create: CreateTableExpr,
+        catalog: &str,
+        schema: &str,
+    ) -> Result<u32, Error> {
+        self.handle(
+            Request::Ddl(api::v1::DdlRequest {
+                expr: Some(api::v1::ddl_request::Expr::CreateTable(create)),
+            }),
+            catalog,
+            schema,
+            &mut None,
+        )
+        .await
+    }
+
+    /// Handle a request to frontend
+    pub(crate) async fn handle(
+        &self,
+        req: api::v1::greptime_request::Request,
+        catalog: &str,
+        schema: &str,
+        peer_desc: &mut Option<PeerDesc>,
+    ) -> Result<u32, Error> {
         match self {
-            Self::Standalone { database_client } => Ok(database_client.clone()),
-            Self::Distributed { meta_client: _ } => self.get_last_active_frontend().await,
+            FrontendClient::Distributed { .. } => {
+                let db = self.get_latest_active_frontend(catalog, schema).await?;
+
+                *peer_desc = Some(PeerDesc::Dist {
+                    peer: db.peer.clone(),
+                });
+
+                db.database
+                    .handle_with_retry(req.clone(), GRPC_MAX_RETRIES)
+                    .await
+                    .with_context(|_| InvalidRequestSnafu {
+                        context: format!("Failed to handle request at {:?}: {:?}", db.peer, req),
+                    })
+            }
+            FrontendClient::Standalone { database_client } => {
+                let ctx = QueryContextBuilder::default()
+                    .current_catalog(catalog.to_string())
+                    .current_schema(schema.to_string())
+                    .build();
+                let ctx = Arc::new(ctx);
+                {
+                    let database_client = {
+                        database_client
+                            .lock()
+                            .map_err(|e| {
+                                UnexpectedSnafu {
+                                    reason: format!("Failed to lock database client: {e}"),
+                                }
+                                .build()
+                            })?
+                            .as_ref()
+                            .context(UnexpectedSnafu {
+                                reason: "Standalone's frontend instance is not set",
+                            })?
+                            .upgrade()
+                            .context(UnexpectedSnafu {
+                                reason: "Failed to upgrade database client",
+                            })?
+                    };
+                    let resp: common_query::Output = database_client
+                        .do_query(req.clone(), ctx)
+                        .await
+                        .map_err(BoxedError::new)
+                        .context(ExternalSnafu)?;
+                    match resp.data {
+                        common_query::OutputData::AffectedRows(rows) => {
+                            Ok(rows.try_into().map_err(|_| {
+                                UnexpectedSnafu {
+                                    reason: format!("Failed to convert rows to u32: {}", rows),
+                                }
+                                .build()
+                            })?)
+                        }
+                        _ => UnexpectedSnafu {
+                            reason: "Unexpected output data",
+                        }
+                        .fail(),
+                    }
+                }
+            }
+        }
+    }
+}
+
+/// Describe a peer of frontend
+#[derive(Debug, Default)]
+pub(crate) enum PeerDesc {
+    /// Distributed mode's frontend peer address
+    Dist {
+        /// frontend peer address
+        peer: Peer,
+    },
+    /// Standalone mode
+    #[default]
+    Standalone,
+}
+
+impl std::fmt::Display for PeerDesc {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            PeerDesc::Dist { peer } => write!(f, "{}", peer.addr),
+            PeerDesc::Standalone => write!(f, "standalone"),
         }
     }
 }

src/flow/src/batching_mode/state.rs

@@ -22,13 +22,14 @@ use common_telemetry::tracing::warn;
 use common_time::Timestamp;
 use datatypes::value::Value;
 use session::context::QueryContextRef;
-use snafu::ResultExt;
+use snafu::{OptionExt, ResultExt};
 use tokio::sync::oneshot;
 use tokio::time::Instant;
 
 use crate::batching_mode::task::BatchingTask;
+use crate::batching_mode::time_window::TimeWindowExpr;
 use crate::batching_mode::MIN_REFRESH_DURATION;
-use crate::error::{DatatypesSnafu, InternalSnafu, TimeSnafu};
+use crate::error::{DatatypesSnafu, InternalSnafu, TimeSnafu, UnexpectedSnafu};
 use crate::{Error, FlowId};
 
 /// The state of the [`BatchingTask`].
@@ -46,6 +47,8 @@ pub struct TaskState {
     exec_state: ExecState,
     /// Shutdown receiver
     pub(crate) shutdown_rx: oneshot::Receiver<()>,
+    /// Task handle
+    pub(crate) task_handle: Option<tokio::task::JoinHandle<()>>,
 }
 impl TaskState {
     pub fn new(query_ctx: QueryContextRef, shutdown_rx: oneshot::Receiver<()>) -> Self {
@@ -56,6 +59,7 @@ impl TaskState {
             dirty_time_windows: Default::default(),
             exec_state: ExecState::Idle,
             shutdown_rx,
+            task_handle: None,
         }
     }
 
@@ -70,7 +74,11 @@ impl TaskState {
     /// wait for at least `last_query_duration`, at most `max_timeout` to start next query
     ///
     /// if have more dirty time window, exec next query immediately
-    pub fn get_next_start_query_time(&self, max_timeout: Option<Duration>) -> Instant {
+    pub fn get_next_start_query_time(
+        &self,
+        flow_id: FlowId,
+        max_timeout: Option<Duration>,
+    ) -> Instant {
         let next_duration = max_timeout
             .unwrap_or(self.last_query_duration)
             .min(self.last_query_duration);
@@ -80,6 +88,12 @@ impl TaskState {
         if self.dirty_time_windows.windows.is_empty() {
             self.last_update_time + next_duration
         } else {
+            debug!(
+                "Flow id = {}, still have {} dirty time window({:?}), execute immediately",
+                flow_id,
+                self.dirty_time_windows.windows.len(),
+                self.dirty_time_windows.windows
+            );
             Instant::now()
         }
     }
@@ -115,6 +129,15 @@ impl DirtyTimeWindows {
         }
     }
 
+    pub fn add_window(&mut self, start: Timestamp, end: Option<Timestamp>) {
+        self.windows.insert(start, end);
+    }
+
+    /// Clean all dirty time windows, useful when can't found time window expr
+    pub fn clean(&mut self) {
+        self.windows.clear();
+    }
+
     /// Generate all filter expressions consuming all time windows
     pub fn gen_filter_exprs(
         &mut self,
@@ -177,6 +200,18 @@ impl DirtyTimeWindows {
 
         let mut expr_lst = vec![];
         for (start, end) in first_nth.into_iter() {
+            // align using time window exprs
+            let (start, end) = if let Some(ctx) = task_ctx {
+                let Some(time_window_expr) = &ctx.config.time_window_expr else {
+                    UnexpectedSnafu {
+                        reason: "time_window_expr is not set",
+                    }
+                    .fail()?
+                };
+                self.align_time_window(start, end, time_window_expr)?
+            } else {
+                (start, end)
+            };
             debug!(
                 "Time window start: {:?}, end: {:?}",
                 start.to_iso8601_string(),
@@ -199,6 +234,30 @@ impl DirtyTimeWindows {
         Ok(expr)
     }
 
+    fn align_time_window(
+        &self,
+        start: Timestamp,
+        end: Option<Timestamp>,
+        time_window_expr: &TimeWindowExpr,
+    ) -> Result<(Timestamp, Option<Timestamp>), Error> {
+        let align_start = time_window_expr.eval(start)?.0.context(UnexpectedSnafu {
+            reason: format!(
+                "Failed to align start time {:?} with time window expr {:?}",
+                start, time_window_expr
+            ),
+        })?;
+        let align_end = end
+            .and_then(|end| {
+                time_window_expr
+                    .eval(end)
+                    // if after aligned, end is the same, then use end(because it's already aligned) else use aligned end
+                    .map(|r| if r.0 == Some(end) { r.0 } else { r.1 })
+                    .transpose()
+            })
+            .transpose()?;
+        Ok((align_start, align_end))
+    }
+
     /// Merge time windows that overlaps or get too close
     pub fn merge_dirty_time_windows(
         &mut self,
@@ -287,8 +346,12 @@ enum ExecState {
 #[cfg(test)]
 mod test {
     use pretty_assertions::assert_eq;
+    use session::context::QueryContext;
 
     use super::*;
+    use crate::batching_mode::time_window::find_time_window_expr;
+    use crate::batching_mode::utils::sql_to_df_plan;
+    use crate::test_utils::create_test_query_engine;
 
     #[test]
     fn test_merge_dirty_time_windows() {
@@ -404,4 +467,59 @@ mod test {
             assert_eq!(expected_filter_expr, to_sql.as_deref());
         }
     }
+
+    #[tokio::test]
+    async fn test_align_time_window() {
+        type TimeWindow = (Timestamp, Option<Timestamp>);
+        struct TestCase {
+            sql: String,
+            aligns: Vec<(TimeWindow, TimeWindow)>,
+        }
+        let testcases: Vec<TestCase> = vec![TestCase{
+            sql: "SELECT date_bin(INTERVAL '5 second', ts) AS time_window FROM numbers_with_ts GROUP BY time_window;".to_string(),
+            aligns: vec![
+                ((Timestamp::new_second(3), None), (Timestamp::new_second(0), None)),
+                ((Timestamp::new_second(8), None), (Timestamp::new_second(5), None)),
+                ((Timestamp::new_second(8), Some(Timestamp::new_second(10))), (Timestamp::new_second(5), Some(Timestamp::new_second(10)))),
+                ((Timestamp::new_second(8), Some(Timestamp::new_second(9))), (Timestamp::new_second(5), Some(Timestamp::new_second(10)))),
+            ],
+        }];
+
+        let query_engine = create_test_query_engine();
+        let ctx = QueryContext::arc();
+        for TestCase { sql, aligns } in testcases {
+            let plan = sql_to_df_plan(ctx.clone(), query_engine.clone(), &sql, true)
+                .await
+                .unwrap();
+
+            let (column_name, time_window_expr, _, df_schema) = find_time_window_expr(
+                &plan,
+                query_engine.engine_state().catalog_manager().clone(),
+                ctx.clone(),
+            )
+            .await
+            .unwrap();
+
+            let time_window_expr = time_window_expr
+                .map(|expr| {
+                    TimeWindowExpr::from_expr(
+                        &expr,
+                        &column_name,
+                        &df_schema,
+                        &query_engine.engine_state().session_state(),
+                    )
+                })
+                .transpose()
+                .unwrap()
+                .unwrap();
+
+            let dirty = DirtyTimeWindows::default();
+            for (before_align, expected_after_align) in aligns {
+                let after_align = dirty
+                    .align_time_window(before_align.0, before_align.1, &time_window_expr)
+                    .unwrap();
+                assert_eq!(expected_after_align, after_align);
+            }
+        }
+    }
 }

src/flow/src/batching_mode/task.rs

@@ -12,33 +12,32 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::collections::HashSet;
+use std::collections::{BTreeSet, HashSet};
 use std::ops::Deref;
 use std::sync::{Arc, RwLock};
 use std::time::{Duration, SystemTime, UNIX_EPOCH};
 
 use api::v1::CreateTableExpr;
 use arrow_schema::Fields;
+use catalog::CatalogManagerRef;
 use common_error::ext::BoxedError;
-use common_meta::key::table_name::TableNameKey;
-use common_meta::key::TableMetadataManagerRef;
+use common_query::logical_plan::breakup_insert_plan;
 use common_telemetry::tracing::warn;
 use common_telemetry::{debug, info};
 use common_time::Timestamp;
+use datafusion::optimizer::analyzer::count_wildcard_rule::CountWildcardRule;
+use datafusion::optimizer::AnalyzerRule;
 use datafusion::sql::unparser::expr_to_sql;
-use datafusion_common::tree_node::TreeNode;
+use datafusion_common::tree_node::{Transformed, TreeNode};
 use datafusion_expr::{DmlStatement, LogicalPlan, WriteOp};
 use datatypes::prelude::ConcreteDataType;
-use datatypes::schema::constraint::NOW_FN;
-use datatypes::schema::{ColumnDefaultConstraint, ColumnSchema};
-use datatypes::value::Value;
+use datatypes::schema::{ColumnSchema, Schema};
 use operator::expr_helper::column_schemas_to_defs;
 use query::query_engine::DefaultSerializer;
 use query::QueryEngineRef;
 use session::context::QueryContextRef;
-use snafu::{OptionExt, ResultExt};
+use snafu::{ensure, OptionExt, ResultExt};
 use substrait::{DFLogicalSubstraitConvertor, SubstraitPlan};
-use table::metadata::RawTableMeta;
 use tokio::sync::oneshot;
 use tokio::sync::oneshot::error::TryRecvError;
 use tokio::time::Instant;
@@ -48,14 +47,16 @@ use crate::batching_mode::frontend_client::FrontendClient;
 use crate::batching_mode::state::TaskState;
 use crate::batching_mode::time_window::TimeWindowExpr;
 use crate::batching_mode::utils::{
-    sql_to_df_plan, AddAutoColumnRewriter, AddFilterRewriter, FindGroupByFinalName,
+    get_table_info_df_schema, sql_to_df_plan, AddAutoColumnRewriter, AddFilterRewriter,
+    FindGroupByFinalName,
 };
 use crate::batching_mode::{
     DEFAULT_BATCHING_ENGINE_QUERY_TIMEOUT, MIN_REFRESH_DURATION, SLOW_QUERY_THRESHOLD,
 };
+use crate::df_optimizer::apply_df_optimizer;
 use crate::error::{
-    ConvertColumnSchemaSnafu, DatafusionSnafu, DatatypesSnafu, ExternalSnafu, InvalidRequestSnafu,
-    SubstraitEncodeLogicalPlanSnafu, TableNotFoundMetaSnafu, TableNotFoundSnafu, UnexpectedSnafu,
+    ConvertColumnSchemaSnafu, DatafusionSnafu, ExternalSnafu, InvalidQuerySnafu,
+    SubstraitEncodeLogicalPlanSnafu, UnexpectedSnafu,
 };
 use crate::metrics::{
     METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME, METRIC_FLOW_BATCHING_ENGINE_SLOW_QUERY,
@@ -73,7 +74,7 @@ pub struct TaskConfig {
     pub expire_after: Option<i64>,
     sink_table_name: [String; 3],
     pub source_table_names: HashSet<[String; 3]>,
-    table_meta: TableMetadataManagerRef,
+    catalog_manager: CatalogManagerRef,
 }
 
 #[derive(Clone)]
@@ -93,7 +94,7 @@ impl BatchingTask {
         sink_table_name: [String; 3],
         source_table_names: Vec<[String; 3]>,
         query_ctx: QueryContextRef,
-        table_meta: TableMetadataManagerRef,
+        catalog_manager: CatalogManagerRef,
         shutdown_rx: oneshot::Receiver<()>,
     ) -> Self {
         Self {
@@ -105,12 +106,42 @@ impl BatchingTask {
                 expire_after,
                 sink_table_name,
                 source_table_names: source_table_names.into_iter().collect(),
-                table_meta,
+                catalog_manager,
             }),
             state: Arc::new(RwLock::new(TaskState::new(query_ctx, shutdown_rx))),
         }
     }
 
+    /// mark time window range (now - expire_after, now) as dirty (or (0, now) if expire_after not set)
+    ///
+    /// useful for flush_flow to flush dirty time windows range
+    pub fn mark_all_windows_as_dirty(&self) -> Result<(), Error> {
+        let now = SystemTime::now();
+        let now = Timestamp::new_second(
+            now.duration_since(UNIX_EPOCH)
+                .expect("Time went backwards")
+                .as_secs() as _,
+        );
+        let lower_bound = self
+            .config
+            .expire_after
+            .map(|e| now.sub_duration(Duration::from_secs(e as _)))
+            .transpose()
+            .map_err(BoxedError::new)
+            .context(ExternalSnafu)?
+            .unwrap_or(Timestamp::new_second(0));
+        debug!(
+            "Flow {} mark range ({:?}, {:?}) as dirty",
+            self.config.flow_id, lower_bound, now
+        );
+        self.state
+            .write()
+            .unwrap()
+            .dirty_time_windows
+            .add_window(lower_bound, Some(now));
+        Ok(())
+    }
+
     /// Test execute, for check syntax or such
     pub async fn check_execute(
         &self,
@@ -148,13 +179,8 @@ impl BatchingTask {
 
     async fn is_table_exist(&self, table_name: &[String; 3]) -> Result<bool, Error> {
         self.config
-            .table_meta
-            .table_name_manager()
-            .exists(TableNameKey {
-                catalog: &table_name[0],
-                schema: &table_name[1],
-                table: &table_name[2],
-            })
+            .catalog_manager
+            .table_exists(&table_name[0], &table_name[1], &table_name[2], None)
             .await
             .map_err(BoxedError::new)
             .context(ExternalSnafu)
@@ -166,8 +192,10 @@ impl BatchingTask {
         frontend_client: &Arc<FrontendClient>,
     ) -> Result<Option<(u32, Duration)>, Error> {
         if let Some(new_query) = self.gen_insert_plan(engine).await? {
+            debug!("Generate new query: {:#?}", new_query);
             self.execute_logical_plan(frontend_client, &new_query).await
         } else {
+            debug!("Generate no query");
             Ok(None)
         }
     }
@@ -176,67 +204,35 @@ impl BatchingTask {
         &self,
         engine: &QueryEngineRef,
     ) -> Result<Option<LogicalPlan>, Error> {
-        let full_table_name = self.config.sink_table_name.clone().join(".");
-
-        let table_id = self
-            .config
-            .table_meta
-            .table_name_manager()
-            .get(common_meta::key::table_name::TableNameKey::new(
-                &self.config.sink_table_name[0],
-                &self.config.sink_table_name[1],
-                &self.config.sink_table_name[2],
-            ))
-            .await
-            .with_context(|_| TableNotFoundMetaSnafu {
-                msg: full_table_name.clone(),
-            })?
-            .map(|t| t.table_id())
-            .with_context(|| TableNotFoundSnafu {
-                name: full_table_name.clone(),
-            })?;
-
-        let table = self
-            .config
-            .table_meta
-            .table_info_manager()
-            .get(table_id)
-            .await
-            .with_context(|_| TableNotFoundMetaSnafu {
-                msg: full_table_name.clone(),
-            })?
-            .with_context(|| TableNotFoundSnafu {
-                name: full_table_name.clone(),
-            })?
-            .into_inner();
-
-        let schema: datatypes::schema::Schema = table
-            .table_info
-            .meta
-            .schema
-            .clone()
-            .try_into()
-            .with_context(|_| DatatypesSnafu {
-                extra: format!(
-                    "Failed to convert schema from raw schema, raw_schema={:?}",
-                    table.table_info.meta.schema
-                ),
-            })?;
-
-        let df_schema = Arc::new(schema.arrow_schema().clone().try_into().with_context(|_| {
-            DatafusionSnafu {
-                context: format!(
-                    "Failed to convert arrow schema to datafusion schema, arrow_schema={:?}",
-                    schema.arrow_schema()
-                ),
-            }
-        })?);
+        let (table, df_schema) = get_table_info_df_schema(
+            self.config.catalog_manager.clone(),
+            self.config.sink_table_name.clone(),
+        )
+        .await?;
 
         let new_query = self
-            .gen_query_with_time_window(engine.clone(), &table.table_info.meta)
+            .gen_query_with_time_window(engine.clone(), &table.meta.schema)
             .await?;
 
         let insert_into = if let Some((new_query, _column_cnt)) = new_query {
+            // first check if all columns in input query exists in sink table
+            // since insert into ref to names in record batch generate by given query
+            let table_columns = df_schema
+                .columns()
+                .into_iter()
+                .map(|c| c.name)
+                .collect::<BTreeSet<_>>();
+            for column in new_query.schema().columns() {
+                ensure!(
+                    table_columns.contains(column.name()),
+                    InvalidQuerySnafu {
+                        reason: format!(
+                            "Column {} not found in sink table with columns {:?}",
+                            column, table_columns
+                        ),
+                    }
+                );
+            }
             // update_at& time index placeholder (if exists) should have default value
             LogicalPlan::Dml(DmlStatement::new(
                 datafusion_common::TableReference::Full {
@@ -251,6 +247,9 @@ impl BatchingTask {
         } else {
             return Ok(None);
         };
+        let insert_into = insert_into.recompute_schema().context(DatafusionSnafu {
+            context: "Failed to recompute schema",
+        })?;
         Ok(Some(insert_into))
     }
 
@@ -259,14 +258,11 @@ impl BatchingTask {
         frontend_client: &Arc<FrontendClient>,
         expr: CreateTableExpr,
     ) -> Result<(), Error> {
-        let db_client = frontend_client.get_database_client().await?;
-        db_client
-            .database
-            .create(expr.clone())
-            .await
-            .with_context(|_| InvalidRequestSnafu {
-                context: format!("Failed to create table with expr: {:?}", expr),
-            })?;
+        let catalog = &self.config.sink_table_name[0];
+        let schema = &self.config.sink_table_name[1];
+        frontend_client
+            .create(expr.clone(), catalog, schema)
+            .await?;
         Ok(())
     }
 
@@ -277,27 +273,78 @@ impl BatchingTask {
     ) -> Result<Option<(u32, Duration)>, Error> {
         let instant = Instant::now();
         let flow_id = self.config.flow_id;
-        let db_client = frontend_client.get_database_client().await?;
-        let peer_addr = db_client.peer.addr;
+
         debug!(
-            "Executing flow {flow_id}(expire_after={:?} secs) on {:?} with query {}",
-            self.config.expire_after, peer_addr, &plan
+            "Executing flow {flow_id}(expire_after={:?} secs) with query {}",
+            self.config.expire_after, &plan
         );
 
-        let timer = METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME
-            .with_label_values(&[flow_id.to_string().as_str()])
-            .start_timer();
+        let catalog = &self.config.sink_table_name[0];
+        let schema = &self.config.sink_table_name[1];
 
-        let message = DFLogicalSubstraitConvertor {}
-            .encode(plan, DefaultSerializer)
-            .context(SubstraitEncodeLogicalPlanSnafu)?;
+        // fix all table ref by make it fully qualified, i.e. "table_name" => "catalog_name.schema_name.table_name"
+        let fixed_plan = plan
+            .clone()
+            .transform_down_with_subqueries(|p| {
+                if let LogicalPlan::TableScan(mut table_scan) = p {
+                    let resolved = table_scan.table_name.resolve(catalog, schema);
+                    table_scan.table_name = resolved.into();
+                    Ok(Transformed::yes(LogicalPlan::TableScan(table_scan)))
+                } else {
+                    Ok(Transformed::no(p))
+                }
+            })
+            .with_context(|_| DatafusionSnafu {
+                context: format!("Failed to fix table ref in logical plan, plan={:?}", plan),
+            })?
+            .data;
 
-        let req = api::v1::greptime_request::Request::Query(api::v1::QueryRequest {
-            query: Some(api::v1::query_request::Query::LogicalPlan(message.to_vec())),
-        });
+        let expanded_plan = CountWildcardRule::new()
+            .analyze(fixed_plan.clone(), &Default::default())
+            .with_context(|_| DatafusionSnafu {
+                context: format!(
+                    "Failed to expand wildcard in logical plan, plan={:?}",
+                    fixed_plan
+                ),
+            })?;
 
-        let res = db_client.database.handle(req).await;
-        drop(timer);
+        let plan = expanded_plan;
+        let mut peer_desc = None;
+
+        let res = {
+            let _timer = METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME
+                .with_label_values(&[flow_id.to_string().as_str()])
+                .start_timer();
+
+            // hack and special handling the insert logical plan
+            let req = if let Some((insert_to, insert_plan)) =
+                breakup_insert_plan(&plan, catalog, schema)
+            {
+                let message = DFLogicalSubstraitConvertor {}
+                    .encode(&insert_plan, DefaultSerializer)
+                    .context(SubstraitEncodeLogicalPlanSnafu)?;
+                api::v1::greptime_request::Request::Query(api::v1::QueryRequest {
+                    query: Some(api::v1::query_request::Query::InsertIntoPlan(
+                        api::v1::InsertIntoPlan {
+                            table_name: Some(insert_to),
+                            logical_plan: message.to_vec(),
+                        },
+                    )),
+                })
+            } else {
+                let message = DFLogicalSubstraitConvertor {}
+                    .encode(&plan, DefaultSerializer)
+                    .context(SubstraitEncodeLogicalPlanSnafu)?;
+
+                api::v1::greptime_request::Request::Query(api::v1::QueryRequest {
+                    query: Some(api::v1::query_request::Query::LogicalPlan(message.to_vec())),
+                })
+            };
+
+            frontend_client
+                .handle(req, catalog, schema, &mut peer_desc)
+                .await
+        };
 
         let elapsed = instant.elapsed();
         if let Ok(affected_rows) = &res {
@@ -307,19 +354,23 @@ impl BatchingTask {
             );
         } else if let Err(err) = &res {
             warn!(
-                "Failed to execute Flow {flow_id} on frontend {}, result: {err:?}, elapsed: {:?} with query: {}",
-                peer_addr, elapsed, &plan
+                "Failed to execute Flow {flow_id} on frontend {:?}, result: {err:?}, elapsed: {:?} with query: {}",
+                peer_desc, elapsed, &plan
             );
         }
 
         // record slow query
         if elapsed >= SLOW_QUERY_THRESHOLD {
             warn!(
-                "Flow {flow_id} on frontend {} executed for {:?} before complete, query: {}",
-                peer_addr, elapsed, &plan
+                "Flow {flow_id} on frontend {:?} executed for {:?} before complete, query: {}",
+                peer_desc, elapsed, &plan
             );
             METRIC_FLOW_BATCHING_ENGINE_SLOW_QUERY
-                .with_label_values(&[flow_id.to_string().as_str(), &plan.to_string(), &peer_addr])
+                .with_label_values(&[
+                    flow_id.to_string().as_str(),
+                    &plan.to_string(),
+                    &peer_desc.unwrap_or_default().to_string(),
+                ])
                 .observe(elapsed.as_secs_f64());
         }
 
@@ -328,12 +379,7 @@ impl BatchingTask {
             .unwrap()
             .after_query_exec(elapsed, res.is_ok());
 
-        let res = res.context(InvalidRequestSnafu {
-            context: format!(
-                "Failed to execute query for flow={}: \'{}\'",
-                self.config.flow_id, &plan
-            ),
-        })?;
+        let res = res?;
 
         Ok(Some((res, elapsed)))
     }
@@ -372,7 +418,10 @@ impl BatchingTask {
                             }
                             Err(TryRecvError::Empty) => (),
                         }
-                        state.get_next_start_query_time(Some(DEFAULT_BATCHING_ENGINE_QUERY_TIMEOUT))
+                        state.get_next_start_query_time(
+                            self.config.flow_id,
+                            Some(DEFAULT_BATCHING_ENGINE_QUERY_TIMEOUT),
+                        )
                     };
                     tokio::time::sleep_until(sleep_until).await;
                 }
@@ -386,14 +435,18 @@ impl BatchingTask {
                     continue;
                 }
                 // TODO(discord9): this error should have better place to go, but for now just print error, also more context is needed
-                Err(err) => match new_query {
-                    Some(query) => {
-                        common_telemetry::error!(err; "Failed to execute query for flow={} with query: {query}", self.config.flow_id)
+                Err(err) => {
+                    match new_query {
+                        Some(query) => {
+                            common_telemetry::error!(err; "Failed to execute query for flow={} with query: {query}", self.config.flow_id)
+                        }
+                        None => {
+                            common_telemetry::error!(err; "Failed to generate query for flow={}", self.config.flow_id)
+                        }
                     }
-                    None => {
-                        common_telemetry::error!(err; "Failed to generate query for flow={}", self.config.flow_id)
-                    }
-                },
+                    // also sleep for a little while before try again to prevent flooding logs
+                    tokio::time::sleep(MIN_REFRESH_DURATION).await;
+                }
             }
         }
     }
@@ -418,7 +471,7 @@ impl BatchingTask {
     async fn gen_query_with_time_window(
         &self,
         engine: QueryEngineRef,
-        sink_table_meta: &RawTableMeta,
+        sink_table_schema: &Arc<Schema>,
     ) -> Result<Option<(LogicalPlan, usize)>, Error> {
         let query_ctx = self.state.read().unwrap().query_ctx.clone();
         let start = SystemTime::now();
@@ -477,9 +530,11 @@ impl BatchingTask {
                         debug!(
                             "Flow id = {:?}, can't get window size: precise_lower_bound={expire_time_window_bound:?}, using the same query", self.config.flow_id
                         );
+                        // clean dirty time window too, this could be from create flow's check_execute
+                        self.state.write().unwrap().dirty_time_windows.clean();
 
                         let mut add_auto_column =
-                            AddAutoColumnRewriter::new(sink_table_meta.schema.clone());
+                            AddAutoColumnRewriter::new(sink_table_schema.clone());
                         let plan = self
                             .config
                             .plan
@@ -487,7 +542,10 @@ impl BatchingTask {
                             .clone()
                             .rewrite(&mut add_auto_column)
                             .with_context(|_| DatafusionSnafu {
-                                context: format!("Failed to rewrite plan {:?}", self.config.plan),
+                                context: format!(
+                                    "Failed to rewrite plan:\n {}\n",
+                                    self.config.plan
+                                ),
                             })?
                             .data;
                         let schema_len = plan.schema().fields().len();
@@ -515,18 +573,23 @@ impl BatchingTask {
                 return Ok(None);
             };
 
+            // TODO(discord9): add auto column or not? This might break compatibility for auto created sink table before this, but that's ok right?
+
             let mut add_filter = AddFilterRewriter::new(expr);
-            let mut add_auto_column = AddAutoColumnRewriter::new(sink_table_meta.schema.clone());
-            // make a not optimized plan for clearer unparse
+            let mut add_auto_column = AddAutoColumnRewriter::new(sink_table_schema.clone());
+
             let plan = sql_to_df_plan(query_ctx.clone(), engine.clone(), &self.config.query, false)
                 .await?;
-            plan.clone()
+            let rewrite = plan
+                .clone()
                 .rewrite(&mut add_filter)
                 .and_then(|p| p.data.rewrite(&mut add_auto_column))
                 .with_context(|_| DatafusionSnafu {
-                    context: format!("Failed to rewrite plan {plan:?}"),
+                    context: format!("Failed to rewrite plan:\n {}\n", plan),
                 })?
-                .data
+                .data;
+            // only apply optimize after complex rewrite is done
+            apply_df_optimizer(rewrite).await?
         };
 
         Ok(Some((new_plan, schema_len)))
@@ -534,7 +597,7 @@ impl BatchingTask {
 }
 
 // auto created table have a auto added column `update_at`, and optional have a `AUTO_CREATED_PLACEHOLDER_TS_COL` column for time index placeholder if no timestamp column is specified
-// TODO(discord9): unit test
+// TODO(discord9): for now no default value is set for auto added column for compatibility reason with streaming mode, but this might change in favor of simpler code?
 fn create_table_with_expr(
     plan: &LogicalPlan,
     sink_table_name: &[String; 3],
@@ -558,11 +621,7 @@ fn create_table_with_expr(
         AUTO_CREATED_UPDATE_AT_TS_COL,
         ConcreteDataType::timestamp_millisecond_datatype(),
         true,
-    )
-    .with_default_constraint(Some(ColumnDefaultConstraint::Function(NOW_FN.to_string())))
-    .context(DatatypesSnafu {
-        extra: "Failed to build column `update_at TimestampMillisecond default now()`",
-    })?;
+    );
     column_schemas.push(update_at_schema);
 
     let time_index = if let Some(time_index) = first_time_stamp {
@@ -574,16 +633,7 @@ fn create_table_with_expr(
                 ConcreteDataType::timestamp_millisecond_datatype(),
                 false,
             )
-            .with_time_index(true)
-            .with_default_constraint(Some(ColumnDefaultConstraint::Value(Value::Timestamp(
-                Timestamp::new_millisecond(0),
-            ))))
-            .context(DatatypesSnafu {
-                extra: format!(
-                    "Failed to build column `{} TimestampMillisecond TIME INDEX default 0`",
-                    AUTO_CREATED_PLACEHOLDER_TS_COL
-                ),
-            })?,
+            .with_time_index(true),
         );
         AUTO_CREATED_PLACEHOLDER_TS_COL.to_string()
     };
@@ -675,20 +725,14 @@ mod test {
             AUTO_CREATED_UPDATE_AT_TS_COL,
             ConcreteDataType::timestamp_millisecond_datatype(),
             true,
-        )
-        .with_default_constraint(Some(ColumnDefaultConstraint::Function(NOW_FN.to_string())))
-        .unwrap();
+        );
 
         let ts_placeholder_schema = ColumnSchema::new(
             AUTO_CREATED_PLACEHOLDER_TS_COL,
             ConcreteDataType::timestamp_millisecond_datatype(),
             false,
         )
-        .with_time_index(true)
-        .with_default_constraint(Some(ColumnDefaultConstraint::Value(Value::Timestamp(
-            Timestamp::new_millisecond(0),
-        ))))
-        .unwrap();
+        .with_time_index(true);
 
         let testcases = vec![
             TestCase {

src/flow/src/batching_mode/time_window.rs

@@ -72,6 +72,17 @@ pub struct TimeWindowExpr {
     df_schema: DFSchema,
 }
 
+impl std::fmt::Display for TimeWindowExpr {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("TimeWindowExpr")
+            .field("phy_expr", &self.phy_expr.to_string())
+            .field("column_name", &self.column_name)
+            .field("logical_expr", &self.logical_expr.to_string())
+            .field("df_schema", &self.df_schema)
+            .finish()
+    }
+}
+
 impl TimeWindowExpr {
     pub fn from_expr(
         expr: &Expr,
@@ -256,7 +267,7 @@ fn columnar_to_ts_vector(columnar: &ColumnarValue) -> Result<Vec<Option<Timestam
     Ok(val)
 }
 
-/// Return (the column name of time index column, the time window expr, the expected time unit of time index column, the expr's schema for evaluating the time window)
+/// Return (`the column name of time index column`, `the time window expr`, `the expected time unit of time index column`, `the expr's schema for evaluating the time window`)
 ///
 /// The time window expr is expected to have one input column with Timestamp type, and also return Timestamp type, the time window expr is expected
 /// to be monotonic increasing and appears in the innermost GROUP BY clause
@@ -693,6 +704,28 @@ mod test {
                 ),
                 "SELECT arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)') AS time_window FROM numbers_with_ts WHERE ((ts >= CAST('2025-02-24 10:48:00' AS TIMESTAMP)) AND (ts <= CAST('2025-02-24 10:49:00' AS TIMESTAMP))) GROUP BY arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)')"
             ),
+            // complex time window index with where
+            (
+                "SELECT arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)') AS time_window FROM numbers_with_ts WHERE number in (2, 3, 4) GROUP BY time_window;",
+                Timestamp::new(1740394109, TimeUnit::Second),
+                (
+                    "ts".to_string(),
+                    Some(Timestamp::new(1740394080, TimeUnit::Second)),
+                    Some(Timestamp::new(1740394140, TimeUnit::Second)),
+                ),
+                "SELECT arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)') AS time_window FROM numbers_with_ts WHERE numbers_with_ts.number IN (2, 3, 4) AND ((ts >= CAST('2025-02-24 10:48:00' AS TIMESTAMP)) AND (ts <= CAST('2025-02-24 10:49:00' AS TIMESTAMP))) GROUP BY arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)')"
+            ),
+            // complex time window index with between and
+            (
+                "SELECT arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)') AS time_window FROM numbers_with_ts WHERE number BETWEEN 2 AND 4 GROUP BY time_window;",
+                Timestamp::new(1740394109, TimeUnit::Second),
+                (
+                    "ts".to_string(),
+                    Some(Timestamp::new(1740394080, TimeUnit::Second)),
+                    Some(Timestamp::new(1740394140, TimeUnit::Second)),
+                ),
+                "SELECT arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)') AS time_window FROM numbers_with_ts WHERE (numbers_with_ts.number BETWEEN 2 AND 4) AND ((ts >= CAST('2025-02-24 10:48:00' AS TIMESTAMP)) AND (ts <= CAST('2025-02-24 10:49:00' AS TIMESTAMP))) GROUP BY arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)')"
+            ),
             // no time index
             (
                 "SELECT date_bin('5 minutes', ts) FROM numbers_with_ts;",

src/flow/src/batching_mode/utils.rs

@@ -14,29 +14,63 @@
 
 //! some utils for helping with batching mode
 
-use std::collections::HashSet;
+use std::collections::{BTreeSet, HashSet};
 use std::sync::Arc;
 
+use catalog::CatalogManagerRef;
 use common_error::ext::BoxedError;
-use common_telemetry::{debug, info};
+use common_telemetry::debug;
 use datafusion::error::Result as DfResult;
 use datafusion::logical_expr::Expr;
 use datafusion::sql::unparser::Unparser;
 use datafusion_common::tree_node::{
     Transformed, TreeNodeRecursion, TreeNodeRewriter, TreeNodeVisitor,
 };
-use datafusion_common::DataFusionError;
-use datafusion_expr::{Distinct, LogicalPlan};
-use datatypes::schema::RawSchema;
+use datafusion_common::{DFSchema, DataFusionError, ScalarValue};
+use datafusion_expr::{Distinct, LogicalPlan, Projection};
+use datatypes::schema::SchemaRef;
 use query::parser::QueryLanguageParser;
 use query::QueryEngineRef;
 use session::context::QueryContextRef;
-use snafu::ResultExt;
+use snafu::{OptionExt, ResultExt};
+use table::metadata::TableInfo;
 
 use crate::adapter::AUTO_CREATED_PLACEHOLDER_TS_COL;
 use crate::df_optimizer::apply_df_optimizer;
-use crate::error::{DatafusionSnafu, ExternalSnafu};
-use crate::Error;
+use crate::error::{DatafusionSnafu, ExternalSnafu, TableNotFoundSnafu};
+use crate::{Error, TableName};
+
+pub async fn get_table_info_df_schema(
+    catalog_mr: CatalogManagerRef,
+    table_name: TableName,
+) -> Result<(Arc<TableInfo>, Arc<DFSchema>), Error> {
+    let full_table_name = table_name.clone().join(".");
+    let table = catalog_mr
+        .table(&table_name[0], &table_name[1], &table_name[2], None)
+        .await
+        .map_err(BoxedError::new)
+        .context(ExternalSnafu)?
+        .context(TableNotFoundSnafu {
+            name: &full_table_name,
+        })?;
+    let table_info = table.table_info().clone();
+
+    let schema = table_info.meta.schema.clone();
+
+    let df_schema: Arc<DFSchema> = Arc::new(
+        schema
+            .arrow_schema()
+            .clone()
+            .try_into()
+            .with_context(|_| DatafusionSnafu {
+                context: format!(
+                    "Failed to convert arrow schema to datafusion schema, arrow_schema={:?}",
+                    schema.arrow_schema()
+                ),
+            })?,
+    );
+    Ok((table_info, df_schema))
+}
 
 /// Convert sql to datafusion logical plan
 pub async fn sql_to_df_plan(
@@ -164,14 +198,16 @@ impl TreeNodeVisitor<'_> for FindGroupByFinalName {
 /// (which doesn't necessary need to have exact name just need to be a extra timestamp column)
 /// and `__ts_placeholder`(this column need to have exact this name and be a timestamp)
 /// with values like `now()` and `0`
+///
+/// it also give existing columns alias to column in sink table if needed
 #[derive(Debug)]
 pub struct AddAutoColumnRewriter {
-    pub schema: RawSchema,
+    pub schema: SchemaRef,
     pub is_rewritten: bool,
 }
 
 impl AddAutoColumnRewriter {
-    pub fn new(schema: RawSchema) -> Self {
+    pub fn new(schema: SchemaRef) -> Self {
         Self {
             schema,
             is_rewritten: false,
@@ -181,37 +217,97 @@ impl AddAutoColumnRewriter {
 
 impl TreeNodeRewriter for AddAutoColumnRewriter {
     type Node = LogicalPlan;
-    fn f_down(&mut self, node: Self::Node) -> DfResult<Transformed<Self::Node>> {
+    fn f_down(&mut self, mut node: Self::Node) -> DfResult<Transformed<Self::Node>> {
         if self.is_rewritten {
             return Ok(Transformed::no(node));
         }
 
-        // if is distinct all, go one level down
-        if let LogicalPlan::Distinct(Distinct::All(_)) = node {
-            return Ok(Transformed::no(node));
+        // if is distinct all, wrap it in a projection
+        if let LogicalPlan::Distinct(Distinct::All(_)) = &node {
+            let mut exprs = vec![];
+
+            for field in node.schema().fields().iter() {
+                exprs.push(Expr::Column(datafusion::common::Column::new_unqualified(
+                    field.name(),
+                )));
+            }
+
+            let projection =
+                LogicalPlan::Projection(Projection::try_new(exprs, Arc::new(node.clone()))?);
+
+            node = projection;
+        }
+        // handle table_scan by wrap it in a projection
+        else if let LogicalPlan::TableScan(table_scan) = node {
+            let mut exprs = vec![];
+
+            for field in table_scan.projected_schema.fields().iter() {
+                exprs.push(Expr::Column(datafusion::common::Column::new(
+                    Some(table_scan.table_name.clone()),
+                    field.name(),
+                )));
+            }
+
+            let projection = LogicalPlan::Projection(Projection::try_new(
+                exprs,
+                Arc::new(LogicalPlan::TableScan(table_scan)),
+            )?);
+
+            node = projection;
         }
 
-        // FIXME(discord9): just read plan.expr and do stuffs
-        let mut exprs = node.expressions();
+        // only do rewrite if found the outermost projection
+        let mut exprs = if let LogicalPlan::Projection(project) = &node {
+            project.expr.clone()
+        } else {
+            return Ok(Transformed::no(node));
+        };
+
+        let all_names = self
+            .schema
+            .column_schemas()
+            .iter()
+            .map(|c| c.name.clone())
+            .collect::<BTreeSet<_>>();
+        // first match by position
+        for (idx, expr) in exprs.iter_mut().enumerate() {
+            if !all_names.contains(&expr.qualified_name().1) {
+                if let Some(col_name) = self
+                    .schema
+                    .column_schemas()
+                    .get(idx)
+                    .map(|c| c.name.clone())
+                {
+                    // if the data type mismatched, later check_execute will error out
+                    // hence no need to check it here, beside, optimize pass might be able to cast it
+                    // so checking here is not necessary
+                    *expr = expr.clone().alias(col_name);
+                }
+            }
+        }
 
         // add columns if have different column count
         let query_col_cnt = exprs.len();
-        let table_col_cnt = self.schema.column_schemas.len();
-        info!("query_col_cnt={query_col_cnt}, table_col_cnt={table_col_cnt}");
+        let table_col_cnt = self.schema.column_schemas().len();
+        debug!("query_col_cnt={query_col_cnt}, table_col_cnt={table_col_cnt}");
+
+        let placeholder_ts_expr =
+            datafusion::logical_expr::lit(ScalarValue::TimestampMillisecond(Some(0), None))
+                .alias(AUTO_CREATED_PLACEHOLDER_TS_COL);
+
         if query_col_cnt == table_col_cnt {
-            self.is_rewritten = true;
-            return Ok(Transformed::no(node));
+            // still need to add alias, see below
         } else if query_col_cnt + 1 == table_col_cnt {
-            let last_col_schema = self.schema.column_schemas.last().unwrap();
+            let last_col_schema = self.schema.column_schemas().last().unwrap();
 
             // if time index column is auto created add it
             if last_col_schema.name == AUTO_CREATED_PLACEHOLDER_TS_COL
-                && self.schema.timestamp_index == Some(table_col_cnt - 1)
+                && self.schema.timestamp_index() == Some(table_col_cnt - 1)
             {
-                exprs.push(datafusion::logical_expr::lit(0));
+                exprs.push(placeholder_ts_expr);
             } else if last_col_schema.data_type.is_timestamp() {
                 // is the update at column
-                exprs.push(datafusion::prelude::now());
+                exprs.push(datafusion::prelude::now().alias(&last_col_schema.name));
             } else {
                 // helpful error message
                 return Err(DataFusionError::Plan(format!(
@@ -221,11 +317,11 @@ impl TreeNodeRewriter for AddAutoColumnRewriter {
                 )));
             }
         } else if query_col_cnt + 2 == table_col_cnt {
-            let mut col_iter = self.schema.column_schemas.iter().rev();
+            let mut col_iter = self.schema.column_schemas().iter().rev();
             let last_col_schema = col_iter.next().unwrap();
             let second_last_col_schema = col_iter.next().unwrap();
             if second_last_col_schema.data_type.is_timestamp() {
-                exprs.push(datafusion::prelude::now());
+                exprs.push(datafusion::prelude::now().alias(&second_last_col_schema.name));
             } else {
                 return Err(DataFusionError::Plan(format!(
                     "Expect the second last column in the table to be timestamp column, found column {} with type {:?}",
@@ -235,9 +331,9 @@ impl TreeNodeRewriter for AddAutoColumnRewriter {
             }
 
             if last_col_schema.name == AUTO_CREATED_PLACEHOLDER_TS_COL
-                && self.schema.timestamp_index == Some(table_col_cnt - 1)
+                && self.schema.timestamp_index() == Some(table_col_cnt - 1)
             {
-                exprs.push(datafusion::logical_expr::lit(0));
+                exprs.push(placeholder_ts_expr);
             } else {
                 return Err(DataFusionError::Plan(format!(
                     "Expect timestamp column {}, found {:?}",
@@ -247,7 +343,7 @@ impl TreeNodeRewriter for AddAutoColumnRewriter {
         } else {
             return Err(DataFusionError::Plan(format!(
                     "Expect table have 0,1 or 2 columns more than query columns, found {} query columns {:?}, {} table columns {:?}",
-                    query_col_cnt, node.expressions(), table_col_cnt, self.schema.column_schemas
+                    query_col_cnt, exprs, table_col_cnt, self.schema.column_schemas()
                 )));
         }
 
@@ -255,9 +351,12 @@ impl TreeNodeRewriter for AddAutoColumnRewriter {
         let new_plan = node.with_new_exprs(exprs, node.inputs().into_iter().cloned().collect())?;
         Ok(Transformed::yes(new_plan))
     }
-}
 
-// TODO(discord9): a method to found out the precise time window
+    /// We might add new columns, so we need to recompute the schema
+    fn f_up(&mut self, node: Self::Node) -> DfResult<Transformed<Self::Node>> {
+        node.recompute_schema().map(Transformed::yes)
+    }
+}
 
 /// Find out the `Filter` Node corresponding to innermost(deepest) `WHERE` and add a new filter expr to it
 #[derive(Debug)]
@@ -301,11 +400,15 @@ impl TreeNodeRewriter for AddFilterRewriter {
 
 #[cfg(test)]
 mod test {
+    use std::sync::Arc;
+
     use datafusion_common::tree_node::TreeNode as _;
     use datatypes::prelude::ConcreteDataType;
-    use datatypes::schema::ColumnSchema;
+    use datatypes::schema::{ColumnSchema, Schema};
     use pretty_assertions::assert_eq;
+    use query::query_engine::DefaultSerializer;
     use session::context::QueryContext;
+    use substrait::{DFLogicalSubstraitConvertor, SubstraitPlan};
 
     use super::*;
     use crate::test_utils::create_test_query_engine;
@@ -386,7 +489,7 @@ mod test {
             // add update_at
             (
                 "SELECT number FROM numbers_with_ts",
-                Ok("SELECT numbers_with_ts.number, now() FROM numbers_with_ts"),
+                Ok("SELECT numbers_with_ts.number, now() AS ts FROM numbers_with_ts"),
                 vec![
                     ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
                     ColumnSchema::new(
@@ -400,7 +503,7 @@ mod test {
             // add ts placeholder
             (
                 "SELECT number FROM numbers_with_ts",
-                Ok("SELECT numbers_with_ts.number, 0 FROM numbers_with_ts"),
+                Ok("SELECT numbers_with_ts.number, CAST('1970-01-01 00:00:00' AS TIMESTAMP) AS __ts_placeholder FROM numbers_with_ts"),
                 vec![
                     ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
                     ColumnSchema::new(
@@ -428,7 +531,7 @@ mod test {
             // add update_at and ts placeholder
             (
                 "SELECT number FROM numbers_with_ts",
-                Ok("SELECT numbers_with_ts.number, now(), 0 FROM numbers_with_ts"),
+                Ok("SELECT numbers_with_ts.number, now() AS update_at, CAST('1970-01-01 00:00:00' AS TIMESTAMP) AS __ts_placeholder FROM numbers_with_ts"),
                 vec![
                     ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
                     ColumnSchema::new(
@@ -447,7 +550,7 @@ mod test {
             // add ts placeholder
             (
                 "SELECT number, ts FROM numbers_with_ts",
-                Ok("SELECT numbers_with_ts.number, numbers_with_ts.ts, 0 FROM numbers_with_ts"),
+                Ok("SELECT numbers_with_ts.number, numbers_with_ts.ts AS update_at, CAST('1970-01-01 00:00:00' AS TIMESTAMP) AS __ts_placeholder FROM numbers_with_ts"),
                 vec![
                     ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
                     ColumnSchema::new(
@@ -466,7 +569,7 @@ mod test {
             // add update_at after time index column
             (
                 "SELECT number, ts FROM numbers_with_ts",
-                Ok("SELECT numbers_with_ts.number, numbers_with_ts.ts, now() FROM numbers_with_ts"),
+                Ok("SELECT numbers_with_ts.number, numbers_with_ts.ts, now() AS update_atat FROM numbers_with_ts"),
                 vec![
                     ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
                     ColumnSchema::new(
@@ -528,8 +631,8 @@ mod test {
         let query_engine = create_test_query_engine();
         let ctx = QueryContext::arc();
         for (before, after, column_schemas) in testcases {
-            let raw_schema = RawSchema::new(column_schemas);
-            let mut add_auto_column_rewriter = AddAutoColumnRewriter::new(raw_schema);
+            let schema = Arc::new(Schema::new(column_schemas));
+            let mut add_auto_column_rewriter = AddAutoColumnRewriter::new(schema);
 
             let plan = sql_to_df_plan(ctx.clone(), query_engine.clone(), before, false)
                 .await
@@ -600,4 +703,18 @@ mod test {
             );
         }
     }
+
+    #[tokio::test]
+    async fn test_null_cast() {
+        let query_engine = create_test_query_engine();
+        let ctx = QueryContext::arc();
+        let sql = "SELECT NULL::DOUBLE FROM numbers_with_ts";
+        let plan = sql_to_df_plan(ctx, query_engine.clone(), sql, false)
+            .await
+            .unwrap();
+
+        let _sub_plan = DFLogicalSubstraitConvertor {}
+            .encode(&plan, DefaultSerializer)
+            .unwrap();
+    }
 }

src/flow/src/df_optimizer.rs

@@ -25,7 +25,6 @@ use datafusion::config::ConfigOptions;
 use datafusion::error::DataFusionError;
 use datafusion::functions_aggregate::count::count_udaf;
 use datafusion::functions_aggregate::sum::sum_udaf;
-use datafusion::optimizer::analyzer::count_wildcard_rule::CountWildcardRule;
 use datafusion::optimizer::analyzer::type_coercion::TypeCoercion;
 use datafusion::optimizer::common_subexpr_eliminate::CommonSubexprEliminate;
 use datafusion::optimizer::optimize_projections::OptimizeProjections;
@@ -42,6 +41,7 @@ use datafusion_expr::{
     BinaryExpr, ColumnarValue, Expr, Operator, Projection, ScalarFunctionArgs, ScalarUDFImpl,
     Signature, TypeSignature, Volatility,
 };
+use query::optimizer::count_wildcard::CountWildcardToTimeIndexRule;
 use query::parser::QueryLanguageParser;
 use query::query_engine::DefaultSerializer;
 use query::QueryEngine;
@@ -61,9 +61,9 @@ pub async fn apply_df_optimizer(
 ) -> Result<datafusion_expr::LogicalPlan, Error> {
     let cfg = ConfigOptions::new();
     let analyzer = Analyzer::with_rules(vec![
-        Arc::new(CountWildcardRule::new()),
-        Arc::new(AvgExpandRule::new()),
-        Arc::new(TumbleExpandRule::new()),
+        Arc::new(CountWildcardToTimeIndexRule),
+        Arc::new(AvgExpandRule),
+        Arc::new(TumbleExpandRule),
         Arc::new(CheckGroupByRule::new()),
         Arc::new(TypeCoercion::new()),
     ]);
@@ -128,13 +128,7 @@ pub async fn sql_to_flow_plan(
 }
 
 #[derive(Debug)]
-struct AvgExpandRule {}
-
-impl AvgExpandRule {
-    pub fn new() -> Self {
-        Self {}
-    }
-}
+struct AvgExpandRule;
 
 impl AnalyzerRule for AvgExpandRule {
     fn analyze(
@@ -331,13 +325,7 @@ impl TreeNodeRewriter for ExpandAvgRewriter<'_> {
 
 /// expand tumble in aggr expr to tumble_start and tumble_end with column name like `window_start`
 #[derive(Debug)]
-struct TumbleExpandRule {}
-
-impl TumbleExpandRule {
-    pub fn new() -> Self {
-        Self {}
-    }
-}
+struct TumbleExpandRule;
 
 impl AnalyzerRule for TumbleExpandRule {
     fn analyze(

src/flow/src/engine.rs

@@ -49,6 +49,8 @@ pub trait FlowEngine {
     async fn flush_flow(&self, flow_id: FlowId) -> Result<usize, Error>;
     /// Check if the flow exists
     async fn flow_exist(&self, flow_id: FlowId) -> Result<bool, Error>;
+    /// List all flows
+    async fn list_flows(&self) -> Result<impl IntoIterator<Item = FlowId>, Error>;
     /// Handle the insert requests for the flow
     async fn handle_flow_inserts(
         &self,

src/flow/src/error.rs

@@ -61,6 +61,16 @@ pub enum Error {
         location: Location,
     },
 
+    #[snafu(display(
+        "No available frontend found after timeout: {timeout:?}, context: {context}"
+    ))]
+    NoAvailableFrontend {
+        timeout: std::time::Duration,
+        context: String,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
     #[snafu(display("External error"))]
     External {
         source: BoxedError,
@@ -149,6 +159,13 @@ pub enum Error {
         location: Location,
     },
 
+    #[snafu(display("Unsupported: {reason}"))]
+    Unsupported {
+        reason: String,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
     #[snafu(display("Unsupported temporal filter: {reason}"))]
     UnsupportedTemporalFilter {
         reason: String,
@@ -189,6 +206,25 @@ pub enum Error {
         location: Location,
     },
 
+    #[snafu(display("Illegal check task state: {reason}"))]
+    IllegalCheckTaskState {
+        reason: String,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
+    #[snafu(display(
+        "Failed to sync with check task for flow {} with allow_drop={}",
+        flow_id,
+        allow_drop
+    ))]
+    SyncCheckTask {
+        flow_id: FlowId,
+        allow_drop: bool,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
     #[snafu(display("Failed to start server"))]
     StartServer {
         #[snafu(implicit)]
@@ -270,7 +306,8 @@ impl ErrorExt for Error {
             Self::Eval { .. }
             | Self::JoinTask { .. }
             | Self::Datafusion { .. }
-            | Self::InsertIntoFlow { .. } => StatusCode::Internal,
+            | Self::InsertIntoFlow { .. }
+            | Self::NoAvailableFrontend { .. } => StatusCode::Internal,
             Self::FlowAlreadyExist { .. } => StatusCode::TableAlreadyExists,
             Self::TableNotFound { .. }
             | Self::TableNotFoundMeta { .. }
@@ -280,10 +317,12 @@ impl ErrorExt for Error {
             Self::CreateFlow { .. } | Self::Arrow { .. } | Self::Time { .. } => {
                 StatusCode::EngineExecuteQuery
             }
-            Self::Unexpected { .. } => StatusCode::Unexpected,
-            Self::NotImplemented { .. } | Self::UnsupportedTemporalFilter { .. } => {
-                StatusCode::Unsupported
-            }
+            Self::Unexpected { .. }
+            | Self::SyncCheckTask { .. }
+            | Self::IllegalCheckTaskState { .. } => StatusCode::Unexpected,
+            Self::NotImplemented { .. }
+            | Self::UnsupportedTemporalFilter { .. }
+            | Self::Unsupported { .. } => StatusCode::Unsupported,
             Self::External { source, .. } => source.status_code(),
             Self::Internal { .. } | Self::CacheRequired { .. } => StatusCode::Internal,
             Self::StartServer { source, .. } | Self::ShutdownServer { source, .. } => {

src/flow/src/lib.rs

@@ -43,8 +43,8 @@ mod utils;
 #[cfg(test)]
 mod test_utils;
 
-pub use adapter::{FlowConfig, FlowWorkerManager, FlowWorkerManagerRef, FlownodeOptions};
-pub use batching_mode::frontend_client::FrontendClient;
+pub use adapter::{FlowConfig, FlowStreamingEngineRef, FlownodeOptions, StreamingEngine};
+pub use batching_mode::frontend_client::{FrontendClient, GrpcQueryHandlerWithBoxedError};
 pub(crate) use engine::{CreateFlowArgs, FlowId, TableName};
 pub use error::{Error, Result};
 pub use server::{

src/flow/src/server.rs

@@ -29,6 +29,7 @@ use common_meta::key::TableMetadataManagerRef;
 use common_meta::kv_backend::KvBackendRef;
 use common_meta::node_manager::{Flownode, NodeManagerRef};
 use common_query::Output;
+use common_runtime::JoinHandle;
 use common_telemetry::tracing::info;
 use futures::{FutureExt, TryStreamExt};
 use greptime_proto::v1::flow::{flow_server, FlowRequest, FlowResponse, InsertRequests};
@@ -50,7 +51,10 @@ use tonic::codec::CompressionEncoding;
 use tonic::transport::server::TcpIncoming;
 use tonic::{Request, Response, Status};
 
-use crate::adapter::{create_worker, FlowWorkerManagerRef};
+use crate::adapter::flownode_impl::{FlowDualEngine, FlowDualEngineRef};
+use crate::adapter::{create_worker, FlowStreamingEngineRef};
+use crate::batching_mode::engine::BatchingEngine;
+use crate::engine::FlowEngine;
 use crate::error::{
     to_status_with_last_err, CacheRequiredSnafu, CreateFlowSnafu, ExternalSnafu, FlowNotFoundSnafu,
     ListFlowsSnafu, ParseAddrSnafu, ShutdownServerSnafu, StartServerSnafu, UnexpectedSnafu,
@@ -59,19 +63,20 @@ use crate::heartbeat::HeartbeatTask;
 use crate::metrics::{METRIC_FLOW_PROCESSING_TIME, METRIC_FLOW_ROWS};
 use crate::transform::register_function_to_query_engine;
 use crate::utils::{SizeReportSender, StateReportHandler};
-use crate::{CreateFlowArgs, Error, FlowWorkerManager, FlownodeOptions, FrontendClient};
+use crate::{CreateFlowArgs, Error, FlownodeOptions, FrontendClient, StreamingEngine};
 
 pub const FLOW_NODE_SERVER_NAME: &str = "FLOW_NODE_SERVER";
 /// wrapping flow node manager to avoid orphan rule with Arc<...>
 #[derive(Clone)]
 pub struct FlowService {
-    /// TODO(discord9): replace with dual engine
-    pub manager: FlowWorkerManagerRef,
+    pub dual_engine: FlowDualEngineRef,
 }
 
 impl FlowService {
-    pub fn new(manager: FlowWorkerManagerRef) -> Self {
-        Self { manager }
+    pub fn new(manager: FlowDualEngineRef) -> Self {
+        Self {
+            dual_engine: manager,
+        }
     }
 }
 
@@ -86,7 +91,7 @@ impl flow_server::Flow for FlowService {
             .start_timer();
 
         let request = request.into_inner();
-        self.manager
+        self.dual_engine
             .handle(request)
             .await
             .map_err(|err| {
@@ -126,7 +131,7 @@ impl flow_server::Flow for FlowService {
             .with_label_values(&["in"])
             .inc_by(row_count as u64);
 
-        self.manager
+        self.dual_engine
             .handle_inserts(request)
             .await
             .map(Response::new)
@@ -139,11 +144,16 @@ pub struct FlownodeServer {
     inner: Arc<FlownodeServerInner>,
 }
 
+/// FlownodeServerInner is the inner state of FlownodeServer,
+/// this struct mostly useful for construct/start and stop the
+/// flow node server
 struct FlownodeServerInner {
     /// worker shutdown signal, not to be confused with server_shutdown_tx
     worker_shutdown_tx: Mutex<broadcast::Sender<()>>,
     /// server shutdown signal for shutdown grpc server
     server_shutdown_tx: Mutex<broadcast::Sender<()>>,
+    /// streaming task handler
+    streaming_task_handler: Mutex<Option<JoinHandle<()>>>,
     flow_service: FlowService,
 }
 
@@ -156,16 +166,30 @@ impl FlownodeServer {
                 flow_service,
                 worker_shutdown_tx: Mutex::new(tx),
                 server_shutdown_tx: Mutex::new(server_tx),
+                streaming_task_handler: Mutex::new(None),
             }),
         }
     }
 
     /// Start the background task for streaming computation.
+    ///
+    /// Should be called only after heartbeat is establish, hence can get cluster info
     async fn start_workers(&self) -> Result<(), Error> {
-        let manager_ref = self.inner.flow_service.manager.clone();
-        let _handle = manager_ref
-            .clone()
+        let manager_ref = self.inner.flow_service.dual_engine.clone();
+        let handle = manager_ref
+            .streaming_engine()
             .run_background(Some(self.inner.worker_shutdown_tx.lock().await.subscribe()));
+        self.inner
+            .streaming_task_handler
+            .lock()
+            .await
+            .replace(handle);
+
+        self.inner
+            .flow_service
+            .dual_engine
+            .start_flow_consistent_check_task()
+            .await?;
 
         Ok(())
     }
@@ -176,6 +200,11 @@ impl FlownodeServer {
         if tx.send(()).is_err() {
             info!("Receiver dropped, the flow node server has already shutdown");
         }
+        self.inner
+            .flow_service
+            .dual_engine
+            .stop_flow_consistent_check_task()
+            .await?;
         Ok(())
     }
 }
@@ -272,8 +301,8 @@ impl FlownodeInstance {
         &self.flownode_server
     }
 
-    pub fn flow_worker_manager(&self) -> FlowWorkerManagerRef {
-        self.flownode_server.inner.flow_service.manager.clone()
+    pub fn flow_engine(&self) -> FlowDualEngineRef {
+        self.flownode_server.inner.flow_service.dual_engine.clone()
     }
 
     pub fn setup_services(&mut self, services: ServerHandlers) {
@@ -334,7 +363,6 @@ impl FlownodeBuilder {
             None,
             None,
             None,
-            None,
             false,
             Default::default(),
             self.opts.query.clone(),
@@ -343,12 +371,21 @@ impl FlownodeBuilder {
             self.build_manager(query_engine_factory.query_engine())
                 .await?,
         );
+        let batching = Arc::new(BatchingEngine::new(
+            self.frontend_client.clone(),
+            query_engine_factory.query_engine(),
+            self.flow_metadata_manager.clone(),
+            self.table_meta.clone(),
+            self.catalog_manager.clone(),
+        ));
+        let dual = FlowDualEngine::new(
+            manager.clone(),
+            batching,
+            self.flow_metadata_manager.clone(),
+            self.catalog_manager.clone(),
+        );
 
-        if let Err(err) = self.recover_flows(&manager).await {
-            common_telemetry::error!(err; "Failed to recover flows");
-        }
-
-        let server = FlownodeServer::new(FlowService::new(manager.clone()));
+        let server = FlownodeServer::new(FlowService::new(Arc::new(dual)));
 
         let heartbeat_task = self.heartbeat_task;
 
@@ -365,7 +402,7 @@ impl FlownodeBuilder {
     /// or recover all existing flow tasks if in standalone mode(nodeid is None)
     ///
     /// TODO(discord9): persistent flow tasks with internal state
-    async fn recover_flows(&self, manager: &FlowWorkerManagerRef) -> Result<usize, Error> {
+    async fn recover_flows(&self, manager: &FlowDualEngine) -> Result<usize, Error> {
         let nodeid = self.opts.node_id;
         let to_be_recovered: Vec<_> = if let Some(nodeid) = nodeid {
             let to_be_recover = self
@@ -402,6 +439,7 @@ impl FlownodeBuilder {
         let cnt = to_be_recovered.len();
 
         // TODO(discord9): recover in parallel
+        info!("Recovering {} flows: {:?}", cnt, to_be_recovered);
         for flow_id in to_be_recovered {
             let info = self
                 .flow_metadata_manager
@@ -417,6 +455,7 @@ impl FlownodeBuilder {
                 info.sink_table_name().schema_name.clone(),
                 info.sink_table_name().table_name.clone(),
             ];
+
             let args = CreateFlowArgs {
                 flow_id: flow_id as _,
                 sink_table_name,
@@ -430,14 +469,27 @@ impl FlownodeBuilder {
                 comment: Some(info.comment().clone()),
                 sql: info.raw_sql().clone(),
                 flow_options: info.options().clone(),
-                query_ctx: Some(
-                    QueryContextBuilder::default()
-                        .current_catalog(info.catalog_name().clone())
-                        .build(),
-                ),
+                query_ctx: info
+                    .query_context()
+                    .clone()
+                    .map(|ctx| {
+                        ctx.try_into()
+                            .map_err(BoxedError::new)
+                            .context(ExternalSnafu)
+                    })
+                    .transpose()?
+                    // or use default QueryContext with catalog_name from info
+                    // to keep compatibility with old version
+                    .or_else(|| {
+                        Some(
+                            QueryContextBuilder::default()
+                                .current_catalog(info.catalog_name().to_string())
+                                .build(),
+                        )
+                    }),
             };
             manager
-                .create_flow_inner(args)
+                .create_flow(args)
                 .await
                 .map_err(BoxedError::new)
                 .with_context(|_| CreateFlowSnafu {
@@ -453,7 +505,7 @@ impl FlownodeBuilder {
     async fn build_manager(
         &mut self,
         query_engine: Arc<dyn QueryEngine>,
-    ) -> Result<FlowWorkerManager, Error> {
+    ) -> Result<StreamingEngine, Error> {
         let table_meta = self.table_meta.clone();
 
         register_function_to_query_engine(&query_engine);
@@ -462,7 +514,7 @@ impl FlownodeBuilder {
 
         let node_id = self.opts.node_id.map(|id| id as u32);
 
-        let mut man = FlowWorkerManager::new(node_id, query_engine, table_meta);
+        let mut man = StreamingEngine::new(node_id, query_engine, table_meta);
         for worker_id in 0..num_workers {
             let (tx, rx) = oneshot::channel();
 
@@ -544,6 +596,10 @@ impl<'a> FlownodeServiceBuilder<'a> {
     }
 }
 
+/// Basically a tiny frontend that communicates with datanode, different from [`FrontendClient`] which
+/// connect to a real frontend instead, this is used for flow's streaming engine. And is for simple query.
+///
+/// For heavy query use [`FrontendClient`] which offload computation to frontend, lifting the load from flownode
 #[derive(Clone)]
 pub struct FrontendInvoker {
     inserter: Arc<Inserter>,
@@ -565,7 +621,7 @@ impl FrontendInvoker {
     }
 
     pub async fn build_from(
-        flow_worker_manager: FlowWorkerManagerRef,
+        flow_streaming_engine: FlowStreamingEngineRef,
         catalog_manager: CatalogManagerRef,
         kv_backend: KvBackendRef,
         layered_cache_registry: LayeredCacheRegistryRef,
@@ -600,7 +656,7 @@ impl FrontendInvoker {
             node_manager.clone(),
         ));
 
-        let query_engine = flow_worker_manager.query_engine.clone();
+        let query_engine = flow_streaming_engine.query_engine.clone();
 
         let statement_executor = Arc::new(StatementExecutor::new(
             catalog_manager.clone(),

src/flow/src/test_utils.rs

@@ -153,7 +153,6 @@ pub fn create_test_query_engine() -> Arc<dyn QueryEngine> {
         None,
         None,
         None,
-        None,
         false,
         QueryOptions::default(),
     );

src/flow/src/transform.rs

@@ -270,7 +270,6 @@ mod test {
             None,
             None,
             None,
-            None,
             false,
             QueryOptions::default(),
         );

src/frontend/Cargo.toml

@@ -15,6 +15,7 @@ api.workspace = true
 arc-swap = "1.0"
 async-trait.workspace = true
 auth.workspace = true
+bytes.workspace = true
 cache.workspace = true
 catalog.workspace = true
 client.workspace = true
@@ -39,6 +40,7 @@ datafusion.workspace = true
 datafusion-expr.workspace = true
 datanode.workspace = true
 datatypes.workspace = true
+futures.workspace = true
 humantime-serde.workspace = true
 lazy_static.workspace = true
 log-query.workspace = true
@@ -47,6 +49,7 @@ meta-client.workspace = true
 num_cpus.workspace = true
 opentelemetry-proto.workspace = true
 operator.workspace = true
+otel-arrow-rust.workspace = true
 partition.workspace = true
 pipeline.workspace = true
 prometheus.workspace = true

src/frontend/src/error.rs

@@ -19,6 +19,8 @@ use common_error::define_into_tonic_status;
 use common_error::ext::{BoxedError, ErrorExt};
 use common_error::status_code::StatusCode;
 use common_macro::stack_trace_debug;
+use common_query::error::datafusion_status_code;
+use datafusion::error::DataFusionError;
 use session::ReadPreference;
 use snafu::{Location, Snafu};
 use store_api::storage::RegionId;
@@ -345,7 +347,15 @@ pub enum Error {
     SubstraitDecodeLogicalPlan {
         #[snafu(implicit)]
         location: Location,
-        source: substrait::error::Error,
+        source: common_query::error::Error,
+    },
+
+    #[snafu(display("DataFusionError"))]
+    DataFusion {
+        #[snafu(source)]
+        error: DataFusionError,
+        #[snafu(implicit)]
+        location: Location,
     },
 }
 
@@ -423,6 +433,8 @@ impl ErrorExt for Error {
             Error::TableOperation { source, .. } => source.status_code(),
 
             Error::InFlightWriteBytesExceeded { .. } => StatusCode::RateLimited,
+
+            Error::DataFusion { error, .. } => datafusion_status_code::<Self>(error, None),
         }
     }
 

src/frontend/src/instance.rs

@@ -278,7 +278,7 @@ impl SqlQueryHandler for Instance {
         // plan should be prepared before exec
         // we'll do check there
         self.query_engine
-            .execute(plan, query_ctx)
+            .execute(plan.clone(), query_ctx)
             .await
             .context(ExecLogicalPlanSnafu)
     }

src/frontend/src/instance/builder.rs

@@ -24,11 +24,9 @@ use common_meta::key::flow::FlowMetadataManager;
 use common_meta::key::TableMetadataManager;
 use common_meta::kv_backend::KvBackendRef;
 use common_meta::node_manager::NodeManagerRef;
-use common_meta::snapshot::MetadataSnapshotManager;
 use operator::delete::Deleter;
 use operator::flow::FlowServiceOperator;
 use operator::insert::Inserter;
-use operator::metadata::MetadataSnapshotOperator;
 use operator::procedure::ProcedureServiceOperator;
 use operator::request::Requester;
 use operator::statement::{StatementExecutor, StatementExecutorRef};
@@ -57,7 +55,6 @@ pub struct FrontendBuilder {
     plugins: Option<Plugins>,
     procedure_executor: ProcedureExecutorRef,
     stats: StatementStatistics,
-    metadata_snapshot_manager: Option<MetadataSnapshotManager>,
 }
 
 impl FrontendBuilder {
@@ -80,17 +77,6 @@ impl FrontendBuilder {
             plugins: None,
             procedure_executor,
             stats,
-            metadata_snapshot_manager: None,
-        }
-    }
-
-    pub fn with_metadata_snapshot_manager(
-        self,
-        metadata_snapshot_manager: MetadataSnapshotManager,
-    ) -> Self {
-        Self {
-            metadata_snapshot_manager: Some(metadata_snapshot_manager),
-            ..self
         }
     }
 
@@ -172,17 +158,12 @@ impl FrontendBuilder {
         let flow_metadata_manager = Arc::new(FlowMetadataManager::new(kv_backend.clone()));
         let flow_service = FlowServiceOperator::new(flow_metadata_manager, node_manager.clone());
 
-        let metadata_snapshot_operator = self
-            .metadata_snapshot_manager
-            .map(|manager| Arc::new(MetadataSnapshotOperator::new(manager)) as _);
-
         let query_engine = QueryEngineFactory::new_with_plugins(
             self.catalog_manager.clone(),
             Some(region_query_handler.clone()),
             Some(table_mutation_handler),
             Some(procedure_service_handler),
             Some(Arc::new(flow_service)),
-            metadata_snapshot_operator,
             true,
             plugins.clone(),
             self.options.query.clone(),

src/frontend/src/instance/grpc.rs

@@ -12,29 +12,33 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::sync::Arc;
+
 use api::v1::ddl_request::{Expr as DdlExpr, Expr};
 use api::v1::greptime_request::Request;
 use api::v1::query_request::Query;
-use api::v1::{DeleteRequests, DropFlowExpr, InsertRequests, RowDeleteRequests, RowInsertRequests};
+use api::v1::{
+    DeleteRequests, DropFlowExpr, InsertIntoPlan, InsertRequests, RowDeleteRequests,
+    RowInsertRequests,
+};
 use async_trait::async_trait;
 use auth::{PermissionChecker, PermissionCheckerRef, PermissionReq};
 use common_base::AffectedRows;
+use common_query::logical_plan::add_insert_to_logical_plan;
 use common_query::Output;
 use common_telemetry::tracing::{self};
-use datafusion::execution::SessionStateBuilder;
 use query::parser::PromQuery;
 use servers::interceptor::{GrpcQueryInterceptor, GrpcQueryInterceptorRef};
 use servers::query_handler::grpc::{GrpcQueryHandler, RawRecordBatch};
 use servers::query_handler::sql::SqlQueryHandler;
 use session::context::QueryContextRef;
 use snafu::{ensure, OptionExt, ResultExt};
-use substrait::{DFLogicalSubstraitConvertor, SubstraitPlan};
 use table::table_name::TableName;
 
 use crate::error::{
-    CatalogSnafu, Error, InFlightWriteBytesExceededSnafu, IncompleteGrpcRequestSnafu,
-    NotSupportedSnafu, PermissionSnafu, Result, SubstraitDecodeLogicalPlanSnafu,
-    TableNotFoundSnafu, TableOperationSnafu,
+    CatalogSnafu, DataFusionSnafu, Error, InFlightWriteBytesExceededSnafu,
+    IncompleteGrpcRequestSnafu, NotSupportedSnafu, PermissionSnafu, PlanStatementSnafu, Result,
+    SubstraitDecodeLogicalPlanSnafu, TableNotFoundSnafu, TableOperationSnafu,
 };
 use crate::instance::{attach_timer, Instance};
 use crate::metrics::{
@@ -91,14 +95,31 @@ impl GrpcQueryHandler for Instance {
                     Query::LogicalPlan(plan) => {
                         // this path is useful internally when flownode needs to execute a logical plan through gRPC interface
                         let timer = GRPC_HANDLE_PLAN_ELAPSED.start_timer();
-                        let plan = DFLogicalSubstraitConvertor {}
-                            .decode(&*plan, SessionStateBuilder::default().build())
+
+                        // use dummy catalog to provide table
+                        let plan_decoder = self
+                            .query_engine()
+                            .engine_context(ctx.clone())
+                            .new_plan_decoder()
+                            .context(PlanStatementSnafu)?;
+
+                        let dummy_catalog_list =
+                            Arc::new(catalog::table_source::dummy_catalog::DummyCatalogList::new(
+                                self.catalog_manager().clone(),
+                            ));
+
+                        let logical_plan = plan_decoder
+                            .decode(bytes::Bytes::from(plan), dummy_catalog_list, true)
                             .await
                             .context(SubstraitDecodeLogicalPlanSnafu)?;
-                        let output = SqlQueryHandler::do_exec_plan(self, plan, ctx.clone()).await?;
+                        let output =
+                            SqlQueryHandler::do_exec_plan(self, logical_plan, ctx.clone()).await?;
 
                         attach_timer(output, timer)
                     }
+                    Query::InsertIntoPlan(insert) => {
+                        self.handle_insert_plan(insert, ctx.clone()).await?
+                    }
                     Query::PromRangeQuery(promql) => {
                         let timer = GRPC_HANDLE_PROMQL_ELAPSED.start_timer();
                         let prom_query = PromQuery {
@@ -284,6 +305,91 @@ fn fill_catalog_and_schema_from_context(ddl_expr: &mut DdlExpr, ctx: &QueryConte
 }
 
 impl Instance {
+    async fn handle_insert_plan(
+        &self,
+        insert: InsertIntoPlan,
+        ctx: QueryContextRef,
+    ) -> Result<Output> {
+        let timer = GRPC_HANDLE_PLAN_ELAPSED.start_timer();
+        let table_name = insert.table_name.context(IncompleteGrpcRequestSnafu {
+            err_msg: "'table_name' is absent in InsertIntoPlan",
+        })?;
+
+        // use dummy catalog to provide table
+        let plan_decoder = self
+            .query_engine()
+            .engine_context(ctx.clone())
+            .new_plan_decoder()
+            .context(PlanStatementSnafu)?;
+
+        let dummy_catalog_list =
+            Arc::new(catalog::table_source::dummy_catalog::DummyCatalogList::new(
+                self.catalog_manager().clone(),
+            ));
+
+        // no optimize yet since we still need to add stuff
+        let logical_plan = plan_decoder
+            .decode(
+                bytes::Bytes::from(insert.logical_plan),
+                dummy_catalog_list,
+                false,
+            )
+            .await
+            .context(SubstraitDecodeLogicalPlanSnafu)?;
+
+        let table = self
+            .catalog_manager()
+            .table(
+                &table_name.catalog_name,
+                &table_name.schema_name,
+                &table_name.table_name,
+                None,
+            )
+            .await
+            .context(CatalogSnafu)?
+            .with_context(|| TableNotFoundSnafu {
+                table_name: [
+                    table_name.catalog_name.clone(),
+                    table_name.schema_name.clone(),
+                    table_name.table_name.clone(),
+                ]
+                .join("."),
+            })?;
+
+        let table_info = table.table_info();
+
+        let df_schema = Arc::new(
+            table_info
+                .meta
+                .schema
+                .arrow_schema()
+                .clone()
+                .try_into()
+                .context(DataFusionSnafu)?,
+        );
+
+        let insert_into = add_insert_to_logical_plan(table_name, df_schema, logical_plan)
+            .context(SubstraitDecodeLogicalPlanSnafu)?;
+
+        let engine_ctx = self.query_engine().engine_context(ctx.clone());
+        let state = engine_ctx.state();
+        // Analyze the plan
+        let analyzed_plan = state
+            .analyzer()
+            .execute_and_check(insert_into, state.config_options(), |_, _| {})
+            .context(common_query::error::GeneralDataFusionSnafu)
+            .context(SubstraitDecodeLogicalPlanSnafu)?;
+
+        // Optimize the plan
+        let optimized_plan = state
+            .optimize(&analyzed_plan)
+            .context(common_query::error::GeneralDataFusionSnafu)
+            .context(SubstraitDecodeLogicalPlanSnafu)?;
+
+        let output = SqlQueryHandler::do_exec_plan(self, optimized_plan, ctx.clone()).await?;
+
+        Ok(attach_timer(output, timer))
+    }
     #[tracing::instrument(skip_all)]
     pub async fn handle_inserts(
         &self,

src/frontend/src/server.rs

@@ -27,6 +27,7 @@ use servers::http::{HttpServer, HttpServerBuilder};
 use servers::interceptor::LogIngestInterceptorRef;
 use servers::metrics_handler::MetricsHandler;
 use servers::mysql::server::{MysqlServer, MysqlSpawnConfig, MysqlSpawnRef};
+use servers::otel_arrow::OtelArrowServiceHandler;
 use servers::postgres::PostgresServer;
 use servers::query_handler::grpc::ServerGrpcQueryHandlerAdapter;
 use servers::query_handler::sql::ServerSqlQueryHandlerAdapter;
@@ -162,6 +163,7 @@ where
         let grpc_server = builder
             .database_handler(greptime_request_handler.clone())
             .prometheus_handler(self.instance.clone(), user_provider.clone())
+            .otel_arrow_handler(OtelArrowServiceHandler(self.instance.clone()))
             .flight_handler(Arc::new(greptime_request_handler))
             .build();
         Ok(grpc_server)

src/index/src/fulltext_index/tokenizer.rs

@@ -46,7 +46,11 @@ pub struct ChineseTokenizer;
 
 impl Tokenizer for ChineseTokenizer {
     fn tokenize<'a>(&self, text: &'a str) -> Vec<&'a str> {
-        JIEBA.cut(text, false)
+        if text.is_ascii() {
+            EnglishTokenizer {}.tokenize(text)
+        } else {
+            JIEBA.cut(text, false)
+        }
     }
 }
 

src/meta-srv/src/bootstrap.rs

@@ -66,10 +66,12 @@ use crate::election::postgres::PgElection;
 #[cfg(any(feature = "pg_kvbackend", feature = "mysql_kvbackend"))]
 use crate::election::CANDIDATE_LEASE_SECS;
 use crate::metasrv::builder::MetasrvBuilder;
-use crate::metasrv::{BackendImpl, Metasrv, MetasrvOptions, SelectorRef};
+use crate::metasrv::{BackendImpl, Metasrv, MetasrvOptions, SelectTarget, SelectorRef};
+use crate::node_excluder::NodeExcluderRef;
 use crate::selector::lease_based::LeaseBasedSelector;
 use crate::selector::load_based::LoadBasedSelector;
 use crate::selector::round_robin::RoundRobinSelector;
+use crate::selector::weight_compute::RegionNumsBasedWeightCompute;
 use crate::selector::SelectorType;
 use crate::service::admin;
 use crate::{error, Result};
@@ -294,10 +296,31 @@ pub async fn metasrv_builder(
 
     let in_memory = Arc::new(MemoryKvBackend::new()) as ResettableKvBackendRef;
 
-    let selector = match opts.selector {
-        SelectorType::LoadBased => Arc::new(LoadBasedSelector::default()) as SelectorRef,
-        SelectorType::LeaseBased => Arc::new(LeaseBasedSelector) as SelectorRef,
-        SelectorType::RoundRobin => Arc::new(RoundRobinSelector::default()) as SelectorRef,
+    let node_excluder = plugins
+        .get::<NodeExcluderRef>()
+        .unwrap_or_else(|| Arc::new(Vec::new()) as NodeExcluderRef);
+    let selector = if let Some(selector) = plugins.get::<SelectorRef>() {
+        info!("Using selector from plugins");
+        selector
+    } else {
+        let selector = match opts.selector {
+            SelectorType::LoadBased => Arc::new(LoadBasedSelector::new(
+                RegionNumsBasedWeightCompute,
+                node_excluder,
+            )) as SelectorRef,
+            SelectorType::LeaseBased => {
+                Arc::new(LeaseBasedSelector::new(node_excluder)) as SelectorRef
+            }
+            SelectorType::RoundRobin => Arc::new(RoundRobinSelector::new(
+                SelectTarget::Datanode,
+                node_excluder,
+            )) as SelectorRef,
+        };
+        info!(
+            "Using selector from options, selector type: {}",
+            opts.selector.as_ref()
+        );
+        selector
     };
 
     Ok(MetasrvBuilder::new()

src/meta-srv/src/error.rs

@@ -336,6 +336,13 @@ pub enum Error {
         location: Location,
     },
 
+    #[snafu(display("Region's leader peer changed: {}", msg))]
+    LeaderPeerChanged {
+        msg: String,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
     #[snafu(display("Invalid arguments: {}", err_msg))]
     InvalidArguments {
         err_msg: String,
@@ -914,7 +921,8 @@ impl ErrorExt for Error {
             | Error::ProcedureNotFound { .. }
             | Error::TooManyPartitions { .. }
             | Error::TomlFormat { .. }
-            | Error::HandlerNotFound { .. } => StatusCode::InvalidArguments,
+            | Error::HandlerNotFound { .. }
+            | Error::LeaderPeerChanged { .. } => StatusCode::InvalidArguments,
             Error::LeaseKeyFromUtf8 { .. }
             | Error::LeaseValueFromUtf8 { .. }
             | Error::InvalidRegionKeyFromUtf8 { .. }

src/meta-srv/src/flow_meta_alloc.rs

@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::collections::HashSet;
+
 use common_error::ext::BoxedError;
 use common_meta::ddl::flow_meta::PartitionPeerAllocator;
 use common_meta::peer::Peer;
@@ -40,6 +42,7 @@ impl PartitionPeerAllocator for FlowPeerAllocator {
                 SelectorOptions {
                     min_required_items: partitions,
                     allow_duplication: true,
+                    exclude_peer_ids: HashSet::new(),
                 },
             )
             .await

src/meta-srv/src/lib.rs

@@ -31,6 +31,7 @@ pub mod metasrv;
 pub mod metrics;
 #[cfg(feature = "mock")]
 pub mod mocks;
+pub mod node_excluder;
 pub mod procedure;
 pub mod pubsub;
 pub mod region;

src/meta-srv/src/metasrv.rs

@@ -14,7 +14,7 @@
 
 pub mod builder;
 
-use std::fmt::Display;
+use std::fmt::{self, Display};
 use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::{Arc, Mutex, RwLock};
 use std::time::Duration;
@@ -96,7 +96,7 @@ pub enum BackendImpl {
     MysqlStore,
 }
 
-#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+#[derive(Clone, PartialEq, Serialize, Deserialize)]
 #[serde(default)]
 pub struct MetasrvOptions {
     /// The address the server listens on.
@@ -111,6 +111,11 @@ pub struct MetasrvOptions {
     pub use_memory_store: bool,
     /// Whether to enable region failover.
     pub enable_region_failover: bool,
+    /// Whether to allow region failover on local WAL.
+    ///
+    /// If it's true, the region failover will be allowed even if the local WAL is used.
+    /// Note that this option is not recommended to be set to true, because it may lead to data loss during failover.
+    pub allow_region_failover_on_local_wal: bool,
     /// The HTTP server options.
     pub http: HttpOptions,
     /// The logging options.
@@ -161,6 +166,47 @@ pub struct MetasrvOptions {
     pub node_max_idle_time: Duration,
 }
 
+impl fmt::Debug for MetasrvOptions {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        let mut debug_struct = f.debug_struct("MetasrvOptions");
+        debug_struct
+            .field("bind_addr", &self.bind_addr)
+            .field("server_addr", &self.server_addr)
+            .field("store_addrs", &self.sanitize_store_addrs())
+            .field("selector", &self.selector)
+            .field("use_memory_store", &self.use_memory_store)
+            .field("enable_region_failover", &self.enable_region_failover)
+            .field(
+                "allow_region_failover_on_local_wal",
+                &self.allow_region_failover_on_local_wal,
+            )
+            .field("http", &self.http)
+            .field("logging", &self.logging)
+            .field("procedure", &self.procedure)
+            .field("failure_detector", &self.failure_detector)
+            .field("datanode", &self.datanode)
+            .field("enable_telemetry", &self.enable_telemetry)
+            .field("data_home", &self.data_home)
+            .field("wal", &self.wal)
+            .field("export_metrics", &self.export_metrics)
+            .field("store_key_prefix", &self.store_key_prefix)
+            .field("max_txn_ops", &self.max_txn_ops)
+            .field("flush_stats_factor", &self.flush_stats_factor)
+            .field("tracing", &self.tracing)
+            .field("backend", &self.backend);
+
+        #[cfg(any(feature = "pg_kvbackend", feature = "mysql_kvbackend"))]
+        debug_struct.field("meta_table_name", &self.meta_table_name);
+
+        #[cfg(feature = "pg_kvbackend")]
+        debug_struct.field("meta_election_lock_id", &self.meta_election_lock_id);
+
+        debug_struct
+            .field("node_max_idle_time", &self.node_max_idle_time)
+            .finish()
+    }
+}
+
 const DEFAULT_METASRV_ADDR_PORT: &str = "3002";
 
 impl Default for MetasrvOptions {
@@ -173,6 +219,7 @@ impl Default for MetasrvOptions {
             selector: SelectorType::default(),
             use_memory_store: false,
             enable_region_failover: false,
+            allow_region_failover_on_local_wal: false,
             http: HttpOptions::default(),
             logging: LoggingOptions {
                 dir: format!("{METASRV_HOME}/logs"),
@@ -243,6 +290,13 @@ impl MetasrvOptions {
             common_telemetry::debug!("detect local IP is not supported on Android");
         }
     }
+
+    fn sanitize_store_addrs(&self) -> Vec<String> {
+        self.store_addrs
+            .iter()
+            .map(|addr| common_meta::kv_backend::util::sanitize_connection_string(addr))
+            .collect()
+    }
 }
 
 pub struct MetasrvInfo {

src/meta-srv/src/metasrv/builder.rs

@@ -40,7 +40,8 @@ use common_meta::state_store::KvStateStore;
 use common_meta::wal_options_allocator::{build_kafka_client, build_wal_options_allocator};
 use common_procedure::local::{LocalManager, ManagerConfig};
 use common_procedure::ProcedureManagerRef;
-use snafu::ResultExt;
+use common_telemetry::warn;
+use snafu::{ensure, ResultExt};
 
 use crate::cache_invalidator::MetasrvCacheInvalidator;
 use crate::cluster::{MetaPeerClientBuilder, MetaPeerClientRef};
@@ -190,7 +191,7 @@ impl MetasrvBuilder {
 
         let meta_peer_client = meta_peer_client
             .unwrap_or_else(|| build_default_meta_peer_client(&election, &in_memory));
-        let selector = selector.unwrap_or_else(|| Arc::new(LeaseBasedSelector));
+        let selector = selector.unwrap_or_else(|| Arc::new(LeaseBasedSelector::default()));
         let pushers = Pushers::default();
         let mailbox = build_mailbox(&kv_backend, &pushers);
         let procedure_manager = build_procedure_manager(&options, &kv_backend);
@@ -234,13 +235,17 @@ impl MetasrvBuilder {
             ))
         });
 
+        let flow_selector = Arc::new(RoundRobinSelector::new(
+            SelectTarget::Flownode,
+            Arc::new(Vec::new()),
+        )) as SelectorRef;
+
         let flow_metadata_allocator = {
             // for now flownode just use round-robin selector
-            let flow_selector = RoundRobinSelector::new(SelectTarget::Flownode);
             let flow_selector_ctx = selector_ctx.clone();
             let peer_allocator = Arc::new(FlowPeerAllocator::new(
                 flow_selector_ctx,
-                Arc::new(flow_selector),
+                flow_selector.clone(),
             ));
             let seq = Arc::new(
                 SequenceBuilder::new(FLOW_ID_SEQ, kv_backend.clone())
@@ -272,18 +277,25 @@ impl MetasrvBuilder {
             },
         ));
         let peer_lookup_service = Arc::new(MetaPeerLookupService::new(meta_peer_client.clone()));
+
         if !is_remote_wal && options.enable_region_failover {
-            return error::UnexpectedSnafu {
-                violated: "Region failover is not supported in the local WAL implementation!",
+            ensure!(
+                options.allow_region_failover_on_local_wal,
+                error::UnexpectedSnafu {
+                    violated: "Region failover is not supported in the local WAL implementation! 
+                    If you want to enable region failover for local WAL, please set `allow_region_failover_on_local_wal` to true.",
+                }
+            );
+            if options.allow_region_failover_on_local_wal {
+                warn!("Region failover is force enabled in the local WAL implementation! This may lead to data loss during failover!");
             }
-            .fail();
         }
 
         let (tx, rx) = RegionSupervisor::channel();
         let (region_failure_detector_controller, region_supervisor_ticker): (
             RegionFailureDetectorControllerRef,
             Option<std::sync::Arc<RegionSupervisorTicker>>,
-        ) = if options.enable_region_failover && is_remote_wal {
+        ) = if options.enable_region_failover {
             (
                 Arc::new(RegionFailureDetectorControl::new(tx.clone())) as _,
                 Some(Arc::new(RegionSupervisorTicker::new(
@@ -309,7 +321,7 @@ impl MetasrvBuilder {
         ));
         region_migration_manager.try_start()?;
 
-        let region_failover_handler = if options.enable_region_failover && is_remote_wal {
+        let region_failover_handler = if options.enable_region_failover {
             let region_supervisor = RegionSupervisor::new(
                 rx,
                 options.failure_detector,
@@ -420,7 +432,7 @@ impl MetasrvBuilder {
             meta_peer_client: meta_peer_client.clone(),
             selector,
             // TODO(jeremy): We do not allow configuring the flow selector.
-            flow_selector: Arc::new(RoundRobinSelector::new(SelectTarget::Flownode)),
+            flow_selector,
             handler_group: RwLock::new(None),
             handler_group_builder: Mutex::new(Some(handler_group_builder)),
             election,

src/meta-srv/src/metrics.rs

@@ -62,11 +62,22 @@ lazy_static! {
         register_int_counter!("greptime_meta_region_migration_fail", "meta region migration fail").unwrap();
     // The heartbeat stat memory size histogram.
     pub static ref METRIC_META_HEARTBEAT_STAT_MEMORY_SIZE: Histogram =
-        register_histogram!("greptime_meta_heartbeat_stat_memory_size", "meta heartbeat stat memory size").unwrap();
+        register_histogram!("greptime_meta_heartbeat_stat_memory_size", "meta heartbeat stat memory size", vec![
+            100.0, 500.0, 1000.0, 1500.0, 2000.0, 3000.0, 5000.0, 10000.0, 20000.0
+        ]).unwrap();
     // The heartbeat rate counter.
     pub static ref METRIC_META_HEARTBEAT_RATE: IntCounter =
         register_int_counter!("greptime_meta_heartbeat_rate", "meta heartbeat arrival rate").unwrap();
     /// The remote WAL prune execute counter.
     pub static ref METRIC_META_REMOTE_WAL_PRUNE_EXECUTE: IntCounterVec =
         register_int_counter_vec!("greptime_meta_remote_wal_prune_execute", "meta remote wal prune execute", &["topic_name"]).unwrap();
+    /// The migration stage elapsed histogram.
+    pub static ref METRIC_META_REGION_MIGRATION_STAGE_ELAPSED: HistogramVec = register_histogram_vec!(
+        "greptime_meta_region_migration_stage_elapsed",
+        "meta region migration stage elapsed",
+        &["stage"],
+        // 0.01 ~ 1000
+        exponential_buckets(0.01, 10.0, 7).unwrap(),
+    )
+    .unwrap();
 }

src/meta-srv/src/node_excluder.rs

@@ -0,0 +1,32 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::sync::Arc;
+
+use common_meta::DatanodeId;
+
+pub type NodeExcluderRef = Arc<dyn NodeExcluder>;
+
+/// [NodeExcluder] is used to help decide whether some nodes should be excluded (out of consideration)
+/// in certain situations. For example, in some node selectors.
+pub trait NodeExcluder: Send + Sync {
+    /// Returns the excluded datanode ids.
+    fn excluded_datanode_ids(&self) -> &Vec<DatanodeId>;
+}
+
+impl NodeExcluder for Vec<DatanodeId> {
+    fn excluded_datanode_ids(&self) -> &Vec<DatanodeId> {
+        self
+    }
+}

src/meta-srv/src/procedure/region_migration.rs

@@ -25,7 +25,7 @@ pub(crate) mod update_metadata;
 pub(crate) mod upgrade_candidate_region;
 
 use std::any::Any;
-use std::fmt::Debug;
+use std::fmt::{Debug, Display};
 use std::time::Duration;
 
 use common_error::ext::BoxedError;
@@ -43,7 +43,7 @@ use common_procedure::error::{
     Error as ProcedureError, FromJsonSnafu, Result as ProcedureResult, ToJsonSnafu,
 };
 use common_procedure::{Context as ProcedureContext, LockKey, Procedure, Status, StringKey};
-use common_telemetry::info;
+use common_telemetry::{error, info};
 use manager::RegionMigrationProcedureGuard;
 pub use manager::{
     RegionMigrationManagerRef, RegionMigrationProcedureTask, RegionMigrationProcedureTracker,
@@ -55,9 +55,15 @@ use tokio::time::Instant;
 
 use self::migration_start::RegionMigrationStart;
 use crate::error::{self, Result};
-use crate::metrics::{METRIC_META_REGION_MIGRATION_ERROR, METRIC_META_REGION_MIGRATION_EXECUTE};
+use crate::metrics::{
+    METRIC_META_REGION_MIGRATION_ERROR, METRIC_META_REGION_MIGRATION_EXECUTE,
+    METRIC_META_REGION_MIGRATION_STAGE_ELAPSED,
+};
 use crate::service::mailbox::MailboxRef;
 
+/// The default timeout for region migration.
+pub const DEFAULT_REGION_MIGRATION_TIMEOUT: Duration = Duration::from_secs(120);
+
 /// It's shared in each step and available even after recovering.
 ///
 /// It will only be updated/stored after the Red node has succeeded.
@@ -100,6 +106,82 @@ impl PersistentContext {
     }
 }
 
+/// Metrics of region migration.
+#[derive(Debug, Clone, Default)]
+pub struct Metrics {
+    /// Elapsed time of downgrading region and upgrading region.
+    operations_elapsed: Duration,
+    /// Elapsed time of downgrading leader region.
+    downgrade_leader_region_elapsed: Duration,
+    /// Elapsed time of open candidate region.
+    open_candidate_region_elapsed: Duration,
+    /// Elapsed time of upgrade candidate region.
+    upgrade_candidate_region_elapsed: Duration,
+}
+
+impl Display for Metrics {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(
+            f,
+            "operations_elapsed: {:?}, downgrade_leader_region_elapsed: {:?}, open_candidate_region_elapsed: {:?}, upgrade_candidate_region_elapsed: {:?}",
+            self.operations_elapsed,
+            self.downgrade_leader_region_elapsed,
+            self.open_candidate_region_elapsed,
+            self.upgrade_candidate_region_elapsed
+        )
+    }
+}
+
+impl Metrics {
+    /// Updates the elapsed time of downgrading region and upgrading region.
+    pub fn update_operations_elapsed(&mut self, elapsed: Duration) {
+        self.operations_elapsed += elapsed;
+    }
+
+    /// Updates the elapsed time of downgrading leader region.
+    pub fn update_downgrade_leader_region_elapsed(&mut self, elapsed: Duration) {
+        self.downgrade_leader_region_elapsed += elapsed;
+    }
+
+    /// Updates the elapsed time of open candidate region.
+    pub fn update_open_candidate_region_elapsed(&mut self, elapsed: Duration) {
+        self.open_candidate_region_elapsed += elapsed;
+    }
+
+    /// Updates the elapsed time of upgrade candidate region.
+    pub fn update_upgrade_candidate_region_elapsed(&mut self, elapsed: Duration) {
+        self.upgrade_candidate_region_elapsed += elapsed;
+    }
+}
+
+impl Drop for Metrics {
+    fn drop(&mut self) {
+        if !self.operations_elapsed.is_zero() {
+            METRIC_META_REGION_MIGRATION_STAGE_ELAPSED
+                .with_label_values(&["operations"])
+                .observe(self.operations_elapsed.as_secs_f64());
+        }
+
+        if !self.downgrade_leader_region_elapsed.is_zero() {
+            METRIC_META_REGION_MIGRATION_STAGE_ELAPSED
+                .with_label_values(&["downgrade_leader_region"])
+                .observe(self.downgrade_leader_region_elapsed.as_secs_f64());
+        }
+
+        if !self.open_candidate_region_elapsed.is_zero() {
+            METRIC_META_REGION_MIGRATION_STAGE_ELAPSED
+                .with_label_values(&["open_candidate_region"])
+                .observe(self.open_candidate_region_elapsed.as_secs_f64());
+        }
+
+        if !self.upgrade_candidate_region_elapsed.is_zero() {
+            METRIC_META_REGION_MIGRATION_STAGE_ELAPSED
+                .with_label_values(&["upgrade_candidate_region"])
+                .observe(self.upgrade_candidate_region_elapsed.as_secs_f64());
+        }
+    }
+}
+
 /// It's shared in each step and available in executing (including retrying).
 ///
 /// It will be dropped if the procedure runner crashes.
@@ -129,8 +211,8 @@ pub struct VolatileContext {
     leader_region_last_entry_id: Option<u64>,
     /// The last_entry_id of leader metadata region (Only used for metric engine).
     leader_region_metadata_last_entry_id: Option<u64>,
-    /// Elapsed time of downgrading region and upgrading region.
-    operations_elapsed: Duration,
+    /// Metrics of region migration.
+    metrics: Metrics,
 }
 
 impl VolatileContext {
@@ -228,12 +310,35 @@ impl Context {
     pub fn next_operation_timeout(&self) -> Option<Duration> {
         self.persistent_ctx
             .timeout
-            .checked_sub(self.volatile_ctx.operations_elapsed)
+            .checked_sub(self.volatile_ctx.metrics.operations_elapsed)
     }
 
     /// Updates operations elapsed.
     pub fn update_operations_elapsed(&mut self, instant: Instant) {
-        self.volatile_ctx.operations_elapsed += instant.elapsed();
+        self.volatile_ctx
+            .metrics
+            .update_operations_elapsed(instant.elapsed());
+    }
+
+    /// Updates the elapsed time of downgrading leader region.
+    pub fn update_downgrade_leader_region_elapsed(&mut self, instant: Instant) {
+        self.volatile_ctx
+            .metrics
+            .update_downgrade_leader_region_elapsed(instant.elapsed());
+    }
+
+    /// Updates the elapsed time of open candidate region.
+    pub fn update_open_candidate_region_elapsed(&mut self, instant: Instant) {
+        self.volatile_ctx
+            .metrics
+            .update_open_candidate_region_elapsed(instant.elapsed());
+    }
+
+    /// Updates the elapsed time of upgrade candidate region.
+    pub fn update_upgrade_candidate_region_elapsed(&mut self, instant: Instant) {
+        self.volatile_ctx
+            .metrics
+            .update_upgrade_candidate_region_elapsed(instant.elapsed());
     }
 
     /// Returns address of meta server.
@@ -547,6 +652,14 @@ impl Procedure for RegionMigrationProcedure {
                     .inc();
                 ProcedureError::retry_later(e)
             } else {
+                error!(
+                    e;
+                    "Region migration procedure failed, region_id: {}, from_peer: {}, to_peer: {}, {}",
+                    self.context.region_id(),
+                    self.context.persistent_ctx.from_peer,
+                    self.context.persistent_ctx.to_peer,
+                    self.context.volatile_ctx.metrics,
+                );
                 METRIC_META_REGION_MIGRATION_ERROR
                     .with_label_values(&[name, "external"])
                     .inc();

src/meta-srv/src/procedure/region_migration/close_downgraded_region.rs

@@ -46,7 +46,13 @@ impl State for CloseDowngradedRegion {
             let region_id = ctx.region_id();
             warn!(err; "Failed to close downgraded leader region: {region_id} on datanode {:?}", downgrade_leader_datanode);
         }
-
+        info!(
+            "Region migration is finished: region_id: {}, from_peer: {}, to_peer: {}, {}",
+            ctx.region_id(),
+            ctx.persistent_ctx.from_peer,
+            ctx.persistent_ctx.to_peer,
+            ctx.volatile_ctx.metrics,
+        );
         Ok((Box::new(RegionMigrationEnd), Status::done()))
     }
 

src/meta-srv/src/procedure/region_migration/downgrade_leader_region.rs

@@ -54,6 +54,7 @@ impl Default for DowngradeLeaderRegion {
 #[typetag::serde]
 impl State for DowngradeLeaderRegion {
     async fn next(&mut self, ctx: &mut Context) -> Result<(Box<dyn State>, Status)> {
+        let now = Instant::now();
         // Ensures the `leader_region_lease_deadline` must exist after recovering.
         ctx.volatile_ctx
             .set_leader_region_lease_deadline(Duration::from_secs(REGION_LEASE_SECS));
@@ -77,6 +78,7 @@ impl State for DowngradeLeaderRegion {
                 }
             }
         }
+        ctx.update_downgrade_leader_region_elapsed(now);
 
         Ok((
             Box::new(UpgradeCandidateRegion::default()),
@@ -348,7 +350,8 @@ mod tests {
         let env = TestingEnv::new();
         let mut ctx = env.context_factory().new_context(persistent_context);
         prepare_table_metadata(&ctx, HashMap::default()).await;
-        ctx.volatile_ctx.operations_elapsed = ctx.persistent_ctx.timeout + Duration::from_secs(1);
+        ctx.volatile_ctx.metrics.operations_elapsed =
+            ctx.persistent_ctx.timeout + Duration::from_secs(1);
 
         let err = state.downgrade_region(&mut ctx).await.unwrap_err();
 
@@ -591,7 +594,8 @@ mod tests {
         let mut ctx = env.context_factory().new_context(persistent_context);
         let mailbox_ctx = env.mailbox_context();
         let mailbox = mailbox_ctx.mailbox().clone();
-        ctx.volatile_ctx.operations_elapsed = ctx.persistent_ctx.timeout + Duration::from_secs(1);
+        ctx.volatile_ctx.metrics.operations_elapsed =
+            ctx.persistent_ctx.timeout + Duration::from_secs(1);
 
         let (tx, rx) = tokio::sync::mpsc::channel(1);
         mailbox_ctx

src/meta-srv/src/procedure/region_migration/manager.rs

@@ -267,8 +267,8 @@ impl RegionMigrationManager {
 
         ensure!(
             leader_peer.id == task.from_peer.id,
-            error::InvalidArgumentsSnafu {
-                err_msg: format!(
+            error::LeaderPeerChangedSnafu {
+                msg: format!(
                     "Region's leader peer({}) is not the `from_peer`({}), region: {}",
                     leader_peer.id, task.from_peer.id, task.region_id
                 ),
@@ -507,8 +507,8 @@ mod test {
             .await;
 
         let err = manager.submit_procedure(task).await.unwrap_err();
-        assert_matches!(err, error::Error::InvalidArguments { .. });
-        assert_eq!(err.to_string(), "Invalid arguments: Region's leader peer(3) is not the `from_peer`(1), region: 4398046511105(1024, 1)");
+        assert_matches!(err, error::Error::LeaderPeerChanged { .. });
+        assert_eq!(err.to_string(), "Region's leader peer changed: Region's leader peer(3) is not the `from_peer`(1), region: 4398046511105(1024, 1)");
     }
 
     #[tokio::test]