feat: opendal metrics

Signed-off-by: evenyag <realevenyag@gmail.com>

.github/CODEOWNERS

@@ -4,7 +4,7 @@
 
 * @GreptimeTeam/db-approver
 
-## [Module] Databse Engine
+## [Module] Database Engine
 /src/index @zhongzc
 /src/mito2 @evenyag @v0y4g3r @waynexia
 /src/query @evenyag

.github/actions/build-dev-builder-images/action.yml

@@ -41,7 +41,14 @@ runs:
         username: ${{ inputs.dockerhub-image-registry-username }}
         password: ${{ inputs.dockerhub-image-registry-token }}
 
-    - name: Build and push dev-builder-ubuntu image
+    - name: Set up qemu for multi-platform builds
+      uses: docker/setup-qemu-action@v3
+      with:
+        platforms: linux/amd64,linux/arm64
+        # The latest version will lead to segmentation fault.
+        image: tonistiigi/binfmt:qemu-v7.0.0-28
+
+    - name: Build and push dev-builder-ubuntu image # Build image for amd64 and arm64 platform.
       shell: bash
       if: ${{ inputs.build-dev-builder-ubuntu == 'true' }}
       run: |
@@ -52,7 +59,7 @@ runs:
           IMAGE_NAMESPACE=${{ inputs.dockerhub-image-namespace }} \
           DEV_BUILDER_IMAGE_TAG=${{ inputs.version }}
 
-    - name: Build and push dev-builder-centos image
+    - name: Build and push dev-builder-centos image # Only build image for amd64 platform.
       shell: bash
       if: ${{ inputs.build-dev-builder-centos == 'true' }}
       run: |
@@ -69,8 +76,7 @@ runs:
       run: |
         make dev-builder \
           BASE_IMAGE=android \
+          BUILDX_MULTI_PLATFORM_BUILD=amd64 \
           IMAGE_REGISTRY=${{ inputs.dockerhub-image-registry }} \
           IMAGE_NAMESPACE=${{ inputs.dockerhub-image-namespace }} \
-          DEV_BUILDER_IMAGE_TAG=${{ inputs.version }} && \
-
-        docker push ${{ inputs.dockerhub-image-registry }}/${{ inputs.dockerhub-image-namespace }}/dev-builder-android:${{ inputs.version }}
+          DEV_BUILDER_IMAGE_TAG=${{ inputs.version }}

.github/actions/build-linux-artifacts/action.yml

@@ -52,7 +52,7 @@ runs:
       uses: ./.github/actions/build-greptime-binary
       with:
         base-image: ubuntu
-        features: servers/dashboard,pg_kvbackend
+        features: servers/dashboard
         cargo-profile: ${{ inputs.cargo-profile }}
         artifacts-dir: greptime-linux-${{ inputs.arch }}-${{ inputs.version }}
         version: ${{ inputs.version }}
@@ -70,7 +70,7 @@ runs:
       if: ${{ inputs.arch == 'amd64' && inputs.dev-mode == 'false' }} # Builds greptime for centos if the host machine is amd64.
       with:
         base-image: centos
-        features: servers/dashboard,pg_kvbackend
+        features: servers/dashboard
         cargo-profile: ${{ inputs.cargo-profile }}
         artifacts-dir: greptime-linux-${{ inputs.arch }}-centos-${{ inputs.version }}
         version: ${{ inputs.version }}

.github/actions/build-windows-artifacts/action.yml

@@ -47,7 +47,6 @@ runs:
       shell: pwsh
       run: make test sqlness-test
       env:
-        RUSTUP_WINDOWS_PATH_ADD_BIN: 1 # Workaround for https://github.com/nextest-rs/nextest/issues/1493
         RUST_BACKTRACE: 1
         SQLNESS_OPTS: "--preserve-state"
 

.github/actions/release-cn-artifacts/action.yaml

@@ -64,11 +64,11 @@ inputs:
   upload-max-retry-times:
     description: Max retry times for uploading artifacts to S3
     required: false
-    default: "20"
+    default: "30"
   upload-retry-timeout:
     description: Timeout for uploading artifacts to S3
     required: false
-    default: "30" # minutes
+    default: "120" # minutes
 runs:
   using: composite
   steps:

.github/actions/setup-greptimedb-cluster/action.yml

@@ -8,15 +8,15 @@ inputs:
     default: 2
     description: "Number of Datanode replicas"
   meta-replicas:
-    default: 1
+    default: 2
     description: "Number of Metasrv replicas"
-  image-registry: 
+  image-registry:
     default: "docker.io"
     description: "Image registry"
-  image-repository: 
+  image-repository:
     default: "greptime/greptimedb"
     description: "Image repository"
-  image-tag: 
+  image-tag:
     default: "latest"
     description: 'Image tag'
   etcd-endpoints:
@@ -32,12 +32,12 @@ runs:
   steps:
   - name: Install GreptimeDB operator
     uses: nick-fields/retry@v3
-    with: 
+    with:
       timeout_minutes: 3
       max_attempts: 3
       shell: bash
       command: |
-        helm repo add greptime https://greptimeteam.github.io/helm-charts/ 
+        helm repo add greptime https://greptimeteam.github.io/helm-charts/
         helm repo update
         helm upgrade \
           --install \
@@ -48,10 +48,10 @@ runs:
           --wait-for-jobs
   - name: Install GreptimeDB cluster
     shell: bash
-    run: | 
+    run: |
       helm upgrade \
         --install my-greptimedb \
-        --set meta.etcdEndpoints=${{ inputs.etcd-endpoints }} \
+        --set meta.backendStorage.etcd.endpoints=${{ inputs.etcd-endpoints }} \
         --set meta.enableRegionFailover=${{ inputs.enable-region-failover }} \
         --set image.registry=${{ inputs.image-registry }} \
         --set image.repository=${{ inputs.image-repository }}  \
@@ -59,7 +59,7 @@ runs:
         --set base.podTemplate.main.resources.requests.cpu=50m \
         --set base.podTemplate.main.resources.requests.memory=256Mi \
         --set base.podTemplate.main.resources.limits.cpu=2000m \
-        --set base.podTemplate.main.resources.limits.memory=2Gi \
+        --set base.podTemplate.main.resources.limits.memory=3Gi \
         --set frontend.replicas=${{ inputs.frontend-replicas }} \
         --set datanode.replicas=${{ inputs.datanode-replicas }} \
         --set meta.replicas=${{ inputs.meta-replicas }} \
@@ -72,7 +72,7 @@ runs:
   - name: Wait for GreptimeDB
     shell: bash
     run: |
-      while true; do 
+      while true; do
         PHASE=$(kubectl -n my-greptimedb get gtc my-greptimedb -o jsonpath='{.status.clusterPhase}')
         if [ "$PHASE" == "Running" ]; then
           echo "Cluster is ready"
@@ -86,10 +86,10 @@ runs:
   - name: Print GreptimeDB info
     if: always()
     shell: bash
-    run: | 
+    run: |
       kubectl get all --show-labels -n my-greptimedb
   - name: Describe Nodes
     if: always()
     shell: bash
-    run: | 
+    run: |
       kubectl describe nodes

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

@@ -2,13 +2,14 @@ meta:
   configData: |-
     [runtime]
     global_rt_size = 4
-    
+
     [wal]
     provider = "kafka"
     broker_endpoints = ["kafka.kafka-cluster.svc.cluster.local:9092"]
     num_topics = 3
+    auto_prune_interval = "30s"
+    trigger_flush_threshold = 100
 
-        
     [datanode]
     [datanode.client]
     timeout = "120s"
@@ -21,7 +22,7 @@ datanode:
     [wal]
     provider = "kafka"
     broker_endpoints = ["kafka.kafka-cluster.svc.cluster.local:9092"]
-    linger = "2ms"
+    overwrite_entry_start_id = true
 frontend:
   configData: |-
     [runtime]

.github/actions/start-runner/action.yml

@@ -56,7 +56,7 @@ runs:
 
     - name: Start EC2 runner
       if: startsWith(inputs.runner, 'ec2')
-      uses: machulav/ec2-github-runner@v2
+      uses: machulav/ec2-github-runner@v2.3.8
       id: start-linux-arm64-ec2-runner
       with:
         mode: start

.github/actions/stop-runner/action.yml

@@ -33,7 +33,7 @@ runs:
 
     - name: Stop EC2 runner
       if: ${{ inputs.label && inputs.ec2-instance-id }}
-      uses: machulav/ec2-github-runner@v2
+      uses: machulav/ec2-github-runner@v2.3.8
       with:
         mode: stop
         label: ${{ inputs.label }}

.github/labeler.yaml

@@ -0,0 +1,15 @@
+ci:
+  - changed-files:
+      - any-glob-to-any-file: .github/**
+
+docker:
+  - changed-files:
+      - any-glob-to-any-file: docker/**
+
+documentation:
+  - changed-files:
+      - any-glob-to-any-file: docs/**
+
+dashboard:
+  - changed-files:
+      - any-glob-to-any-file: grafana/**

.github/scripts/check-version.sh

@@ -0,0 +1,42 @@
+#!/bin/bash
+
+# Get current version
+CURRENT_VERSION=$1
+if [ -z "$CURRENT_VERSION" ]; then
+  echo "Error: Failed to get current version"
+  exit 1
+fi
+
+# Get the latest version from GitHub Releases
+API_RESPONSE=$(curl -s "https://api.github.com/repos/GreptimeTeam/greptimedb/releases/latest")
+
+if [ -z "$API_RESPONSE" ] || [ "$(echo "$API_RESPONSE" | jq -r '.message')" = "Not Found" ]; then
+  echo "Error: Failed to fetch latest version from GitHub"
+  exit 1
+fi
+
+# Get the latest version
+LATEST_VERSION=$(echo "$API_RESPONSE" | jq -r '.tag_name')
+
+if [ -z "$LATEST_VERSION" ] || [ "$LATEST_VERSION" = "null" ]; then
+  echo "Error: No valid version found in GitHub releases"
+  exit 1
+fi
+
+# Cleaned up version number format (removed possible 'v' prefix and -nightly suffix)
+CLEAN_CURRENT=$(echo "$CURRENT_VERSION" | sed 's/^v//' | sed 's/-nightly-.*//')
+CLEAN_LATEST=$(echo "$LATEST_VERSION" | sed 's/^v//' | sed 's/-nightly-.*//')
+
+echo "Current version: $CLEAN_CURRENT"
+echo "Latest release version: $CLEAN_LATEST"
+
+# Use sort -V to compare versions
+HIGHER_VERSION=$(printf "%s\n%s" "$CLEAN_CURRENT" "$CLEAN_LATEST" | sort -V | tail -n1)
+
+if [ "$HIGHER_VERSION" = "$CLEAN_CURRENT" ]; then
+  echo "Current version ($CLEAN_CURRENT) is NEWER than or EQUAL to latest ($CLEAN_LATEST)"
+  echo "should-push-latest-tag=true" >> $GITHUB_OUTPUT
+else
+  echo "Current version ($CLEAN_CURRENT) is OLDER than latest ($CLEAN_LATEST)"
+  echo "should-push-latest-tag=false" >> $GITHUB_OUTPUT
+fi

.github/scripts/create-version.sh

@@ -8,24 +8,25 @@ set -e
 # - If it's a nightly build, the version is 'nightly-YYYYMMDD-$(git rev-parse --short HEAD)', like 'nightly-20230712-e5b243c'.
 # create_version ${GIHUB_EVENT_NAME} ${NEXT_RELEASE_VERSION} ${NIGHTLY_RELEASE_PREFIX}
 function create_version() {
-  # Read from envrionment variables.
+  # Read from environment variables.
   if [ -z "$GITHUB_EVENT_NAME" ]; then
-      echo "GITHUB_EVENT_NAME is empty"
+      echo "GITHUB_EVENT_NAME is empty" >&2
       exit 1
   fi
 
   if [ -z "$NEXT_RELEASE_VERSION" ]; then
-      echo "NEXT_RELEASE_VERSION is empty"
-      exit 1
+      echo "NEXT_RELEASE_VERSION is empty, use version from Cargo.toml" >&2
+      # NOTE: Need a `v` prefix for the version string.
+      export NEXT_RELEASE_VERSION=v$(grep '^version = ' Cargo.toml | cut -d '"' -f 2 | head -n 1)
   fi
 
   if [ -z "$NIGHTLY_RELEASE_PREFIX" ]; then
-      echo "NIGHTLY_RELEASE_PREFIX is empty"
+      echo "NIGHTLY_RELEASE_PREFIX is empty" >&2
       exit 1
   fi
 
   # Reuse $NEXT_RELEASE_VERSION to identify whether it's a nightly build.
-  # It will be like 'nigtly-20230808-7d0d8dc6'.
+  # It will be like 'nightly-20230808-7d0d8dc6'.
   if [ "$NEXT_RELEASE_VERSION" = nightly ]; then
     echo "$NIGHTLY_RELEASE_PREFIX-$(date "+%Y%m%d")-$(git rev-parse --short HEAD)"
     exit 0
@@ -35,7 +36,7 @@ function create_version() {
   # It will be like 'dev-2023080819-f0e7216c'.
   if [ "$NEXT_RELEASE_VERSION" = dev ]; then
     if [ -z "$COMMIT_SHA" ]; then
-      echo "COMMIT_SHA is empty in dev build"
+      echo "COMMIT_SHA is empty in dev build" >&2
       exit 1
     fi
     echo "dev-$(date "+%Y%m%d-%s")-$(echo "$COMMIT_SHA" | cut -c1-8)"
@@ -45,7 +46,7 @@ function create_version() {
   # Note: Only output 'version=xxx' to stdout when everything is ok, so that it can be used in GitHub Actions Outputs.
   if [ "$GITHUB_EVENT_NAME" = push ]; then
     if [ -z "$GITHUB_REF_NAME" ]; then
-      echo "GITHUB_REF_NAME is empty in push event"
+      echo "GITHUB_REF_NAME is empty in push event" >&2
       exit 1
     fi
     echo "$GITHUB_REF_NAME"
@@ -54,15 +55,15 @@ function create_version() {
   elif [ "$GITHUB_EVENT_NAME" = schedule ]; then
     echo "$NEXT_RELEASE_VERSION-$NIGHTLY_RELEASE_PREFIX-$(date "+%Y%m%d")"
   else
-    echo "Unsupported GITHUB_EVENT_NAME: $GITHUB_EVENT_NAME"
+    echo "Unsupported GITHUB_EVENT_NAME: $GITHUB_EVENT_NAME" >&2
     exit 1
   fi
 }
 
 # You can run as following examples:
-#  GITHUB_EVENT_NAME=push NEXT_RELEASE_VERSION=v0.4.0 NIGHTLY_RELEASE_PREFIX=nigtly GITHUB_REF_NAME=v0.3.0 ./create-version.sh
-#  GITHUB_EVENT_NAME=workflow_dispatch NEXT_RELEASE_VERSION=v0.4.0 NIGHTLY_RELEASE_PREFIX=nigtly ./create-version.sh
-#  GITHUB_EVENT_NAME=schedule NEXT_RELEASE_VERSION=v0.4.0 NIGHTLY_RELEASE_PREFIX=nigtly ./create-version.sh
-#  GITHUB_EVENT_NAME=schedule NEXT_RELEASE_VERSION=nightly NIGHTLY_RELEASE_PREFIX=nigtly ./create-version.sh
-#  GITHUB_EVENT_NAME=workflow_dispatch COMMIT_SHA=f0e7216c4bb6acce9b29a21ec2d683be2e3f984a NEXT_RELEASE_VERSION=dev NIGHTLY_RELEASE_PREFIX=nigtly ./create-version.sh
+#  GITHUB_EVENT_NAME=push NEXT_RELEASE_VERSION=v0.4.0 NIGHTLY_RELEASE_PREFIX=nightly GITHUB_REF_NAME=v0.3.0 ./create-version.sh
+#  GITHUB_EVENT_NAME=workflow_dispatch NEXT_RELEASE_VERSION=v0.4.0 NIGHTLY_RELEASE_PREFIX=nightly ./create-version.sh
+#  GITHUB_EVENT_NAME=schedule NEXT_RELEASE_VERSION=v0.4.0 NIGHTLY_RELEASE_PREFIX=nightly ./create-version.sh
+#  GITHUB_EVENT_NAME=schedule NEXT_RELEASE_VERSION=nightly NIGHTLY_RELEASE_PREFIX=nightly ./create-version.sh
+#  GITHUB_EVENT_NAME=workflow_dispatch COMMIT_SHA=f0e7216c4bb6acce9b29a21ec2d683be2e3f984a NEXT_RELEASE_VERSION=dev NIGHTLY_RELEASE_PREFIX=nightly ./create-version.sh
 create_version

.github/scripts/deploy-greptimedb.sh

@@ -10,7 +10,7 @@ GREPTIMEDB_IMAGE_TAG=${GREPTIMEDB_IMAGE_TAG:-latest}
 ETCD_CHART="oci://registry-1.docker.io/bitnamicharts/etcd"
 GREPTIME_CHART="https://greptimeteam.github.io/helm-charts/"
 
-# Ceate a cluster with 1 control-plane node and 5 workers.
+# Create a cluster with 1 control-plane node and 5 workers.
 function create_kind_cluster() {
   cat <<EOF | kind create cluster --name "${CLUSTER}" --image kindest/node:"$KUBERNETES_VERSION" --config=-
 kind: Cluster
@@ -68,7 +68,7 @@ function deploy_greptimedb_cluster() {
 
   helm install "$cluster_name" greptime/greptimedb-cluster \
     --set image.tag="$GREPTIMEDB_IMAGE_TAG" \
-    --set meta.etcdEndpoints="etcd.$install_namespace:2379" \
+    --set meta.backendStorage.etcd.endpoints="etcd.$install_namespace:2379" \
     -n "$install_namespace"
 
   # Wait for greptimedb cluster to be ready.
@@ -103,7 +103,7 @@ function deploy_greptimedb_cluster_with_s3_storage() {
 
   helm install "$cluster_name" greptime/greptimedb-cluster -n "$install_namespace" \
     --set image.tag="$GREPTIMEDB_IMAGE_TAG" \
-    --set meta.etcdEndpoints="etcd.$install_namespace:2379" \
+    --set meta.backendStorage.etcd.endpoints="etcd.$install_namespace:2379" \
     --set storage.s3.bucket="$AWS_CI_TEST_BUCKET" \
     --set storage.s3.region="$AWS_REGION" \
     --set storage.s3.root="$DATA_ROOT" \

.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.
+  sed -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/scripts/update-helm-charts-version.sh

@@ -0,0 +1,46 @@
+#!/bin/bash
+
+set -e
+
+VERSION=${VERSION}
+GITHUB_TOKEN=${GITHUB_TOKEN}
+
+update_helm_charts_version() {
+  # Configure Git configs.
+  git config --global user.email update-helm-charts-version@greptime.com
+  git config --global user.name update-helm-charts-version
+
+  # Clone helm-charts repository.
+  git clone "https://x-access-token:${GITHUB_TOKEN}@github.com/GreptimeTeam/helm-charts.git"
+  cd helm-charts
+
+  # Set default remote for gh CLI
+  gh repo set-default GreptimeTeam/helm-charts
+
+  # Checkout a new branch.
+  BRANCH_NAME="chore/greptimedb-${VERSION}"
+  git checkout -b $BRANCH_NAME
+
+  # Update version.
+  make update-version CHART=greptimedb-cluster VERSION=${VERSION}
+  make update-version CHART=greptimedb-standalone VERSION=${VERSION}
+
+  # Update docs.
+  make docs
+
+  # Commit the changes.
+  git add .
+  git commit -s -m "chore: Update GreptimeDB version to ${VERSION}"
+  git push origin $BRANCH_NAME
+
+  # Create a Pull Request.
+  gh pr create \
+    --title "chore: Update GreptimeDB version to ${VERSION}" \
+    --body "This PR updates the GreptimeDB version." \
+    --base main \
+    --head $BRANCH_NAME \
+    --reviewer zyy17 \
+    --reviewer daviderli614
+}
+
+update_helm_charts_version

.github/scripts/update-homebrew-greptme-version.sh

@@ -0,0 +1,42 @@
+#!/bin/bash
+
+set -e
+
+VERSION=${VERSION}
+GITHUB_TOKEN=${GITHUB_TOKEN}
+
+update_homebrew_greptime_version() {
+  # Configure Git configs.
+  git config --global user.email update-greptime-version@greptime.com
+  git config --global user.name update-greptime-version
+
+  # Clone helm-charts repository.
+  git clone "https://x-access-token:${GITHUB_TOKEN}@github.com/GreptimeTeam/homebrew-greptime.git"
+  cd homebrew-greptime
+
+  # Set default remote for gh CLI
+  gh repo set-default GreptimeTeam/homebrew-greptime
+
+  # Checkout a new branch.
+  BRANCH_NAME="chore/greptimedb-${VERSION}"
+  git checkout -b $BRANCH_NAME
+
+  # Update version.
+  make update-greptime-version VERSION=${VERSION}
+
+  # Commit the changes.
+  git add .
+  git commit -s -m "chore: Update GreptimeDB version to ${VERSION}"
+  git push origin $BRANCH_NAME
+
+  # Create a Pull Request.
+  gh pr create \
+    --title "chore: Update GreptimeDB version to ${VERSION}" \
+    --body "This PR updates the GreptimeDB version." \
+    --base main \
+    --head $BRANCH_NAME \
+    --reviewer zyy17 \
+    --reviewer daviderli614
+}
+
+update_homebrew_greptime_version

.github/scripts/upload-artifacts-to-s3.sh

@@ -41,7 +41,7 @@ function upload_artifacts() {
 # Updates the latest version information in AWS S3 if UPDATE_VERSION_INFO is true.
 function update_version_info() {
   if [ "$UPDATE_VERSION_INFO" == "true" ]; then
-    # If it's the officail release(like v1.0.0, v1.0.1, v1.0.2, etc.), update latest-version.txt.
+    # If it's the official release(like v1.0.0, v1.0.1, v1.0.2, etc.), update latest-version.txt.
     if [[ "$VERSION" =~ ^v[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
       echo "Updating latest-version.txt"
       echo "$VERSION" > latest-version.txt

.github/workflows/apidoc.yml

@@ -14,7 +14,7 @@ name: Build API docs
 
 jobs:
   apidoc:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v4
       with:

.github/workflows/dev-build.yml

@@ -16,11 +16,11 @@ on:
         description: The runner uses to build linux-amd64 artifacts
         default: ec2-c6i.4xlarge-amd64
         options:
-          - ubuntu-20.04
-          - ubuntu-20.04-8-cores
-          - ubuntu-20.04-16-cores
-          - ubuntu-20.04-32-cores
-          - ubuntu-20.04-64-cores
+          - ubuntu-22.04
+          - ubuntu-22.04-8-cores
+          - ubuntu-22.04-16-cores
+          - ubuntu-22.04-32-cores
+          - ubuntu-22.04-64-cores
           - ec2-c6i.xlarge-amd64 # 4C8G
           - ec2-c6i.2xlarge-amd64 # 8C16G
           - ec2-c6i.4xlarge-amd64 # 16C32G
@@ -55,6 +55,11 @@ on:
         description: Build and push images to DockerHub and ACR
         required: false
         default: true
+      upload_artifacts_to_s3:
+        type: boolean
+        description: Whether upload artifacts to s3
+        required: false
+        default: false
       cargo_profile:
         type: choice
         description: The cargo profile to use in building GreptimeDB.
@@ -83,7 +88,7 @@ jobs:
   allocate-runners:
     name: Allocate runners
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     outputs:
       linux-amd64-runner: ${{ steps.start-linux-amd64-runner.outputs.label }}
       linux-arm64-runner: ${{ steps.start-linux-arm64-runner.outputs.label }}
@@ -218,7 +223,7 @@ jobs:
       build-linux-amd64-artifacts,
       build-linux-arm64-artifacts,
     ]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     outputs:
       build-result: ${{ steps.set-build-result.outputs.build-result }}
     steps:
@@ -239,6 +244,13 @@ jobs:
           push-latest-tag: false # Don't push the latest tag to registry.
           dev-mode: true # Only build the standard images.
 
+      - name: Echo Docker image tag to step summary
+        run: |
+          echo "## Docker Image Tag" >> $GITHUB_STEP_SUMMARY
+          echo "Image Tag: \`${{ needs.allocate-runners.outputs.version }}\`" >> $GITHUB_STEP_SUMMARY
+          echo "Full Image Name: \`docker.io/${{ vars.IMAGE_NAMESPACE }}/${{ vars.DEV_BUILD_IMAGE_NAME }}:${{ needs.allocate-runners.outputs.version }}\`" >> $GITHUB_STEP_SUMMARY
+          echo "Pull Command: \`docker pull docker.io/${{ vars.IMAGE_NAMESPACE }}/${{ vars.DEV_BUILD_IMAGE_NAME }}:${{ needs.allocate-runners.outputs.version }}\`" >> $GITHUB_STEP_SUMMARY
+
       - name: Set build result
         id: set-build-result
         run: |
@@ -251,7 +263,7 @@ jobs:
       allocate-runners,
       release-images-to-dockerhub,
     ]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     continue-on-error: true
     steps:
       - uses: actions/checkout@v4
@@ -274,7 +286,7 @@ jobs:
           aws-cn-access-key-id: ${{ secrets.AWS_CN_ACCESS_KEY_ID }}
           aws-cn-secret-access-key: ${{ secrets.AWS_CN_SECRET_ACCESS_KEY }}
           aws-cn-region: ${{ vars.AWS_RELEASE_BUCKET_REGION }}
-          upload-to-s3: false
+          upload-to-s3: ${{ inputs.upload_artifacts_to_s3 }}
           dev-mode: true                     # Only build the standard images(exclude centos images).
           push-latest-tag: false             # Don't push the latest tag to registry.
           update-version-info: false         # Don't update the version info in S3.
@@ -283,7 +295,7 @@ jobs:
     name: Stop linux-amd64 runner
     # Only run this job when the runner is allocated.
     if: ${{ always() }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     needs: [
       allocate-runners,
       build-linux-amd64-artifacts,
@@ -309,7 +321,7 @@ jobs:
     name: Stop linux-arm64 runner
     # Only run this job when the runner is allocated.
     if: ${{ always() }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     needs: [
       allocate-runners,
       build-linux-arm64-artifacts,
@@ -337,7 +349,7 @@ jobs:
     needs: [
       release-images-to-dockerhub
     ]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     permissions:
       issues: write
 

.github/workflows/develop.yml

@@ -22,8 +22,9 @@ concurrency:
 
 jobs:
   check-typos-and-docs:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Check typos and docs
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
         with:
@@ -36,7 +37,8 @@ jobs:
           || (echo "'config/config.md' is not up-to-date, please run 'make config-docs'." && exit 1)
 
   license-header-check:
-    runs-on: ubuntu-20.04
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
+    runs-on: ubuntu-latest
     name: Check License Header
     steps:
       - uses: actions/checkout@v4
@@ -45,11 +47,12 @@ jobs:
       - uses: korandoru/hawkeye@v5
 
   check:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Check
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: [ ubuntu-20.04 ]
+        os: [ ubuntu-latest ]
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
@@ -71,8 +74,9 @@ jobs:
         run: cargo check --locked --workspace --all-targets
 
   toml:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Toml Check
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
@@ -85,11 +89,12 @@ jobs:
         run: taplo format --check
 
   build:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Build GreptimeDB binaries
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: [ ubuntu-20.04 ]
+        os: [ ubuntu-latest ]
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
@@ -127,6 +132,7 @@ jobs:
           version: current
 
   fuzztest:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Fuzz Test
     needs: build
     runs-on: ubuntu-latest
@@ -183,11 +189,13 @@ jobs:
           max-total-time: 120
 
   unstable-fuzztest:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Unstable Fuzz Test
     needs: build-greptime-ci
     runs-on: ubuntu-latest
     timeout-minutes: 60
     strategy:
+      fail-fast: false
       matrix:
         target: [ "unstable_fuzz_create_table_standalone" ]
     steps:
@@ -215,12 +223,12 @@ jobs:
         run: |
           sudo apt update && sudo apt install -y libfuzzer-14-dev
           cargo install cargo-fuzz cargo-gc-bin --force
-      - name: Download pre-built binariy
+      - name: Download pre-built binary
         uses: actions/download-artifact@v4
         with:
           name: bin
           path: .
-      - name: Unzip bianry
+      - name: Unzip binary
         run: |
           tar -xvf ./bin.tar.gz
           rm ./bin.tar.gz
@@ -242,13 +250,19 @@ jobs:
           name: unstable-fuzz-logs
           path: /tmp/unstable-greptime/
           retention-days: 3
+      - name: Describe pods
+        if: failure()
+        shell: bash
+        run: |
+          kubectl describe pod -n my-greptimedb
 
   build-greptime-ci:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Build GreptimeDB binary (profile-CI)
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: [ ubuntu-20.04 ]
+        os: [ ubuntu-latest ]
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
@@ -267,7 +281,7 @@ jobs:
       - name: Install cargo-gc-bin
         shell: bash
         run: cargo install cargo-gc-bin --force
-      - name: Build greptime bianry
+      - name: Build greptime binary
         shell: bash
         # `cargo gc` will invoke `cargo build` with specified args
         run: cargo gc --profile ci -- --bin greptime --features pg_kvbackend
@@ -285,11 +299,13 @@ jobs:
           version: current
 
   distributed-fuzztest:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Fuzz Test (Distributed, ${{ matrix.mode.name }}, ${{ matrix.target }})
     runs-on: ubuntu-latest
     needs:  build-greptime-ci
     timeout-minutes: 60
     strategy:
+      fail-fast: false
       matrix:
         target: [ "fuzz_create_table", "fuzz_alter_table", "fuzz_create_database", "fuzz_create_logical_table", "fuzz_alter_logical_table", "fuzz_insert", "fuzz_insert_logical_table" ]
         mode:
@@ -319,9 +335,9 @@ jobs:
         name: Setup Minio
         uses: ./.github/actions/setup-minio
       - if: matrix.mode.kafka
-        name: Setup Kafka cluser
+        name: Setup Kafka cluster
         uses: ./.github/actions/setup-kafka-cluster
-      - name: Setup Etcd cluser
+      - name: Setup Etcd cluster
         uses: ./.github/actions/setup-etcd-cluster
       # Prepares for fuzz tests
       - uses: arduino/setup-protoc@v3
@@ -394,6 +410,11 @@ jobs:
         shell: bash
         run: |
           kubectl describe nodes
+      - name: Describe pod
+        if: failure()
+        shell: bash
+        run: |
+          kubectl describe pod -n my-greptimedb
       - name: Export kind logs
         if: failure()
         shell: bash
@@ -416,11 +437,13 @@ jobs:
           docker system prune -f
 
   distributed-fuzztest-with-chaos:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Fuzz Test with Chaos (Distributed, ${{ matrix.mode.name }}, ${{ matrix.target }})
     runs-on: ubuntu-latest
     needs:  build-greptime-ci
     timeout-minutes: 60
     strategy:
+      fail-fast: false
       matrix:
         target: ["fuzz_migrate_mito_regions", "fuzz_migrate_metric_regions", "fuzz_failover_mito_regions", "fuzz_failover_metric_regions"]
         mode:
@@ -465,9 +488,9 @@ jobs:
         name: Setup Minio
         uses: ./.github/actions/setup-minio
       - if: matrix.mode.kafka
-        name: Setup Kafka cluser
+        name: Setup Kafka cluster
         uses: ./.github/actions/setup-kafka-cluster
-      - name: Setup Etcd cluser
+      - name: Setup Etcd cluster
         uses: ./.github/actions/setup-etcd-cluster
       # Prepares for fuzz tests
       - uses: arduino/setup-protoc@v3
@@ -541,6 +564,11 @@ jobs:
         shell: bash
         run: |
           kubectl describe nodes
+      - name: Describe pods
+        if: failure()
+        shell: bash
+        run: |
+          kubectl describe pod -n my-greptimedb
       - name: Export kind logs
         if: failure()
         shell: bash
@@ -563,12 +591,14 @@ jobs:
           docker system prune -f
 
   sqlness:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Sqlness Test (${{ matrix.mode.name }})
     needs: build
     runs-on: ${{ matrix.os }}
     strategy:
+      fail-fast: false
       matrix:
-        os: [ ubuntu-20.04 ]
+        os: [ ubuntu-latest ]
         mode:
           - name: "Basic"
             opts: ""
@@ -576,7 +606,7 @@ jobs:
           - name: "Remote WAL"
             opts: "-w kafka -k 127.0.0.1:9092"
             kafka: true
-          - name: "Pg Kvbackend"
+          - name: "PostgreSQL KvBackend"
             opts: "--setup-pg"
             kafka: false
     timeout-minutes: 60
@@ -606,8 +636,9 @@ jobs:
           retention-days: 3
 
   fmt:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Rustfmt
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
@@ -623,8 +654,9 @@ jobs:
         run: make fmt-check
 
   clippy:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Clippy
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
@@ -648,6 +680,7 @@ jobs:
         run: make clippy
 
   conflict-check:
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     name: Check for conflict
     runs-on: ubuntu-latest
     steps:
@@ -658,7 +691,7 @@ jobs:
         uses: olivernybroe/action-conflict-finder@v4.0
 
   test:
-    if: github.event_name != 'merge_group'
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' && github.event_name != 'merge_group' }}
     runs-on: ubuntu-22.04-arm
     timeout-minutes: 60
     needs:  [conflict-check, clippy, fmt]
@@ -673,7 +706,7 @@ jobs:
       - name: Install toolchain
         uses: actions-rust-lang/setup-rust-toolchain@v1
         with:
-            cache: false
+          cache: false
       - name: Rust Cache
         uses: Swatinem/rust-cache@v2
         with:
@@ -704,13 +737,14 @@ jobs:
           GT_MINIO_ENDPOINT_URL: http://127.0.0.1:9000
           GT_ETCD_ENDPOINTS: http://127.0.0.1:2379
           GT_POSTGRES_ENDPOINTS: postgres://greptimedb:admin@127.0.0.1:5432/postgres
+          GT_MYSQL_ENDPOINTS: mysql://greptimedb:admin@127.0.0.1:3306/mysql
           GT_KAFKA_ENDPOINTS: 127.0.0.1:9092
           GT_KAFKA_SASL_ENDPOINTS: 127.0.0.1:9093
           UNITTEST_LOG_DIR: "__unittest_logs"
 
   coverage:
-    if: github.event_name == 'merge_group'
-    runs-on: ubuntu-20.04-8-cores
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' && github.event_name == 'merge_group' }}
+    runs-on: ubuntu-22.04-8-cores
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
@@ -755,6 +789,7 @@ jobs:
           GT_MINIO_ENDPOINT_URL: http://127.0.0.1:9000
           GT_ETCD_ENDPOINTS: http://127.0.0.1:2379
           GT_POSTGRES_ENDPOINTS: postgres://greptimedb:admin@127.0.0.1:5432/postgres
+          GT_MYSQL_ENDPOINTS: mysql://greptimedb:admin@127.0.0.1:3306/mysql
           GT_KAFKA_ENDPOINTS: 127.0.0.1:9092
           GT_KAFKA_SASL_ENDPOINTS: 127.0.0.1:9093
           UNITTEST_LOG_DIR: "__unittest_logs"
@@ -768,9 +803,10 @@ jobs:
           verbose: true
 
   # compat:
+  #   if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
   #   name: Compatibility Test
   #   needs: build
-  #   runs-on: ubuntu-20.04
+  #   runs-on: ubuntu-22.04
   #   timeout-minutes: 60
   #   steps:
   #     - uses: actions/checkout@v4

.github/workflows/docbot.yml

@@ -9,7 +9,7 @@ concurrency:
 
 jobs:
   docbot:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     permissions:
       pull-requests: write
       contents: read

.github/workflows/docs.yml

@@ -31,7 +31,7 @@ name: CI
 jobs:
   typos:
     name: Spell Check with Typos
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
         with:
@@ -39,7 +39,7 @@ jobs:
       - uses: crate-ci/typos@master
 
   license-header-check:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     name: Check License Header
     steps:
       - uses: actions/checkout@v4
@@ -49,29 +49,29 @@ jobs:
 
   check:
     name: Check
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
       - run: 'echo "No action required"'
 
   fmt:
     name: Rustfmt
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
       - run: 'echo "No action required"'
 
   clippy:
     name: Clippy
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
       - run: 'echo "No action required"'
 
   coverage:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
       - run: 'echo "No action required"'
 
   test:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
       - run: 'echo "No action required"'
 
@@ -80,7 +80,7 @@ jobs:
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: [ ubuntu-20.04 ]
+        os: [ ubuntu-latest ]
         mode:
           - name: "Basic"
           - name: "Remote WAL"

.github/workflows/nightly-build.yml

@@ -14,11 +14,11 @@ on:
         description: The runner uses to build linux-amd64 artifacts
         default: ec2-c6i.4xlarge-amd64
         options:
-          - ubuntu-20.04
-          - ubuntu-20.04-8-cores
-          - ubuntu-20.04-16-cores
-          - ubuntu-20.04-32-cores
-          - ubuntu-20.04-64-cores
+          - ubuntu-22.04
+          - ubuntu-22.04-8-cores
+          - ubuntu-22.04-16-cores
+          - ubuntu-22.04-32-cores
+          - ubuntu-22.04-64-cores
           - ec2-c6i.xlarge-amd64 # 4C8G
           - ec2-c6i.2xlarge-amd64 # 8C16G
           - ec2-c6i.4xlarge-amd64 # 16C32G
@@ -70,7 +70,7 @@ jobs:
   allocate-runners:
     name: Allocate runners
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     outputs:
       linux-amd64-runner: ${{ steps.start-linux-amd64-runner.outputs.label }}
       linux-arm64-runner: ${{ steps.start-linux-arm64-runner.outputs.label }}
@@ -182,7 +182,7 @@ jobs:
       build-linux-amd64-artifacts,
       build-linux-arm64-artifacts,
     ]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     outputs:
       nightly-build-result: ${{ steps.set-nightly-build-result.outputs.nightly-build-result }}
     steps:
@@ -214,7 +214,7 @@ jobs:
       allocate-runners,
       release-images-to-dockerhub,
     ]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     # When we push to ACR, it's easy to fail due to some unknown network issues.
     # However, we don't want to fail the whole workflow because of this.
     # The ACR have daily sync with DockerHub, so don't worry about the image not being updated.
@@ -249,7 +249,7 @@ jobs:
     name: Stop linux-amd64 runner
     # Only run this job when the runner is allocated.
     if: ${{ always() }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     needs: [
       allocate-runners,
       build-linux-amd64-artifacts,
@@ -275,7 +275,7 @@ jobs:
     name: Stop linux-arm64 runner
     # Only run this job when the runner is allocated.
     if: ${{ always() }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     needs: [
       allocate-runners,
       build-linux-arm64-artifacts,
@@ -303,7 +303,7 @@ jobs:
     needs: [
       release-images-to-dockerhub
     ]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     permissions:
       issues: write
     env:

.github/workflows/nightly-ci.yml

@@ -13,7 +13,7 @@ jobs:
   sqlness-test:
     name: Run sqlness test
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
-    runs-on: ubuntu-22.04
+    runs-on: ubuntu-latest
     steps:
       - name: Checkout
         uses: actions/checkout@v4
@@ -107,7 +107,6 @@ jobs:
           CARGO_BUILD_RUSTFLAGS: "-C linker=lld-link"
           RUST_BACKTRACE: 1
           CARGO_INCREMENTAL: 0
-          RUSTUP_WINDOWS_PATH_ADD_BIN: 1 # Workaround for https://github.com/nextest-rs/nextest/issues/1493
           GT_S3_BUCKET: ${{ vars.AWS_CI_TEST_BUCKET }}
           GT_S3_ACCESS_KEY_ID: ${{ secrets.AWS_CI_TEST_ACCESS_KEY_ID }}
           GT_S3_ACCESS_KEY: ${{ secrets.AWS_CI_TEST_SECRET_ACCESS_KEY }}
@@ -118,22 +117,22 @@ jobs:
     name: Run clean build on Linux
     runs-on: ubuntu-latest
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
-    timeout-minutes: 60
+    timeout-minutes: 45
     steps:
       - uses: actions/checkout@v4
         with:
           fetch-depth: 0
           persist-credentials: false
-      - uses: cachix/install-nix-action@v27
-        with:
-          nix_path: nixpkgs=channel:nixos-24.11
-      - run: nix develop --command cargo build
+      - uses: cachix/install-nix-action@v31
+      - run: nix develop --command cargo check --bin greptime
+        env:
+          CARGO_BUILD_RUSTFLAGS: "-C link-arg=-fuse-ld=mold"
 
   check-status:
     name: Check status
     needs: [sqlness-test, sqlness-windows, test-on-windows]
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     outputs:
       check-result: ${{ steps.set-check-result.outputs.check-result }}
     steps:
@@ -146,7 +145,7 @@ jobs:
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' && always() }} # Not requiring successful dependent jobs, always run.
     name: Send notification to Greptime team
     needs: [check-status]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     env:
       SLACK_WEBHOOK_URL: ${{ secrets.SLACK_WEBHOOK_URL_DEVELOP_CHANNEL }}
     steps:

.github/workflows/pr-labeling.yaml

@@ -0,0 +1,42 @@
+name: 'PR Labeling'
+
+on:
+  pull_request_target:
+    types:
+      - opened
+      - synchronize
+      - reopened
+
+permissions:
+  contents: read
+  pull-requests: write
+  issues: write
+
+jobs:
+  labeler:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout sources
+        uses: actions/checkout@v4
+
+      - uses: actions/labeler@v5
+        with:
+          configuration-path: ".github/labeler.yaml"
+          repo-token: "${{ secrets.GITHUB_TOKEN }}"
+
+  size-label:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: pascalgn/size-label-action@v0.5.5
+        env:
+          GITHUB_TOKEN: "${{ secrets.GITHUB_TOKEN }}"
+        with:
+          sizes: >
+            {
+              "0": "XS",
+              "100": "S",
+              "300": "M",
+              "1000": "L",
+              "1500": "XL",
+              "2000": "XXL"
+            }

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

@@ -24,12 +24,20 @@ 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.
-    runs-on: ubuntu-20.04-16-cores
+    # 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 }}
     steps:
@@ -57,13 +65,13 @@ 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
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     needs: [
       release-dev-builder-images
     ]
@@ -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 }}
@@ -148,7 +156,7 @@ jobs:
 
   release-dev-builder-images-cn: # Note: Be careful issue: https://github.com/containers/skopeo/issues/1874 and we decide to use the latest stable skopeo container.
     name: Release dev builder images to CN region
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     needs: [
       release-dev-builder-images
     ]
@@ -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 }}

.github/workflows/release.yml

@@ -18,11 +18,11 @@ on:
         description: The runner uses to build linux-amd64 artifacts
         default: ec2-c6i.4xlarge-amd64
         options:
-          - ubuntu-20.04
-          - ubuntu-20.04-8-cores
-          - ubuntu-20.04-16-cores
-          - ubuntu-20.04-32-cores
-          - ubuntu-20.04-64-cores
+          - ubuntu-22.04
+          - ubuntu-22.04-8-cores
+          - ubuntu-22.04-16-cores
+          - ubuntu-22.04-32-cores
+          - ubuntu-22.04-64-cores
           - ec2-c6i.xlarge-amd64 # 4C8G
           - ec2-c6i.2xlarge-amd64 # 8C16G
           - ec2-c6i.4xlarge-amd64 # 16C32G
@@ -88,16 +88,14 @@ env:
   # Controls whether to run tests, include unit-test, integration-test and sqlness.
   DISABLE_RUN_TESTS: ${{ inputs.skip_test || vars.DEFAULT_SKIP_TEST }}
 
-  # The scheduled version is '${{ env.NEXT_RELEASE_VERSION }}-nightly-YYYYMMDD', like v0.2.0-nigthly-20230313;
+  # The scheduled version is '${{ env.NEXT_RELEASE_VERSION }}-nightly-YYYYMMDD', like v0.2.0-nightly-20230313;
   NIGHTLY_RELEASE_PREFIX: nightly
-  # Note: The NEXT_RELEASE_VERSION should be modified manually by every formal release.
-  NEXT_RELEASE_VERSION: v0.12.0
 
 jobs:
   allocate-runners:
     name: Allocate runners
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     outputs:
       linux-amd64-runner: ${{ steps.start-linux-amd64-runner.outputs.label }}
       linux-arm64-runner: ${{ steps.start-linux-arm64-runner.outputs.label }}
@@ -112,6 +110,8 @@ jobs:
 
       # The 'version' use as the global tag name of the release workflow.
       version: ${{ steps.create-version.outputs.version }}
+
+      should-push-latest-tag: ${{ steps.check-version.outputs.should-push-latest-tag }}
     steps:
       - name: Checkout
         uses: actions/checkout@v4
@@ -126,7 +126,7 @@ jobs:
 
       # The create-version will create a global variable named 'version' in the global workflows.
       # - If it's a tag push release, the version is the tag name(${{ github.ref_name }});
-      # - If it's a scheduled release, the version is '${{ env.NEXT_RELEASE_VERSION }}-nightly-$buildTime', like v0.2.0-nigthly-20230313;
+      # - If it's a scheduled release, the version is '${{ env.NEXT_RELEASE_VERSION }}-nightly-$buildTime', like v0.2.0-nightly-20230313;
       # - If it's a manual release, the version is '${{ env.NEXT_RELEASE_VERSION }}-<short-git-sha>-YYYYMMDDSS', like v0.2.0-e5b243c-2023071245;
       - name: Create version
         id: create-version
@@ -135,9 +135,13 @@ jobs:
         env:
           GITHUB_EVENT_NAME: ${{ github.event_name }}
           GITHUB_REF_NAME: ${{ github.ref_name }}
-          NEXT_RELEASE_VERSION: ${{ env.NEXT_RELEASE_VERSION }}
           NIGHTLY_RELEASE_PREFIX: ${{ env.NIGHTLY_RELEASE_PREFIX }}
 
+      - name: Check version
+        id: check-version
+        run: |
+          ./.github/scripts/check-version.sh "${{ steps.create-version.outputs.version }}"
+
       - name: Allocate linux-amd64 runner
         if: ${{ inputs.build_linux_amd64_artifacts || github.event_name == 'push' || github.event_name == 'schedule' }}
         uses: ./.github/actions/start-runner
@@ -299,7 +303,7 @@ jobs:
       build-linux-amd64-artifacts,
       build-linux-arm64-artifacts,
     ]
-    runs-on: ubuntu-2004-16-cores
+    runs-on: ubuntu-latest
     outputs:
       build-image-result: ${{ steps.set-build-image-result.outputs.build-image-result }}
     steps:
@@ -317,7 +321,7 @@ jobs:
           image-registry-username: ${{ secrets.DOCKERHUB_USERNAME }}
           image-registry-password: ${{ secrets.DOCKERHUB_TOKEN }}
           version: ${{ needs.allocate-runners.outputs.version }}
-          push-latest-tag: true
+          push-latest-tag: ${{ needs.allocate-runners.outputs.should-push-latest-tag == 'true' && github.ref_type == 'tag' && !contains(github.ref_name, 'nightly') && github.event_name != 'schedule' }}
 
       - name: Set build image result
         id: set-build-image-result
@@ -335,7 +339,7 @@ jobs:
       build-windows-artifacts,
       release-images-to-dockerhub,
     ]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest-16-cores
     # When we push to ACR, it's easy to fail due to some unknown network issues.
     # However, we don't want to fail the whole workflow because of this.
     # The ACR have daily sync with DockerHub, so don't worry about the image not being updated.
@@ -364,7 +368,7 @@ jobs:
           dev-mode: false
           upload-to-s3: true
           update-version-info: true
-          push-latest-tag: true
+          push-latest-tag: ${{ needs.allocate-runners.outputs.should-push-latest-tag == 'true' && github.ref_type == 'tag' && !contains(github.ref_name, 'nightly') && github.event_name != 'schedule' }}
 
   publish-github-release:
     name: Create GitHub release and upload artifacts
@@ -377,7 +381,7 @@ jobs:
       build-windows-artifacts,
       release-images-to-dockerhub,
     ]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
         with:
@@ -391,12 +395,12 @@ jobs:
 
   ### Stop runners ###
   # It's very necessary to split the job of releasing runners into 'stop-linux-amd64-runner' and 'stop-linux-arm64-runner'.
-  # Because we can terminate the specified EC2 instance immediately after the job is finished without uncessary waiting.
+  # Because we can terminate the specified EC2 instance immediately after the job is finished without unnecessary waiting.
   stop-linux-amd64-runner: # It's always run as the last job in the workflow to make sure that the runner is released.
     name: Stop linux-amd64 runner
     # Only run this job when the runner is allocated.
     if: ${{ always() }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     needs: [
       allocate-runners,
       build-linux-amd64-artifacts,
@@ -422,7 +426,7 @@ jobs:
     name: Stop linux-arm64 runner
     # Only run this job when the runner is allocated.
     if: ${{ always() }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     needs: [
       allocate-runners,
       build-linux-arm64-artifacts,
@@ -444,11 +448,11 @@ jobs:
           aws-region: ${{ vars.EC2_RUNNER_REGION }}
           github-token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
 
-  bump-doc-version:
-    name: Bump doc version
+  bump-downstream-repo-versions:
+    name: Bump downstream repo versions
     if: ${{ github.event_name == 'push' || github.event_name == 'schedule' }}
-    needs: [allocate-runners]
-    runs-on: ubuntu-20.04
+    needs: [allocate-runners, publish-github-release]
+    runs-on: ubuntu-latest
     # Permission reference: https://docs.github.com/en/actions/using-jobs/assigning-permissions-to-jobs
     permissions:
       issues: write # Allows the action to create issues for cyborg.
@@ -459,13 +463,58 @@ jobs:
           fetch-depth: 0
           persist-credentials: false
       - uses: ./.github/actions/setup-cyborg
-      - name: Bump doc version
+      - name: Bump downstream repo versions
         working-directory: cyborg
-        run: pnpm tsx bin/bump-doc-version.ts
+        run: pnpm tsx bin/bump-versions.ts
         env:
+          TARGET_REPOS: website,docs,demo
           VERSION: ${{ needs.allocate-runners.outputs.version }}
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          WEBSITE_REPO_TOKEN: ${{ secrets.WEBSITE_REPO_TOKEN }}
           DOCS_REPO_TOKEN: ${{ secrets.DOCS_REPO_TOKEN }}
+          DEMO_REPO_TOKEN: ${{ secrets.DEMO_REPO_TOKEN }}
+
+  bump-helm-charts-version:
+    name: Bump helm charts version
+    if: ${{ github.ref_type == 'tag' && !contains(github.ref_name, 'nightly') && github.event_name != 'schedule' }}
+    needs: [allocate-runners, publish-github-release]
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+      pull-requests: write
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Bump helm charts version
+        env:
+          GITHUB_TOKEN: ${{ secrets.HELM_CHARTS_REPO_TOKEN }}
+          VERSION: ${{ needs.allocate-runners.outputs.version }}
+        run: |
+          ./.github/scripts/update-helm-charts-version.sh
+
+  bump-homebrew-greptime-version:
+    name: Bump homebrew greptime version
+    if: ${{ github.ref_type == 'tag' && !contains(github.ref_name, 'nightly') && github.event_name != 'schedule' }}
+    needs: [allocate-runners, publish-github-release]
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+      pull-requests: write
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Bump homebrew greptime version
+        env:
+          GITHUB_TOKEN: ${{ secrets.HOMEBREW_GREPTIME_REPO_TOKEN }}
+          VERSION: ${{ needs.allocate-runners.outputs.version }}
+        run: |
+          ./.github/scripts/update-homebrew-greptme-version.sh
 
   notification:
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' && (github.event_name == 'push' || github.event_name == 'schedule') && always() }}
@@ -475,7 +524,7 @@ jobs:
       build-macos-artifacts,
       build-windows-artifacts,
     ]
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     # Permission reference: https://docs.github.com/en/actions/using-jobs/assigning-permissions-to-jobs
     permissions:
       issues: write # Allows the action to create issues for cyborg.

.github/workflows/semantic-pull-request.yml

@@ -11,14 +11,17 @@ concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
   cancel-in-progress: true
 
+permissions:
+  issues: write
+  contents: write
+  pull-requests: write
+
 jobs:
   check:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     timeout-minutes: 10
     steps:
       - uses: actions/checkout@v4
-        with:
-          persist-credentials: false
       - uses: ./.github/actions/setup-cyborg
       - name: Check Pull Request
         working-directory: cyborg

.gitignore

@@ -54,3 +54,10 @@ tests-fuzz/corpus/
 # Nix
 .direnv
 .envrc
+
+## default data home
+greptimedb_data
+
+# github
+!/.github
+

Cargo.lock

@@ -185,7 +185,7 @@ checksum = "d301b3b94cb4b2f23d7917810addbbaff90738e0ca2be692bd027e70d7e0330c"
 
 [[package]]
 name = "api"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-base",
  "common-decimal",
@@ -710,7 +710,7 @@ dependencies = [
 
 [[package]]
 name = "auth"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -1324,7 +1324,7 @@ dependencies = [
 
 [[package]]
 name = "cache"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "catalog",
  "common-error",
@@ -1348,7 +1348,7 @@ checksum = "37b2a672a2cb129a2e41c10b1224bb368f9f37a2b16b612598138befd7b37eb5"
 
 [[package]]
 name = "catalog"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arrow",
@@ -1661,7 +1661,7 @@ checksum = "1462739cb27611015575c0c11df5df7601141071f07518d56fcc1be504cbec97"
 
 [[package]]
 name = "cli"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "auth",
@@ -1703,7 +1703,7 @@ dependencies = [
  "session",
  "snafu 0.8.5",
  "store-api",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "table",
  "tempfile",
  "tokio",
@@ -1712,7 +1712,7 @@ dependencies = [
 
 [[package]]
 name = "client"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arc-swap",
@@ -1739,7 +1739,7 @@ dependencies = [
  "rand",
  "serde_json",
  "snafu 0.8.5",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "substrait 0.37.3",
  "tokio",
  "tokio-stream",
@@ -1780,7 +1780,7 @@ dependencies = [
 
 [[package]]
 name = "cmd"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "auth",
@@ -1791,6 +1791,7 @@ dependencies = [
  "clap 4.5.19",
  "cli",
  "client",
+ "colored",
  "common-base",
  "common-catalog",
  "common-config",
@@ -1825,7 +1826,10 @@ dependencies = [
  "mito2",
  "moka",
  "nu-ansi-term",
+ "object-store",
+ "parquet",
  "plugins",
+ "pprof",
  "prometheus",
  "prost 0.13.3",
  "query",
@@ -1841,7 +1845,7 @@ dependencies = [
  "similar-asserts",
  "snafu 0.8.5",
  "store-api",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "table",
  "temp-env",
  "tempfile",
@@ -1858,6 +1862,16 @@ version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d3fd119d74b830634cea2a0f58bbd0d54540518a14397557951e79340abc28c0"
 
+[[package]]
+name = "colored"
+version = "2.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "117725a109d387c937a1533ce01b450cbde6b88abceea8473c4d7a85853cda3c"
+dependencies = [
+ "lazy_static",
+ "windows-sys 0.59.0",
+]
+
 [[package]]
 name = "combine"
 version = "4.6.7"
@@ -1887,7 +1901,7 @@ checksum = "55b672471b4e9f9e95499ea597ff64941a309b2cdbffcc46f2cc5e2d971fd335"
 
 [[package]]
 name = "common-base"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "anymap2",
  "async-trait",
@@ -1909,11 +1923,11 @@ dependencies = [
 
 [[package]]
 name = "common-catalog"
-version = "0.12.0"
+version = "0.12.2"
 
 [[package]]
 name = "common-config"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-base",
  "common-error",
@@ -1938,7 +1952,7 @@ dependencies = [
 
 [[package]]
 name = "common-datasource"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arrow",
  "arrow-schema",
@@ -1974,7 +1988,7 @@ dependencies = [
 
 [[package]]
 name = "common-decimal"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "bigdecimal 0.4.5",
  "common-error",
@@ -1987,7 +2001,7 @@ dependencies = [
 
 [[package]]
 name = "common-error"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "http 1.1.0",
  "snafu 0.8.5",
@@ -1997,7 +2011,7 @@ dependencies = [
 
 [[package]]
 name = "common-frontend"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "common-error",
@@ -2007,7 +2021,7 @@ dependencies = [
 
 [[package]]
 name = "common-function"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "api",
@@ -2055,7 +2069,7 @@ dependencies = [
 
 [[package]]
 name = "common-greptimedb-telemetry"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "common-runtime",
@@ -2072,7 +2086,7 @@ dependencies = [
 
 [[package]]
 name = "common-grpc"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arrow-flight",
@@ -2100,7 +2114,7 @@ dependencies = [
 
 [[package]]
 name = "common-grpc-expr"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "common-base",
@@ -2119,7 +2133,7 @@ dependencies = [
 
 [[package]]
 name = "common-macro"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arc-swap",
  "common-query",
@@ -2133,7 +2147,7 @@ dependencies = [
 
 [[package]]
 name = "common-mem-prof"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-error",
  "common-macro",
@@ -2146,7 +2160,7 @@ dependencies = [
 
 [[package]]
 name = "common-meta"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "anymap2",
  "api",
@@ -2206,7 +2220,7 @@ dependencies = [
 
 [[package]]
 name = "common-options"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-grpc",
  "humantime-serde",
@@ -2215,11 +2229,11 @@ dependencies = [
 
 [[package]]
 name = "common-plugins"
-version = "0.12.0"
+version = "0.12.2"
 
 [[package]]
 name = "common-pprof"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-error",
  "common-macro",
@@ -2231,7 +2245,7 @@ dependencies = [
 
 [[package]]
 name = "common-procedure"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-stream",
  "async-trait",
@@ -2258,7 +2272,7 @@ dependencies = [
 
 [[package]]
 name = "common-procedure-test"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "common-procedure",
@@ -2266,7 +2280,7 @@ dependencies = [
 
 [[package]]
 name = "common-query"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -2292,7 +2306,7 @@ dependencies = [
 
 [[package]]
 name = "common-recordbatch"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arc-swap",
  "common-error",
@@ -2311,7 +2325,7 @@ dependencies = [
 
 [[package]]
 name = "common-runtime"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "clap 4.5.19",
@@ -2341,7 +2355,7 @@ dependencies = [
 
 [[package]]
 name = "common-telemetry"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "atty",
  "backtrace",
@@ -2369,7 +2383,7 @@ dependencies = [
 
 [[package]]
 name = "common-test-util"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "client",
  "common-query",
@@ -2381,7 +2395,7 @@ dependencies = [
 
 [[package]]
 name = "common-time"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arrow",
  "chrono",
@@ -2399,7 +2413,7 @@ dependencies = [
 
 [[package]]
 name = "common-version"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "build-data",
  "const_format",
@@ -2409,7 +2423,7 @@ dependencies = [
 
 [[package]]
 name = "common-wal"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-base",
  "common-error",
@@ -3340,7 +3354,7 @@ dependencies = [
 
 [[package]]
 name = "datanode"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arrow-flight",
@@ -3392,7 +3406,7 @@ dependencies = [
  "session",
  "snafu 0.8.5",
  "store-api",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "table",
  "tokio",
  "toml 0.8.19",
@@ -3401,7 +3415,7 @@ dependencies = [
 
 [[package]]
 name = "datatypes"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -4045,7 +4059,7 @@ dependencies = [
 
 [[package]]
 name = "file-engine"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -4155,7 +4169,7 @@ checksum = "8bf7cc16383c4b8d58b9905a8509f02926ce3058053c056376248d958c9df1e8"
 
 [[package]]
 name = "flow"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arrow",
@@ -4216,7 +4230,7 @@ dependencies = [
  "snafu 0.8.5",
  "store-api",
  "strum 0.25.0",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "table",
  "tokio",
  "tonic 0.12.3",
@@ -4271,7 +4285,7 @@ checksum = "6c2141d6d6c8512188a7891b4b01590a45f6dac67afb4f255c4124dbb86d4eaa"
 
 [[package]]
 name = "frontend"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arc-swap",
@@ -5539,7 +5553,7 @@ dependencies = [
 
 [[package]]
 name = "index"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "asynchronous-codec",
@@ -6331,7 +6345,7 @@ checksum = "a7a70ba024b9dc04c27ea2f0c0548feb474ec5c54bba33a7f72f873a39d07b24"
 
 [[package]]
 name = "log-query"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "chrono",
  "common-error",
@@ -6343,7 +6357,7 @@ dependencies = [
 
 [[package]]
 name = "log-store"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-stream",
  "async-trait",
@@ -6636,7 +6650,7 @@ dependencies = [
 
 [[package]]
 name = "meta-client"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -6663,7 +6677,7 @@ dependencies = [
 
 [[package]]
 name = "meta-srv"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -6749,7 +6763,7 @@ dependencies = [
 
 [[package]]
 name = "metric-engine"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "aquamarine",
@@ -6847,7 +6861,7 @@ dependencies = [
 
 [[package]]
 name = "mito2"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "aquamarine",
@@ -7544,7 +7558,7 @@ dependencies = [
 
 [[package]]
 name = "object-store"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "anyhow",
  "bytes",
@@ -7793,7 +7807,7 @@ dependencies = [
 
 [[package]]
 name = "operator"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "api",
@@ -7841,7 +7855,7 @@ dependencies = [
  "sql",
  "sqlparser 0.52.0 (git+https://github.com/GreptimeTeam/sqlparser-rs.git?rev=71dd86058d2af97b9925093d40c4e03360403170)",
  "store-api",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "table",
  "tokio",
  "tokio-util",
@@ -8078,7 +8092,7 @@ dependencies = [
 
 [[package]]
 name = "partition"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -8346,7 +8360,7 @@ checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"
 
 [[package]]
 name = "pipeline"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "api",
@@ -8486,7 +8500,7 @@ dependencies = [
 
 [[package]]
 name = "plugins"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "auth",
  "clap 4.5.19",
@@ -8748,7 +8762,7 @@ dependencies = [
 
 [[package]]
 name = "promql"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "async-trait",
@@ -8993,7 +9007,7 @@ dependencies = [
 
 [[package]]
 name = "puffin"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-compression 0.4.13",
  "async-trait",
@@ -9034,7 +9048,7 @@ dependencies = [
 
 [[package]]
 name = "query"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "api",
@@ -9099,7 +9113,7 @@ dependencies = [
  "sqlparser 0.52.0 (git+https://github.com/GreptimeTeam/sqlparser-rs.git?rev=71dd86058d2af97b9925093d40c4e03360403170)",
  "statrs",
  "store-api",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "table",
  "tokio",
  "tokio-stream",
@@ -10444,7 +10458,7 @@ dependencies = [
 
 [[package]]
 name = "servers"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "api",
@@ -10561,7 +10575,7 @@ dependencies = [
 
 [[package]]
 name = "session"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arc-swap",
@@ -10870,7 +10884,7 @@ dependencies = [
 
 [[package]]
 name = "sql"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "chrono",
@@ -10924,7 +10938,7 @@ dependencies = [
 
 [[package]]
 name = "sqlness-runner"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "clap 4.5.19",
@@ -11241,7 +11255,7 @@ dependencies = [
 
 [[package]]
 name = "store-api"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "aquamarine",
@@ -11371,7 +11385,7 @@ dependencies = [
 
 [[package]]
 name = "substrait"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "bytes",
@@ -11552,7 +11566,7 @@ dependencies = [
 
 [[package]]
 name = "table"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -11803,7 +11817,7 @@ checksum = "3369f5ac52d5eb6ab48c6b4ffdc8efbcad6b89c765749064ba298f2c68a16a76"
 
 [[package]]
 name = "tests-fuzz"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arbitrary",
  "async-trait",
@@ -11847,7 +11861,7 @@ dependencies = [
 
 [[package]]
 name = "tests-integration"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arrow-flight",
@@ -11913,7 +11927,7 @@ dependencies = [
  "sql",
  "sqlx",
  "store-api",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "table",
  "tempfile",
  "time",

Cargo.toml

@@ -67,7 +67,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "0.12.0"
+version = "0.12.2"
 edition = "2021"
 license = "Apache-2.0"
 

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). |
+| `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. |
 | `runtime.global_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |

config/datanode.example.toml

@@ -231,6 +231,7 @@ overwrite_entry_start_id = false
 # secret_access_key = "123456"
 # endpoint = "https://s3.amazonaws.com"
 # region = "us-west-2"
+# enable_virtual_host_style = false
 
 # Example of using Oss as the storage.
 # [storage]

config/metasrv.example.toml

@@ -50,6 +50,9 @@ use_memory_store = false
 ## - Using shared storage (e.g., s3).
 enable_region_failover = false
 
+## Max allowed idle time before removing node info from metasrv memory.
+node_max_idle_time = "24hours"
+
 ## Whether to enable greptimedb telemetry. Enabled by default.
 #+ enable_telemetry = true
 

config/standalone.example.toml

@@ -318,6 +318,7 @@ retry_delay = "500ms"
 # secret_access_key = "123456"
 # endpoint = "https://s3.amazonaws.com"
 # region = "us-west-2"
+# enable_virtual_host_style = false
 
 # Example of using Oss as the storage.
 # [storage]

cyborg/bin/bump-versions.ts

@@ -0,0 +1,156 @@
+/*
+ * 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.
+ */
+
+import * as core from "@actions/core";
+import {obtainClient} from "@/common";
+
+interface RepoConfig {
+  tokenEnv: string;
+  repo: string;
+  workflowLogic: (version: string) => [string, string] | null;
+}
+
+const REPO_CONFIGS: Record<string, RepoConfig> = {
+  website: {
+    tokenEnv: "WEBSITE_REPO_TOKEN",
+    repo: "website",
+    workflowLogic: (version: string) => {
+      // Skip nightly versions for website
+      if (version.includes('nightly')) {
+        console.log('Nightly version detected for website, skipping workflow trigger.');
+        return null;
+      }
+      return ['bump-patch-version.yml', version];
+    }
+  },
+  demo: {
+    tokenEnv: "DEMO_REPO_TOKEN",
+    repo: "demo-scene",
+    workflowLogic: (version: string) => {
+      // Skip nightly versions for demo
+      if (version.includes('nightly')) {
+        console.log('Nightly version detected for demo, skipping workflow trigger.');
+        return null;
+      }
+      return ['bump-patch-version.yml', version];
+    }
+  },
+  docs: {
+    tokenEnv: "DOCS_REPO_TOKEN",
+    repo: "docs",
+    workflowLogic: (version: string) => {
+      // Check if it's a nightly version
+      if (version.includes('nightly')) {
+        return ['bump-nightly-version.yml', version];
+      }
+
+      const parts = version.split('.');
+      if (parts.length !== 3) {
+        throw new Error('Invalid version format');
+      }
+
+      // If patch version (last number) is 0, it's a major version
+      // Return only major.minor version
+      if (parts[2] === '0') {
+        return ['bump-version.yml', `${parts[0]}.${parts[1]}`];
+      }
+
+      // Otherwise it's a patch version, use full version
+      return ['bump-patch-version.yml', version];
+    }
+  }
+};
+
+async function triggerWorkflow(repoConfig: RepoConfig, workflowId: string, version: string) {
+  const client = obtainClient(repoConfig.tokenEnv);
+  try {
+    await client.rest.actions.createWorkflowDispatch({
+      owner: "GreptimeTeam",
+      repo: repoConfig.repo,
+      workflow_id: workflowId,
+      ref: "main",
+      inputs: {
+        version,
+      },
+    });
+    console.log(`Successfully triggered ${workflowId} workflow for ${repoConfig.repo} with version ${version}`);
+  } catch (error) {
+    core.setFailed(`Failed to trigger workflow for ${repoConfig.repo}: ${error.message}`);
+    throw error;
+  }
+}
+
+async function processRepo(repoName: string, version: string) {
+  const repoConfig = REPO_CONFIGS[repoName];
+  if (!repoConfig) {
+    throw new Error(`Unknown repository: ${repoName}`);
+  }
+
+  try {
+    const workflowResult = repoConfig.workflowLogic(version);
+    if (workflowResult === null) {
+      // Skip this repo (e.g., nightly version for website)
+      return;
+    }
+
+    const [workflowId, apiVersion] = workflowResult;
+    await triggerWorkflow(repoConfig, workflowId, apiVersion);
+  } catch (error) {
+    core.setFailed(`Error processing ${repoName} with version ${version}: ${error.message}`);
+    throw error;
+  }
+}
+
+async function main() {
+  const version = process.env.VERSION;
+  if (!version) {
+    core.setFailed("VERSION environment variable is required");
+    process.exit(1);
+  }
+
+  // Remove 'v' prefix if exists
+  const cleanVersion = version.startsWith('v') ? version.slice(1) : version;
+
+  // Get target repositories from environment variable
+  // Default to both if not specified
+  const targetRepos = process.env.TARGET_REPOS?.split(',').map(repo => repo.trim()) || ['website', 'docs'];
+
+  console.log(`Processing version ${cleanVersion} for repositories: ${targetRepos.join(', ')}`);
+
+  const errors: string[] = [];
+
+  // Process each repository
+  for (const repo of targetRepos) {
+    try {
+      await processRepo(repo, cleanVersion);
+    } catch (error) {
+      errors.push(`${repo}: ${error.message}`);
+    }
+  }
+
+  if (errors.length > 0) {
+    core.setFailed(`Failed to process some repositories: ${errors.join('; ')}`);
+    process.exit(1);
+  }
+
+  console.log('All repositories processed successfully');
+}
+
+// Execute main function
+main().catch((error) => {
+  core.setFailed(`Unexpected error: ${error.message}`);
+  process.exit(1);
+});

cyborg/bin/follow-up-docs-issue.ts

@@ -55,12 +55,25 @@ async function main() {
         await client.rest.issues.addLabels({
             owner, repo, issue_number: number, labels: [labelDocsRequired],
         })
+
+        // Get available assignees for the docs repo
+        const assigneesResponse = await docsClient.rest.issues.listAssignees({
+            owner: 'GreptimeTeam',
+            repo: 'docs',
+        })
+        const validAssignees = assigneesResponse.data.map(assignee => assignee.login)
+        core.info(`Available assignees: ${validAssignees.join(', ')}`)
+
+        // Check if the actor is a valid assignee, otherwise fallback to fengjiachun
+        const assignee = validAssignees.includes(actor) ? actor : 'fengjiachun'
+        core.info(`Assigning issue to: ${assignee}`)
+
         await docsClient.rest.issues.create({
             owner: 'GreptimeTeam',
             repo: 'docs',
             title: `Update docs for ${title}`,
             body: `A document change request is generated from ${html_url}`,
-            assignee: actor,
+            assignee: assignee,
         }).then((res) => {
             core.info(`Created issue ${res.data}`)
         })

scripts/install.sh

@@ -53,6 +53,54 @@ get_arch_type() {
   esac
 }
 
+# Verify SHA256 checksum
+verify_sha256() {
+  file="$1"
+  expected_sha256="$2"
+
+  if command -v sha256sum >/dev/null 2>&1; then
+    actual_sha256=$(sha256sum "$file" | cut -d' ' -f1)
+  elif command -v shasum >/dev/null 2>&1; then
+    actual_sha256=$(shasum -a 256 "$file" | cut -d' ' -f1)
+  else
+    echo "Warning: No SHA256 verification tool found (sha256sum or shasum). Skipping checksum verification."
+    return 0
+  fi
+
+  if [ "$actual_sha256" = "$expected_sha256" ]; then
+    echo "SHA256 checksum verified successfully."
+    return 0
+  else
+    echo "Error: SHA256 checksum verification failed!"
+    echo "Expected: $expected_sha256"
+    echo "Actual: $actual_sha256"
+    return 1
+  fi
+}
+
+# Prompt for user confirmation (compatible with different shells)
+prompt_confirmation() {
+  message="$1"
+  printf "%s (y/N): " "$message"
+
+  # Try to read user input, fallback if read fails
+  answer=""
+  if read answer </dev/tty 2>/dev/null; then
+    case "$answer" in
+      [Yy]|[Yy][Ee][Ss])
+        return 0
+        ;;
+      *)
+        return 1
+        ;;
+    esac
+  else
+    echo ""
+    echo "Cannot read user input. Defaulting to No."
+    return 1
+  fi
+}
+
 download_artifact() {
   if [ -n "${OS_TYPE}" ] && [ -n "${ARCH_TYPE}" ]; then
     # Use the latest stable released version.
@@ -71,17 +119,104 @@ download_artifact() {
     fi
 
     echo "Downloading ${BIN}, OS: ${OS_TYPE}, Arch: ${ARCH_TYPE}, Version: ${VERSION}"
-    PACKAGE_NAME="${BIN}-${OS_TYPE}-${ARCH_TYPE}-${VERSION}.tar.gz"
+    PKG_NAME="${BIN}-${OS_TYPE}-${ARCH_TYPE}-${VERSION}"
+    PACKAGE_NAME="${PKG_NAME}.tar.gz"
+    SHA256_FILE="${PKG_NAME}.sha256sum"
 
     if [ -n "${PACKAGE_NAME}" ]; then
-      wget "https://github.com/${GITHUB_ORG}/${GITHUB_REPO}/releases/download/${VERSION}/${PACKAGE_NAME}"
+      # Check if files already exist and prompt for override
+      if [ -f "${PACKAGE_NAME}" ]; then
+        echo "File ${PACKAGE_NAME} already exists."
+        if prompt_confirmation "Do you want to override it?"; then
+          echo "Overriding existing file..."
+          rm -f "${PACKAGE_NAME}"
+        else
+          echo "Skipping download. Using existing file."
+        fi
+      fi
+
+      if [ -f "${BIN}" ]; then
+        echo "Binary ${BIN} already exists."
+        if prompt_confirmation "Do you want to override it?"; then
+          echo "Will override existing binary..."
+          rm -f "${BIN}"
+        else
+          echo "Installation cancelled."
+          exit 0
+        fi
+      fi
+
+      # Download package if not exists
+      if [ ! -f "${PACKAGE_NAME}" ]; then
+        echo "Downloading ${PACKAGE_NAME}..."
+        # Use curl instead of wget for better compatibility
+        if command -v curl >/dev/null 2>&1; then
+          if ! curl -L -o "${PACKAGE_NAME}" "https://github.com/${GITHUB_ORG}/${GITHUB_REPO}/releases/download/${VERSION}/${PACKAGE_NAME}"; then
+            echo "Error: Failed to download ${PACKAGE_NAME}"
+            exit 1
+          fi
+        elif command -v wget >/dev/null 2>&1; then
+          if ! wget -O "${PACKAGE_NAME}" "https://github.com/${GITHUB_ORG}/${GITHUB_REPO}/releases/download/${VERSION}/${PACKAGE_NAME}"; then
+            echo "Error: Failed to download ${PACKAGE_NAME}"
+            exit 1
+          fi
+        else
+          echo "Error: Neither curl nor wget is available for downloading."
+          exit 1
+        fi
+      fi
+
+      # Download and verify SHA256 checksum
+      echo "Downloading SHA256 checksum..."
+      sha256_download_success=0
+      if command -v curl >/dev/null 2>&1; then
+        if curl -L -s -o "${SHA256_FILE}" "https://github.com/${GITHUB_ORG}/${GITHUB_REPO}/releases/download/${VERSION}/${SHA256_FILE}" 2>/dev/null; then
+          sha256_download_success=1
+        fi
+      elif command -v wget >/dev/null 2>&1; then
+        if wget -q -O "${SHA256_FILE}" "https://github.com/${GITHUB_ORG}/${GITHUB_REPO}/releases/download/${VERSION}/${SHA256_FILE}" 2>/dev/null; then
+          sha256_download_success=1
+        fi
+      fi
+
+      if [ $sha256_download_success -eq 1 ] && [ -f "${SHA256_FILE}" ]; then
+        expected_sha256=$(cat "${SHA256_FILE}" | cut -d' ' -f1)
+        if [ -n "$expected_sha256" ]; then
+          if ! verify_sha256 "${PACKAGE_NAME}" "${expected_sha256}"; then
+            echo "SHA256 verification failed. Removing downloaded file."
+            rm -f "${PACKAGE_NAME}" "${SHA256_FILE}"
+            exit 1
+          fi
+        else
+          echo "Warning: Could not parse SHA256 checksum from file."
+        fi
+        rm -f "${SHA256_FILE}"
+      else
+        echo "Warning: Could not download SHA256 checksum file. Skipping verification."
+      fi
 
       # Extract the binary and clean the rest.
-      tar xvf "${PACKAGE_NAME}" && \
-      mv "${PACKAGE_NAME%.tar.gz}/${BIN}" "${PWD}" && \
-      rm -r "${PACKAGE_NAME}" && \
-      rm -r "${PACKAGE_NAME%.tar.gz}" && \
-      echo "Run './${BIN} --help' to get started"
+      echo "Extracting ${PACKAGE_NAME}..."
+      if ! tar xf "${PACKAGE_NAME}"; then
+        echo "Error: Failed to extract ${PACKAGE_NAME}"
+        exit 1
+      fi
+
+      # Find the binary in the extracted directory
+      extracted_dir="${PACKAGE_NAME%.tar.gz}"
+      if [ -f "${extracted_dir}/${BIN}" ]; then
+        mv "${extracted_dir}/${BIN}" "${PWD}/"
+        rm -f "${PACKAGE_NAME}"
+        rm -rf "${extracted_dir}"
+        chmod +x "${BIN}"
+        echo "Installation completed successfully!"
+        echo "Run './${BIN} --help' to get started"
+      else
+        echo "Error: Binary ${BIN} not found in extracted archive"
+        rm -f "${PACKAGE_NAME}"
+        rm -rf "${extracted_dir}"
+        exit 1
+      fi
     fi
   fi
 }

src/client/src/lib.rs

@@ -16,6 +16,7 @@
 
 mod client;
 pub mod client_manager;
+#[cfg(feature = "testing")]
 mod database;
 pub mod error;
 pub mod flow;
@@ -33,6 +34,7 @@ pub use common_recordbatch::{RecordBatches, SendableRecordBatchStream};
 use snafu::OptionExt;
 
 pub use self::client::Client;
+#[cfg(feature = "testing")]
 pub use self::database::Database;
 pub use self::error::{Error, Result};
 use crate::error::{IllegalDatabaseResponseSnafu, ServerSnafu};

src/cmd/Cargo.toml

@@ -9,6 +9,10 @@ default-run = "greptime"
 name = "greptime"
 path = "src/bin/greptime.rs"
 
+[[bin]]
+name = "objbench"
+path = "src/bin/objbench.rs"
+
 [features]
 default = ["servers/pprof", "servers/mem-prof"]
 tokio-console = ["common-telemetry/tokio-console"]
@@ -20,6 +24,7 @@ workspace = true
 async-trait.workspace = true
 auth.workspace = true
 base64.workspace = true
+colored = "2.0"
 cache.workspace = true
 catalog.workspace = true
 chrono.workspace = true
@@ -55,6 +60,9 @@ futures.workspace = true
 human-panic = "2.0"
 humantime.workspace = true
 lazy_static.workspace = true
+object-store.workspace = true
+parquet = "53"
+pprof = "0.14"
 meta-client.workspace = true
 meta-srv.workspace = true
 metric-engine.workspace = true

src/cmd/src/bin/greptime.rs

@@ -21,6 +21,8 @@ use cmd::{cli, datanode, flownode, frontend, metasrv, standalone, App};
 use common_version::version;
 use servers::install_ring_crypto_provider;
 
+pub mod objbench;
+
 #[derive(Parser)]
 #[command(name = "greptime", author, version, long_version = version(), about)]
 #[command(propagate_version = true)]

src/cmd/src/bin/objbench.rs

@@ -0,0 +1,602 @@
+// Copyright 2025 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::path::{Path, PathBuf};
+use std::time::Instant;
+
+use clap::Parser;
+use cmd::error::{self, Result};
+use colored::Colorize;
+use datanode::config::ObjectStoreConfig;
+use mito2::config::{FulltextIndexConfig, MitoConfig, Mode};
+use mito2::read::Source;
+use mito2::sst::file::{FileHandle, FileId, FileMeta};
+use mito2::sst::file_purger::{FilePurger, FilePurgerRef, PurgeRequest};
+use mito2::sst::parquet::{WriteOptions, PARQUET_METADATA_KEY};
+use mito2::{build_access_layer, Metrics, OperationType, SstWriteRequest};
+use object_store::ObjectStore;
+use serde::{Deserialize, Serialize};
+use store_api::metadata::{RegionMetadata, RegionMetadataRef};
+
+#[tokio::main]
+pub async fn main() {
+    // common_telemetry::init_default_ut_logging();
+    let cmd = Command::parse();
+    if let Err(e) = cmd.run().await {
+        eprintln!("{}: {}", "Error".red().bold(), e);
+        std::process::exit(1);
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Default)]
+#[serde(default)]
+pub struct StorageConfigWrapper {
+    storage: StorageConfig,
+}
+
+/// Storage engine config
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Default)]
+#[serde(default)]
+pub struct StorageConfig {
+    /// The working directory of database
+    pub data_home: String,
+    #[serde(flatten)]
+    pub store: ObjectStoreConfig,
+}
+
+#[derive(Debug, Parser)]
+pub struct Command {
+    /// Path to the object-store config file (TOML). Must deserialize into datanode::config::ObjectStoreConfig.
+    #[clap(long, value_name = "FILE")]
+    pub config: PathBuf,
+
+    /// Source SST file path in object-store (e.g. "region_dir/<uuid>.parquet").
+    #[clap(long, value_name = "PATH")]
+    pub source: String,
+
+    /// Target SST file path in object-store; its parent directory is used as destination region dir.
+    #[clap(long, value_name = "PATH")]
+    pub target: String,
+
+    /// Verbose output
+    #[clap(short, long, default_value_t = false)]
+    pub verbose: bool,
+
+    /// Output file path for pprof flamegraph (enables profiling)
+    #[clap(long, value_name = "FILE")]
+    pub pprof_file: Option<PathBuf>,
+}
+
+impl Command {
+    pub async fn run(&self) -> Result<()> {
+        if self.verbose {
+            common_telemetry::init_default_ut_logging();
+        }
+
+        println!("{}", "Starting objbench...".cyan().bold());
+
+        // Build object store from config
+        let cfg_str = std::fs::read_to_string(&self.config).map_err(|e| {
+            error::IllegalConfigSnafu {
+                msg: format!("failed to read config {}: {e}", self.config.display()),
+            }
+            .build()
+        })?;
+        let store_cfg: StorageConfigWrapper = toml::from_str(&cfg_str).map_err(|e| {
+            error::IllegalConfigSnafu {
+                msg: format!("failed to parse config {}: {e}", self.config.display()),
+            }
+            .build()
+        })?;
+
+        let object_store = build_object_store(&store_cfg.storage).await?;
+        println!("{} Object store initialized", "✓".green());
+
+        // Prepare source identifiers
+        let (src_region_dir, src_file_id) = split_sst_path(&self.source)?;
+        println!("{} Source path parsed: {}", "✓".green(), self.source);
+
+        // Load parquet metadata to extract RegionMetadata and file stats
+        println!("{}", "Loading parquet metadata...".yellow());
+        let file_size = object_store
+            .stat(&self.source)
+            .await
+            .map_err(|e| {
+                error::IllegalConfigSnafu {
+                    msg: format!("stat failed: {e}"),
+                }
+                .build()
+            })?
+            .content_length();
+        let parquet_meta = load_parquet_metadata(object_store.clone(), &self.source, file_size)
+            .await
+            .map_err(|e| {
+                error::IllegalConfigSnafu {
+                    msg: format!("read parquet metadata failed: {e}"),
+                }
+                .build()
+            })?;
+
+        let region_meta = extract_region_metadata(&self.source, &parquet_meta)?;
+        let num_rows = parquet_meta.file_metadata().num_rows() as u64;
+        let num_row_groups = parquet_meta.num_row_groups() as u64;
+
+        println!(
+            "{} Metadata loaded - rows: {}, size: {} bytes",
+            "✓".green(),
+            num_rows,
+            file_size
+        );
+
+        // Build a FileHandle for the source file
+        let file_meta = FileMeta {
+            region_id: region_meta.region_id,
+            file_id: src_file_id,
+            time_range: Default::default(),
+            level: 0,
+            file_size,
+            available_indexes: Default::default(),
+            index_file_size: 0,
+            num_rows,
+            num_row_groups,
+            sequence: None,
+        };
+        let src_handle = FileHandle::new(file_meta, new_noop_file_purger());
+
+        // Build the reader for a single file via ParquetReaderBuilder
+        println!("{}", "Building reader...".yellow());
+        let (_src_access_layer, _cache_manager) =
+            build_access_layer_simple(src_region_dir.clone(), object_store.clone()).await?;
+        let reader_build_start = Instant::now();
+        let reader = mito2::sst::parquet::reader::ParquetReaderBuilder::new(
+            src_region_dir.clone(),
+            src_handle.clone(),
+            object_store.clone(),
+        )
+        .expected_metadata(Some(region_meta.clone()))
+        .build()
+        .await
+        .map_err(|e| {
+            error::IllegalConfigSnafu {
+                msg: format!("build reader failed: {e}"),
+            }
+            .build()
+        })?;
+
+        let reader_build_elapsed = reader_build_start.elapsed();
+        let total_rows = reader.parquet_metadata().file_metadata().num_rows();
+        println!("{} Reader built in {:?}", "✓".green(), reader_build_elapsed);
+
+        // Prepare target access layer for writing
+        println!("{}", "Preparing target access layer...".yellow());
+        let (tgt_access_layer, tgt_cache_manager) =
+            build_access_layer_simple(self.target.clone(), object_store.clone()).await?;
+
+        // Build write request
+        let fulltext_index_config = FulltextIndexConfig {
+            create_on_compaction: Mode::Disable,
+            ..Default::default()
+        };
+        let write_opts = WriteOptions::default();
+        let write_req = SstWriteRequest {
+            op_type: OperationType::Compact,
+            metadata: region_meta,
+            source: Source::Reader(Box::new(reader)),
+            cache_manager: tgt_cache_manager,
+            storage: None,
+            max_sequence: None,
+            index_options: Default::default(),
+            inverted_index_config: MitoConfig::default().inverted_index,
+            fulltext_index_config,
+            bloom_filter_index_config: MitoConfig::default().bloom_filter_index,
+        };
+
+        // Write SST
+        println!("{}", "Writing SST...".yellow());
+        let mut metrics = Metrics::default();
+
+        // Start profiling if pprof_file is specified
+        #[cfg(unix)]
+        let profiler_guard = if self.pprof_file.is_some() {
+            println!("{} Starting profiling...", "⚡".yellow());
+            Some(
+                pprof::ProfilerGuardBuilder::default()
+                    .frequency(99)
+                    .blocklist(&["libc", "libgcc", "pthread", "vdso"])
+                    .build()
+                    .map_err(|e| {
+                        error::IllegalConfigSnafu {
+                            msg: format!("Failed to start profiler: {e}"),
+                        }
+                        .build()
+                    })?,
+            )
+        } else {
+            None
+        };
+
+        #[cfg(not(unix))]
+        if self.pprof_file.is_some() {
+            eprintln!(
+                "{}: Profiling is not supported on this platform",
+                "Warning".yellow()
+            );
+        }
+
+        let write_start = Instant::now();
+        let infos = tgt_access_layer
+            .write_sst(write_req, &write_opts, &mut metrics)
+            .await
+            .map_err(|e| {
+                error::IllegalConfigSnafu {
+                    msg: format!("write_sst failed: {e}"),
+                }
+                .build()
+            })?;
+
+        let write_elapsed = write_start.elapsed();
+
+        // Stop profiling and generate flamegraph if enabled
+        #[cfg(unix)]
+        if let (Some(guard), Some(pprof_file)) = (profiler_guard, &self.pprof_file) {
+            println!("{} Generating flamegraph...", "🔥".yellow());
+            match guard.report().build() {
+                Ok(report) => {
+                    let mut flamegraph_data = Vec::new();
+                    if let Err(e) = report.flamegraph(&mut flamegraph_data) {
+                        eprintln!(
+                            "{}: Failed to generate flamegraph: {}",
+                            "Warning".yellow(),
+                            e
+                        );
+                    } else if let Err(e) = std::fs::write(pprof_file, flamegraph_data) {
+                        eprintln!(
+                            "{}: Failed to write flamegraph to {}: {}",
+                            "Warning".yellow(),
+                            pprof_file.display(),
+                            e
+                        );
+                    } else {
+                        println!(
+                            "{} Flamegraph saved to {}",
+                            "✓".green(),
+                            pprof_file.display().to_string().cyan()
+                        );
+                    }
+                }
+                Err(e) => {
+                    eprintln!(
+                        "{}: Failed to generate pprof report: {}",
+                        "Warning".yellow(),
+                        e
+                    );
+                }
+            }
+        }
+        assert_eq!(infos.len(), 1);
+        let dst_file_id = infos[0].file_id;
+        let dst_file_path = format!("{}{}", self.target, dst_file_id.as_parquet(),);
+
+        // Report results with ANSI colors
+        println!("\n{} {}", "Write complete!".green().bold(), "✓".green());
+        println!("  {}: {}", "Destination file".bold(), dst_file_path.cyan());
+        println!("  {}: {}", "Rows".bold(), total_rows.to_string().cyan());
+        println!(
+            "  {}: {}",
+            "File size".bold(),
+            format!("{} bytes", file_size).cyan()
+        );
+        println!(
+            "  {}: {:?}",
+            "Reader build time".bold(),
+            reader_build_elapsed
+        );
+        println!("  {}: {:?}", "Total time".bold(), write_elapsed);
+
+        // Print metrics in a formatted way
+        println!(
+            "  {}: {:?}, sum: {:?}",
+            "Metrics".bold(),
+            metrics,
+            metrics.sum()
+        );
+
+        // Print infos
+        println!("  {}: {:?}", "Index".bold(), infos[0].index_metadata);
+
+        // Cleanup
+        println!("\n{}", "Cleaning up...".yellow());
+        object_store.delete(&dst_file_path).await.map_err(|e| {
+            error::IllegalConfigSnafu {
+                msg: format!("Failed to delete dest file {}: {}", dst_file_path, e),
+            }
+            .build()
+        })?;
+        println!("{} Temporary file deleted", "✓".green());
+
+        println!("\n{}", "Benchmark completed successfully!".green().bold());
+        Ok(())
+    }
+}
+
+fn split_sst_path(path: &str) -> Result<(String, FileId)> {
+    let p = Path::new(path);
+    let file_name = p.file_name().and_then(|s| s.to_str()).ok_or_else(|| {
+        error::IllegalConfigSnafu {
+            msg: "invalid source path".to_string(),
+        }
+        .build()
+    })?;
+    let uuid_str = file_name.strip_suffix(".parquet").ok_or_else(|| {
+        error::IllegalConfigSnafu {
+            msg: "expect .parquet file".to_string(),
+        }
+        .build()
+    })?;
+    let file_id = FileId::parse_str(uuid_str).map_err(|e| {
+        error::IllegalConfigSnafu {
+            msg: format!("invalid file id: {e}"),
+        }
+        .build()
+    })?;
+    let parent = p
+        .parent()
+        .and_then(|s| s.to_str())
+        .unwrap_or("")
+        .to_string();
+    Ok((parent, file_id))
+}
+
+fn extract_region_metadata(
+    file_path: &str,
+    meta: &parquet::file::metadata::ParquetMetaData,
+) -> Result<RegionMetadataRef> {
+    use parquet::format::KeyValue;
+    let kvs: Option<&Vec<KeyValue>> = meta.file_metadata().key_value_metadata();
+    let Some(kvs) = kvs else {
+        return Err(error::IllegalConfigSnafu {
+            msg: format!("{file_path}: missing parquet key_value metadata"),
+        }
+        .build());
+    };
+    let json = kvs
+        .iter()
+        .find(|kv| kv.key == PARQUET_METADATA_KEY)
+        .and_then(|kv| kv.value.as_ref())
+        .ok_or_else(|| {
+            error::IllegalConfigSnafu {
+                msg: format!("{file_path}: key {PARQUET_METADATA_KEY} not found or empty"),
+            }
+            .build()
+        })?;
+    let region: RegionMetadata = RegionMetadata::from_json(json).map_err(|e| {
+        error::IllegalConfigSnafu {
+            msg: format!("invalid region metadata json: {e}"),
+        }
+        .build()
+    })?;
+    Ok(std::sync::Arc::new(region))
+}
+
+async fn build_object_store(sc: &StorageConfig) -> Result<ObjectStore> {
+    use datanode::config::ObjectStoreConfig::*;
+    let oss = &sc.store;
+    match oss {
+        File(_) => {
+            use object_store::services::Fs;
+            let builder = Fs::default().root(&sc.data_home);
+            Ok(ObjectStore::new(builder)
+                .map_err(|e| {
+                    error::IllegalConfigSnafu {
+                        msg: format!("init fs backend failed: {e}"),
+                    }
+                    .build()
+                })?
+                .finish())
+        }
+        S3(s3) => {
+            use common_base::secrets::ExposeSecret;
+            use object_store::services::S3;
+            use object_store::util;
+            let root = util::normalize_dir(&s3.root);
+            let mut builder = S3::default()
+                .root(&root)
+                .bucket(&s3.bucket)
+                .access_key_id(s3.access_key_id.expose_secret())
+                .secret_access_key(s3.secret_access_key.expose_secret());
+            if let Some(ep) = &s3.endpoint {
+                builder = builder.endpoint(ep);
+            }
+            if let Some(region) = &s3.region {
+                builder = builder.region(region);
+            }
+            if s3.enable_virtual_host_style {
+                builder = builder.enable_virtual_host_style();
+            }
+            Ok(ObjectStore::new(builder)
+                .map_err(|e| {
+                    error::IllegalConfigSnafu {
+                        msg: format!("init s3 backend failed: {e}"),
+                    }
+                    .build()
+                })?
+                .finish())
+        }
+        Oss(oss) => {
+            use common_base::secrets::ExposeSecret;
+            use object_store::services::Oss;
+            use object_store::util;
+            let root = util::normalize_dir(&oss.root);
+            let builder = Oss::default()
+                .root(&root)
+                .bucket(&oss.bucket)
+                .endpoint(&oss.endpoint)
+                .access_key_id(oss.access_key_id.expose_secret())
+                .access_key_secret(oss.access_key_secret.expose_secret());
+            Ok(ObjectStore::new(builder)
+                .map_err(|e| {
+                    error::IllegalConfigSnafu {
+                        msg: format!("init oss backend failed: {e}"),
+                    }
+                    .build()
+                })?
+                .finish())
+        }
+        Azblob(az) => {
+            use common_base::secrets::ExposeSecret;
+            use object_store::services::Azblob;
+            use object_store::util;
+            let root = util::normalize_dir(&az.root);
+            let mut builder = Azblob::default()
+                .root(&root)
+                .container(&az.container)
+                .endpoint(&az.endpoint)
+                .account_name(az.account_name.expose_secret())
+                .account_key(az.account_key.expose_secret());
+            if let Some(token) = &az.sas_token {
+                builder = builder.sas_token(token);
+            }
+            Ok(ObjectStore::new(builder)
+                .map_err(|e| {
+                    error::IllegalConfigSnafu {
+                        msg: format!("init azblob backend failed: {e}"),
+                    }
+                    .build()
+                })?
+                .finish())
+        }
+        Gcs(gcs) => {
+            use common_base::secrets::ExposeSecret;
+            use object_store::services::Gcs;
+            use object_store::util;
+            let root = util::normalize_dir(&gcs.root);
+            let builder = Gcs::default()
+                .root(&root)
+                .bucket(&gcs.bucket)
+                .scope(&gcs.scope)
+                .credential_path(gcs.credential_path.expose_secret())
+                .credential(gcs.credential.expose_secret())
+                .endpoint(&gcs.endpoint);
+            Ok(ObjectStore::new(builder)
+                .map_err(|e| {
+                    error::IllegalConfigSnafu {
+                        msg: format!("init gcs backend failed: {e}"),
+                    }
+                    .build()
+                })?
+                .finish())
+        }
+    }
+}
+
+async fn build_access_layer_simple(
+    region_dir: String,
+    object_store: ObjectStore,
+) -> Result<(
+    std::sync::Arc<mito2::AccessLayer>,
+    std::sync::Arc<mito2::CacheManager>,
+)> {
+    // Minimal index aux path setup
+    let mut mito_cfg = MitoConfig::default();
+    // Use a temporary directory as aux path
+    let data_home = std::env::temp_dir().join("greptime_objbench");
+    let _ = std::fs::create_dir_all(&data_home);
+    let _ = mito_cfg.index.sanitize(
+        data_home.to_str().unwrap_or("/tmp"),
+        &mito_cfg.inverted_index,
+    );
+    let access_layer = build_access_layer(&region_dir, object_store, &mito_cfg)
+        .await
+        .map_err(|e| {
+            error::IllegalConfigSnafu {
+                msg: format!("build_access_layer failed: {e}"),
+            }
+            .build()
+        })?;
+    Ok((
+        access_layer,
+        std::sync::Arc::new(mito2::CacheManager::default()),
+    ))
+}
+
+fn new_noop_file_purger() -> FilePurgerRef {
+    #[derive(Debug)]
+    struct Noop;
+    impl FilePurger for Noop {
+        fn send_request(&self, _request: PurgeRequest) {}
+    }
+    std::sync::Arc::new(Noop)
+}
+
+async fn load_parquet_metadata(
+    object_store: ObjectStore,
+    path: &str,
+    file_size: u64,
+) -> std::result::Result<
+    parquet::file::metadata::ParquetMetaData,
+    Box<dyn std::error::Error + Send + Sync>,
+> {
+    use parquet::file::metadata::ParquetMetaDataReader;
+    use parquet::file::FOOTER_SIZE;
+    let actual_size = if file_size == 0 {
+        object_store.stat(path).await?.content_length()
+    } else {
+        file_size
+    };
+    if actual_size < FOOTER_SIZE as u64 {
+        return Err("file too small".into());
+    }
+    let prefetch: u64 = 64 * 1024;
+    let start = actual_size.saturating_sub(prefetch);
+    let buffer = object_store
+        .read_with(path)
+        .range(start..actual_size)
+        .await?
+        .to_vec();
+    let buffer_len = buffer.len();
+    let mut footer = [0; 8];
+    footer.copy_from_slice(&buffer[buffer_len - FOOTER_SIZE..]);
+    let metadata_len = ParquetMetaDataReader::decode_footer(&footer)? as u64;
+    if actual_size - (FOOTER_SIZE as u64) < metadata_len {
+        return Err("invalid footer/metadata length".into());
+    }
+    if (metadata_len as usize) <= buffer_len - FOOTER_SIZE {
+        let metadata_start = buffer_len - metadata_len as usize - FOOTER_SIZE;
+        let meta = ParquetMetaDataReader::decode_metadata(
+            &buffer[metadata_start..buffer_len - FOOTER_SIZE],
+        )?;
+        Ok(meta)
+    } else {
+        let metadata_start = actual_size - metadata_len - FOOTER_SIZE as u64;
+        let data = object_store
+            .read_with(path)
+            .range(metadata_start..(actual_size - FOOTER_SIZE as u64))
+            .await?
+            .to_vec();
+        let meta = ParquetMetaDataReader::decode_metadata(&data)?;
+        Ok(meta)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::StorageConfigWrapper;
+
+    #[test]
+    fn test_decode() {
+        let cfg = std::fs::read_to_string("/home/lei/datanode-bulk.toml").unwrap();
+        let storage: StorageConfigWrapper = toml::from_str(&cfg).unwrap();
+        println!("{:?}", storage);
+    }
+}

src/cmd/src/flownode.rs

@@ -32,7 +32,7 @@ use common_meta::key::TableMetadataManager;
 use common_telemetry::info;
 use common_telemetry::logging::TracingOptions;
 use common_version::{short_version, version};
-use flow::{FlownodeBuilder, FlownodeInstance, FrontendClient, FrontendInvoker};
+use flow::{FlownodeBuilder, FlownodeInstance, FrontendInvoker};
 use meta_client::{MetaClientOptions, MetaClientType};
 use servers::Mode;
 use snafu::{OptionExt, ResultExt};
@@ -317,8 +317,6 @@ impl StartCommand {
             Arc::new(executor),
         );
 
-        let frontend_client = FrontendClient::from_meta_client(meta_client.clone());
-
         let flow_metadata_manager = Arc::new(FlowMetadataManager::new(cached_meta_backend.clone()));
         let flownode_builder = FlownodeBuilder::new(
             opts,
@@ -326,7 +324,6 @@ impl StartCommand {
             table_metadata_manager,
             catalog_manager.clone(),
             flow_metadata_manager,
-            Arc::new(frontend_client),
         )
         .with_heartbeat_task(heartbeat_task);
 

src/cmd/src/standalone.rs

@@ -54,10 +54,7 @@ use datanode::config::{DatanodeOptions, ProcedureConfig, RegionEngineConfig, Sto
 use datanode::datanode::{Datanode, DatanodeBuilder};
 use datanode::region_server::RegionServer;
 use file_engine::config::EngineConfig as FileEngineConfig;
-use flow::{
-    FlowConfig, FlowWorkerManager, FlownodeBuilder, FlownodeOptions, FrontendClient,
-    FrontendInvoker,
-};
+use flow::{FlowConfig, FlowWorkerManager, FlownodeBuilder, FlownodeOptions, FrontendInvoker};
 use frontend::frontend::FrontendOptions;
 use frontend::instance::builder::FrontendBuilder;
 use frontend::instance::{FrontendInstance, Instance as FeInstance, StandaloneDatanodeManager};
@@ -536,16 +533,12 @@ impl StartCommand {
             flow: opts.flow.clone(),
             ..Default::default()
         };
-
-        let fe_server_addr = fe_opts.grpc.bind_addr.clone();
-        let frontend_client = FrontendClient::from_static_grpc_addr(fe_server_addr);
         let flow_builder = FlownodeBuilder::new(
             flownode_options,
             plugins.clone(),
             table_metadata_manager.clone(),
             catalog_manager.clone(),
             flow_metadata_manager.clone(),
-            Arc::new(frontend_client),
         );
         let flownode = Arc::new(
             flow_builder

src/common/meta/src/cluster.rs

@@ -57,12 +57,10 @@ pub trait ClusterInfo {
 }
 
 /// The key of [NodeInfo] in the storage. The format is `__meta_cluster_node_info-{cluster_id}-{role}-{node_id}`.
-///
-/// This key cannot be used to describe the `Metasrv` because the `Metasrv` does not have
-/// a `cluster_id`, it serves multiple clusters.
 #[derive(Debug, Clone, Copy, Eq, Hash, PartialEq, Serialize, Deserialize)]
 pub struct NodeInfoKey {
     /// The cluster id.
+    // todo(hl): remove cluster_id as it is not assigned anywhere.
     pub cluster_id: ClusterId,
     /// The role of the node. It can be `[Role::Datanode]` or `[Role::Frontend]`.
     pub role: Role,
@@ -232,8 +230,8 @@ impl TryFrom<Vec<u8>> for NodeInfoKey {
     }
 }
 
-impl From<NodeInfoKey> for Vec<u8> {
-    fn from(key: NodeInfoKey) -> Self {
+impl From<&NodeInfoKey> for Vec<u8> {
+    fn from(key: &NodeInfoKey) -> Self {
         format!(
             "{}-{}-{}-{}",
             CLUSTER_NODE_INFO_PREFIX,
@@ -315,7 +313,7 @@ mod tests {
             node_id: 2,
         };
 
-        let key_bytes: Vec<u8> = key.into();
+        let key_bytes: Vec<u8> = (&key).into();
         let new_key: NodeInfoKey = key_bytes.try_into().unwrap();
 
         assert_eq!(1, new_key.cluster_id);

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

@@ -343,7 +343,6 @@ pub enum FlowType {
 impl FlowType {
     pub const RECORDING_RULE: &str = "recording_rule";
     pub const STREAMING: &str = "streaming";
-    pub const FLOW_TYPE_KEY: &str = "flow_type";
 }
 
 impl Default for FlowType {
@@ -399,8 +398,7 @@ impl From<&CreateFlowData> for CreateRequest {
         };
 
         let flow_type = value.flow_type.unwrap_or_default().to_string();
-        req.flow_options
-            .insert(FlowType::FLOW_TYPE_KEY.to_string(), flow_type);
+        req.flow_options.insert("flow_type".to_string(), flow_type);
         req
     }
 }
@@ -432,7 +430,7 @@ impl From<&CreateFlowData> for (FlowInfoValue, Vec<(FlowPartitionId, FlowRouteVa
             .collect::<Vec<_>>();
 
         let flow_type = value.flow_type.unwrap_or_default().to_string();
-        options.insert(FlowType::FLOW_TYPE_KEY.to_string(), flow_type);
+        options.insert("flow_type".to_string(), flow_type);
 
         let flow_info = FlowInfoValue {
             source_table_ids: value.source_table_ids.clone(),

src/common/meta/src/lib.rs

@@ -34,6 +34,7 @@ pub mod kv_backend;
 pub mod leadership_notifier;
 pub mod lock_key;
 pub mod metrics;
+pub mod node_expiry_listener;
 pub mod node_manager;
 pub mod peer;
 pub mod range_stream;

src/common/meta/src/node_expiry_listener.rs

@@ -0,0 +1,152 @@
+// 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::Mutex;
+use std::time::Duration;
+
+use common_telemetry::{debug, error, info, warn};
+use tokio::task::JoinHandle;
+use tokio::time::{interval, MissedTickBehavior};
+
+use crate::cluster::{NodeInfo, NodeInfoKey};
+use crate::error;
+use crate::kv_backend::ResettableKvBackendRef;
+use crate::leadership_notifier::LeadershipChangeListener;
+use crate::rpc::store::RangeRequest;
+use crate::rpc::KeyValue;
+
+/// [NodeExpiryListener] periodically checks all node info in memory and removes
+/// expired node info to prevent memory leak.
+pub struct NodeExpiryListener {
+    handle: Mutex<Option<JoinHandle<()>>>,
+    max_idle_time: Duration,
+    in_memory: ResettableKvBackendRef,
+}
+
+impl Drop for NodeExpiryListener {
+    fn drop(&mut self) {
+        self.stop();
+    }
+}
+
+impl NodeExpiryListener {
+    pub fn new(max_idle_time: Duration, in_memory: ResettableKvBackendRef) -> Self {
+        Self {
+            handle: Mutex::new(None),
+            max_idle_time,
+            in_memory,
+        }
+    }
+
+    async fn start(&self) {
+        let mut handle = self.handle.lock().unwrap();
+        if handle.is_none() {
+            let in_memory = self.in_memory.clone();
+
+            let max_idle_time = self.max_idle_time;
+            let ticker_loop = tokio::spawn(async move {
+                // Run clean task every minute.
+                let mut interval = interval(Duration::from_secs(60));
+                interval.set_missed_tick_behavior(MissedTickBehavior::Skip);
+                loop {
+                    interval.tick().await;
+                    if let Err(e) = Self::clean_expired_nodes(&in_memory, max_idle_time).await {
+                        error!(e; "Failed to clean expired node");
+                    }
+                }
+            });
+            *handle = Some(ticker_loop);
+        }
+    }
+
+    fn stop(&self) {
+        if let Some(handle) = self.handle.lock().unwrap().take() {
+            handle.abort();
+            info!("Node expiry listener stopped")
+        }
+    }
+
+    /// Cleans expired nodes from memory.
+    async fn clean_expired_nodes(
+        in_memory: &ResettableKvBackendRef,
+        max_idle_time: Duration,
+    ) -> error::Result<()> {
+        let node_keys = Self::list_expired_nodes(in_memory, max_idle_time).await?;
+        for key in node_keys {
+            let key_bytes: Vec<u8> = (&key).into();
+            if let Err(e) = in_memory.delete(&key_bytes, false).await {
+                warn!(e; "Failed to delete expired node: {:?}", key_bytes);
+            } else {
+                debug!("Deleted expired node key: {:?}", key);
+            }
+        }
+        Ok(())
+    }
+
+    /// Lists expired nodes that have been inactive more than `max_idle_time`.
+    async fn list_expired_nodes(
+        in_memory: &ResettableKvBackendRef,
+        max_idle_time: Duration,
+    ) -> error::Result<impl Iterator<Item = NodeInfoKey>> {
+        let prefix = NodeInfoKey::key_prefix_with_cluster_id(0);
+        let req = RangeRequest::new().with_prefix(prefix);
+        let current_time_millis = common_time::util::current_time_millis();
+        let resp = in_memory.range(req).await?;
+        Ok(resp
+            .kvs
+            .into_iter()
+            .filter_map(move |KeyValue { key, value }| {
+                let Ok(info) = NodeInfo::try_from(value).inspect_err(|e| {
+                    warn!(e; "Unrecognized node info value");
+                }) else {
+                    return None;
+                };
+                if (current_time_millis - info.last_activity_ts) > max_idle_time.as_millis() as i64
+                {
+                    NodeInfoKey::try_from(key)
+                        .inspect_err(|e| {
+                            warn!(e; "Unrecognized node info key: {:?}", info.peer);
+                        })
+                        .ok()
+                        .inspect(|node_key| {
+                            debug!("Found expired node: {:?}", node_key);
+                        })
+                } else {
+                    None
+                }
+            }))
+    }
+}
+
+#[async_trait::async_trait]
+impl LeadershipChangeListener for NodeExpiryListener {
+    fn name(&self) -> &str {
+        "NodeExpiryListener"
+    }
+
+    async fn on_leader_start(&self) -> error::Result<()> {
+        self.start().await;
+        info!(
+            "On leader start, node expiry listener started with max idle time: {:?}",
+            self.max_idle_time
+        );
+        Ok(())
+    }
+
+    async fn on_leader_stop(&self) -> error::Result<()> {
+        self.stop();
+        info!("On leader stop, node expiry listener stopped");
+        Ok(())
+    }
+}

src/datanode/src/config.rs

@@ -171,6 +171,10 @@ pub struct S3Config {
     pub secret_access_key: SecretString,
     pub endpoint: Option<String>,
     pub region: Option<String>,
+    /// Enable virtual host style so that opendal will send API requests in virtual host style instead of path style.
+    /// By default, opendal will send API to https://s3.us-east-1.amazonaws.com/bucket_name
+    /// Enabled, opendal will send API to https://bucket_name.s3.us-east-1.amazonaws.com
+    pub enable_virtual_host_style: bool,
     #[serde(flatten)]
     pub cache: ObjectStorageCacheConfig,
     pub http_client: HttpClientConfig,
@@ -185,6 +189,7 @@ impl PartialEq for S3Config {
             && self.secret_access_key.expose_secret() == other.secret_access_key.expose_secret()
             && self.endpoint == other.endpoint
             && self.region == other.region
+            && self.enable_virtual_host_style == other.enable_virtual_host_style
             && self.cache == other.cache
             && self.http_client == other.http_client
     }
@@ -289,6 +294,7 @@ impl Default for S3Config {
             root: String::default(),
             access_key_id: SecretString::from(String::default()),
             secret_access_key: SecretString::from(String::default()),
+            enable_virtual_host_style: false,
             endpoint: Option::default(),
             region: Option::default(),
             cache: ObjectStorageCacheConfig::default(),

src/datanode/src/store/s3.rs

@@ -41,10 +41,13 @@ pub(crate) async fn new_s3_object_store(s3_config: &S3Config) -> Result<ObjectSt
 
     if s3_config.endpoint.is_some() {
         builder = builder.endpoint(s3_config.endpoint.as_ref().unwrap());
-    };
+    }
     if s3_config.region.is_some() {
         builder = builder.region(s3_config.region.as_ref().unwrap());
-    };
+    }
+    if s3_config.enable_virtual_host_style {
+        builder = builder.enable_virtual_host_style();
+    }
 
     Ok(ObjectStore::new(builder)
         .context(error::InitBackendSnafu)?

src/datatypes/src/lib.rs

@@ -32,5 +32,5 @@ pub mod types;
 pub mod value;
 pub mod vectors;
 
-pub use arrow;
+pub use arrow::{self, compute};
 pub use error::{Error, Result};

src/flow/src/adapter.rs

@@ -49,13 +49,12 @@ pub(crate) use crate::adapter::node_context::FlownodeContext;
 use crate::adapter::refill::RefillTask;
 use crate::adapter::table_source::ManagedTableSource;
 use crate::adapter::util::relation_desc_to_column_schemas_with_fallback;
-pub(crate) use crate::adapter::worker::{create_worker, WorkerHandle};
+pub(crate) use crate::adapter::worker::{create_worker, Worker, WorkerHandle};
 use crate::compute::ErrCollector;
 use crate::df_optimizer::sql_to_flow_plan;
 use crate::error::{EvalSnafu, ExternalSnafu, InternalSnafu, InvalidQuerySnafu, UnexpectedSnafu};
 use crate::expr::Batch;
 use crate::metrics::{METRIC_FLOW_INSERT_ELAPSED, METRIC_FLOW_ROWS, METRIC_FLOW_RUN_INTERVAL_MS};
-use crate::recording_rules::RecordingRuleEngine;
 use crate::repr::{self, DiffRow, RelationDesc, Row, BATCH_SIZE};
 
 mod flownode_impl;
@@ -172,8 +171,6 @@ pub struct FlowWorkerManager {
     flush_lock: RwLock<()>,
     /// receive a oneshot sender to send state size report
     state_report_handler: RwLock<Option<StateReportHandler>>,
-    /// engine for recording rule
-    rule_engine: RecordingRuleEngine,
 }
 
 /// Building FlownodeManager
@@ -188,7 +185,6 @@ impl FlowWorkerManager {
         node_id: Option<u32>,
         query_engine: Arc<dyn QueryEngine>,
         table_meta: TableMetadataManagerRef,
-        rule_engine: RecordingRuleEngine,
     ) -> Self {
         let srv_map = ManagedTableSource::new(
             table_meta.table_info_manager().clone(),
@@ -211,7 +207,6 @@ impl FlowWorkerManager {
             node_id,
             flush_lock: RwLock::new(()),
             state_report_handler: RwLock::new(None),
-            rule_engine,
         }
     }
 
@@ -220,6 +215,25 @@ impl FlowWorkerManager {
         self
     }
 
+    /// Create a flownode manager with one worker
+    pub fn new_with_workers<'s>(
+        node_id: Option<u32>,
+        query_engine: Arc<dyn QueryEngine>,
+        table_meta: TableMetadataManagerRef,
+        num_workers: usize,
+    ) -> (Self, Vec<Worker<'s>>) {
+        let mut zelf = Self::new(node_id, query_engine, table_meta);
+
+        let workers: Vec<_> = (0..num_workers)
+            .map(|_| {
+                let (handle, worker) = create_worker();
+                zelf.add_worker_handle(handle);
+                worker
+            })
+            .collect();
+        (zelf, workers)
+    }
+
     /// add a worker handler to manager, meaning this corresponding worker is under it's manage
     pub fn add_worker_handle(&mut self, handle: WorkerHandle) {
         self.worker_handles.push(handle);
@@ -737,11 +751,7 @@ pub struct CreateFlowArgs {
 /// Create&Remove flow
 impl FlowWorkerManager {
     /// remove a flow by it's id
-    #[allow(unreachable_code)]
     pub async fn remove_flow(&self, flow_id: FlowId) -> Result<(), Error> {
-        // TODO(discord9): reroute some back to streaming engine later
-        return self.rule_engine.remove_flow(flow_id).await;
-
         for handle in self.worker_handles.iter() {
             if handle.contains_flow(flow_id).await? {
                 handle.remove_flow(flow_id).await?;
@@ -757,10 +767,8 @@ 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, unreachable_code)]
+    #[allow(clippy::too_many_arguments)]
     pub async fn create_flow(&self, args: CreateFlowArgs) -> Result<Option<FlowId>, Error> {
-        // TODO(discord9): reroute some back to streaming engine later
-        return self.rule_engine.create_flow(args).await;
         let CreateFlowArgs {
             flow_id,
             sink_table_name,

src/flow/src/adapter/flownode_impl.rs

@@ -133,7 +133,7 @@ impl Flownode for FlowWorkerManager {
                     .map_err(to_meta_err(snafu::location!()))?;
 
                 debug!(
-                    "Done to flush flow_id={:?} with {} input rows flushed, {} rows sended and {} output rows flushed",
+                    "Done to flush flow_id={:?} with {} input rows flushed, {} rows sent and {} output rows flushed",
                     flow_id, flushed_input_rows, rows_send, row
                 );
                 Ok(FlowResponse {
@@ -153,10 +153,7 @@ impl Flownode for FlowWorkerManager {
         }
     }
 
-    #[allow(unreachable_code, unused)]
     async fn handle_inserts(&self, request: InsertRequests) -> Result<FlowResponse> {
-        return Ok(Default::default());
-
         // using try_read to ensure two things:
         // 1. flush wouldn't happen until inserts before it is inserted
         // 2. inserts happening concurrently with flush wouldn't be block by flush

src/flow/src/error.rs

@@ -16,7 +16,6 @@
 
 use std::any::Any;
 
-use arrow_schema::ArrowError;
 use common_error::ext::BoxedError;
 use common_error::{define_into_tonic_status, from_err_code_msg_to_header};
 use common_macro::stack_trace_debug;
@@ -157,15 +156,6 @@ pub enum Error {
         location: Location,
     },
 
-    #[snafu(display("Arrow error: {raw:?} in context: {context}"))]
-    Arrow {
-        #[snafu(source)]
-        raw: ArrowError,
-        context: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
     #[snafu(display("Datafusion error: {raw:?} in context: {context}"))]
     Datafusion {
         #[snafu(source)]
@@ -240,7 +230,6 @@ impl ErrorExt for Error {
         match self {
             Self::Eval { .. }
             | Self::JoinTask { .. }
-            | Self::Arrow { .. }
             | Self::Datafusion { .. }
             | Self::InsertIntoFlow { .. } => StatusCode::Internal,
             Self::FlowAlreadyExist { .. } => StatusCode::TableAlreadyExists,

src/flow/src/expr/utils.rs

@@ -238,7 +238,6 @@ mod test {
 
         for (sql, current, expected) in &testcases {
             let plan = sql_to_substrait(engine.clone(), sql).await;
-
             let mut ctx = create_test_ctx();
             let flow_plan = TypedPlan::from_substrait_plan(&mut ctx, &plan)
                 .await

src/flow/src/heartbeat.rs

@@ -14,6 +14,7 @@
 
 //! Send heartbeat from flownode to metasrv
 
+use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::Arc;
 
 use api::v1::meta::{HeartbeatRequest, Peer};
@@ -24,7 +25,7 @@ use common_meta::heartbeat::handler::{
 use common_meta::heartbeat::mailbox::{HeartbeatMailbox, MailboxRef, OutgoingMessage};
 use common_meta::heartbeat::utils::outgoing_message_to_mailbox_message;
 use common_meta::key::flow::flow_state::FlowStat;
-use common_telemetry::{debug, error, info};
+use common_telemetry::{debug, error, info, warn};
 use greptime_proto::v1::meta::NodeInfo;
 use meta_client::client::{HeartbeatSender, HeartbeatStream, MetaClient};
 use servers::addrs;
@@ -65,6 +66,7 @@ pub struct HeartbeatTask {
     report_interval: Duration,
     retry_interval: Duration,
     resp_handler_executor: HeartbeatResponseHandlerExecutorRef,
+    running: Arc<AtomicBool>,
     query_stat_size: Option<SizeReportSender>,
 }
 
@@ -87,11 +89,25 @@ impl HeartbeatTask {
             report_interval: heartbeat_opts.interval,
             retry_interval: heartbeat_opts.retry_interval,
             resp_handler_executor,
+            running: Arc::new(AtomicBool::new(false)),
             query_stat_size: None,
         }
     }
 
     pub async fn start(&self) -> Result<(), Error> {
+        if self
+            .running
+            .compare_exchange(false, true, Ordering::AcqRel, Ordering::Acquire)
+            .is_err()
+        {
+            warn!("Heartbeat task started multiple times");
+            return Ok(());
+        }
+
+        self.create_streams().await
+    }
+
+    async fn create_streams(&self) -> Result<(), Error> {
         info!("Start to establish the heartbeat connection to metasrv.");
         let (req_sender, resp_stream) = self
             .meta_client
@@ -114,6 +130,13 @@ impl HeartbeatTask {
 
     pub fn shutdown(&self) {
         info!("Close heartbeat task for flownode");
+        if self
+            .running
+            .compare_exchange(true, false, Ordering::AcqRel, Ordering::Acquire)
+            .is_err()
+        {
+            warn!("Call close heartbeat task multiple times");
+        }
     }
 
     fn new_heartbeat_request(
@@ -191,6 +214,7 @@ impl HeartbeatTask {
                         if let Some(message) = message {
                             Self::new_heartbeat_request(&heartbeat_request, Some(message), &latest_report)
                         } else {
+                            warn!("Sender has been dropped, exiting the heartbeat loop");
                             // Receives None that means Sender was dropped, we need to break the current loop
                             break
                         }
@@ -232,7 +256,11 @@ impl HeartbeatTask {
                             error!(e; "Error while handling heartbeat response");
                         }
                     }
-                    Ok(None) => break,
+                    Ok(None) => {
+                        warn!("Heartbeat response stream closed");
+                        capture_self.start_with_retry(retry_interval).await;
+                        break;
+                    }
                     Err(e) => {
                         error!(e; "Occur error while reading heartbeat response");
                         capture_self.start_with_retry(retry_interval).await;
@@ -258,7 +286,7 @@ impl HeartbeatTask {
 
             info!("Try to re-establish the heartbeat connection to metasrv.");
 
-            if self.start().await.is_ok() {
+            if self.create_streams().await.is_ok() {
                 break;
             }
         }

src/flow/src/lib.rs

@@ -33,7 +33,6 @@ mod expr;
 pub mod heartbeat;
 mod metrics;
 mod plan;
-mod recording_rules;
 mod repr;
 mod server;
 mod transform;
@@ -44,5 +43,4 @@ mod test_utils;
 
 pub use adapter::{FlowConfig, FlowWorkerManager, FlowWorkerManagerRef, FlownodeOptions};
 pub use error::{Error, Result};
-pub use recording_rules::FrontendClient;
 pub use server::{FlownodeBuilder, FlownodeInstance, FlownodeServer, FrontendInvoker};

src/flow/src/metrics.rs

@@ -28,32 +28,6 @@ lazy_static! {
         &["table_id"]
     )
     .unwrap();
-    pub static ref METRIC_FLOW_RULE_ENGINE_QUERY_TIME: HistogramVec = register_histogram_vec!(
-        "greptime_flow_rule_engine_query_time",
-        "flow rule engine query time",
-        &["flow_id"],
-        vec![
-            0.0,
-            1.,
-            3.,
-            5.,
-            10.,
-            20.,
-            30.,
-            60.,
-            2. * 60.,
-            5. * 60.,
-            10. * 60.
-        ]
-    )
-    .unwrap();
-    pub static ref METRIC_FLOW_RULE_ENGINE_SLOW_QUERY: HistogramVec = register_histogram_vec!(
-        "greptime_flow_rule_engine_slow_query",
-        "flow rule engine slow query",
-        &["flow_id", "sql", "peer"],
-        vec![60., 2. * 60., 3. * 60., 5. * 60., 10. * 60.]
-    )
-    .unwrap();
     pub static ref METRIC_FLOW_RUN_INTERVAL_MS: IntGauge =
         register_int_gauge!("greptime_flow_run_interval_ms", "flow run interval in ms").unwrap();
     pub static ref METRIC_FLOW_ROWS: IntCounterVec = register_int_counter_vec!(

src/flow/src/recording_rules.rs

@@ -1,744 +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.
-
-//! Run flow as recording rule which is time-window-aware normal query triggered every tick set by user
-
-mod engine;
-mod frontend_client;
-
-use std::collections::HashSet;
-use std::sync::Arc;
-
-use common_error::ext::BoxedError;
-use common_recordbatch::DfRecordBatch;
-use common_time::timestamp::TimeUnit;
-use common_time::Timestamp;
-use datafusion::error::Result as DfResult;
-use datafusion::logical_expr::Expr;
-use datafusion::physical_planner::{DefaultPhysicalPlanner, PhysicalPlanner};
-use datafusion::prelude::SessionContext;
-use datafusion::sql::unparser::Unparser;
-use datafusion_common::tree_node::{Transformed, TreeNode, TreeNodeRecursion, TreeNodeRewriter};
-use datafusion_common::{Column, DFSchema, TableReference};
-use datafusion_expr::LogicalPlan;
-use datafusion_physical_expr::PhysicalExprRef;
-use datatypes::prelude::{ConcreteDataType, DataType};
-use datatypes::value::Value;
-use datatypes::vectors::{
-    TimestampMicrosecondVector, TimestampMillisecondVector, TimestampNanosecondVector,
-    TimestampSecondVector, Vector,
-};
-pub use engine::RecordingRuleEngine;
-pub use frontend_client::FrontendClient;
-use query::parser::QueryLanguageParser;
-use query::QueryEngineRef;
-use session::context::QueryContextRef;
-use snafu::{ensure, OptionExt, ResultExt};
-
-use crate::df_optimizer::apply_df_optimizer;
-use crate::error::{ArrowSnafu, DatafusionSnafu, DatatypesSnafu, ExternalSnafu, UnexpectedSnafu};
-use crate::Error;
-
-/// Convert sql to datafusion logical plan
-pub async fn sql_to_df_plan(
-    query_ctx: QueryContextRef,
-    engine: QueryEngineRef,
-    sql: &str,
-    optimize: bool,
-) -> Result<LogicalPlan, Error> {
-    let stmt = QueryLanguageParser::parse_sql(sql, &query_ctx)
-        .map_err(BoxedError::new)
-        .context(ExternalSnafu)?;
-    let plan = engine
-        .planner()
-        .plan(&stmt, query_ctx)
-        .await
-        .map_err(BoxedError::new)
-        .context(ExternalSnafu)?;
-    let plan = if optimize {
-        apply_df_optimizer(plan).await?
-    } else {
-        plan
-    };
-    Ok(plan)
-}
-
-/// Find nearest lower bound for time `current` in given `plan` for the time window expr.
-/// i.e. for time window expr being `date_bin(INTERVAL '5 minutes', ts) as time_window` and `current="2021-07-01 00:01:01.000"`,
-/// return `Some("2021-07-01 00:00:00.000")`
-/// if `plan` doesn't contain a `TIME INDEX` column, return `None`
-///
-/// Time window expr is a expr that:
-/// 1. ref only to a time index column
-/// 2. is monotonic increasing
-/// 3. show up in GROUP BY clause
-///
-/// note this plan should only contain one TableScan
-pub async fn find_plan_time_window_bound(
-    plan: &LogicalPlan,
-    current: Timestamp,
-    query_ctx: QueryContextRef,
-    engine: QueryEngineRef,
-) -> Result<(String, Option<Timestamp>, Option<Timestamp>), Error> {
-    // TODO(discord9): find the expr that do time window
-    let catalog_man = engine.engine_state().catalog_manager();
-
-    let mut table_name = None;
-    // first find the table source in the logical plan
-    plan.apply(|plan| {
-        let LogicalPlan::TableScan(table_scan) = plan else {
-            return Ok(TreeNodeRecursion::Continue);
-        };
-        table_name = Some(table_scan.table_name.clone());
-        Ok(TreeNodeRecursion::Stop)
-    })
-    .with_context(|_| DatafusionSnafu {
-        context: format!("Can't find table source in plan {plan:?}"),
-    })?;
-    let Some(table_name) = table_name else {
-        UnexpectedSnafu {
-            reason: format!("Can't find table source in plan {plan:?}"),
-        }
-        .fail()?
-    };
-
-    let current_schema = query_ctx.current_schema();
-
-    let catalog_name = table_name.catalog().unwrap_or(query_ctx.current_catalog());
-    let schema_name = table_name.schema().unwrap_or(&current_schema);
-    let table_name = table_name.table();
-
-    let Some(table_ref) = catalog_man
-        .table(catalog_name, schema_name, table_name, Some(&query_ctx))
-        .await
-        .map_err(BoxedError::new)
-        .context(ExternalSnafu)?
-    else {
-        UnexpectedSnafu {
-            reason: format!(
-                "Can't find table {table_name:?} in catalog {catalog_name:?}/{schema_name:?}"
-            ),
-        }
-        .fail()?
-    };
-
-    let schema = &table_ref.table_info().meta.schema;
-
-    let ts_index = schema.timestamp_column().context(UnexpectedSnafu {
-        reason: format!("Can't find timestamp column in table {table_name:?}"),
-    })?;
-
-    let ts_col_name = ts_index.name.clone();
-
-    let expected_time_unit = ts_index.data_type.as_timestamp().with_context(|| UnexpectedSnafu {
-        reason: format!(
-            "Expected timestamp column {ts_col_name:?} in table {table_name:?} to be timestamp, but got {ts_index:?}"
-        ),
-    })?.unit();
-
-    let ts_columns: HashSet<_> = HashSet::from_iter(vec![
-        format!("{catalog_name}.{schema_name}.{table_name}.{ts_col_name}"),
-        format!("{schema_name}.{table_name}.{ts_col_name}"),
-        format!("{table_name}.{ts_col_name}"),
-        format!("{ts_col_name}"),
-    ]);
-    let ts_columns: HashSet<_> = ts_columns
-        .into_iter()
-        .map(Column::from_qualified_name)
-        .collect();
-
-    let ts_columns_ref: HashSet<&Column> = ts_columns.iter().collect();
-
-    // find the time window expr which refers to the time index column
-    let mut time_window_expr: Option<Expr> = None;
-    let find_time_window_expr = |plan: &LogicalPlan| {
-        let LogicalPlan::Aggregate(aggregate) = plan else {
-            return Ok(TreeNodeRecursion::Continue);
-        };
-
-        for group_expr in &aggregate.group_expr {
-            let refs = group_expr.column_refs();
-            if refs.len() != 1 {
-                continue;
-            }
-            let ref_col = refs.iter().next().unwrap();
-            if ts_columns_ref.contains(ref_col) {
-                time_window_expr = Some(group_expr.clone());
-                break;
-            }
-        }
-
-        Ok(TreeNodeRecursion::Stop)
-    };
-    plan.apply(find_time_window_expr)
-        .with_context(|_| DatafusionSnafu {
-            context: format!("Can't find time window expr in plan {plan:?}"),
-        })?;
-
-    let arrow_schema = Arc::new(arrow_schema::Schema::new(vec![arrow_schema::Field::new(
-        ts_col_name.clone(),
-        ts_index.data_type.as_arrow_type(),
-        false,
-    )]));
-
-    let df_schema = DFSchema::from_field_specific_qualified_schema(
-        vec![Some(TableReference::bare(table_name))],
-        &arrow_schema,
-    )
-    .with_context(|_e| DatafusionSnafu {
-        context: format!("Failed to create DFSchema from arrow schema {arrow_schema:?}"),
-    })?;
-
-    // cast current to ts_index's type
-    let new_current = current
-        .convert_to(expected_time_unit)
-        .with_context(|| UnexpectedSnafu {
-            reason: format!("Failed to cast current timestamp {current:?} to {expected_time_unit}"),
-        })?;
-
-    // if no time_window_expr is found, return None
-    if let Some(time_window_expr) = time_window_expr {
-        let lower_bound =
-            find_expr_time_window_lower_bound(&time_window_expr, &df_schema, new_current)?;
-        let upper_bound =
-            find_expr_time_window_upper_bound(&time_window_expr, &df_schema, new_current)?;
-        Ok((ts_col_name, lower_bound, upper_bound))
-    } else {
-        Ok((ts_col_name, None, None))
-    }
-}
-
-/// Find the lower bound of time window in given `expr` and `current` timestamp.
-///
-/// i.e. for `current="2021-07-01 00:01:01.000"` and `expr=date_bin(INTERVAL '5 minutes', ts) as time_window` and `ts_col=ts`,
-/// return `Some("2021-07-01 00:00:00.000")` since it's the lower bound
-/// of current time window given the current timestamp
-///
-/// if return None, meaning this time window have no lower bound
-fn find_expr_time_window_lower_bound(
-    expr: &Expr,
-    df_schema: &DFSchema,
-    current: Timestamp,
-) -> Result<Option<Timestamp>, Error> {
-    use std::cmp::Ordering;
-
-    let phy_planner = DefaultPhysicalPlanner::default();
-
-    let phy_expr: PhysicalExprRef = phy_planner
-        .create_physical_expr(expr, df_schema, &SessionContext::new().state())
-        .with_context(|_e| DatafusionSnafu {
-            context: format!(
-                "Failed to create physical expression from {expr:?} using {df_schema:?}"
-            ),
-        })?;
-
-    let cur_time_window = eval_ts_to_ts(&phy_expr, df_schema, current)?;
-    if cur_time_window == current {
-        return Ok(Some(current));
-    }
-
-    // search to find the lower bound
-    let mut offset: i64 = 1;
-    let lower_bound;
-    let mut upper_bound = Some(current);
-    // first expontial probe to found a range for binary search
-    loop {
-        let Some(next_val) = current.value().checked_sub(offset) else {
-            // no lower bound
-            return Ok(None);
-        };
-
-        let prev_time_probe = common_time::Timestamp::new(next_val, current.unit());
-
-        let prev_time_window = eval_ts_to_ts(&phy_expr, df_schema, prev_time_probe)?;
-
-        match prev_time_window.cmp(&cur_time_window) {
-            Ordering::Less => {
-                lower_bound = Some(prev_time_probe);
-                break;
-            }
-            Ordering::Equal => {
-                upper_bound = Some(prev_time_probe);
-            }
-            Ordering::Greater => {
-                UnexpectedSnafu {
-                    reason: format!(
-                        "Unsupported time window expression, expect monotonic increasing for time window expression {expr:?}"
-                    ),
-                }
-                .fail()?
-            }
-        }
-
-        let Some(new_offset) = offset.checked_mul(2) else {
-            // no lower bound
-            return Ok(None);
-        };
-        offset = new_offset;
-    }
-
-    // binary search for the exact lower bound
-
-    ensure!(lower_bound.map(|v|v.unit())==upper_bound.map(|v|v.unit()), UnexpectedSnafu{
-        reason: format!(" unit mismatch for time window expression {expr:?}, found {lower_bound:?} and {upper_bound:?}"),
-    });
-
-    let input_time_unit = lower_bound
-        .context(UnexpectedSnafu {
-            reason: "should have lower bound",
-        })?
-        .unit();
-
-    let mut low = lower_bound
-        .context(UnexpectedSnafu {
-            reason: "should have lower bound",
-        })?
-        .value();
-    let mut high = upper_bound
-        .context(UnexpectedSnafu {
-            reason: "should have upper bound",
-        })?
-        .value();
-
-    while low < high {
-        let mid = (low + high) / 2;
-        let mid_probe = common_time::Timestamp::new(mid, input_time_unit);
-        let mid_time_window = eval_ts_to_ts(&phy_expr, df_schema, mid_probe)?;
-
-        match mid_time_window.cmp(&cur_time_window) {
-            Ordering::Less => low = mid + 1,
-            Ordering::Equal => high = mid,
-            Ordering::Greater => UnexpectedSnafu {
-                reason: format!("Binary search failed for time window expression {expr:?}"),
-            }
-            .fail()?,
-        }
-    }
-
-    let final_lower_bound_for_time_window = common_time::Timestamp::new(low, input_time_unit);
-
-    Ok(Some(final_lower_bound_for_time_window))
-}
-
-/// Find the upper bound for time window expression
-fn find_expr_time_window_upper_bound(
-    expr: &Expr,
-    df_schema: &DFSchema,
-    current: Timestamp,
-) -> Result<Option<Timestamp>, Error> {
-    use std::cmp::Ordering;
-
-    let phy_planner = DefaultPhysicalPlanner::default();
-
-    let phy_expr: PhysicalExprRef = phy_planner
-        .create_physical_expr(expr, df_schema, &SessionContext::new().state())
-        .with_context(|_e| DatafusionSnafu {
-            context: format!(
-                "Failed to create physical expression from {expr:?} using {df_schema:?}"
-            ),
-        })?;
-
-    let cur_time_window = eval_ts_to_ts(&phy_expr, df_schema, current)?;
-
-    // search to find the lower bound
-    let mut offset: i64 = 1;
-    let mut lower_bound = Some(current);
-    let upper_bound;
-    // first expontial probe to found a range for binary search
-    loop {
-        let Some(next_val) = current.value().checked_add(offset) else {
-            // no upper bound if overflow
-            return Ok(None);
-        };
-
-        let next_time_probe = common_time::Timestamp::new(next_val, current.unit());
-
-        let next_time_window = eval_ts_to_ts(&phy_expr, df_schema, next_time_probe)?;
-
-        match next_time_window.cmp(&cur_time_window) {
-            Ordering::Less => {UnexpectedSnafu {
-                reason: format!(
-                    "Unsupported time window expression, expect monotonic increasing for time window expression {expr:?}"
-                ),
-            }
-            .fail()?
-            }
-            Ordering::Equal => {
-                lower_bound = Some(next_time_probe);
-            }
-            Ordering::Greater => {
-                upper_bound = Some(next_time_probe);
-                break
-            }
-        }
-
-        let Some(new_offset) = offset.checked_mul(2) else {
-            // no upper bound if overflow
-            return Ok(None);
-        };
-        offset = new_offset;
-    }
-
-    // binary search for the exact upper bound
-
-    ensure!(lower_bound.map(|v|v.unit())==upper_bound.map(|v|v.unit()), UnexpectedSnafu{
-        reason: format!(" unit mismatch for time window expression {expr:?}, found {lower_bound:?} and {upper_bound:?}"),
-    });
-
-    let output_unit = upper_bound
-        .context(UnexpectedSnafu {
-            reason: "should have lower bound",
-        })?
-        .unit();
-
-    let mut low = lower_bound
-        .context(UnexpectedSnafu {
-            reason: "should have lower bound",
-        })?
-        .value();
-    let mut high = upper_bound
-        .context(UnexpectedSnafu {
-            reason: "should have upper bound",
-        })?
-        .value();
-    while low < high {
-        let mid = (low + high) / 2;
-        let mid_probe = common_time::Timestamp::new(mid, output_unit);
-        let mid_time_window = eval_ts_to_ts(&phy_expr, df_schema, mid_probe)?;
-
-        match mid_time_window.cmp(&cur_time_window) {
-            Ordering::Less => UnexpectedSnafu {
-                reason: format!("Binary search failed for time window expression {expr:?}"),
-            }
-            .fail()?,
-            Ordering::Equal => low = mid + 1,
-            Ordering::Greater => high = mid,
-        }
-    }
-
-    let final_upper_bound_for_time_window = common_time::Timestamp::new(high, output_unit);
-
-    Ok(Some(final_upper_bound_for_time_window))
-}
-
-fn eval_ts_to_ts(
-    phy: &PhysicalExprRef,
-    df_schema: &DFSchema,
-    input_value: Timestamp,
-) -> Result<Timestamp, Error> {
-    let ts_vector = match input_value.unit() {
-        TimeUnit::Second => {
-            TimestampSecondVector::from_vec(vec![input_value.value()]).to_arrow_array()
-        }
-        TimeUnit::Millisecond => {
-            TimestampMillisecondVector::from_vec(vec![input_value.value()]).to_arrow_array()
-        }
-        TimeUnit::Microsecond => {
-            TimestampMicrosecondVector::from_vec(vec![input_value.value()]).to_arrow_array()
-        }
-        TimeUnit::Nanosecond => {
-            TimestampNanosecondVector::from_vec(vec![input_value.value()]).to_arrow_array()
-        }
-    };
-
-    let rb = DfRecordBatch::try_new(df_schema.inner().clone(), vec![ts_vector.clone()])
-        .with_context(|_| ArrowSnafu {
-            context: format!("Failed to create record batch from {df_schema:?} and {ts_vector:?}"),
-        })?;
-
-    let eval_res = phy.evaluate(&rb).with_context(|_| DatafusionSnafu {
-        context: format!("Failed to evaluate physical expression {phy:?} on {rb:?}"),
-    })?;
-
-    let val = match eval_res {
-        datafusion_expr::ColumnarValue::Array(array) => {
-            let ty = array.data_type();
-            let ty = ConcreteDataType::from_arrow_type(ty);
-            let time_unit = if let ConcreteDataType::Timestamp(ty) = ty {
-                ty.unit()
-            } else {
-                return UnexpectedSnafu {
-                    reason: format!("Physical expression {phy:?} evaluated to non-timestamp type"),
-                }
-                .fail();
-            };
-
-            match time_unit {
-                TimeUnit::Second => TimestampSecondVector::try_from_arrow_array(array.clone())
-                    .with_context(|_| DatatypesSnafu {
-                        extra: format!("Failed to create vector from arrow array {array:?}"),
-                    })?
-                    .get(0),
-                TimeUnit::Millisecond => {
-                    TimestampMillisecondVector::try_from_arrow_array(array.clone())
-                        .with_context(|_| DatatypesSnafu {
-                            extra: format!("Failed to create vector from arrow array {array:?}"),
-                        })?
-                        .get(0)
-                }
-                TimeUnit::Microsecond => {
-                    TimestampMicrosecondVector::try_from_arrow_array(array.clone())
-                        .with_context(|_| DatatypesSnafu {
-                            extra: format!("Failed to create vector from arrow array {array:?}"),
-                        })?
-                        .get(0)
-                }
-                TimeUnit::Nanosecond => {
-                    TimestampNanosecondVector::try_from_arrow_array(array.clone())
-                        .with_context(|_| DatatypesSnafu {
-                            extra: format!("Failed to create vector from arrow array {array:?}"),
-                        })?
-                        .get(0)
-                }
-            }
-        }
-        datafusion_expr::ColumnarValue::Scalar(scalar) => Value::try_from(scalar.clone())
-            .with_context(|_| DatatypesSnafu {
-                extra: format!("Failed to convert scalar {scalar:?} to value"),
-            })?,
-    };
-
-    if let Value::Timestamp(ts) = val {
-        Ok(ts)
-    } else {
-        UnexpectedSnafu {
-            reason: format!("Expected timestamp in expression {phy:?} but got {val:?}"),
-        }
-        .fail()?
-    }
-}
-
-// TODO(discord9): a method to found out the precise time window
-
-/// Find out the `Filter` Node corresponding to outermost `WHERE` and add a new filter expr to it
-#[derive(Debug)]
-pub struct AddFilterRewriter {
-    extra_filter: Expr,
-    is_rewritten: bool,
-}
-
-impl AddFilterRewriter {
-    fn new(filter: Expr) -> Self {
-        Self {
-            extra_filter: filter,
-            is_rewritten: false,
-        }
-    }
-}
-
-impl TreeNodeRewriter for AddFilterRewriter {
-    type Node = LogicalPlan;
-    fn f_down(&mut self, node: Self::Node) -> DfResult<Transformed<Self::Node>> {
-        if self.is_rewritten {
-            return Ok(Transformed::no(node));
-        }
-        match node {
-            LogicalPlan::Filter(mut filter) if !filter.having => {
-                filter.predicate = filter.predicate.and(self.extra_filter.clone());
-                self.is_rewritten = true;
-                Ok(Transformed::yes(LogicalPlan::Filter(filter)))
-            }
-            LogicalPlan::TableScan(_) => {
-                // add a new filter
-                let filter =
-                    datafusion_expr::Filter::try_new(self.extra_filter.clone(), Arc::new(node))?;
-                self.is_rewritten = true;
-                Ok(Transformed::yes(LogicalPlan::Filter(filter)))
-            }
-            _ => Ok(Transformed::no(node)),
-        }
-    }
-}
-
-fn df_plan_to_sql(plan: &LogicalPlan) -> Result<String, Error> {
-    let unparser = Unparser::default();
-    let sql = unparser
-        .plan_to_sql(plan)
-        .with_context(|_e| DatafusionSnafu {
-            context: format!("Failed to unparse logical plan {plan:?}"),
-        })?;
-    Ok(sql.to_string())
-}
-
-#[cfg(test)]
-mod test {
-    use datafusion_common::tree_node::TreeNode;
-    use pretty_assertions::assert_eq;
-    use session::context::QueryContext;
-
-    use super::{sql_to_df_plan, *};
-    use crate::recording_rules::{df_plan_to_sql, AddFilterRewriter};
-    use crate::test_utils::create_test_query_engine;
-
-    #[tokio::test]
-    async fn test_add_filter() {
-        let testcases = vec![
-            (
-                "SELECT number FROM numbers_with_ts GROUP BY number","SELECT numbers_with_ts.number FROM numbers_with_ts WHERE (number > 4) GROUP BY numbers_with_ts.number"
-            ),
-            (
-                "SELECT number FROM numbers_with_ts WHERE number < 2 OR number >10",
-                "SELECT numbers_with_ts.number FROM numbers_with_ts WHERE (((numbers_with_ts.number < 2) OR (numbers_with_ts.number > 10)) AND (number > 4))"
-            ),
-            (
-                "SELECT date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window",
-                "SELECT date_bin('5 minutes', numbers_with_ts.ts) AS time_window FROM numbers_with_ts WHERE (number > 4) GROUP BY date_bin('5 minutes', numbers_with_ts.ts)"
-            )
-        ];
-        use datafusion_expr::{col, lit};
-        let query_engine = create_test_query_engine();
-        let ctx = QueryContext::arc();
-
-        for (before, after) in testcases {
-            let sql = before;
-            let plan = sql_to_df_plan(ctx.clone(), query_engine.clone(), sql, false)
-                .await
-                .unwrap();
-
-            let mut add_filter = AddFilterRewriter::new(col("number").gt(lit(4u32)));
-            let plan = plan.rewrite(&mut add_filter).unwrap().data;
-            let new_sql = df_plan_to_sql(&plan).unwrap();
-            assert_eq!(after, new_sql);
-        }
-    }
-
-    #[tokio::test]
-    async fn test_plan_time_window_lower_bound() {
-        use datafusion_expr::{col, lit};
-        let query_engine = create_test_query_engine();
-        let ctx = QueryContext::arc();
-
-        let testcases = [
-            // same alias is not same column
-            (
-                "SELECT arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)') AS ts FROM numbers_with_ts GROUP BY ts;",
-                Timestamp::new(1740394109, TimeUnit::Second),
-                (
-                    "ts".to_string(),
-                    Some(Timestamp::new(1740394109000, TimeUnit::Millisecond)),
-                    Some(Timestamp::new(1740394109001, TimeUnit::Millisecond)),
-                ),
-                "SELECT arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)') AS ts FROM numbers_with_ts WHERE ((ts >= CAST('2025-02-24 10:48:29' AS TIMESTAMP)) AND (ts <= CAST('2025-02-24 10:48:29.001' AS TIMESTAMP))) GROUP BY numbers_with_ts.ts"
-            ),
-            // complex time window index
-            (
-                "SELECT arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)') AS time_window FROM numbers_with_ts 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 ((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;",
-                Timestamp::new(23, TimeUnit::Millisecond),
-                ("ts".to_string(), None, None),
-                "SELECT date_bin('5 minutes', ts) FROM numbers_with_ts;"
-            ),
-            // time index
-            (
-                "SELECT date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window;",
-                Timestamp::new(23, TimeUnit::Nanosecond),
-                (
-                    "ts".to_string(),
-                    Some(Timestamp::new(0, TimeUnit::Millisecond)),
-                    Some(Timestamp::new(300000, TimeUnit::Millisecond)),
-                ),
-                "SELECT date_bin('5 minutes', numbers_with_ts.ts) AS time_window FROM numbers_with_ts WHERE ((ts >= CAST('1970-01-01 00:00:00' AS TIMESTAMP)) AND (ts <= CAST('1970-01-01 00:05:00' AS TIMESTAMP))) GROUP BY date_bin('5 minutes', numbers_with_ts.ts)"
-            ),
-            // on spot
-            (
-                "SELECT date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window;",
-                Timestamp::new(0, TimeUnit::Nanosecond),
-                (
-                    "ts".to_string(),
-                    Some(Timestamp::new(0, TimeUnit::Millisecond)),
-                    Some(Timestamp::new(300000, TimeUnit::Millisecond)),
-                ),
-                "SELECT date_bin('5 minutes', numbers_with_ts.ts) AS time_window FROM numbers_with_ts WHERE ((ts >= CAST('1970-01-01 00:00:00' AS TIMESTAMP)) AND (ts <= CAST('1970-01-01 00:05:00' AS TIMESTAMP))) GROUP BY date_bin('5 minutes', numbers_with_ts.ts)"
-            ),
-            // different time unit
-            (
-                "SELECT date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window;",
-                Timestamp::new(23_000_000, TimeUnit::Nanosecond),
-                (
-                    "ts".to_string(),
-                    Some(Timestamp::new(0, TimeUnit::Millisecond)),
-                    Some(Timestamp::new(300000, TimeUnit::Millisecond)),
-                ),
-                "SELECT date_bin('5 minutes', numbers_with_ts.ts) AS time_window FROM numbers_with_ts WHERE ((ts >= CAST('1970-01-01 00:00:00' AS TIMESTAMP)) AND (ts <= CAST('1970-01-01 00:05:00' AS TIMESTAMP))) GROUP BY date_bin('5 minutes', numbers_with_ts.ts)"
-            ),
-            // time index with other fields
-            (
-                "SELECT sum(number) as sum_up, date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window;",
-                Timestamp::new(23, TimeUnit::Millisecond),
-                (
-                    "ts".to_string(),
-                    Some(Timestamp::new(0, TimeUnit::Millisecond)),
-                    Some(Timestamp::new(300000, TimeUnit::Millisecond)),
-                ),
-                "SELECT sum(numbers_with_ts.number) AS sum_up, date_bin('5 minutes', numbers_with_ts.ts) AS time_window FROM numbers_with_ts WHERE ((ts >= CAST('1970-01-01 00:00:00' AS TIMESTAMP)) AND (ts <= CAST('1970-01-01 00:05:00' AS TIMESTAMP))) GROUP BY date_bin('5 minutes', numbers_with_ts.ts)"
-            ),
-            // time index with other pks
-            (
-                "SELECT number, date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window, number;",
-                Timestamp::new(23, TimeUnit::Millisecond),
-                (
-                    "ts".to_string(),
-                    Some(Timestamp::new(0, TimeUnit::Millisecond)),
-                    Some(Timestamp::new(300000, TimeUnit::Millisecond)),
-                ),
-                "SELECT numbers_with_ts.number, date_bin('5 minutes', numbers_with_ts.ts) AS time_window FROM numbers_with_ts WHERE ((ts >= CAST('1970-01-01 00:00:00' AS TIMESTAMP)) AND (ts <= CAST('1970-01-01 00:05:00' AS TIMESTAMP))) GROUP BY date_bin('5 minutes', numbers_with_ts.ts), numbers_with_ts.number"
-            ),
-        ];
-
-        for (sql, current, expected, unparsed) in testcases {
-            let plan = sql_to_df_plan(ctx.clone(), query_engine.clone(), sql, true)
-                .await
-                .unwrap();
-
-            let real =
-                find_plan_time_window_bound(&plan, current, ctx.clone(), query_engine.clone())
-                    .await
-                    .unwrap();
-            assert_eq!(expected, real);
-
-            let plan = sql_to_df_plan(ctx.clone(), query_engine.clone(), sql, false)
-                .await
-                .unwrap();
-            let (col_name, lower, upper) = real;
-            let new_sql = if lower.is_some() {
-                let to_df_literal = |value| {
-                    let value = Value::from(value);
-
-                    value.try_to_scalar_value(&value.data_type()).unwrap()
-                };
-                let lower = to_df_literal(lower.unwrap());
-                let upper = to_df_literal(upper.unwrap());
-                let expr = col(&col_name)
-                    .gt_eq(lit(lower))
-                    .and(col(&col_name).lt_eq(lit(upper)));
-                let mut add_filter = AddFilterRewriter::new(expr);
-                let plan = plan.rewrite(&mut add_filter).unwrap().data;
-                df_plan_to_sql(&plan).unwrap()
-            } else {
-                sql.to_string()
-            };
-            assert_eq!(unparsed, new_sql);
-        }
-    }
-}

src/flow/src/recording_rules/engine.rs

@@ -1,407 +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::collections::BTreeMap;
-use std::sync::Arc;
-use std::time::{Duration, SystemTime, UNIX_EPOCH};
-
-use common_meta::ddl::create_flow::FlowType;
-use common_telemetry::tracing::warn;
-use common_telemetry::{debug, info};
-use common_time::Timestamp;
-use datafusion_common::tree_node::TreeNode;
-use datatypes::value::Value;
-use query::QueryEngineRef;
-use session::context::QueryContextRef;
-use snafu::{ensure, ResultExt};
-use tokio::sync::oneshot::error::TryRecvError;
-use tokio::sync::{oneshot, RwLock};
-use tokio::time::Instant;
-
-use super::frontend_client::FrontendClient;
-use super::{df_plan_to_sql, AddFilterRewriter};
-use crate::adapter::{CreateFlowArgs, FlowId};
-use crate::error::{DatafusionSnafu, DatatypesSnafu, FlowAlreadyExistSnafu, UnexpectedSnafu};
-use crate::metrics::{METRIC_FLOW_RULE_ENGINE_QUERY_TIME, METRIC_FLOW_RULE_ENGINE_SLOW_QUERY};
-use crate::recording_rules::{find_plan_time_window_bound, sql_to_df_plan};
-use crate::Error;
-
-/// TODO(discord9): make those constants configurable
-/// The default rule engine query timeout is 10 minutes
-pub const DEFAULT_RULE_ENGINE_QUERY_TIMEOUT: Duration = Duration::from_secs(10 * 60);
-
-/// will output a warn log for any query that runs for more that 1 minutes, and also every 1 minutes when that query is still running
-pub const SLOW_QUERY_THRESHOLD: Duration = Duration::from_secs(60);
-
-/// TODO(discord9): determine how to configure refresh rate
-pub struct RecordingRuleEngine {
-    tasks: RwLock<BTreeMap<FlowId, RecordingRuleTask>>,
-    shutdown_txs: RwLock<BTreeMap<FlowId, oneshot::Sender<()>>>,
-    frontend_client: Arc<FrontendClient>,
-    engine: QueryEngineRef,
-}
-
-impl RecordingRuleEngine {
-    pub fn new(frontend_client: Arc<FrontendClient>, engine: QueryEngineRef) -> Self {
-        Self {
-            tasks: Default::default(),
-            shutdown_txs: Default::default(),
-            frontend_client,
-            engine,
-        }
-    }
-}
-
-const MIN_REFRESH_DURATION: Duration = Duration::new(5, 0);
-
-impl RecordingRuleEngine {
-    pub async fn create_flow(&self, args: CreateFlowArgs) -> Result<Option<FlowId>, Error> {
-        let CreateFlowArgs {
-            flow_id,
-            sink_table_name,
-            source_table_ids: _,
-            create_if_not_exists,
-            or_replace,
-            expire_after,
-            comment: _,
-            sql,
-            flow_options,
-            query_ctx,
-        } = args;
-
-        // or replace logic
-        {
-            let is_exist = self.tasks.read().await.contains_key(&flow_id);
-            match (create_if_not_exists, or_replace, is_exist) {
-                // if replace, ignore that old flow exists
-                (_, true, true) => {
-                    info!("Replacing flow with id={}", flow_id);
-                }
-                (false, false, true) => FlowAlreadyExistSnafu { id: flow_id }.fail()?,
-                // already exists, and not replace, return None
-                (true, false, true) => {
-                    info!("Flow with id={} already exists, do nothing", flow_id);
-                    return Ok(None);
-                }
-
-                // continue as normal
-                (_, _, false) => (),
-            }
-        }
-
-        let flow_type = flow_options.get(FlowType::FLOW_TYPE_KEY);
-
-        ensure!(
-            flow_type == Some(&FlowType::RecordingRule.to_string()) || flow_type.is_none(),
-            UnexpectedSnafu {
-                reason: format!("Flow type is not RecordingRule nor None, got {flow_type:?}")
-            }
-        );
-
-        let Some(query_ctx) = query_ctx else {
-            UnexpectedSnafu {
-                reason: "Query context is None".to_string(),
-            }
-            .fail()?
-        };
-
-        let (tx, rx) = oneshot::channel();
-        let task = RecordingRuleTask::new(
-            flow_id,
-            &sql,
-            expire_after,
-            sink_table_name,
-            Arc::new(query_ctx),
-            rx,
-        );
-
-        let task_inner = task.clone();
-        let engine = self.engine.clone();
-        let frontend = self.frontend_client.clone();
-
-        // TODO(discord9): also save handle & use time wheel or what for better
-        let _handle = common_runtime::spawn_global(async move {
-            match task_inner.start_executing(engine, frontend).await {
-                Ok(()) => info!("Flow {} shutdown", task_inner.flow_id),
-                Err(err) => common_telemetry::error!(
-                    "Flow {} encounter unrecoverable error: {err:?}",
-                    task_inner.flow_id
-                ),
-            }
-        });
-
-        // TODO(discord9): deal with replace logic
-        let replaced_old_task_opt = self.tasks.write().await.insert(flow_id, task);
-        drop(replaced_old_task_opt);
-
-        self.shutdown_txs.write().await.insert(flow_id, tx);
-
-        Ok(Some(flow_id))
-    }
-
-    pub async fn remove_flow(&self, flow_id: FlowId) -> Result<(), Error> {
-        if self.tasks.write().await.remove(&flow_id).is_none() {
-            warn!("Flow {flow_id} not found in tasks")
-        }
-        let Some(tx) = self.shutdown_txs.write().await.remove(&flow_id) else {
-            UnexpectedSnafu {
-                reason: format!("Can't found shutdown tx for flow {flow_id}"),
-            }
-            .fail()?
-        };
-        if tx.send(()).is_err() {
-            warn!("Fail to shutdown flow {flow_id} due to receiver already dropped, maybe flow {flow_id} is already dropped?")
-        }
-        Ok(())
-    }
-}
-
-#[derive(Debug, Clone)]
-pub struct RecordingRuleTask {
-    flow_id: FlowId,
-    query: String,
-    /// in seconds
-    expire_after: Option<i64>,
-    sink_table_name: [String; 3],
-    state: Arc<RwLock<RecordingRuleState>>,
-}
-
-impl RecordingRuleTask {
-    pub fn new(
-        flow_id: FlowId,
-        query: &str,
-        expire_after: Option<i64>,
-        sink_table_name: [String; 3],
-        query_ctx: QueryContextRef,
-        shutdown_rx: oneshot::Receiver<()>,
-    ) -> Self {
-        Self {
-            flow_id,
-            query: query.to_string(),
-            expire_after,
-            sink_table_name,
-            state: Arc::new(RwLock::new(RecordingRuleState::new(query_ctx, shutdown_rx))),
-        }
-    }
-}
-impl RecordingRuleTask {
-    /// This should be called in a new tokio task
-    pub async fn start_executing(
-        &self,
-        engine: QueryEngineRef,
-        frontend_client: Arc<FrontendClient>,
-    ) -> Result<(), Error> {
-        // only first query don't need upper bound
-        let mut is_first = true;
-
-        loop {
-            // FIXME(discord9): test if need upper bound also works
-            let new_query = self
-                .gen_query_with_time_window(engine.clone(), false)
-                .await?;
-
-            let insert_into = format!(
-                "INSERT INTO {}.{}.{} {}",
-                self.sink_table_name[0],
-                self.sink_table_name[1],
-                self.sink_table_name[2],
-                new_query
-            );
-
-            if is_first {
-                is_first = false;
-            }
-
-            let instant = Instant::now();
-            let flow_id = self.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.expire_after, peer_addr, &insert_into
-            );
-
-            let timer = METRIC_FLOW_RULE_ENGINE_QUERY_TIME
-                .with_label_values(&[flow_id.to_string().as_str()])
-                .start_timer();
-
-            let res = db_client.database.sql(&insert_into).await;
-            drop(timer);
-
-            let elapsed = instant.elapsed();
-            if let Ok(res1) = &res {
-                debug!(
-                    "Flow {flow_id} executed, result: {res1:?}, elapsed: {:?}",
-                    elapsed
-                );
-            } else if let Err(res) = &res {
-                warn!(
-                    "Failed to execute Flow {flow_id} on frontend {}, result: {res:?}, elapsed: {:?} with query: {}",
-                    peer_addr, elapsed, &insert_into
-                );
-            }
-
-            // record slow query
-            if elapsed >= SLOW_QUERY_THRESHOLD {
-                warn!(
-                    "Flow {flow_id} on frontend {} executed for {:?} before complete, query: {}",
-                    peer_addr, elapsed, &insert_into
-                );
-                METRIC_FLOW_RULE_ENGINE_SLOW_QUERY
-                    .with_label_values(&[flow_id.to_string().as_str(), &insert_into, &peer_addr])
-                    .observe(elapsed.as_secs_f64());
-            }
-
-            self.state
-                .write()
-                .await
-                .after_query_exec(elapsed, res.is_ok());
-
-            let sleep_until = {
-                let mut state = self.state.write().await;
-                match state.shutdown_rx.try_recv() {
-                    Ok(()) => break Ok(()),
-                    Err(TryRecvError::Closed) => {
-                        warn!("Unexpected shutdown flow {flow_id}, shutdown anyway");
-                        break Ok(());
-                    }
-                    Err(TryRecvError::Empty) => (),
-                }
-                state.get_next_start_query_time(None)
-            };
-            tokio::time::sleep_until(sleep_until).await;
-        }
-    }
-
-    async fn gen_query_with_time_window(
-        &self,
-        engine: QueryEngineRef,
-        need_upper_bound: bool,
-    ) -> Result<String, Error> {
-        let query_ctx = self.state.read().await.query_ctx.clone();
-        let start = SystemTime::now();
-        let since_the_epoch = start
-            .duration_since(UNIX_EPOCH)
-            .expect("Time went backwards");
-        let low_bound = self
-            .expire_after
-            .map(|e| since_the_epoch.as_secs() - e as u64);
-
-        let Some(low_bound) = low_bound else {
-            return Ok(self.query.clone());
-        };
-
-        let low_bound = Timestamp::new_second(low_bound as i64);
-
-        let plan = sql_to_df_plan(query_ctx.clone(), engine.clone(), &self.query, true).await?;
-
-        let (col_name, lower, upper) =
-            find_plan_time_window_bound(&plan, low_bound, query_ctx.clone(), engine.clone())
-                .await?;
-
-        let new_sql = {
-            let to_df_literal = |value| -> Result<_, Error> {
-                let value = Value::from(value);
-                let value = value
-                    .try_to_scalar_value(&value.data_type())
-                    .with_context(|_| DatatypesSnafu {
-                        extra: format!("Failed to convert to scalar value: {}", value),
-                    })?;
-                Ok(value)
-            };
-            let lower = lower.map(to_df_literal).transpose()?;
-            let upper = upper.map(to_df_literal).transpose()?.and_then(|u| {
-                if need_upper_bound {
-                    Some(u)
-                } else {
-                    None
-                }
-            });
-            let expr = {
-                use datafusion_expr::{col, lit};
-                match (lower, upper) {
-                    (Some(l), Some(u)) => col(&col_name)
-                        .gt_eq(lit(l))
-                        .and(col(&col_name).lt_eq(lit(u))),
-                    (Some(l), None) => col(&col_name).gt_eq(lit(l)),
-                    (None, Some(u)) => col(&col_name).lt(lit(u)),
-                    // no time window, direct return
-                    (None, None) => return Ok(self.query.clone()),
-                }
-            };
-
-            let mut add_filter = AddFilterRewriter::new(expr);
-            // make a not optimized plan for clearer unparse
-            let plan =
-                sql_to_df_plan(query_ctx.clone(), engine.clone(), &self.query, false).await?;
-            let plan = plan
-                .clone()
-                .rewrite(&mut add_filter)
-                .with_context(|_| DatafusionSnafu {
-                    context: format!("Failed to rewrite plan {plan:?}"),
-                })?
-                .data;
-            df_plan_to_sql(&plan)?
-        };
-
-        Ok(new_sql)
-    }
-}
-
-#[derive(Debug)]
-pub struct RecordingRuleState {
-    query_ctx: QueryContextRef,
-    /// last query complete time
-    last_update_time: Instant,
-    /// last time query duration
-    last_query_duration: Duration,
-    exec_state: ExecState,
-    shutdown_rx: oneshot::Receiver<()>,
-}
-
-impl RecordingRuleState {
-    pub fn new(query_ctx: QueryContextRef, shutdown_rx: oneshot::Receiver<()>) -> Self {
-        Self {
-            query_ctx,
-            last_update_time: Instant::now(),
-            last_query_duration: Duration::from_secs(0),
-            exec_state: ExecState::Idle,
-            shutdown_rx,
-        }
-    }
-
-    /// called after last query is done
-    /// `is_succ` indicate whether the last query is successful
-    pub fn after_query_exec(&mut self, elapsed: Duration, _is_succ: bool) {
-        self.exec_state = ExecState::Idle;
-        self.last_query_duration = elapsed;
-        self.last_update_time = Instant::now();
-    }
-
-    /// wait for at least `last_query_duration`, at most `max_timeout` to start next query
-    pub fn get_next_start_query_time(&self, max_timeout: Option<Duration>) -> Instant {
-        let next_duration = max_timeout
-            .unwrap_or(self.last_query_duration)
-            .min(self.last_query_duration);
-        let next_duration = next_duration.max(MIN_REFRESH_DURATION);
-
-        self.last_update_time + next_duration
-    }
-}
-
-#[derive(Debug, Clone)]
-enum ExecState {
-    Idle,
-    Executing,
-}

src/flow/src/recording_rules/frontend_client.rs

@@ -1,150 +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.
-
-//! Frontend client to run flow as recording rule which is time-window-aware normal query triggered every tick set by user
-
-use std::sync::Arc;
-
-use client::{Client, Database, DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
-use common_error::ext::BoxedError;
-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 meta_client::client::MetaClient;
-use snafu::ResultExt;
-
-use crate::error::{ExternalSnafu, UnexpectedSnafu};
-use crate::recording_rules::engine::DEFAULT_RULE_ENGINE_QUERY_TIMEOUT;
-use crate::Error;
-
-fn default_channel_mgr() -> ChannelManager {
-    let cfg = ChannelConfig::new().timeout(DEFAULT_RULE_ENGINE_QUERY_TIMEOUT);
-    ChannelManager::with_config(cfg)
-}
-
-fn client_from_urls(addrs: Vec<String>) -> Client {
-    Client::with_manager_and_urls(default_channel_mgr(), addrs)
-}
-
-/// A simple frontend client able to execute sql using grpc protocol
-#[derive(Debug)]
-pub enum FrontendClient {
-    Distributed {
-        meta_client: Arc<MetaClient>,
-    },
-    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,
-    },
-}
-
-#[derive(Debug, Clone)]
-pub struct DatabaseWithPeer {
-    pub database: Database,
-    pub peer: Peer,
-}
-
-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),
-        }
-    }
-}
-
-impl FrontendClient {
-    async fn scan_for_frontend(&self) -> Result<Vec<(NodeInfoKey, NodeInfo)>, Error> {
-        let Self::Distributed { meta_client, .. } = self else {
-            return Ok(vec![]);
-        };
-        let cluster_client = meta_client
-            .cluster_client()
-            .map_err(BoxedError::new)
-            .context(ExternalSnafu)?;
-        let cluster_id = meta_client.id().0;
-        let prefix = NodeInfoKey::key_prefix_with_role(cluster_id, Role::Frontend);
-        let req = RangeRequest::new().with_prefix(prefix);
-        let resp = cluster_client
-            .range(req)
-            .await
-            .map_err(BoxedError::new)
-            .context(ExternalSnafu)?;
-        let mut res = Vec::with_capacity(resp.kvs.len());
-        for kv in resp.kvs {
-            let key = NodeInfoKey::try_from(kv.key)
-                .map_err(BoxedError::new)
-                .context(ExternalSnafu)?;
-
-            let val = NodeInfo::try_from(kv.value)
-                .map_err(BoxedError::new)
-                .context(ExternalSnafu)?;
-            res.push((key, val));
-        }
-        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 last_activity_ts = i64::MIN;
-        let mut peer = None;
-        for (_key, val) in frontends.iter() {
-            if val.last_activity_ts > last_activity_ts {
-                last_activity_ts = val.last_activity_ts;
-                peer = Some(val.peer.clone());
-            }
-        }
-        let Some(peer) = peer else {
-            UnexpectedSnafu {
-                reason: format!("No frontend available: {:?}", frontends),
-            }
-            .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))
-    }
-
-    /// Get a database client, and possibly update it before returning.
-    pub async fn get_database_client(&self) -> Result<DatabaseWithPeer, Error> {
-        match self {
-            Self::Standalone { database_client } => Ok(database_client.clone()),
-            Self::Distributed { meta_client: _ } => self.get_last_active_frontend().await,
-        }
-    }
-}

src/flow/src/server.rs

@@ -57,7 +57,6 @@ use crate::error::{
 };
 use crate::heartbeat::HeartbeatTask;
 use crate::metrics::{METRIC_FLOW_PROCESSING_TIME, METRIC_FLOW_ROWS};
-use crate::recording_rules::{FrontendClient, RecordingRuleEngine};
 use crate::transform::register_function_to_query_engine;
 use crate::utils::{SizeReportSender, StateReportHandler};
 use crate::{Error, FlowWorkerManager, FlownodeOptions};
@@ -246,7 +245,6 @@ impl FlownodeInstance {
         self.server.shutdown().await.context(ShutdownServerSnafu)?;
 
         if let Some(task) = &self.heartbeat_task {
-            info!("Close heartbeat task for flownode");
             task.shutdown();
         }
 
@@ -273,8 +271,6 @@ pub struct FlownodeBuilder {
     heartbeat_task: Option<HeartbeatTask>,
     /// receive a oneshot sender to send state size report
     state_report_handler: Option<StateReportHandler>,
-    /// Client to send sql to frontend
-    frontend_client: Arc<FrontendClient>,
 }
 
 impl FlownodeBuilder {
@@ -285,7 +281,6 @@ impl FlownodeBuilder {
         table_meta: TableMetadataManagerRef,
         catalog_manager: CatalogManagerRef,
         flow_metadata_manager: FlowMetadataManagerRef,
-        frontend_client: Arc<FrontendClient>,
     ) -> Self {
         Self {
             opts,
@@ -295,7 +290,6 @@ impl FlownodeBuilder {
             flow_metadata_manager,
             heartbeat_task: None,
             state_report_handler: None,
-            frontend_client,
         }
     }
 
@@ -453,10 +447,7 @@ impl FlownodeBuilder {
 
         let node_id = self.opts.node_id.map(|id| id as u32);
 
-        let rule_engine =
-            RecordingRuleEngine::new(self.frontend_client.clone(), query_engine.clone());
-
-        let mut man = FlowWorkerManager::new(node_id, query_engine, table_meta, rule_engine);
+        let mut man = FlowWorkerManager::new(node_id, query_engine, table_meta);
         for worker_id in 0..num_workers {
             let (tx, rx) = oneshot::channel();
 

src/flow/src/test_utils.rs

@@ -86,8 +86,7 @@ pub fn create_test_query_engine() -> Arc<dyn QueryEngine> {
 
     let schema = vec![
         datatypes::schema::ColumnSchema::new("number", CDT::uint32_datatype(), false),
-        datatypes::schema::ColumnSchema::new("ts", CDT::timestamp_millisecond_datatype(), false)
-            .with_time_index(true),
+        datatypes::schema::ColumnSchema::new("ts", CDT::timestamp_millisecond_datatype(), false),
     ];
     let mut columns = vec![];
     let numbers = (1..=10).collect_vec();

src/frontend/src/heartbeat.rs

@@ -23,7 +23,7 @@ use common_meta::heartbeat::handler::{
 };
 use common_meta::heartbeat::mailbox::{HeartbeatMailbox, MailboxRef, OutgoingMessage};
 use common_meta::heartbeat::utils::outgoing_message_to_mailbox_message;
-use common_telemetry::{debug, error, info};
+use common_telemetry::{debug, error, info, warn};
 use meta_client::client::{HeartbeatSender, HeartbeatStream, MetaClient};
 use servers::addrs;
 use servers::heartbeat_options::HeartbeatOptions;
@@ -42,8 +42,8 @@ use crate::metrics::{HEARTBEAT_RECV_COUNT, HEARTBEAT_SENT_COUNT};
 pub struct HeartbeatTask {
     peer_addr: String,
     meta_client: Arc<MetaClient>,
-    report_interval: u64,
-    retry_interval: u64,
+    report_interval: Duration,
+    retry_interval: Duration,
     resp_handler_executor: HeartbeatResponseHandlerExecutorRef,
     start_time_ms: u64,
 }
@@ -58,8 +58,8 @@ impl HeartbeatTask {
         HeartbeatTask {
             peer_addr: addrs::resolve_addr(&opts.grpc.bind_addr, Some(&opts.grpc.server_addr)),
             meta_client,
-            report_interval: heartbeat_opts.interval.as_millis() as u64,
-            retry_interval: heartbeat_opts.retry_interval.as_millis() as u64,
+            report_interval: heartbeat_opts.interval,
+            retry_interval: heartbeat_opts.retry_interval,
             resp_handler_executor,
             start_time_ms: common_time::util::current_time_millis() as u64,
         }
@@ -103,13 +103,15 @@ impl HeartbeatTask {
                             HEARTBEAT_RECV_COUNT.with_label_values(&["success"]).inc();
                         }
                     }
-                    Ok(None) => break,
+                    Ok(None) => {
+                        warn!("Heartbeat response stream closed");
+                        capture_self.start_with_retry(retry_interval).await;
+                        break;
+                    }
                     Err(e) => {
                         HEARTBEAT_RECV_COUNT.with_label_values(&["error"]).inc();
                         error!(e; "Occur error while reading heartbeat response");
-                        capture_self
-                            .start_with_retry(Duration::from_millis(retry_interval))
-                            .await;
+                        capture_self.start_with_retry(retry_interval).await;
 
                         break;
                     }
@@ -177,12 +179,13 @@ impl HeartbeatTask {
                         if let Some(message) = message {
                             Self::new_heartbeat_request(&heartbeat_request, Some(message))
                         } else {
+                            warn!("Sender has been dropped, exiting the heartbeat loop");
                             // Receives None that means Sender was dropped, we need to break the current loop
                             break
                         }
                     }
                     _ = &mut sleep => {
-                        sleep.as_mut().reset(Instant::now() + Duration::from_millis(report_interval));
+                        sleep.as_mut().reset(Instant::now() + report_interval);
                        Self::new_heartbeat_request(&heartbeat_request, None)
                     }
                 };

src/index/src/bloom_filter/applier.rs

@@ -42,7 +42,16 @@ impl BloomFilterApplier {
     ) -> Result<Vec<Range<usize>>> {
         let rows_per_segment = self.meta.rows_per_segment as usize;
         let start_seg = search_range.start / rows_per_segment;
-        let end_seg = search_range.end.div_ceil(rows_per_segment);
+        let mut end_seg = search_range.end.div_ceil(rows_per_segment);
+
+        if end_seg == self.meta.segment_loc_indices.len() + 1 {
+            // In a previous version, there was a bug where if the last segment was all null,
+            // this segment would not be written into the index. This caused the slice
+            // `self.meta.segment_loc_indices[start_seg..end_seg]` to go out of bounds due to
+            // the missing segment. Since the `search` function does not search for nulls,
+            // we can simply ignore the last segment in this buggy scenario.
+            end_seg -= 1;
+        }
 
         let locs = &self.meta.segment_loc_indices[start_seg..end_seg];
 

src/index/src/bloom_filter/creator.rs

@@ -64,6 +64,9 @@ pub struct BloomFilterCreator {
     /// Storage for finalized Bloom filters.
     finalized_bloom_filters: FinalizedBloomFilterStorage,
 
+    /// Row count that finalized so far.
+    finalized_row_count: usize,
+
     /// Global memory usage of the bloom filter creator.
     global_memory_usage: Arc<AtomicUsize>,
 }
@@ -96,6 +99,7 @@ impl BloomFilterCreator {
                 global_memory_usage,
                 global_memory_usage_threshold,
             ),
+            finalized_row_count: 0,
         }
     }
 
@@ -136,6 +140,7 @@ impl BloomFilterCreator {
 
             if self.accumulated_row_count % self.rows_per_segment == 0 {
                 self.finalize_segment().await?;
+                self.finalized_row_count = self.accumulated_row_count;
             }
         }
 
@@ -161,6 +166,7 @@ impl BloomFilterCreator {
 
         if self.accumulated_row_count % self.rows_per_segment == 0 {
             self.finalize_segment().await?;
+            self.finalized_row_count = self.accumulated_row_count;
         }
 
         Ok(())
@@ -168,7 +174,7 @@ impl BloomFilterCreator {
 
     /// Finalizes any remaining segments and writes the bloom filters and metadata to the provided writer.
     pub async fn finish(&mut self, mut writer: impl AsyncWrite + Unpin) -> Result<()> {
-        if !self.cur_seg_distinct_elems.is_empty() {
+        if self.accumulated_row_count > self.finalized_row_count {
             self.finalize_segment().await?;
         }
 
@@ -406,4 +412,35 @@ mod tests {
             assert!(bf.contains(&b"f"));
         }
     }
+
+    #[tokio::test]
+    async fn test_final_seg_all_null() {
+        let mut writer = Cursor::new(Vec::new());
+        let mut creator = BloomFilterCreator::new(
+            2,
+            Arc::new(MockExternalTempFileProvider::new()),
+            Arc::new(AtomicUsize::new(0)),
+            None,
+        );
+
+        creator
+            .push_n_row_elems(4, vec![b"a".to_vec(), b"b".to_vec()])
+            .await
+            .unwrap();
+        creator.push_row_elems(Vec::new()).await.unwrap();
+
+        creator.finish(&mut writer).await.unwrap();
+
+        let bytes = writer.into_inner();
+        let total_size = bytes.len();
+        let meta_size_offset = total_size - 4;
+        let meta_size = u32::from_le_bytes((&bytes[meta_size_offset..]).try_into().unwrap());
+
+        let meta_bytes = &bytes[total_size - meta_size as usize - 4..total_size - 4];
+        let meta = BloomFilterMeta::decode(meta_bytes).unwrap();
+
+        assert_eq!(meta.rows_per_segment, 2);
+        assert_eq!(meta.segment_count, 3);
+        assert_eq!(meta.row_count, 5);
+    }
 }

src/meta-client/src/client.rs

@@ -112,7 +112,6 @@ impl MetaClientBuilder {
             .enable_store()
             .enable_heartbeat()
             .enable_procedure()
-            .enable_access_cluster_info()
     }
 
     pub fn enable_heartbeat(self) -> Self {

src/meta-srv/src/handler/collect_cluster_info_handler.rs

@@ -157,7 +157,7 @@ fn extract_base_info(request: &HeartbeatRequest) -> Option<(NodeInfoKey, Peer, P
 }
 
 async fn put_into_memory_store(ctx: &mut Context, key: NodeInfoKey, value: NodeInfo) -> Result<()> {
-    let key = key.into();
+    let key = (&key).into();
     let value = value.try_into().context(InvalidClusterInfoFormatSnafu)?;
     let put_req = PutRequest {
         key,

src/meta-srv/src/metasrv.rs

@@ -32,6 +32,7 @@ use common_meta::kv_backend::{KvBackendRef, ResettableKvBackend, ResettableKvBac
 use common_meta::leadership_notifier::{
     LeadershipChangeNotifier, LeadershipChangeNotifierCustomizerRef,
 };
+use common_meta::node_expiry_listener::NodeExpiryListener;
 use common_meta::peer::Peer;
 use common_meta::region_keeper::MemoryRegionKeeperRef;
 use common_meta::wal_options_allocator::WalOptionsAllocatorRef;
@@ -151,6 +152,8 @@ pub struct MetasrvOptions {
     #[cfg(feature = "pg_kvbackend")]
     /// Lock id for meta kv election. Only effect when using pg_kvbackend.
     pub meta_election_lock_id: u64,
+    #[serde(with = "humantime_serde")]
+    pub node_max_idle_time: Duration,
 }
 
 const DEFAULT_METASRV_ADDR_PORT: &str = "3002";
@@ -192,6 +195,7 @@ impl Default for MetasrvOptions {
             meta_table_name: DEFAULT_META_TABLE_NAME.to_string(),
             #[cfg(feature = "pg_kvbackend")]
             meta_election_lock_id: DEFAULT_META_ELECTION_LOCK_ID,
+            node_max_idle_time: Duration::from_secs(24 * 60 * 60),
         }
     }
 }
@@ -442,6 +446,10 @@ impl Metasrv {
             leadership_change_notifier.add_listener(self.wal_options_allocator.clone());
             leadership_change_notifier
                 .add_listener(Arc::new(ProcedureManagerListenerAdapter(procedure_manager)));
+            leadership_change_notifier.add_listener(Arc::new(NodeExpiryListener::new(
+                self.options.node_max_idle_time,
+                self.in_memory.clone(),
+            )));
             if let Some(region_supervisor_ticker) = &self.region_supervisor_ticker {
                 leadership_change_notifier.add_listener(region_supervisor_ticker.clone() as _);
             }

src/meta-srv/src/service/heartbeat.rs

@@ -27,10 +27,9 @@ use snafu::OptionExt;
 use tokio::sync::mpsc;
 use tokio::sync::mpsc::Sender;
 use tokio_stream::wrappers::ReceiverStream;
-use tonic::{Request, Response, Streaming};
+use tonic::{Request, Response, Status, Streaming};
 
-use crate::error;
-use crate::error::Result;
+use crate::error::{self, Result};
 use crate::handler::{HeartbeatHandlerGroup, Pusher, PusherId};
 use crate::metasrv::{Context, Metasrv};
 use crate::metrics::METRIC_META_HEARTBEAT_RECV;
@@ -68,13 +67,15 @@ impl heartbeat_server::Heartbeat for Metasrv {
                         };
 
                         if pusher_id.is_none() {
-                            pusher_id = register_pusher(&handler_group, header, tx.clone()).await;
+                            pusher_id =
+                                Some(register_pusher(&handler_group, header, tx.clone()).await);
                         }
                         if let Some(k) = &pusher_id {
                             METRIC_META_HEARTBEAT_RECV.with_label_values(&[&k.to_string()]);
                         } else {
                             METRIC_META_HEARTBEAT_RECV.with_label_values(&["none"]);
                         }
+
                         let res = handler_group
                             .handle(req, ctx.clone())
                             .await
@@ -107,6 +108,12 @@ impl heartbeat_server::Heartbeat for Metasrv {
 
                 if is_not_leader {
                     warn!("Quit because it is no longer the leader");
+                    let _ = tx
+                        .send(Err(Status::aborted(format!(
+                            "The requested metasrv node is not leader, node addr: {}",
+                            ctx.server_addr
+                        ))))
+                        .await;
                     break;
                 }
             }
@@ -173,13 +180,13 @@ async fn register_pusher(
     handler_group: &HeartbeatHandlerGroup,
     header: &RequestHeader,
     sender: Sender<std::result::Result<HeartbeatResponse, tonic::Status>>,
-) -> Option<PusherId> {
+) -> PusherId {
     let role = header.role();
     let id = get_node_id(header);
     let pusher_id = PusherId::new(role, id);
     let pusher = Pusher::new(sender, header);
     handler_group.register_pusher(pusher_id, pusher).await;
-    Some(pusher_id)
+    pusher_id
 }
 
 #[cfg(test)]

src/mito2/src/access_layer.rs

@@ -13,6 +13,7 @@
 // limitations under the License.
 
 use std::sync::Arc;
+use std::time::Duration;
 
 use object_store::services::Fs;
 use object_store::util::{join_dir, with_instrument_layers};
@@ -42,6 +43,29 @@ pub type AccessLayerRef = Arc<AccessLayer>;
 /// SST write results.
 pub type SstInfoArray = SmallVec<[SstInfo; 2]>;
 
+#[derive(Debug, Default)]
+pub struct Metrics {
+    pub read: Duration,
+    pub write: Duration,
+    pub convert: Duration,
+    pub index_update: Duration,
+    pub index_finish: Duration,
+    pub close: Duration,
+    pub num_series: usize,
+
+    // SST Opendal metrics.
+    pub opendal_create_cost: Duration,
+    pub opendal_num_writes: usize,
+    pub opendal_write_cost: Duration,
+    pub opendal_complete_cost: Duration,
+}
+
+impl Metrics {
+    pub fn sum(&self) -> Duration {
+        self.read + self.write + self.convert + self.index_update + self.index_finish + self.close
+    }
+}
+
 /// A layer to access SST files under the same directory.
 pub struct AccessLayer {
     region_dir: String,
@@ -121,10 +145,11 @@ impl AccessLayer {
     /// Writes a SST with specific `file_id` and `metadata` to the layer.
     ///
     /// Returns the info of the SST. If no data written, returns None.
-    pub(crate) async fn write_sst(
+    pub async fn write_sst(
         &self,
         request: SstWriteRequest,
         write_opts: &WriteOptions,
+        metrics: &mut Metrics,
     ) -> Result<SstInfoArray> {
         let region_id = request.metadata.region_id;
         let cache_manager = request.cache_manager.clone();
@@ -167,9 +192,16 @@ impl AccessLayer {
                 path_provider,
             )
             .await;
-            writer
-                .write_all(request.source, request.max_sequence, write_opts)
-                .await?
+            let sst_info = writer
+                .write_all(request.source, request.max_sequence, write_opts, metrics)
+                .await?;
+            let opendal_metrics = writer.opendal_metrics_val();
+            metrics.opendal_create_cost += opendal_metrics.create_cost;
+            metrics.opendal_num_writes += opendal_metrics.num_writes;
+            metrics.opendal_write_cost += opendal_metrics.write_cost;
+            metrics.opendal_complete_cost += opendal_metrics.complete_cost;
+
+            sst_info
         };
 
         // Put parquet metadata to cache manager.
@@ -189,28 +221,53 @@ impl AccessLayer {
     }
 }
 
+/// Helper to build an [AccessLayerRef] with internal index managers.
+///
+/// This is a convenience constructor intended for tooling that needs to
+/// interact with SSTs without wiring all indexing internals manually.
+pub async fn build_access_layer(
+    region_dir: &str,
+    object_store: ObjectStore,
+    config: &crate::config::MitoConfig,
+) -> Result<AccessLayerRef> {
+    let puffin_manager_factory = PuffinManagerFactory::new(
+        &config.index.aux_path,
+        config.index.staging_size.as_bytes(),
+        Some(config.index.write_buffer_size.as_bytes() as _),
+        config.index.staging_ttl,
+    )
+    .await?;
+    let intermediate_manager = IntermediateManager::init_fs(&config.index.aux_path).await?;
+    Ok(Arc::new(AccessLayer::new(
+        region_dir,
+        object_store,
+        puffin_manager_factory,
+        intermediate_manager,
+    )))
+}
+
 /// `OperationType` represents the origin of the `SstWriteRequest`.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
-pub(crate) enum OperationType {
+pub enum OperationType {
     Flush,
     Compact,
 }
 
 /// Contents to build a SST.
-pub(crate) struct SstWriteRequest {
-    pub(crate) op_type: OperationType,
-    pub(crate) metadata: RegionMetadataRef,
-    pub(crate) source: Source,
-    pub(crate) cache_manager: CacheManagerRef,
+pub struct SstWriteRequest {
+    pub op_type: OperationType,
+    pub metadata: RegionMetadataRef,
+    pub source: Source,
+    pub cache_manager: CacheManagerRef,
     #[allow(dead_code)]
-    pub(crate) storage: Option<String>,
-    pub(crate) max_sequence: Option<SequenceNumber>,
+    pub storage: Option<String>,
+    pub max_sequence: Option<SequenceNumber>,
 
     /// Configs for index
-    pub(crate) index_options: IndexOptions,
-    pub(crate) inverted_index_config: InvertedIndexConfig,
-    pub(crate) fulltext_index_config: FulltextIndexConfig,
-    pub(crate) bloom_filter_index_config: BloomFilterConfig,
+    pub index_options: IndexOptions,
+    pub inverted_index_config: InvertedIndexConfig,
+    pub fulltext_index_config: FulltextIndexConfig,
+    pub bloom_filter_index_config: BloomFilterConfig,
 }
 
 pub(crate) async fn new_fs_cache_store(root: &str) -> Result<ObjectStore> {

src/mito2/src/cache/write_cache.rs

@@ -40,6 +40,7 @@ use crate::sst::index::IndexerBuilderImpl;
 use crate::sst::parquet::writer::ParquetWriter;
 use crate::sst::parquet::WriteOptions;
 use crate::sst::{DEFAULT_WRITE_BUFFER_SIZE, DEFAULT_WRITE_CONCURRENCY};
+use crate::Metrics;
 
 /// A cache for uploading files to remote object stores.
 ///
@@ -140,7 +141,12 @@ impl WriteCache {
         .await;
 
         let sst_info = writer
-            .write_all(write_request.source, write_request.max_sequence, write_opts)
+            .write_all(
+                write_request.source,
+                write_request.max_sequence,
+                write_opts,
+                &mut Metrics::default(),
+            )
             .await?;
 
         timer.stop_and_record();

src/mito2/src/compaction/compactor.rs

@@ -27,7 +27,7 @@ use snafu::{OptionExt, ResultExt};
 use store_api::metadata::RegionMetadataRef;
 use store_api::storage::RegionId;
 
-use crate::access_layer::{AccessLayer, AccessLayerRef, OperationType, SstWriteRequest};
+use crate::access_layer::{AccessLayer, AccessLayerRef, Metrics, OperationType, SstWriteRequest};
 use crate::cache::{CacheManager, CacheManagerRef};
 use crate::compaction::picker::{new_picker, PickerOutput};
 use crate::compaction::{find_ttl, CompactionSstReaderBuilder};
@@ -340,6 +340,7 @@ impl Compactor for DefaultCompactor {
                             bloom_filter_index_config,
                         },
                         &write_opts,
+                        &mut Metrics::default(),
                     )
                     .await?
                     .into_iter()

src/mito2/src/flush.rs

@@ -25,7 +25,7 @@ use store_api::storage::RegionId;
 use strum::IntoStaticStr;
 use tokio::sync::{mpsc, watch};
 
-use crate::access_layer::{AccessLayerRef, OperationType, SstWriteRequest};
+use crate::access_layer::{AccessLayerRef, Metrics, OperationType, SstWriteRequest};
 use crate::cache::CacheManagerRef;
 use crate::config::MitoConfig;
 use crate::error::{
@@ -366,7 +366,7 @@ impl RegionFlushTask {
 
             let ssts_written = self
                 .access_layer
-                .write_sst(write_request, &write_opts)
+                .write_sst(write_request, &write_opts, &mut Metrics::default())
                 .await?;
             if ssts_written.is_empty() {
                 // No data written.

src/mito2/src/lib.rs

@@ -44,6 +44,12 @@ mod time_provider;
 pub mod wal;
 mod worker;
 
+// Public re-exports for tooling convenience
+pub use access_layer::{
+    build_access_layer, AccessLayer, AccessLayerRef, Metrics, OperationType, SstWriteRequest,
+};
+pub use cache::{CacheManager, CacheManagerRef};
+
 #[cfg_attr(doc, aquamarine::aquamarine)]
 /// # Mito developer document
 ///

src/mito2/src/sst/parquet.rs

@@ -109,6 +109,7 @@ mod tests {
         new_batch_with_binary, new_source, sst_file_handle, sst_region_metadata,
     };
     use crate::test_util::{check_reader_result, TestEnv};
+    use crate::Metrics;
 
     const FILE_DIR: &str = "/";
 
@@ -165,7 +166,7 @@ mod tests {
         .await;
 
         let info = writer
-            .write_all(source, None, &write_opts)
+            .write_all(source, None, &write_opts, &mut Metrics::default())
             .await
             .unwrap()
             .remove(0);
@@ -222,7 +223,7 @@ mod tests {
         .await;
 
         writer
-            .write_all(source, None, &write_opts)
+            .write_all(source, None, &write_opts, &mut Metrics::default())
             .await
             .unwrap()
             .remove(0);
@@ -293,7 +294,7 @@ mod tests {
         .await;
 
         let sst_info = writer
-            .write_all(source, None, &write_opts)
+            .write_all(source, None, &write_opts, &mut Metrics::default())
             .await
             .unwrap()
             .remove(0);
@@ -334,7 +335,7 @@ mod tests {
         )
         .await;
         writer
-            .write_all(source, None, &write_opts)
+            .write_all(source, None, &write_opts, &mut Metrics::default())
             .await
             .unwrap()
             .remove(0);
@@ -389,7 +390,7 @@ mod tests {
         )
         .await;
         writer
-            .write_all(source, None, &write_opts)
+            .write_all(source, None, &write_opts, &mut Metrics::default())
             .await
             .unwrap()
             .remove(0);
@@ -427,7 +428,7 @@ mod tests {
         .await;
 
         writer
-            .write_all(source, None, &write_opts)
+            .write_all(source, None, &write_opts, &mut Metrics::default())
             .await
             .unwrap()
             .remove(0);

src/mito2/src/sst/parquet/reader.rs

@@ -1117,7 +1117,6 @@ impl ParquetReader {
         self.context.read_format().metadata()
     }
 
-    #[cfg(test)]
     pub fn parquet_metadata(&self) -> Arc<ParquetMetaData> {
         self.context.reader_builder().parquet_meta.clone()
     }

src/mito2/src/sst/parquet/writer.rs

@@ -17,14 +17,19 @@
 use std::future::Future;
 use std::pin::Pin;
 use std::sync::atomic::{AtomicUsize, Ordering};
-use std::sync::Arc;
+use std::sync::{Arc, Mutex};
 use std::task::{Context, Poll};
+use std::time::{Duration, Instant};
 
+use bytes::Bytes;
 use common_time::Timestamp;
 use datatypes::arrow::datatypes::SchemaRef;
-use object_store::{FuturesAsyncWriter, ObjectStore};
+use futures::future::BoxFuture;
+use object_store::{FuturesAsyncWriter, ObjectStore, Writer};
+use parquet::arrow::async_writer::AsyncFileWriter;
 use parquet::arrow::AsyncArrowWriter;
 use parquet::basic::{Compression, Encoding, ZstdLevel};
+use parquet::errors::ParquetError;
 use parquet::file::metadata::KeyValue;
 use parquet::file::properties::{WriterProperties, WriterPropertiesBuilder};
 use parquet::schema::types::ColumnPath;
@@ -45,12 +50,13 @@ use crate::sst::parquet::format::WriteFormat;
 use crate::sst::parquet::helper::parse_parquet_metadata;
 use crate::sst::parquet::{SstInfo, WriteOptions, PARQUET_METADATA_KEY};
 use crate::sst::{DEFAULT_WRITE_BUFFER_SIZE, DEFAULT_WRITE_CONCURRENCY};
+use crate::Metrics;
 
 /// Parquet SST writer.
 pub struct ParquetWriter<F: WriterFactory, I: IndexerBuilder, P: FilePathProvider> {
     /// Path provider that creates SST and index file paths according to file id.
     path_provider: P,
-    writer: Option<AsyncArrowWriter<SizeAwareWriter<F::Writer>>>,
+    writer: Option<AsyncArrowWriter<OpenDalWriter>>,
     /// Current active file id.
     current_file: FileId,
     writer_factory: F,
@@ -61,11 +67,18 @@ pub struct ParquetWriter<F: WriterFactory, I: IndexerBuilder, P: FilePathProvide
     /// Current active indexer.
     current_indexer: Option<Indexer>,
     bytes_written: Arc<AtomicUsize>,
+    opendal_metrics: Arc<Mutex<OpenDalMetrics>>,
 }
 
 pub trait WriterFactory {
     type Writer: AsyncWrite + Send + Unpin;
     fn create(&mut self, file_path: &str) -> impl Future<Output = Result<Self::Writer>>;
+
+    fn create_opendal(
+        &mut self,
+        file_path: &str,
+        size: Arc<AtomicUsize>,
+    ) -> impl Future<Output = Result<OpenDalWriter>>;
 }
 
 pub struct ObjectStoreWriterFactory {
@@ -84,6 +97,22 @@ impl WriterFactory for ObjectStoreWriterFactory {
             .map(|v| v.into_futures_async_write().compat_write())
             .context(OpenDalSnafu)
     }
+
+    async fn create_opendal(
+        &mut self,
+        file_path: &str,
+        size: Arc<AtomicUsize>,
+    ) -> Result<OpenDalWriter> {
+        let writer = self
+            .object_store
+            .writer_with(file_path)
+            .chunk(DEFAULT_WRITE_BUFFER_SIZE.as_bytes() as usize)
+            .concurrent(DEFAULT_WRITE_CONCURRENCY)
+            .await
+            .context(OpenDalSnafu)?;
+
+        Ok(OpenDalWriter::new(writer, size))
+    }
 }
 
 impl<I, P> ParquetWriter<ObjectStoreWriterFactory, I, P>
@@ -105,6 +134,10 @@ where
         )
         .await
     }
+
+    pub fn opendal_metrics_val(&self) -> OpenDalMetrics {
+        self.opendal_metrics.lock().unwrap().clone()
+    }
 }
 
 impl<F, I, P> ParquetWriter<F, I, P>
@@ -132,6 +165,7 @@ where
             indexer_builder,
             current_indexer: Some(indexer),
             bytes_written: Arc::new(AtomicUsize::new(0)),
+            opendal_metrics: Arc::new(Mutex::new(OpenDalMetrics::default())),
         }
     }
 
@@ -156,20 +190,33 @@ where
         mut source: Source,
         override_sequence: Option<SequenceNumber>, // override the `sequence` field from `Source`
         opts: &WriteOptions,
+        metrics: &mut Metrics,
     ) -> Result<SstInfoArray> {
         let write_format =
             WriteFormat::new(self.metadata.clone()).with_override_sequence(override_sequence);
         let mut stats = SourceStats::default();
+        let mut last_key = None;
 
         while let Some(res) = self
-            .write_next_batch(&mut source, &write_format, opts)
+            .write_next_batch(&mut source, &write_format, opts, metrics)
             .await
             .transpose()
         {
             match res {
                 Ok(mut batch) => {
+                    if let Some(last) = &last_key {
+                        if last != batch.primary_key() {
+                            metrics.num_series += 1;
+                            last_key = Some(batch.primary_key().to_vec());
+                        }
+                    } else {
+                        metrics.num_series += 1;
+                    }
+
                     stats.update(&batch);
+                    let index_start = Instant::now();
                     self.get_or_create_indexer().await.update(&mut batch).await;
+                    metrics.index_update += index_start.elapsed();
                 }
                 Err(e) => {
                     self.get_or_create_indexer().await.abort().await;
@@ -178,7 +225,9 @@ where
             }
         }
 
+        let index_finish_start = Instant::now();
         let index_output = self.get_or_create_indexer().await.finish().await;
+        metrics.index_finish += index_finish_start.elapsed();
 
         if stats.num_rows == 0 {
             return Ok(smallvec![]);
@@ -189,9 +238,10 @@ where
             return Ok(smallvec![]);
         };
 
+        let close_start = Instant::now();
         arrow_writer.flush().await.context(WriteParquetSnafu)?;
-
         let file_meta = arrow_writer.close().await.context(WriteParquetSnafu)?;
+        metrics.close += close_start.elapsed();
         let file_size = self.bytes_written.load(Ordering::Relaxed) as u64;
 
         // Safety: num rows > 0 so we must have min/max.
@@ -238,17 +288,25 @@ where
         source: &mut Source,
         write_format: &WriteFormat,
         opts: &WriteOptions,
+        metrics: &mut Metrics,
     ) -> Result<Option<Batch>> {
+        let read_start = Instant::now();
         let Some(batch) = source.next_batch().await? else {
             return Ok(None);
         };
+        metrics.read += read_start.elapsed();
 
+        let convert_start = Instant::now();
         let arrow_batch = write_format.convert_batch(&batch)?;
+        metrics.convert += convert_start.elapsed();
+
+        let write_start = Instant::now();
         self.maybe_init_writer(write_format.arrow_schema(), opts)
             .await?
             .write(&arrow_batch)
             .await
             .context(WriteParquetSnafu)?;
+        metrics.write += write_start.elapsed();
         Ok(Some(batch))
     }
 
@@ -256,7 +314,7 @@ where
         &mut self,
         schema: &SchemaRef,
         opts: &WriteOptions,
-    ) -> Result<&mut AsyncArrowWriter<SizeAwareWriter<F::Writer>>> {
+    ) -> Result<&mut AsyncArrowWriter<OpenDalWriter>> {
         if let Some(ref mut w) = self.writer {
             Ok(w)
         } else {
@@ -274,10 +332,17 @@ where
             let writer_props = props_builder.build();
 
             let sst_file_path = self.path_provider.build_sst_file_path(self.current_file);
-            let writer = SizeAwareWriter::new(
-                self.writer_factory.create(&sst_file_path).await?,
-                self.bytes_written.clone(),
-            );
+            // let writer = SizeAwareWriter::new(
+            //     self.writer_factory.create(&sst_file_path).await?,
+            //     self.bytes_written.clone(),
+            // );
+            let create_start = Instant::now();
+            let mut writer = self
+                .writer_factory
+                .create_opendal(&sst_file_path, self.bytes_written.clone())
+                .await?;
+            self.opendal_metrics.lock().unwrap().create_cost += create_start.elapsed();
+            writer = writer.with_metrics(self.opendal_metrics.clone());
             let arrow_writer =
                 AsyncArrowWriter::try_new(writer, schema.clone(), Some(writer_props))
                     .context(WriteParquetSnafu)?;
@@ -317,6 +382,78 @@ impl SourceStats {
     }
 }
 
+#[derive(Default, Debug, Clone)]
+pub(crate) struct OpenDalMetrics {
+    pub(crate) create_cost: Duration,
+    pub(crate) num_writes: usize,
+    pub(crate) write_cost: Duration,
+    pub(crate) complete_cost: Duration,
+}
+
+/// Workaround for [AsyncArrowWriter] does not provide a method to
+/// get total bytes written after close.
+pub struct OpenDalWriter {
+    inner: Writer,
+    size: Arc<AtomicUsize>,
+    metrics: Option<Arc<Mutex<OpenDalMetrics>>>,
+}
+
+impl OpenDalWriter {
+    fn new(inner: Writer, size: Arc<AtomicUsize>) -> Self {
+        Self {
+            inner,
+            size: size.clone(),
+            metrics: None,
+        }
+    }
+
+    fn with_metrics(mut self, metrics: Arc<Mutex<OpenDalMetrics>>) -> Self {
+        self.metrics = Some(metrics);
+        self
+    }
+}
+
+impl AsyncFileWriter for OpenDalWriter {
+    fn write(&mut self, bs: Bytes) -> BoxFuture<'_, Result<(), ParquetError>> {
+        let write_start = Instant::now();
+        let size = self.size.clone();
+        let metrics = self.metrics.clone();
+        Box::pin(async move {
+            let bytes_written = bs.len();
+            self.inner
+                .write(bs)
+                .await
+                .map_err(|err| ParquetError::External(Box::new(err)))?;
+
+            size.fetch_add(bytes_written, Ordering::Relaxed);
+            if let Some(metrics) = metrics {
+                let mut m = metrics.lock().unwrap();
+                m.num_writes += 1;
+                m.write_cost += write_start.elapsed();
+            }
+            Ok(())
+        })
+    }
+
+    fn complete(&mut self) -> BoxFuture<'_, Result<(), ParquetError>> {
+        let complete_start = Instant::now();
+        let metrics = self.metrics.clone();
+        Box::pin(async move {
+            self.inner
+                .close()
+                .await
+                .map(|_| ())
+                .map_err(|err| ParquetError::External(Box::new(err)))?;
+
+            if let Some(metrics) = metrics {
+                let mut m = metrics.lock().unwrap();
+                m.complete_cost += complete_start.elapsed();
+            }
+            Ok(())
+        })
+    }
+}
+
 /// Workaround for [AsyncArrowWriter] does not provide a method to
 /// get total bytes written after close.
 struct SizeAwareWriter<W> {

src/operator/src/insert.rs

@@ -68,7 +68,6 @@ pub struct Inserter {
     catalog_manager: CatalogManagerRef,
     partition_manager: PartitionRuleManagerRef,
     node_manager: NodeManagerRef,
-    #[allow(unused)]
     table_flownode_set_cache: TableFlownodeSetCacheRef,
 }
 
@@ -336,11 +335,9 @@ impl Inserter {
 
         let InstantAndNormalInsertRequests {
             normal_requests,
-            instant_requests: _,
+            instant_requests,
         } = requests;
 
-        // TODO(discord9): mirror some
-        /*
         // Mirror requests for source table to flownode asynchronously
         let flow_mirror_task = FlowMirrorTask::new(
             &self.table_flownode_set_cache,
@@ -350,7 +347,7 @@ impl Inserter {
                 .chain(instant_requests.requests.iter()),
         )
         .await?;
-        flow_mirror_task.detach(self.node_manager.clone())?;*/
+        flow_mirror_task.detach(self.node_manager.clone())?;
 
         // Write requests to datanode and wait for response
         let write_tasks = self
@@ -820,14 +817,12 @@ struct CreateAlterTableResult {
     table_infos: HashMap<TableId, Arc<TableInfo>>,
 }
 
-#[allow(unused)]
 struct FlowMirrorTask {
     requests: HashMap<Peer, RegionInsertRequests>,
     num_rows: usize,
 }
 
 impl FlowMirrorTask {
-    #[allow(unused)]
     async fn new(
         cache: &TableFlownodeSetCacheRef,
         requests: impl Iterator<Item = &RegionInsertRequest>,
@@ -901,7 +896,6 @@ impl FlowMirrorTask {
         })
     }
 
-    #[allow(unused)]
     fn detach(self, node_manager: NodeManagerRef) -> Result<()> {
         crate::metrics::DIST_MIRROR_PENDING_ROW_COUNT.add(self.num_rows as i64);
         for (peer, inserts) in self.requests {

src/promql/src/extension_plan/histogram_fold.rs

@@ -583,7 +583,8 @@ impl HistogramFoldStream {
                     .expect("field column should not be nullable");
                 counters.push(counter);
             }
-            let result = Self::evaluate_row(self.quantile, &bucket, &counters)?;
+            // ignore invalid data
+            let result = Self::evaluate_row(self.quantile, &bucket, &counters).unwrap_or(f64::NAN);
             self.output_buffer[self.field_column_index].push_value_ref(ValueRef::from(result));
             cursor += bucket_num;
             remaining_rows -= bucket_num;
@@ -672,7 +673,7 @@ impl HistogramFoldStream {
         if bucket.len() <= 1 {
             return Ok(f64::NAN);
         }
-        if *bucket.last().unwrap() != f64::INFINITY {
+        if bucket.last().unwrap().is_finite() {
             return Err(DataFusionError::Execution(
                 "last bucket should be +Inf".to_string(),
             ));
@@ -692,8 +693,8 @@ impl HistogramFoldStream {
         }
 
         // check input value
-        debug_assert!(bucket.windows(2).all(|w| w[0] <= w[1]));
-        debug_assert!(counter.windows(2).all(|w| w[0] <= w[1]));
+        debug_assert!(bucket.windows(2).all(|w| w[0] <= w[1]), "{bucket:?}");
+        debug_assert!(counter.windows(2).all(|w| w[0] <= w[1]), "{counter:?}");
 
         let total = *counter.last().unwrap();
         let expected_pos = total * quantile;

src/promql/src/functions.rs

@@ -21,6 +21,7 @@ mod idelta;
 mod predict_linear;
 mod quantile;
 mod resets;
+mod round;
 #[cfg(test)]
 mod test_util;
 
@@ -39,6 +40,7 @@ pub use idelta::IDelta;
 pub use predict_linear::PredictLinear;
 pub use quantile::QuantileOverTime;
 pub use resets::Resets;
+pub use round::Round;
 
 pub(crate) fn extract_array(columnar_value: &ColumnarValue) -> Result<ArrayRef, DataFusionError> {
     if let ColumnarValue::Array(array) = columnar_value {

src/promql/src/functions/round.rs

@@ -0,0 +1,105 @@
+// 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 datafusion::error::DataFusionError;
+use datafusion_expr::{create_udf, ColumnarValue, ScalarUDF, Volatility};
+use datatypes::arrow::array::AsArray;
+use datatypes::arrow::datatypes::{DataType, Float64Type};
+use datatypes::compute;
+
+use crate::functions::extract_array;
+
+pub struct Round {
+    nearest: f64,
+}
+
+impl Round {
+    fn new(nearest: f64) -> Self {
+        Self { nearest }
+    }
+
+    pub const fn name() -> &'static str {
+        "prom_round"
+    }
+
+    fn input_type() -> Vec<DataType> {
+        vec![DataType::Float64]
+    }
+
+    pub fn return_type() -> DataType {
+        DataType::Float64
+    }
+
+    pub fn scalar_udf(nearest: f64) -> ScalarUDF {
+        create_udf(
+            Self::name(),
+            Self::input_type(),
+            Self::return_type(),
+            Volatility::Immutable,
+            Arc::new(move |input: &_| Self::new(nearest).calc(input)) as _,
+        )
+    }
+
+    fn calc(&self, input: &[ColumnarValue]) -> Result<ColumnarValue, DataFusionError> {
+        assert_eq!(input.len(), 1);
+
+        let value_array = extract_array(&input[0])?;
+
+        if self.nearest == 0.0 {
+            let values = value_array.as_primitive::<Float64Type>();
+            let result = compute::unary::<_, _, Float64Type>(values, |a| a.round());
+            Ok(ColumnarValue::Array(Arc::new(result) as _))
+        } else {
+            let values = value_array.as_primitive::<Float64Type>();
+            let nearest = self.nearest;
+            let result =
+                compute::unary::<_, _, Float64Type>(values, |a| ((a / nearest).round() * nearest));
+            Ok(ColumnarValue::Array(Arc::new(result) as _))
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use datatypes::arrow::array::Float64Array;
+
+    use super::*;
+
+    fn test_round_f64(value: Vec<f64>, nearest: f64, expected: Vec<f64>) {
+        let round_udf = Round::scalar_udf(nearest);
+        let input = vec![ColumnarValue::Array(Arc::new(Float64Array::from(value)))];
+        let result = round_udf.invoke_batch(&input, 1).unwrap();
+        let result_array = extract_array(&result).unwrap();
+        assert_eq!(result_array.len(), 1);
+        assert_eq!(
+            result_array.as_primitive::<Float64Type>().values(),
+            &expected
+        );
+    }
+
+    #[test]
+    fn test_round() {
+        test_round_f64(vec![123.456], 0.001, vec![123.456]);
+        test_round_f64(vec![123.456], 0.01, vec![123.46000000000001]);
+        test_round_f64(vec![123.456], 0.1, vec![123.5]);
+        test_round_f64(vec![123.456], 0.0, vec![123.0]);
+        test_round_f64(vec![123.456], 1.0, vec![123.0]);
+        test_round_f64(vec![123.456], 10.0, vec![120.0]);
+        test_round_f64(vec![123.456], 100.0, vec![100.0]);
+        test_round_f64(vec![123.456], 105.0, vec![105.0]);
+        test_round_f64(vec![123.456], 1000.0, vec![0.0]);
+    }
+}

src/query/src/promql/planner.rs

@@ -52,7 +52,7 @@ use promql::extension_plan::{
 use promql::functions::{
     AbsentOverTime, AvgOverTime, Changes, CountOverTime, Delta, Deriv, HoltWinters, IDelta,
     Increase, LastOverTime, MaxOverTime, MinOverTime, PredictLinear, PresentOverTime,
-    QuantileOverTime, Rate, Resets, StddevOverTime, StdvarOverTime, SumOverTime,
+    QuantileOverTime, Rate, Resets, Round, StddevOverTime, StdvarOverTime, SumOverTime,
 };
 use promql_parser::label::{MatchOp, Matcher, Matchers, METRIC_NAME};
 use promql_parser::parser::token::TokenType;
@@ -200,10 +200,9 @@ impl PromPlanner {
             PromExpr::Paren(ParenExpr { expr }) => {
                 self.prom_expr_to_plan(expr, session_state).await?
             }
-            PromExpr::Subquery(SubqueryExpr { .. }) => UnsupportedExprSnafu {
-                name: "Prom Subquery",
+            PromExpr::Subquery(expr) => {
+                self.prom_subquery_expr_to_plan(session_state, expr).await?
             }
-            .fail()?,
             PromExpr::NumberLiteral(lit) => self.prom_number_lit_to_plan(lit)?,
             PromExpr::StringLiteral(lit) => self.prom_string_lit_to_plan(lit)?,
             PromExpr::VectorSelector(selector) => {
@@ -218,6 +217,48 @@ impl PromPlanner {
         Ok(res)
     }
 
+    async fn prom_subquery_expr_to_plan(
+        &mut self,
+        session_state: &SessionState,
+        subquery_expr: &SubqueryExpr,
+    ) -> Result<LogicalPlan> {
+        let SubqueryExpr {
+            expr, range, step, ..
+        } = subquery_expr;
+
+        let current_interval = self.ctx.interval;
+        if let Some(step) = step {
+            self.ctx.interval = step.as_millis() as _;
+        }
+        let current_start = self.ctx.start;
+        self.ctx.start -= range.as_millis() as i64 - self.ctx.interval;
+        let input = self.prom_expr_to_plan(expr, session_state).await?;
+        self.ctx.interval = current_interval;
+        self.ctx.start = current_start;
+
+        ensure!(!range.is_zero(), ZeroRangeSelectorSnafu);
+        let range_ms = range.as_millis() as _;
+        self.ctx.range = Some(range_ms);
+
+        let manipulate = RangeManipulate::new(
+            self.ctx.start,
+            self.ctx.end,
+            self.ctx.interval,
+            range_ms,
+            self.ctx
+                .time_index_column
+                .clone()
+                .expect("time index should be set in `setup_context`"),
+            self.ctx.field_columns.clone(),
+            input,
+        )
+        .context(DataFusionPlanningSnafu)?;
+
+        Ok(LogicalPlan::Extension(Extension {
+            node: Arc::new(manipulate),
+        }))
+    }
+
     async fn prom_aggr_expr_to_plan(
         &mut self,
         session_state: &SessionState,
@@ -441,6 +482,7 @@ impl PromPlanner {
                     // if left plan or right plan tag is empty, means case like `scalar(...) + host` or `host + scalar(...)`
                     // under this case we only join on time index
                     left_context.tag_columns.is_empty() || right_context.tag_columns.is_empty(),
+                    modifier,
                 )?;
                 let join_plan_schema = join_plan.schema().clone();
 
@@ -1468,6 +1510,20 @@ impl PromPlanner {
 
                 ScalarFunc::GeneratedExpr
             }
+            "round" => {
+                let nearest = match other_input_exprs.pop_front() {
+                    Some(DfExpr::Literal(ScalarValue::Float64(Some(t)))) => t,
+                    Some(DfExpr::Literal(ScalarValue::Int64(Some(t)))) => t as f64,
+                    None => 0.0,
+                    other => UnexpectedPlanExprSnafu {
+                        desc: format!("expected f64 literal as t, but found {:?}", other),
+                    }
+                    .fail()?,
+                };
+
+                ScalarFunc::DataFusionUdf(Arc::new(Round::scalar_udf(nearest)))
+            }
+
             _ => {
                 if let Some(f) = session_state.scalar_functions().get(func.name) {
                     ScalarFunc::DataFusionBuiltin(f.clone())
@@ -1674,7 +1730,7 @@ impl PromPlanner {
         ensure!(
             !src_labels.is_empty(),
             FunctionInvalidArgumentSnafu {
-                fn_name: "label_join",
+                fn_name: "label_join"
             }
         );
 
@@ -2121,24 +2177,49 @@ impl PromPlanner {
         left_time_index_column: Option<String>,
         right_time_index_column: Option<String>,
         only_join_time_index: bool,
+        modifier: &Option<BinModifier>,
     ) -> Result<LogicalPlan> {
         let mut left_tag_columns = if only_join_time_index {
-            vec![]
+            BTreeSet::new()
         } else {
             self.ctx
                 .tag_columns
                 .iter()
-                .map(Column::from_name)
-                .collect::<Vec<_>>()
+                .cloned()
+                .collect::<BTreeSet<_>>()
         };
         let mut right_tag_columns = left_tag_columns.clone();
 
+        // apply modifier
+        if let Some(modifier) = modifier {
+            // apply label modifier
+            if let Some(matching) = &modifier.matching {
+                match matching {
+                    // keeps columns mentioned in `on`
+                    LabelModifier::Include(on) => {
+                        let mask = on.labels.iter().cloned().collect::<BTreeSet<_>>();
+                        left_tag_columns = left_tag_columns.intersection(&mask).cloned().collect();
+                        right_tag_columns =
+                            right_tag_columns.intersection(&mask).cloned().collect();
+                    }
+                    // removes columns memtioned in `ignoring`
+                    LabelModifier::Exclude(ignoring) => {
+                        // doesn't check existence of label
+                        for label in &ignoring.labels {
+                            let _ = left_tag_columns.remove(label);
+                            let _ = right_tag_columns.remove(label);
+                        }
+                    }
+                }
+            }
+        }
+
         // push time index column if it exists
         if let (Some(left_time_index_column), Some(right_time_index_column)) =
             (left_time_index_column, right_time_index_column)
         {
-            left_tag_columns.push(Column::from_name(left_time_index_column));
-            right_tag_columns.push(Column::from_name(right_time_index_column));
+            left_tag_columns.insert(left_time_index_column);
+            right_tag_columns.insert(right_time_index_column);
         }
 
         let right = LogicalPlanBuilder::from(right)
@@ -2154,7 +2235,16 @@ impl PromPlanner {
             .join(
                 right,
                 JoinType::Inner,
-                (left_tag_columns, right_tag_columns),
+                (
+                    left_tag_columns
+                        .into_iter()
+                        .map(Column::from_name)
+                        .collect::<Vec<_>>(),
+                    right_tag_columns
+                        .into_iter()
+                        .map(Column::from_name)
+                        .collect::<Vec<_>>(),
+                ),
                 None,
             )
             .context(DataFusionPlanningSnafu)?
@@ -3340,6 +3430,59 @@ mod test {
         indie_query_plan_compare(query, expected).await;
     }
 
+    #[tokio::test]
+    async fn test_hash_join() {
+        let mut eval_stmt = EvalStmt {
+            expr: PromExpr::NumberLiteral(NumberLiteral { val: 1.0 }),
+            start: UNIX_EPOCH,
+            end: UNIX_EPOCH
+                .checked_add(Duration::from_secs(100_000))
+                .unwrap(),
+            interval: Duration::from_secs(5),
+            lookback_delta: Duration::from_secs(1),
+        };
+
+        let case = r#"http_server_requests_seconds_sum{uri="/accounts/login"} / ignoring(kubernetes_pod_name,kubernetes_namespace) http_server_requests_seconds_count{uri="/accounts/login"}"#;
+
+        let prom_expr = parser::parse(case).unwrap();
+        eval_stmt.expr = prom_expr;
+        let table_provider = build_test_table_provider_with_fields(
+            &[
+                (
+                    DEFAULT_SCHEMA_NAME.to_string(),
+                    "http_server_requests_seconds_sum".to_string(),
+                ),
+                (
+                    DEFAULT_SCHEMA_NAME.to_string(),
+                    "http_server_requests_seconds_count".to_string(),
+                ),
+            ],
+            &["uri", "kubernetes_namespace", "kubernetes_pod_name"],
+        )
+        .await;
+        // Should be ok
+        let plan = PromPlanner::stmt_to_plan(table_provider, &eval_stmt, &build_session_state())
+            .await
+            .unwrap();
+        let expected = r#"Projection: http_server_requests_seconds_count.uri, http_server_requests_seconds_count.kubernetes_namespace, http_server_requests_seconds_count.kubernetes_pod_name, http_server_requests_seconds_count.greptime_timestamp, http_server_requests_seconds_sum.greptime_value / http_server_requests_seconds_count.greptime_value AS http_server_requests_seconds_sum.greptime_value / http_server_requests_seconds_count.greptime_value
+  Inner Join: http_server_requests_seconds_sum.greptime_timestamp = http_server_requests_seconds_count.greptime_timestamp, http_server_requests_seconds_sum.uri = http_server_requests_seconds_count.uri
+    SubqueryAlias: http_server_requests_seconds_sum
+      PromInstantManipulate: range=[0..100000000], lookback=[1000], interval=[5000], time index=[greptime_timestamp]
+        PromSeriesNormalize: offset=[0], time index=[greptime_timestamp], filter NaN: [false]
+          PromSeriesDivide: tags=["uri", "kubernetes_namespace", "kubernetes_pod_name"]
+            Sort: http_server_requests_seconds_sum.uri DESC NULLS LAST, http_server_requests_seconds_sum.kubernetes_namespace DESC NULLS LAST, http_server_requests_seconds_sum.kubernetes_pod_name DESC NULLS LAST, http_server_requests_seconds_sum.greptime_timestamp DESC NULLS LAST
+              Filter: http_server_requests_seconds_sum.uri = Utf8("/accounts/login") AND http_server_requests_seconds_sum.greptime_timestamp >= TimestampMillisecond(-1000, None) AND http_server_requests_seconds_sum.greptime_timestamp <= TimestampMillisecond(100001000, None)
+                TableScan: http_server_requests_seconds_sum
+    SubqueryAlias: http_server_requests_seconds_count
+      PromInstantManipulate: range=[0..100000000], lookback=[1000], interval=[5000], time index=[greptime_timestamp]
+        PromSeriesNormalize: offset=[0], time index=[greptime_timestamp], filter NaN: [false]
+          PromSeriesDivide: tags=["uri", "kubernetes_namespace", "kubernetes_pod_name"]
+            Sort: http_server_requests_seconds_count.uri DESC NULLS LAST, http_server_requests_seconds_count.kubernetes_namespace DESC NULLS LAST, http_server_requests_seconds_count.kubernetes_pod_name DESC NULLS LAST, http_server_requests_seconds_count.greptime_timestamp DESC NULLS LAST
+              Filter: http_server_requests_seconds_count.uri = Utf8("/accounts/login") AND http_server_requests_seconds_count.greptime_timestamp >= TimestampMillisecond(-1000, None) AND http_server_requests_seconds_count.greptime_timestamp <= TimestampMillisecond(100001000, None)
+                TableScan: http_server_requests_seconds_count"#;
+        assert_eq!(plan.to_string(), expected);
+    }
+
     #[tokio::test]
     async fn test_nested_histogram_quantile() {
         let mut eval_stmt = EvalStmt {

tests-integration/src/standalone.rs

@@ -40,7 +40,7 @@ use common_procedure::options::ProcedureConfig;
 use common_procedure::ProcedureManagerRef;
 use common_wal::config::{DatanodeWalConfig, MetasrvWalConfig};
 use datanode::datanode::DatanodeBuilder;
-use flow::{FlownodeBuilder, FrontendClient};
+use flow::FlownodeBuilder;
 use frontend::instance::builder::FrontendBuilder;
 use frontend::instance::{FrontendInstance, Instance, StandaloneDatanodeManager};
 use meta_srv::metasrv::{FLOW_ID_SEQ, TABLE_ID_SEQ};
@@ -164,15 +164,12 @@ impl GreptimeDbStandaloneBuilder {
             Some(procedure_manager.clone()),
         );
 
-        let fe_server_addr = opts.frontend_options().grpc.bind_addr.clone();
-        let frontend_client = FrontendClient::from_static_grpc_addr(fe_server_addr);
         let flow_builder = FlownodeBuilder::new(
             Default::default(),
             plugins.clone(),
             table_metadata_manager.clone(),
             catalog_manager.clone(),
             flow_metadata_manager.clone(),
-            Arc::new(frontend_client),
         );
         let flownode = Arc::new(flow_builder.build().await.unwrap());
 

tests-integration/tests/http.rs

@@ -1070,6 +1070,7 @@ fn drop_lines_with_inconsistent_results(input: String) -> String {
         "root =",
         "endpoint =",
         "region =",
+        "enable_virtual_host_style =",
         "cache_path =",
         "cache_capacity =",
         "sas_token =",

tests/cases/standalone/common/promql/round_fn.result

@@ -0,0 +1,81 @@
+create table cache_hit (
+    ts timestamp time index,
+    job string,
+    greptime_value double,
+    primary key (job)
+);
+
+Affected Rows: 0
+
+insert into cache_hit values
+    (3000, "read", 123.45),
+    (3000, "write", 234.567),
+    (4000, "read", 345.678),
+    (4000, "write", 456.789);
+
+Affected Rows: 4
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit, 0.01);
+
++---------------------+----------------------------+-------+
+| ts                  | prom_round(greptime_value) | job   |
++---------------------+----------------------------+-------+
+| 1970-01-01T00:00:03 | 123.45                     | read  |
+| 1970-01-01T00:00:03 | 234.57                     | write |
+| 1970-01-01T00:00:04 | 345.68                     | read  |
+| 1970-01-01T00:00:04 | 456.79                     | write |
++---------------------+----------------------------+-------+
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit, 0.1);
+
++---------------------+----------------------------+-------+
+| ts                  | prom_round(greptime_value) | job   |
++---------------------+----------------------------+-------+
+| 1970-01-01T00:00:03 | 123.5                      | read  |
+| 1970-01-01T00:00:03 | 234.60000000000002         | write |
+| 1970-01-01T00:00:04 | 345.70000000000005         | read  |
+| 1970-01-01T00:00:04 | 456.8                      | write |
++---------------------+----------------------------+-------+
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit, 1.0);
+
++---------------------+----------------------------+-------+
+| ts                  | prom_round(greptime_value) | job   |
++---------------------+----------------------------+-------+
+| 1970-01-01T00:00:03 | 123.0                      | read  |
+| 1970-01-01T00:00:03 | 235.0                      | write |
+| 1970-01-01T00:00:04 | 346.0                      | read  |
+| 1970-01-01T00:00:04 | 457.0                      | write |
++---------------------+----------------------------+-------+
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit);
+
++---------------------+----------------------------+-------+
+| ts                  | prom_round(greptime_value) | job   |
++---------------------+----------------------------+-------+
+| 1970-01-01T00:00:03 | 123.0                      | read  |
+| 1970-01-01T00:00:03 | 235.0                      | write |
+| 1970-01-01T00:00:04 | 346.0                      | read  |
+| 1970-01-01T00:00:04 | 457.0                      | write |
++---------------------+----------------------------+-------+
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit, 10.0);
+
++---------------------+----------------------------+-------+
+| ts                  | prom_round(greptime_value) | job   |
++---------------------+----------------------------+-------+
+| 1970-01-01T00:00:03 | 120.0                      | read  |
+| 1970-01-01T00:00:03 | 230.0                      | write |
+| 1970-01-01T00:00:04 | 350.0                      | read  |
+| 1970-01-01T00:00:04 | 460.0                      | write |
++---------------------+----------------------------+-------+
+
+drop table cache_hit;
+
+Affected Rows: 0
+

tests/cases/standalone/common/promql/round_fn.sql

@@ -0,0 +1,30 @@
+
+create table cache_hit (
+    ts timestamp time index,
+    job string,
+    greptime_value double,
+    primary key (job)
+);
+
+insert into cache_hit values
+    (3000, "read", 123.45),
+    (3000, "write", 234.567),
+    (4000, "read", 345.678),
+    (4000, "write", 456.789);
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit, 0.01);
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit, 0.1);
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit, 1.0);
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit);
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') round(cache_hit, 10.0);
+
+drop table cache_hit;

tests/cases/standalone/common/promql/set_operation.result

@@ -638,3 +638,78 @@ drop table cache_miss;
 
 Affected Rows: 0
 
+create table cache_hit_with_null_label (
+    ts timestamp time index,
+    job string,
+    null_label string null,
+    greptime_value double,
+    primary key (job, null_label)
+);
+
+Affected Rows: 0
+
+create table cache_miss_with_null_label (
+    ts timestamp time index,
+    job string,
+    null_label string null,
+    greptime_value double,
+    primary key (job, null_label)
+);
+
+Affected Rows: 0
+
+insert into cache_hit_with_null_label values
+    (3000, "read", null, 1.0),
+    (3000, "write", null, 2.0),
+    (4000, "read", null, 3.0),
+    (4000, "write", null, 4.0);
+
+Affected Rows: 4
+
+insert into cache_miss_with_null_label values
+    (3000, "read", null, 1.0),
+    (3000, "write", null, 2.0),
+    (4000, "read", null, 1.0),
+    (4000, "write", null, 2.0);
+
+Affected Rows: 4
+
+-- SQLNESS SORT_RESULT 3 1
+-- null!=null, so it will returns the empty set.
+tql eval (3, 4, '1s') cache_hit_with_null_label / (cache_miss_with_null_label + cache_hit_with_null_label);
+
+++
+++
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') cache_hit_with_null_label / ignoring(null_label) (cache_miss_with_null_label + ignoring(null_label) cache_hit_with_null_label);
+
++-------+------------+---------------------+---------------------------------------------------------------------------------------------------------------+
+| job   | null_label | ts                  | lhs.greptime_value / rhs.cache_miss_with_null_label.greptime_value + cache_hit_with_null_label.greptime_value |
++-------+------------+---------------------+---------------------------------------------------------------------------------------------------------------+
+| read  |            | 1970-01-01T00:00:03 | 0.5                                                                                                           |
+| read  |            | 1970-01-01T00:00:04 | 0.75                                                                                                          |
+| write |            | 1970-01-01T00:00:03 | 0.5                                                                                                           |
+| write |            | 1970-01-01T00:00:04 | 0.6666666666666666                                                                                            |
++-------+------------+---------------------+---------------------------------------------------------------------------------------------------------------+
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') cache_hit_with_null_label / on(job) (cache_miss_with_null_label + on(job) cache_hit_with_null_label);
+
++-------+------------+---------------------+---------------------------------------------------------------------------------------------------------------+
+| job   | null_label | ts                  | lhs.greptime_value / rhs.cache_miss_with_null_label.greptime_value + cache_hit_with_null_label.greptime_value |
++-------+------------+---------------------+---------------------------------------------------------------------------------------------------------------+
+| read  |            | 1970-01-01T00:00:03 | 0.5                                                                                                           |
+| read  |            | 1970-01-01T00:00:04 | 0.75                                                                                                          |
+| write |            | 1970-01-01T00:00:03 | 0.5                                                                                                           |
+| write |            | 1970-01-01T00:00:04 | 0.6666666666666666                                                                                            |
++-------+------------+---------------------+---------------------------------------------------------------------------------------------------------------+
+
+drop table cache_hit_with_null_label;
+
+Affected Rows: 0
+
+drop table cache_miss_with_null_label;
+
+Affected Rows: 0
+

tests/cases/standalone/common/promql/set_operation.sql

@@ -295,3 +295,45 @@ tql eval (3, 4, '1s') cache_hit / (cache_miss + cache_hit);
 drop table cache_hit;
 
 drop table cache_miss;
+
+create table cache_hit_with_null_label (
+    ts timestamp time index,
+    job string,
+    null_label string null,
+    greptime_value double,
+    primary key (job, null_label)
+);
+
+create table cache_miss_with_null_label (
+    ts timestamp time index,
+    job string,
+    null_label string null,
+    greptime_value double,
+    primary key (job, null_label)
+);
+
+insert into cache_hit_with_null_label values
+    (3000, "read", null, 1.0),
+    (3000, "write", null, 2.0),
+    (4000, "read", null, 3.0),
+    (4000, "write", null, 4.0);
+
+insert into cache_miss_with_null_label values
+    (3000, "read", null, 1.0),
+    (3000, "write", null, 2.0),
+    (4000, "read", null, 1.0),
+    (4000, "write", null, 2.0);
+
+-- SQLNESS SORT_RESULT 3 1
+-- null!=null, so it will returns the empty set.
+tql eval (3, 4, '1s') cache_hit_with_null_label / (cache_miss_with_null_label + cache_hit_with_null_label);
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') cache_hit_with_null_label / ignoring(null_label) (cache_miss_with_null_label + ignoring(null_label) cache_hit_with_null_label);
+
+-- SQLNESS SORT_RESULT 3 1
+tql eval (3, 4, '1s') cache_hit_with_null_label / on(job) (cache_miss_with_null_label + on(job) cache_hit_with_null_label);
+
+drop table cache_hit_with_null_label;
+
+drop table cache_miss_with_null_label;

tests/cases/standalone/common/promql/simple_histogram.result

@@ -295,3 +295,40 @@ drop table histogram3_bucket;
 
 Affected Rows: 0
 
+-- test with invalid data (unaligned buckets)
+create table histogram4_bucket (
+    ts timestamp time index,
+    le string,
+    s string,
+    val double,
+    primary key (s, le),
+);
+
+Affected Rows: 0
+
+insert into histogram4_bucket values
+    (2900000, "0.1", "a", 0),
+    (2900000, "1", "a", 10),
+    (2900000, "5", "a", 20),
+    (2900000, "+Inf", "a", 150),
+    (3000000, "0.1", "a", 50),
+    (3000000, "1", "a", 70),
+    (3000000, "5", "a", 120),
+    -- INF here is missing
+;
+
+Affected Rows: 7
+
+tql eval (2900, 3000, '100s') histogram_quantile(0.9, histogram4_bucket);
+
++---------------------+---+-----+
+| ts                  | s | val |
++---------------------+---+-----+
+| 1970-01-01T00:48:20 | a | 5.0 |
+| 1970-01-01T00:50:00 | a | 5.0 |
++---------------------+---+-----+
+
+drop table histogram4_bucket;
+
+Affected Rows: 0
+

tests/cases/standalone/common/promql/simple_histogram.sql

@@ -163,3 +163,27 @@ insert into histogram3_bucket values
 tql eval (3000, 3005, '3s') histogram_quantile(0.5, sum by(le, s) (rate(histogram3_bucket[5m])));
 
 drop table histogram3_bucket;
+
+-- test with invalid data (unaligned buckets)
+create table histogram4_bucket (
+    ts timestamp time index,
+    le string,
+    s string,
+    val double,
+    primary key (s, le),
+);
+
+insert into histogram4_bucket values
+    (2900000, "0.1", "a", 0),
+    (2900000, "1", "a", 10),
+    (2900000, "5", "a", 20),
+    (2900000, "+Inf", "a", 150),
+    (3000000, "0.1", "a", 50),
+    (3000000, "1", "a", 70),
+    (3000000, "5", "a", 120),
+    -- INF here is missing
+;
+
+tql eval (2900, 3000, '100s') histogram_quantile(0.9, histogram4_bucket);
+
+drop table histogram4_bucket;

tests/cases/standalone/common/promql/subquery.result

@@ -0,0 +1,65 @@
+create table metric_total (
+    ts timestamp time index,
+    val double,
+);
+
+Affected Rows: 0
+
+insert into metric_total values
+    (0, 1),
+    (10000, 2);
+
+Affected Rows: 2
+
+tql eval (10, 10, '1s') sum_over_time(metric_total[50s:10s]);
+
++---------------------+----------------------------------+
+| ts                  | prom_sum_over_time(ts_range,val) |
++---------------------+----------------------------------+
+| 1970-01-01T00:00:10 | 3.0                              |
++---------------------+----------------------------------+
+
+tql eval (10, 10, '1s') sum_over_time(metric_total[50s:5s]);
+
++---------------------+----------------------------------+
+| ts                  | prom_sum_over_time(ts_range,val) |
++---------------------+----------------------------------+
+| 1970-01-01T00:00:10 | 4.0                              |
++---------------------+----------------------------------+
+
+tql eval (300, 300, '1s') sum_over_time(metric_total[50s:10s]);
+
++---------------------+----------------------------------+
+| ts                  | prom_sum_over_time(ts_range,val) |
++---------------------+----------------------------------+
+| 1970-01-01T00:05:00 | 10.0                             |
++---------------------+----------------------------------+
+
+tql eval (359, 359, '1s') sum_over_time(metric_total[60s:10s]);
+
++---------------------+----------------------------------+
+| ts                  | prom_sum_over_time(ts_range,val) |
++---------------------+----------------------------------+
+| 1970-01-01T00:05:59 | 2.0                              |
++---------------------+----------------------------------+
+
+tql eval (10, 10, '1s') rate(metric_total[20s:10s]);
+
++---------------------+----------------------------+
+| ts                  | prom_rate(ts_range,val,ts) |
++---------------------+----------------------------+
+| 1970-01-01T00:00:10 | 0.1                        |
++---------------------+----------------------------+
+
+tql eval (20, 20, '1s') rate(metric_total[20s:5s]);
+
++---------------------+----------------------------+
+| ts                  | prom_rate(ts_range,val,ts) |
++---------------------+----------------------------+
+| 1970-01-01T00:00:20 | 0.06666666666666667        |
++---------------------+----------------------------+
+
+drop table metric_total;
+
+Affected Rows: 0
+

tests/cases/standalone/common/promql/subquery.sql

@@ -0,0 +1,22 @@
+create table metric_total (
+    ts timestamp time index,
+    val double,
+);
+
+insert into metric_total values
+    (0, 1),
+    (10000, 2);
+
+tql eval (10, 10, '1s') sum_over_time(metric_total[50s:10s]);
+
+tql eval (10, 10, '1s') sum_over_time(metric_total[50s:5s]);
+
+tql eval (300, 300, '1s') sum_over_time(metric_total[50s:10s]);
+
+tql eval (359, 359, '1s') sum_over_time(metric_total[60s:10s]);
+
+tql eval (10, 10, '1s') rate(metric_total[20s:10s]);
+
+tql eval (20, 20, '1s') rate(metric_total[20s:5s]);
+
+drop table metric_total;

typos.toml

@@ -4,6 +4,10 @@ ue = "ue"
 worl = "worl"
 ot = "ot"
 unqualifed = "unqualifed"
+typ = "typ"
+varidic = "varidic"
+typs = "typs"
+varadic = "varadic"
 
 [files]
 extend-exclude = [