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 }} && \
+          DEV_BUILDER_IMAGE_TAG=${{ inputs.version }}
-
-        docker push ${{ inputs.dockerhub-image-registry }}/${{ inputs.dockerhub-image-namespace }}/dev-builder-android:${{ 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 "$NEXT_RELEASE_VERSION" ]; then
-      echo "NEXT_RELEASE_VERSION is empty"
+      echo "NEXT_RELEASE_VERSION is empty, use version from Cargo.toml" >&2
-      exit 1
+      # NOTE: Need a `v` prefix for the version string.
+      export NEXT_RELEASE_VERSION=v$(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=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=nigtly ./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=nigtly ./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=nigtly ./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=nigtly ./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-22.04
-          - ubuntu-20.04-8-cores
+          - ubuntu-22.04-8-cores
-          - ubuntu-20.04-16-cores
+          - ubuntu-22.04-16-cores
-          - ubuntu-20.04-32-cores
+          - ubuntu-22.04-32-cores
-          - ubuntu-20.04-64-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'
+    if: ${{ github.repository == 'GreptimeTeam/greptimedb' && github.event_name == 'merge_group' }}
-    runs-on: ubuntu-20.04-8-cores
+    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

@@ -3,9 +3,13 @@ on:
   pull_request_target:
     types: [opened, edited]
 
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
 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-22.04
-          - ubuntu-20.04-8-cores
+          - ubuntu-22.04-8-cores
-          - ubuntu-20.04-16-cores
+          - ubuntu-22.04-16-cores
-          - ubuntu-20.04-32-cores
+          - ubuntu-22.04-32-cores
-          - ubuntu-20.04-64-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
+      - uses: cachix/install-nix-action@v31
-        with:
+      - run: nix develop --command cargo check --bin greptime
-          nix_path: nixpkgs=channel:nixos-24.11
+        env:
-      - run: nix develop --command cargo build
+          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.
+    # The jobs are triggered by the following events:
-    runs-on: ubuntu-20.04-16-cores
+    # 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-ubuntu: ${{ inputs.release_dev_builder_ubuntu_image || github.event_name == 'push' }}
-          build-dev-builder-centos: ${{ inputs.release_dev_builder_centos_image }}
+          build-dev-builder-centos: ${{ inputs.release_dev_builder_centos_image || github.event_name == 'push' }}
-          build-dev-builder-android: ${{ inputs.release_dev_builder_android_image }}
+          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-22.04
-          - ubuntu-20.04-8-cores
+          - ubuntu-22.04-8-cores
-          - ubuntu-20.04-16-cores
+          - ubuntu-22.04-16-cores
-          - ubuntu-20.04-32-cores
+          - ubuntu-22.04-32-cores
-          - ubuntu-20.04-64-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:
+  bump-downstream-repo-versions:
-    name: Bump doc version
+    name: Bump downstream repo versions
     if: ${{ github.event_name == 'push' || github.event_name == 'schedule' }}
-    needs: [allocate-runners]
+    needs: [allocate-runners, publish-github-release]
-    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.
@@ -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

@@ -7,14 +7,21 @@ on:
       - reopened
       - edited
 
+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
+

AUTHOR.md

@@ -3,30 +3,28 @@
 ## Individual Committers (in alphabetical order)
 
 * [CookiePieWw](https://github.com/CookiePieWw)
-* [KKould](https://github.com/KKould)
-* [NiwakaDev](https://github.com/NiwakaDev)
 * [etolbakov](https://github.com/etolbakov)
 * [irenjj](https://github.com/irenjj)
-* [tisonkun](https://github.com/tisonkun)
+* [KKould](https://github.com/KKould)
 * [Lanqing Yang](https://github.com/lyang24)
+* [NiwakaDev](https://github.com/NiwakaDev)
+* [tisonkun](https://github.com/tisonkun)
+
 
 ## Team Members (in alphabetical order)
 
-* [Breeze-P](https://github.com/Breeze-P)
-* [GrepTime](https://github.com/GrepTime)
-* [MichaelScofield](https://github.com/MichaelScofield)
-* [Wenjie0329](https://github.com/Wenjie0329)
-* [WenyXu](https://github.com/WenyXu)
-* [ZonaHex](https://github.com/ZonaHex)
 * [apdong2022](https://github.com/apdong2022)
 * [beryl678](https://github.com/beryl678)
+* [Breeze-P](https://github.com/Breeze-P)
 * [daviderli614](https://github.com/daviderli614)
 * [discord9](https://github.com/discord9)
 * [evenyag](https://github.com/evenyag)
 * [fengjiachun](https://github.com/fengjiachun)
 * [fengys1996](https://github.com/fengys1996)
+* [GrepTime](https://github.com/GrepTime)
 * [holalengyu](https://github.com/holalengyu)
 * [killme2008](https://github.com/killme2008)
+* [MichaelScofield](https://github.com/MichaelScofield)
 * [nicecui](https://github.com/nicecui)
 * [paomian](https://github.com/paomian)
 * [shuiyisong](https://github.com/shuiyisong)
@@ -34,11 +32,14 @@
 * [sunng87](https://github.com/sunng87)
 * [v0y4g3r](https://github.com/v0y4g3r)
 * [waynexia](https://github.com/waynexia)
+* [Wenjie0329](https://github.com/Wenjie0329)
+* [WenyXu](https://github.com/WenyXu)
 * [xtang](https://github.com/xtang)
 * [zhaoyingnan01](https://github.com/zhaoyingnan01)
 * [zhongzc](https://github.com/zhongzc)
+* [ZonaHex](https://github.com/ZonaHex)
 * [zyy17](https://github.com/zyy17)
 
 ## All Contributors
 
-[![All Contributors](https://contrib.rocks/image?repo=GreptimeTeam/greptimedb)](https://github.com/GreptimeTeam/greptimedb/graphs/contributors)
+To see the full list of contributors, please visit our [Contributors page](https://github.com/GreptimeTeam/greptimedb/graphs/contributors)

Cargo.lock

@@ -185,7 +185,7 @@ checksum = "d301b3b94cb4b2f23d7917810addbbaff90738e0ca2be692bd027e70d7e0330c"
 
 [[package]]
 name = "api"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-base",
  "common-decimal",
@@ -432,7 +432,7 @@ dependencies = [
  "arrow-schema",
  "chrono",
  "half",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "lexical-core",
  "num",
  "serde",
@@ -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",
@@ -1475,7 +1475,7 @@ version = "0.13.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6026d8cd82ada8bbcfe337805dd1eb6afdc9e80fa4d57e977b3a36315e0c5525"
 dependencies = [
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "lazy_static",
  "num-traits",
  "regex",
@@ -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",
@@ -2007,12 +2021,14 @@ dependencies = [
 
 [[package]]
 name = "common-function"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
+ "ahash 0.8.11",
  "api",
  "approx 0.5.1",
  "arc-swap",
  "async-trait",
+ "bincode",
  "common-base",
  "common-catalog",
  "common-error",
@@ -2030,6 +2046,7 @@ dependencies = [
  "geo-types",
  "geohash",
  "h3o",
+ "hyperloglogplus",
  "jsonb",
  "nalgebra 0.33.2",
  "num",
@@ -2046,12 +2063,13 @@ dependencies = [
  "store-api",
  "table",
  "tokio",
+ "uddsketch",
  "wkt",
 ]
 
 [[package]]
 name = "common-greptimedb-telemetry"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "common-runtime",
@@ -2068,7 +2086,7 @@ dependencies = [
 
 [[package]]
 name = "common-grpc"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arrow-flight",
@@ -2096,7 +2114,7 @@ dependencies = [
 
 [[package]]
 name = "common-grpc-expr"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "common-base",
@@ -2115,7 +2133,7 @@ dependencies = [
 
 [[package]]
 name = "common-macro"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arc-swap",
  "common-query",
@@ -2129,7 +2147,7 @@ dependencies = [
 
 [[package]]
 name = "common-mem-prof"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-error",
  "common-macro",
@@ -2142,7 +2160,7 @@ dependencies = [
 
 [[package]]
 name = "common-meta"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "anymap2",
  "api",
@@ -2202,7 +2220,7 @@ dependencies = [
 
 [[package]]
 name = "common-options"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-grpc",
  "humantime-serde",
@@ -2211,11 +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",
@@ -2227,7 +2245,7 @@ dependencies = [
 
 [[package]]
 name = "common-procedure"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-stream",
  "async-trait",
@@ -2254,7 +2272,7 @@ dependencies = [
 
 [[package]]
 name = "common-procedure-test"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "common-procedure",
@@ -2262,7 +2280,7 @@ dependencies = [
 
 [[package]]
 name = "common-query"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -2288,7 +2306,7 @@ dependencies = [
 
 [[package]]
 name = "common-recordbatch"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arc-swap",
  "common-error",
@@ -2307,7 +2325,7 @@ dependencies = [
 
 [[package]]
 name = "common-runtime"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "clap 4.5.19",
@@ -2337,7 +2355,7 @@ dependencies = [
 
 [[package]]
 name = "common-telemetry"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "atty",
  "backtrace",
@@ -2365,7 +2383,7 @@ dependencies = [
 
 [[package]]
 name = "common-test-util"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "client",
  "common-query",
@@ -2377,7 +2395,7 @@ dependencies = [
 
 [[package]]
 name = "common-time"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arrow",
  "chrono",
@@ -2395,7 +2413,7 @@ dependencies = [
 
 [[package]]
 name = "common-version"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "build-data",
  "const_format",
@@ -2405,7 +2423,7 @@ dependencies = [
 
 [[package]]
 name = "common-wal"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "common-base",
  "common-error",
@@ -2972,7 +2990,7 @@ dependencies = [
  "chrono",
  "half",
  "hashbrown 0.14.5",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "libc",
  "object_store",
  "parquet",
@@ -3032,7 +3050,7 @@ dependencies = [
  "datafusion-functions-aggregate-common",
  "datafusion-functions-window-common",
  "datafusion-physical-expr-common",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "paste",
  "recursive",
  "serde_json",
@@ -3154,7 +3172,7 @@ dependencies = [
  "datafusion-physical-expr-common",
  "datafusion-physical-plan",
  "half",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "log",
  "parking_lot 0.12.3",
  "paste",
@@ -3205,7 +3223,7 @@ dependencies = [
  "datafusion-common",
  "datafusion-expr",
  "datafusion-physical-expr",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "itertools 0.13.0",
  "log",
  "recursive",
@@ -3230,7 +3248,7 @@ dependencies = [
  "datafusion-physical-expr-common",
  "half",
  "hashbrown 0.14.5",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "itertools 0.13.0",
  "log",
  "paste",
@@ -3289,7 +3307,7 @@ dependencies = [
  "futures",
  "half",
  "hashbrown 0.14.5",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "itertools 0.13.0",
  "log",
  "once_cell",
@@ -3309,7 +3327,7 @@ dependencies = [
  "arrow-schema",
  "datafusion-common",
  "datafusion-expr",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "log",
  "recursive",
  "regex",
@@ -3336,7 +3354,7 @@ dependencies = [
 
 [[package]]
 name = "datanode"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arrow-flight",
@@ -3388,7 +3406,7 @@ dependencies = [
  "session",
  "snafu 0.8.5",
  "store-api",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "table",
  "tokio",
  "toml 0.8.19",
@@ -3397,7 +3415,7 @@ dependencies = [
 
 [[package]]
 name = "datatypes"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -4041,7 +4059,7 @@ dependencies = [
 
 [[package]]
 name = "file-engine"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -4151,7 +4169,7 @@ checksum = "8bf7cc16383c4b8d58b9905a8509f02926ce3058053c056376248d958c9df1e8"
 
 [[package]]
 name = "flow"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arrow",
@@ -4212,7 +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",
@@ -4267,7 +4285,7 @@ checksum = "6c2141d6d6c8512188a7891b4b01590a45f6dac67afb4f255c4124dbb86d4eaa"
 
 [[package]]
 name = "frontend"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arc-swap",
@@ -4695,7 +4713,7 @@ dependencies = [
 [[package]]
 name = "greptime-proto"
 version = "0.1.0"
-source = "git+https://github.com/GreptimeTeam/greptime-proto.git?rev=a25adc8a01340231121646d8f0a29d0e92f45461#a25adc8a01340231121646d8f0a29d0e92f45461"
+source = "git+https://github.com/GreptimeTeam/greptime-proto.git?rev=072ce580502e015df1a6b03a185b60309a7c2a7a#072ce580502e015df1a6b03a185b60309a7c2a7a"
 dependencies = [
  "prost 0.13.3",
  "serde",
@@ -4718,7 +4736,7 @@ dependencies = [
  "futures-sink",
  "futures-util",
  "http 0.2.12",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "slab",
  "tokio",
  "tokio-util",
@@ -4737,7 +4755,7 @@ dependencies = [
  "futures-core",
  "futures-sink",
  "http 1.1.0",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "slab",
  "tokio",
  "tokio-util",
@@ -5287,6 +5305,15 @@ dependencies = [
  "tracing",
 ]
 
+[[package]]
+name = "hyperloglogplus"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "621debdf94dcac33e50475fdd76d34d5ea9c0362a834b9db08c3024696c1fbe3"
+dependencies = [
+ "serde",
+]
+
 [[package]]
 name = "i_float"
 version = "1.3.1"
@@ -5526,7 +5553,7 @@ dependencies = [
 
 [[package]]
 name = "index"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "asynchronous-codec",
@@ -5575,9 +5602,9 @@ dependencies = [
 
 [[package]]
 name = "indexmap"
-version = "2.6.0"
+version = "2.7.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "707907fe3c25f5424cce2cb7e1cbcafee6bdbe735ca90ef77c29e84591e5b9da"
+checksum = "8c9c992b02b5b4c94ea26e32fe5bccb7aa7d9f390ab5c1221ff895bc7ea8b652"
 dependencies = [
  "equivalent",
  "hashbrown 0.15.2",
@@ -5591,7 +5618,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "232929e1d75fe899576a3d5c7416ad0d88dbfbb3c3d6aa00873a7408a50ddb88"
 dependencies = [
  "ahash 0.8.11",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "is-terminal",
  "itoa",
  "log",
@@ -5938,7 +5965,7 @@ version = "0.4.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4ee7893dab2e44ae5f9d0173f26ff4aa327c10b01b06a72b52dd9405b628640d"
 dependencies = [
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
 ]
 
 [[package]]
@@ -6318,7 +6345,7 @@ checksum = "a7a70ba024b9dc04c27ea2f0c0548feb474ec5c54bba33a7f72f873a39d07b24"
 
 [[package]]
 name = "log-query"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "chrono",
  "common-error",
@@ -6330,7 +6357,7 @@ dependencies = [
 
 [[package]]
 name = "log-store"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-stream",
  "async-trait",
@@ -6421,7 +6448,7 @@ dependencies = [
  "cactus",
  "cfgrammar",
  "filetime",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "lazy_static",
  "lrtable",
  "num-traits",
@@ -6623,7 +6650,7 @@ dependencies = [
 
 [[package]]
 name = "meta-client"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -6650,7 +6677,7 @@ dependencies = [
 
 [[package]]
 name = "meta-srv"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -6736,7 +6763,7 @@ dependencies = [
 
 [[package]]
 name = "metric-engine"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "aquamarine",
@@ -6834,7 +6861,7 @@ dependencies = [
 
 [[package]]
 name = "mito2"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "aquamarine",
@@ -7531,7 +7558,7 @@ dependencies = [
 
 [[package]]
 name = "object-store"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "anyhow",
  "bytes",
@@ -7662,7 +7689,7 @@ checksum = "1e32339a5dc40459130b3bd269e9892439f55b33e772d2a9d402a789baaf4e8a"
 dependencies = [
  "futures-core",
  "futures-sink",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "js-sys",
  "once_cell",
  "pin-project-lite",
@@ -7780,7 +7807,7 @@ dependencies = [
 
 [[package]]
 name = "operator"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "api",
@@ -7828,7 +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",
@@ -8065,7 +8092,7 @@ dependencies = [
 
 [[package]]
 name = "partition"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -8234,7 +8261,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b4c5cc86750666a3ed20bdaf5ca2a0344f9c67674cae0515bec2da16fbaa47db"
 dependencies = [
  "fixedbitset",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
 ]
 
 [[package]]
@@ -8333,7 +8360,7 @@ checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"
 
 [[package]]
 name = "pipeline"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "api",
@@ -8473,7 +8500,7 @@ dependencies = [
 
 [[package]]
 name = "plugins"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "auth",
  "clap 4.5.19",
@@ -8735,7 +8762,7 @@ dependencies = [
 
 [[package]]
 name = "promql"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "async-trait",
@@ -8980,7 +9007,7 @@ dependencies = [
 
 [[package]]
 name = "puffin"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-compression 0.4.13",
  "async-trait",
@@ -9021,7 +9048,7 @@ dependencies = [
 
 [[package]]
 name = "query"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "api",
@@ -9086,7 +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",
@@ -10325,7 +10352,7 @@ version = "1.0.137"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "930cfb6e6abf99298aaad7d29abbef7a9999a9a8806a40088f55f0dcec03146b"
 dependencies = [
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "itoa",
  "memchr",
  "ryu",
@@ -10396,7 +10423,7 @@ dependencies = [
  "chrono",
  "hex",
  "indexmap 1.9.3",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "serde",
  "serde_derive",
  "serde_json",
@@ -10422,7 +10449,7 @@ version = "0.9.34+deprecated"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6a8b1a1a2ebf674015cc02edccce75287f1a0130d394307b36743c2f5d504b47"
 dependencies = [
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "itoa",
  "ryu",
  "serde",
@@ -10431,7 +10458,7 @@ dependencies = [
 
 [[package]]
 name = "servers"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "ahash 0.8.11",
  "api",
@@ -10483,6 +10510,7 @@ dependencies = [
  "humantime",
  "humantime-serde",
  "hyper 1.4.1",
+ "indexmap 2.7.1",
  "influxdb_line_protocol",
  "itertools 0.10.5",
  "json5",
@@ -10547,7 +10575,7 @@ dependencies = [
 
 [[package]]
 name = "session"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arc-swap",
@@ -10856,7 +10884,7 @@ dependencies = [
 
 [[package]]
 name = "sql"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "chrono",
@@ -10893,12 +10921,12 @@ dependencies = [
 [[package]]
 name = "sqlness"
 version = "0.6.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
+source = "git+https://github.com/CeresDB/sqlness.git?rev=bb91f31ff58993e07ea89845791235138283a24c#bb91f31ff58993e07ea89845791235138283a24c"
-checksum = "308a7338f2211813d6e9da117e9b9b7aee5d072872d11a934002fd2bd4ab5276"
 dependencies = [
  "async-trait",
  "derive_builder 0.11.2",
  "duration-str",
+ "futures",
  "minijinja",
  "prettydiff",
  "regex",
@@ -10910,7 +10938,7 @@ dependencies = [
 
 [[package]]
 name = "sqlness-runner"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "clap 4.5.19",
@@ -10924,6 +10952,7 @@ dependencies = [
  "hex",
  "local-ip-address",
  "mysql",
+ "num_cpus",
  "reqwest",
  "serde",
  "serde_json",
@@ -11023,7 +11052,7 @@ dependencies = [
  "futures-util",
  "hashbrown 0.15.2",
  "hashlink",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "log",
  "memchr",
  "once_cell",
@@ -11226,7 +11255,7 @@ dependencies = [
 
 [[package]]
 name = "store-api"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "aquamarine",
@@ -11356,7 +11385,7 @@ dependencies = [
 
 [[package]]
 name = "substrait"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "async-trait",
  "bytes",
@@ -11537,7 +11566,7 @@ dependencies = [
 
 [[package]]
 name = "table"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "async-trait",
@@ -11788,7 +11817,7 @@ checksum = "3369f5ac52d5eb6ab48c6b4ffdc8efbcad6b89c765749064ba298f2c68a16a76"
 
 [[package]]
 name = "tests-fuzz"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "arbitrary",
  "async-trait",
@@ -11832,7 +11861,7 @@ dependencies = [
 
 [[package]]
 name = "tests-integration"
-version = "0.12.0"
+version = "0.12.2"
 dependencies = [
  "api",
  "arrow-flight",
@@ -11898,7 +11927,7 @@ dependencies = [
  "sql",
  "sqlx",
  "store-api",
- "substrait 0.12.0",
+ "substrait 0.12.2",
  "table",
  "tempfile",
  "time",
@@ -12319,7 +12348,7 @@ version = "0.19.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1b5bb770da30e5cbfde35a2d7b9b8a2c4b8ef89548a7a6aeab5c9a576e3e7421"
 dependencies = [
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "toml_datetime",
  "winnow 0.5.40",
 ]
@@ -12330,7 +12359,7 @@ version = "0.22.22"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4ae48d6208a266e853d946088ed816055e556cc6028c5e8e2b84d9fa5dd7c7f5"
 dependencies = [
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "serde",
  "serde_spanned",
  "toml_datetime",
@@ -12468,7 +12497,7 @@ dependencies = [
  "futures-core",
  "futures-util",
  "hdrhistogram",
- "indexmap 2.6.0",
+ "indexmap 2.7.1",
  "pin-project-lite",
  "slab",
  "sync_wrapper 1.0.1",
@@ -12956,6 +12985,14 @@ version = "0.1.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2896d95c02a80c6d6a5d6e953d479f5ddf2dfdb6a244441010e373ac0fb88971"
 
+[[package]]
+name = "uddsketch"
+version = "0.1.0"
+source = "git+https://github.com/GreptimeTeam/timescaledb-toolkit.git?rev=84828fe8fb494a6a61412a3da96517fc80f7bb20#84828fe8fb494a6a61412a3da96517fc80f7bb20"
+dependencies = [
+ "serde",
+]
+
 [[package]]
 name = "unescaper"
 version = "0.1.5"

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"
 
@@ -129,7 +129,7 @@ etcd-client = "0.14"
 fst = "0.4.7"
 futures = "0.3"
 futures-util = "0.3"
-greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "a25adc8a01340231121646d8f0a29d0e92f45461" }
+greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "072ce580502e015df1a6b03a185b60309a7c2a7a" }
 hex = "0.4"
 http = "1"
 humantime = "2.1"

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}" && \
+      echo "Extracting ${PACKAGE_NAME}..."
-      mv "${PACKAGE_NAME%.tar.gz}/${BIN}" "${PWD}" && \
+      if ! tar xf "${PACKAGE_NAME}"; then
-      rm -r "${PACKAGE_NAME}" && \
+        echo "Error: Failed to extract ${PACKAGE_NAME}"
-      rm -r "${PACKAGE_NAME%.tar.gz}" && \
+        exit 1
-      echo "Run './${BIN} --help' to get started"
+      fi
+
+      # Find the binary in the extracted directory
+      extracted_dir="${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/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/common/function/Cargo.toml

@@ -12,9 +12,11 @@ default = ["geo"]
 geo = ["geohash", "h3o", "s2", "wkt", "geo-types", "dep:geo"]
 
 [dependencies]
+ahash = "0.8"
 api.workspace = true
 arc-swap = "1.0"
 async-trait.workspace = true
+bincode = "1.3"
 common-base.workspace = true
 common-catalog.workspace = true
 common-error.workspace = true
@@ -32,6 +34,7 @@ geo = { version = "0.29", optional = true }
 geo-types = { version = "0.7", optional = true }
 geohash = { version = "0.13", optional = true }
 h3o = { version = "0.6", optional = true }
+hyperloglogplus = "0.4"
 jsonb.workspace = true
 nalgebra.workspace = true
 num = "0.4"
@@ -47,6 +50,7 @@ sql.workspace = true
 statrs = "0.16"
 store-api.workspace = true
 table.workspace = true
+uddsketch = { git = "https://github.com/GreptimeTeam/timescaledb-toolkit.git", rev = "84828fe8fb494a6a61412a3da96517fc80f7bb20" }
 wkt = { version = "0.11", optional = true }
 
 [dev-dependencies]

src/common/function/src/aggr.rs

@@ -0,0 +1,20 @@
+// 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.
+
+mod hll;
+mod uddsketch_state;
+
+pub(crate) use hll::HllStateType;
+pub use hll::{HllState, HLL_MERGE_NAME, HLL_NAME};
+pub use uddsketch_state::{UddSketchState, UDDSKETCH_STATE_NAME};

src/common/function/src/aggr/hll.rs

@@ -0,0 +1,319 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::sync::Arc;
+
+use common_query::prelude::*;
+use common_telemetry::trace;
+use datafusion::arrow::array::ArrayRef;
+use datafusion::common::cast::{as_binary_array, as_string_array};
+use datafusion::common::not_impl_err;
+use datafusion::error::{DataFusionError, Result as DfResult};
+use datafusion::logical_expr::function::AccumulatorArgs;
+use datafusion::logical_expr::{Accumulator as DfAccumulator, AggregateUDF};
+use datafusion::prelude::create_udaf;
+use datatypes::arrow::datatypes::DataType;
+use hyperloglogplus::{HyperLogLog, HyperLogLogPlus};
+
+use crate::utils::FixedRandomState;
+
+pub const HLL_NAME: &str = "hll";
+pub const HLL_MERGE_NAME: &str = "hll_merge";
+
+const DEFAULT_PRECISION: u8 = 14;
+
+pub(crate) type HllStateType = HyperLogLogPlus<String, FixedRandomState>;
+
+pub struct HllState {
+    hll: HllStateType,
+}
+
+impl std::fmt::Debug for HllState {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "HllState<Opaque>")
+    }
+}
+
+impl Default for HllState {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl HllState {
+    pub fn new() -> Self {
+        Self {
+            // Safety: the DEFAULT_PRECISION is fixed and valid
+            hll: HllStateType::new(DEFAULT_PRECISION, FixedRandomState::new()).unwrap(),
+        }
+    }
+
+    /// Create a UDF for the `hll` function.
+    ///
+    /// `hll` accepts a string column and aggregates the
+    /// values into a HyperLogLog state.
+    pub fn state_udf_impl() -> AggregateUDF {
+        create_udaf(
+            HLL_NAME,
+            vec![DataType::Utf8],
+            Arc::new(DataType::Binary),
+            Volatility::Immutable,
+            Arc::new(Self::create_accumulator),
+            Arc::new(vec![DataType::Binary]),
+        )
+    }
+
+    /// Create a UDF for the `hll_merge` function.
+    ///
+    /// `hll_merge` accepts a binary column of states generated by `hll`
+    /// and merges them into a single state.
+    pub fn merge_udf_impl() -> AggregateUDF {
+        create_udaf(
+            HLL_MERGE_NAME,
+            vec![DataType::Binary],
+            Arc::new(DataType::Binary),
+            Volatility::Immutable,
+            Arc::new(Self::create_merge_accumulator),
+            Arc::new(vec![DataType::Binary]),
+        )
+    }
+
+    fn update(&mut self, value: &str) {
+        self.hll.insert(value);
+    }
+
+    fn merge(&mut self, raw: &[u8]) {
+        if let Ok(serialized) = bincode::deserialize::<HllStateType>(raw) {
+            if let Ok(()) = self.hll.merge(&serialized) {
+                return;
+            }
+        }
+        trace!("Warning: Failed to merge HyperLogLog from {:?}", raw);
+    }
+
+    fn create_accumulator(acc_args: AccumulatorArgs) -> DfResult<Box<dyn DfAccumulator>> {
+        let data_type = acc_args.exprs[0].data_type(acc_args.schema)?;
+
+        match data_type {
+            DataType::Utf8 => Ok(Box::new(HllState::new())),
+            other => not_impl_err!("{HLL_NAME} does not support data type: {other}"),
+        }
+    }
+
+    fn create_merge_accumulator(acc_args: AccumulatorArgs) -> DfResult<Box<dyn DfAccumulator>> {
+        let data_type = acc_args.exprs[0].data_type(acc_args.schema)?;
+
+        match data_type {
+            DataType::Binary => Ok(Box::new(HllState::new())),
+            other => not_impl_err!("{HLL_MERGE_NAME} does not support data type: {other}"),
+        }
+    }
+}
+
+impl DfAccumulator for HllState {
+    fn update_batch(&mut self, values: &[ArrayRef]) -> DfResult<()> {
+        let array = &values[0];
+
+        match array.data_type() {
+            DataType::Utf8 => {
+                let string_array = as_string_array(array)?;
+                for value in string_array.iter().flatten() {
+                    self.update(value);
+                }
+            }
+            DataType::Binary => {
+                let binary_array = as_binary_array(array)?;
+                for v in binary_array.iter().flatten() {
+                    self.merge(v);
+                }
+            }
+            _ => {
+                return not_impl_err!(
+                    "HLL functions do not support data type: {}",
+                    array.data_type()
+                )
+            }
+        }
+
+        Ok(())
+    }
+
+    fn evaluate(&mut self) -> DfResult<ScalarValue> {
+        Ok(ScalarValue::Binary(Some(
+            bincode::serialize(&self.hll).map_err(|e| {
+                DataFusionError::Internal(format!("Failed to serialize HyperLogLog: {}", e))
+            })?,
+        )))
+    }
+
+    fn size(&self) -> usize {
+        std::mem::size_of_val(&self.hll)
+    }
+
+    fn state(&mut self) -> DfResult<Vec<ScalarValue>> {
+        Ok(vec![ScalarValue::Binary(Some(
+            bincode::serialize(&self.hll).map_err(|e| {
+                DataFusionError::Internal(format!("Failed to serialize HyperLogLog: {}", e))
+            })?,
+        ))])
+    }
+
+    fn merge_batch(&mut self, states: &[ArrayRef]) -> DfResult<()> {
+        let array = &states[0];
+        let binary_array = as_binary_array(array)?;
+        for v in binary_array.iter().flatten() {
+            self.merge(v);
+        }
+
+        Ok(())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use datafusion::arrow::array::{BinaryArray, StringArray};
+
+    use super::*;
+
+    #[test]
+    fn test_hll_basic() {
+        let mut state = HllState::new();
+        state.update("1");
+        state.update("2");
+        state.update("3");
+
+        let result = state.evaluate().unwrap();
+        if let ScalarValue::Binary(Some(bytes)) = result {
+            let mut hll: HllStateType = bincode::deserialize(&bytes).unwrap();
+            assert_eq!(hll.count().trunc() as u32, 3);
+        } else {
+            panic!("Expected binary scalar value");
+        }
+    }
+
+    #[test]
+    fn test_hll_roundtrip() {
+        let mut state = HllState::new();
+        state.update("1");
+        state.update("2");
+
+        // Serialize
+        let serialized = state.evaluate().unwrap();
+
+        // Create new state and merge the serialized data
+        let mut new_state = HllState::new();
+        if let ScalarValue::Binary(Some(bytes)) = &serialized {
+            new_state.merge(bytes);
+
+            // Verify the merged state matches original
+            let result = new_state.evaluate().unwrap();
+            if let ScalarValue::Binary(Some(new_bytes)) = result {
+                let mut original: HllStateType = bincode::deserialize(bytes).unwrap();
+                let mut merged: HllStateType = bincode::deserialize(&new_bytes).unwrap();
+                assert_eq!(original.count(), merged.count());
+            } else {
+                panic!("Expected binary scalar value");
+            }
+        } else {
+            panic!("Expected binary scalar value");
+        }
+    }
+
+    #[test]
+    fn test_hll_batch_update() {
+        let mut state = HllState::new();
+
+        // Test string values
+        let str_values = vec!["a", "b", "c", "d", "e", "f", "g", "h", "i"];
+        let str_array = Arc::new(StringArray::from(str_values)) as ArrayRef;
+        state.update_batch(&[str_array]).unwrap();
+
+        let result = state.evaluate().unwrap();
+        if let ScalarValue::Binary(Some(bytes)) = result {
+            let mut hll: HllStateType = bincode::deserialize(&bytes).unwrap();
+            assert_eq!(hll.count().trunc() as u32, 9);
+        } else {
+            panic!("Expected binary scalar value");
+        }
+    }
+
+    #[test]
+    fn test_hll_merge_batch() {
+        let mut state1 = HllState::new();
+        state1.update("1");
+        let state1_binary = state1.evaluate().unwrap();
+
+        let mut state2 = HllState::new();
+        state2.update("2");
+        let state2_binary = state2.evaluate().unwrap();
+
+        let mut merged_state = HllState::new();
+        if let (ScalarValue::Binary(Some(bytes1)), ScalarValue::Binary(Some(bytes2))) =
+            (&state1_binary, &state2_binary)
+        {
+            let binary_array = Arc::new(BinaryArray::from(vec![
+                bytes1.as_slice(),
+                bytes2.as_slice(),
+            ])) as ArrayRef;
+            merged_state.merge_batch(&[binary_array]).unwrap();
+
+            let result = merged_state.evaluate().unwrap();
+            if let ScalarValue::Binary(Some(bytes)) = result {
+                let mut hll: HllStateType = bincode::deserialize(&bytes).unwrap();
+                assert_eq!(hll.count().trunc() as u32, 2);
+            } else {
+                panic!("Expected binary scalar value");
+            }
+        } else {
+            panic!("Expected binary scalar values");
+        }
+    }
+
+    #[test]
+    fn test_hll_merge_function() {
+        // Create two HLL states with different values
+        let mut state1 = HllState::new();
+        state1.update("1");
+        state1.update("2");
+        let state1_binary = state1.evaluate().unwrap();
+
+        let mut state2 = HllState::new();
+        state2.update("2");
+        state2.update("3");
+        let state2_binary = state2.evaluate().unwrap();
+
+        // Create a merge state and merge both states
+        let mut merge_state = HllState::new();
+        if let (ScalarValue::Binary(Some(bytes1)), ScalarValue::Binary(Some(bytes2))) =
+            (&state1_binary, &state2_binary)
+        {
+            let binary_array = Arc::new(BinaryArray::from(vec![
+                bytes1.as_slice(),
+                bytes2.as_slice(),
+            ])) as ArrayRef;
+            merge_state.update_batch(&[binary_array]).unwrap();
+
+            let result = merge_state.evaluate().unwrap();
+            if let ScalarValue::Binary(Some(bytes)) = result {
+                let mut hll: HllStateType = bincode::deserialize(&bytes).unwrap();
+                // Should have 3 unique values: "1", "2", "3"
+                assert_eq!(hll.count().trunc() as u32, 3);
+            } else {
+                panic!("Expected binary scalar value");
+            }
+        } else {
+            panic!("Expected binary scalar values");
+        }
+    }
+}

src/common/function/src/aggr/uddsketch_state.rs

@@ -0,0 +1,307 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::sync::Arc;
+
+use common_query::prelude::*;
+use common_telemetry::trace;
+use datafusion::common::cast::{as_binary_array, as_primitive_array};
+use datafusion::common::not_impl_err;
+use datafusion::error::{DataFusionError, Result as DfResult};
+use datafusion::logical_expr::function::AccumulatorArgs;
+use datafusion::logical_expr::{Accumulator as DfAccumulator, AggregateUDF};
+use datafusion::physical_plan::expressions::Literal;
+use datafusion::prelude::create_udaf;
+use datatypes::arrow::array::ArrayRef;
+use datatypes::arrow::datatypes::{DataType, Float64Type};
+use uddsketch::{SketchHashKey, UDDSketch};
+
+pub const UDDSKETCH_STATE_NAME: &str = "uddsketch_state";
+
+#[derive(Debug)]
+pub struct UddSketchState {
+    uddsketch: UDDSketch,
+}
+
+impl UddSketchState {
+    pub fn new(bucket_size: u64, error_rate: f64) -> Self {
+        Self {
+            uddsketch: UDDSketch::new(bucket_size, error_rate),
+        }
+    }
+
+    pub fn udf_impl() -> AggregateUDF {
+        create_udaf(
+            UDDSKETCH_STATE_NAME,
+            vec![DataType::Int64, DataType::Float64, DataType::Float64],
+            Arc::new(DataType::Binary),
+            Volatility::Immutable,
+            Arc::new(|args| {
+                let (bucket_size, error_rate) = downcast_accumulator_args(args)?;
+                Ok(Box::new(UddSketchState::new(bucket_size, error_rate)))
+            }),
+            Arc::new(vec![DataType::Binary]),
+        )
+    }
+
+    fn update(&mut self, value: f64) {
+        self.uddsketch.add_value(value);
+    }
+
+    fn merge(&mut self, raw: &[u8]) {
+        if let Ok(uddsketch) = bincode::deserialize::<UDDSketch>(raw) {
+            if uddsketch.count() != 0 {
+                self.uddsketch.merge_sketch(&uddsketch);
+            }
+        } else {
+            trace!("Warning: Failed to deserialize UDDSketch from {:?}", raw);
+        }
+    }
+}
+
+fn downcast_accumulator_args(args: AccumulatorArgs) -> DfResult<(u64, f64)> {
+    let bucket_size = match args.exprs[0]
+        .as_any()
+        .downcast_ref::<Literal>()
+        .map(|lit| lit.value())
+    {
+        Some(ScalarValue::Int64(Some(value))) => *value as u64,
+        _ => {
+            return not_impl_err!(
+                "{} not supported for bucket size: {}",
+                UDDSKETCH_STATE_NAME,
+                &args.exprs[0]
+            )
+        }
+    };
+
+    let error_rate = match args.exprs[1]
+        .as_any()
+        .downcast_ref::<Literal>()
+        .map(|lit| lit.value())
+    {
+        Some(ScalarValue::Float64(Some(value))) => *value,
+        _ => {
+            return not_impl_err!(
+                "{} not supported for error rate: {}",
+                UDDSKETCH_STATE_NAME,
+                &args.exprs[1]
+            )
+        }
+    };
+
+    Ok((bucket_size, error_rate))
+}
+
+impl DfAccumulator for UddSketchState {
+    fn update_batch(&mut self, values: &[ArrayRef]) -> DfResult<()> {
+        let array = &values[2]; // the third column is data value
+        let f64_array = as_primitive_array::<Float64Type>(array)?;
+        for v in f64_array.iter().flatten() {
+            self.update(v);
+        }
+
+        Ok(())
+    }
+
+    fn evaluate(&mut self) -> DfResult<ScalarValue> {
+        Ok(ScalarValue::Binary(Some(
+            bincode::serialize(&self.uddsketch).map_err(|e| {
+                DataFusionError::Internal(format!("Failed to serialize UDDSketch: {}", e))
+            })?,
+        )))
+    }
+
+    fn size(&self) -> usize {
+        // Base size of UDDSketch struct fields
+        let mut total_size = std::mem::size_of::<f64>() * 3 + // alpha, gamma, values_sum
+                            std::mem::size_of::<u32>() +      // compactions
+                            std::mem::size_of::<u64>() * 2; // max_buckets, num_values
+
+        // Size of buckets (SketchHashMap)
+        // Each bucket entry contains:
+        // - SketchHashKey (enum with i64/Zero/Invalid variants)
+        // - SketchHashEntry (count: u64, next: SketchHashKey)
+        let bucket_entry_size = std::mem::size_of::<SketchHashKey>() + // key
+                               std::mem::size_of::<u64>() +            // count
+                               std::mem::size_of::<SketchHashKey>(); // next
+
+        total_size += self.uddsketch.current_buckets_count() * bucket_entry_size;
+
+        total_size
+    }
+
+    fn state(&mut self) -> DfResult<Vec<ScalarValue>> {
+        Ok(vec![ScalarValue::Binary(Some(
+            bincode::serialize(&self.uddsketch).map_err(|e| {
+                DataFusionError::Internal(format!("Failed to serialize UDDSketch: {}", e))
+            })?,
+        ))])
+    }
+
+    fn merge_batch(&mut self, states: &[ArrayRef]) -> DfResult<()> {
+        let array = &states[0];
+        let binary_array = as_binary_array(array)?;
+        for v in binary_array.iter().flatten() {
+            self.merge(v);
+        }
+
+        Ok(())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use datafusion::arrow::array::{BinaryArray, Float64Array};
+
+    use super::*;
+
+    #[test]
+    fn test_uddsketch_state_basic() {
+        let mut state = UddSketchState::new(10, 0.01);
+        state.update(1.0);
+        state.update(2.0);
+        state.update(3.0);
+
+        let result = state.evaluate().unwrap();
+        if let ScalarValue::Binary(Some(bytes)) = result {
+            let deserialized: UDDSketch = bincode::deserialize(&bytes).unwrap();
+            assert_eq!(deserialized.count(), 3);
+        } else {
+            panic!("Expected binary scalar value");
+        }
+    }
+
+    #[test]
+    fn test_uddsketch_state_roundtrip() {
+        let mut state = UddSketchState::new(10, 0.01);
+        state.update(1.0);
+        state.update(2.0);
+
+        // Serialize
+        let serialized = state.evaluate().unwrap();
+
+        // Create new state and merge the serialized data
+        let mut new_state = UddSketchState::new(10, 0.01);
+        if let ScalarValue::Binary(Some(bytes)) = &serialized {
+            new_state.merge(bytes);
+
+            // Verify the merged state matches original by comparing deserialized values
+            let original_sketch: UDDSketch = bincode::deserialize(bytes).unwrap();
+            let new_result = new_state.evaluate().unwrap();
+            if let ScalarValue::Binary(Some(new_bytes)) = new_result {
+                let new_sketch: UDDSketch = bincode::deserialize(&new_bytes).unwrap();
+                assert_eq!(original_sketch.count(), new_sketch.count());
+                assert_eq!(original_sketch.sum(), new_sketch.sum());
+                assert_eq!(original_sketch.mean(), new_sketch.mean());
+                assert_eq!(original_sketch.max_error(), new_sketch.max_error());
+                // Compare a few quantiles to ensure statistical equivalence
+                for q in [0.1, 0.5, 0.9].iter() {
+                    assert!(
+                        (original_sketch.estimate_quantile(*q) - new_sketch.estimate_quantile(*q))
+                            .abs()
+                            < 1e-10,
+                        "Quantile {} mismatch: original={}, new={}",
+                        q,
+                        original_sketch.estimate_quantile(*q),
+                        new_sketch.estimate_quantile(*q)
+                    );
+                }
+            } else {
+                panic!("Expected binary scalar value");
+            }
+        } else {
+            panic!("Expected binary scalar value");
+        }
+    }
+
+    #[test]
+    fn test_uddsketch_state_batch_update() {
+        let mut state = UddSketchState::new(10, 0.01);
+        let values = vec![1.0f64, 2.0, 3.0];
+        let array = Arc::new(Float64Array::from(values)) as ArrayRef;
+
+        state
+            .update_batch(&[array.clone(), array.clone(), array])
+            .unwrap();
+
+        let result = state.evaluate().unwrap();
+        if let ScalarValue::Binary(Some(bytes)) = result {
+            let deserialized: UDDSketch = bincode::deserialize(&bytes).unwrap();
+            assert_eq!(deserialized.count(), 3);
+        } else {
+            panic!("Expected binary scalar value");
+        }
+    }
+
+    #[test]
+    fn test_uddsketch_state_merge_batch() {
+        let mut state1 = UddSketchState::new(10, 0.01);
+        state1.update(1.0);
+        let state1_binary = state1.evaluate().unwrap();
+
+        let mut state2 = UddSketchState::new(10, 0.01);
+        state2.update(2.0);
+        let state2_binary = state2.evaluate().unwrap();
+
+        let mut merged_state = UddSketchState::new(10, 0.01);
+        if let (ScalarValue::Binary(Some(bytes1)), ScalarValue::Binary(Some(bytes2))) =
+            (&state1_binary, &state2_binary)
+        {
+            let binary_array = Arc::new(BinaryArray::from(vec![
+                bytes1.as_slice(),
+                bytes2.as_slice(),
+            ])) as ArrayRef;
+            merged_state.merge_batch(&[binary_array]).unwrap();
+
+            let result = merged_state.evaluate().unwrap();
+            if let ScalarValue::Binary(Some(bytes)) = result {
+                let deserialized: UDDSketch = bincode::deserialize(&bytes).unwrap();
+                assert_eq!(deserialized.count(), 2);
+            } else {
+                panic!("Expected binary scalar value");
+            }
+        } else {
+            panic!("Expected binary scalar values");
+        }
+    }
+
+    #[test]
+    fn test_uddsketch_state_size() {
+        let mut state = UddSketchState::new(10, 0.01);
+        let initial_size = state.size();
+
+        // Add some values to create buckets
+        state.update(1.0);
+        state.update(2.0);
+        state.update(3.0);
+
+        let size_with_values = state.size();
+        assert!(
+            size_with_values > initial_size,
+            "Size should increase after adding values: initial={}, with_values={}",
+            initial_size,
+            size_with_values
+        );
+
+        // Verify size increases with more buckets
+        state.update(10.0); // This should create a new bucket
+        assert!(
+            state.size() > size_with_values,
+            "Size should increase after adding new bucket: prev={}, new={}",
+            size_with_values,
+            state.size()
+        );
+    }
+}

src/common/function/src/function_registry.rs

@@ -22,10 +22,12 @@ use crate::function::{AsyncFunctionRef, FunctionRef};
 use crate::scalars::aggregate::{AggregateFunctionMetaRef, AggregateFunctions};
 use crate::scalars::date::DateFunction;
 use crate::scalars::expression::ExpressionFunction;
+use crate::scalars::hll_count::HllCalcFunction;
 use crate::scalars::json::JsonFunction;
 use crate::scalars::matches::MatchesFunction;
 use crate::scalars::math::MathFunction;
 use crate::scalars::timestamp::TimestampFunction;
+use crate::scalars::uddsketch_calc::UddSketchCalcFunction;
 use crate::scalars::vector::VectorFunction;
 use crate::system::SystemFunction;
 use crate::table::TableFunction;
@@ -105,6 +107,8 @@ pub static FUNCTION_REGISTRY: Lazy<Arc<FunctionRegistry>> = Lazy::new(|| {
     TimestampFunction::register(&function_registry);
     DateFunction::register(&function_registry);
     ExpressionFunction::register(&function_registry);
+    UddSketchCalcFunction::register(&function_registry);
+    HllCalcFunction::register(&function_registry);
 
     // Aggregate functions
     AggregateFunctions::register(&function_registry);

src/common/function/src/lib.rs

@@ -21,6 +21,7 @@ pub mod scalars;
 mod system;
 mod table;
 
+pub mod aggr;
 pub mod function;
 pub mod function_registry;
 pub mod handlers;

src/common/function/src/scalars.rs

@@ -22,7 +22,9 @@ pub mod matches;
 pub mod math;
 pub mod vector;
 
+pub(crate) mod hll_count;
 #[cfg(test)]
 pub(crate) mod test;
 pub(crate) mod timestamp;
+pub(crate) mod uddsketch_calc;
 pub mod udf;

src/common/function/src/scalars/hll_count.rs

@@ -0,0 +1,175 @@
+// 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.
+
+//! Implementation of the scalar function `hll_count`.
+
+use std::fmt;
+use std::fmt::Display;
+use std::sync::Arc;
+
+use common_query::error::{DowncastVectorSnafu, InvalidFuncArgsSnafu, Result};
+use common_query::prelude::{Signature, Volatility};
+use datatypes::data_type::ConcreteDataType;
+use datatypes::prelude::Vector;
+use datatypes::scalars::{ScalarVector, ScalarVectorBuilder};
+use datatypes::vectors::{BinaryVector, MutableVector, UInt64VectorBuilder, VectorRef};
+use hyperloglogplus::HyperLogLog;
+use snafu::OptionExt;
+
+use crate::aggr::HllStateType;
+use crate::function::{Function, FunctionContext};
+use crate::function_registry::FunctionRegistry;
+
+const NAME: &str = "hll_count";
+
+/// HllCalcFunction implements the scalar function `hll_count`.
+///
+/// It accepts one argument:
+/// 1. The serialized HyperLogLogPlus state, as produced by the aggregator (binary).
+///
+/// For each row, it deserializes the sketch and returns the estimated cardinality.
+#[derive(Debug, Default)]
+pub struct HllCalcFunction;
+
+impl HllCalcFunction {
+    pub fn register(registry: &FunctionRegistry) {
+        registry.register(Arc::new(HllCalcFunction));
+    }
+}
+
+impl Display for HllCalcFunction {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "{}", NAME.to_ascii_uppercase())
+    }
+}
+
+impl Function for HllCalcFunction {
+    fn name(&self) -> &str {
+        NAME
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::uint64_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        // Only argument: HyperLogLogPlus state (binary)
+        Signature::exact(
+            vec![ConcreteDataType::binary_datatype()],
+            Volatility::Immutable,
+        )
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        if columns.len() != 1 {
+            return InvalidFuncArgsSnafu {
+                err_msg: format!("hll_count expects 1 argument, got {}", columns.len()),
+            }
+            .fail();
+        }
+
+        let hll_vec = columns[0]
+            .as_any()
+            .downcast_ref::<BinaryVector>()
+            .with_context(|| DowncastVectorSnafu {
+                err_msg: format!("expect BinaryVector, got {}", columns[0].vector_type_name()),
+            })?;
+        let len = hll_vec.len();
+        let mut builder = UInt64VectorBuilder::with_capacity(len);
+
+        for i in 0..len {
+            let hll_opt = hll_vec.get_data(i);
+
+            if hll_opt.is_none() {
+                builder.push_null();
+                continue;
+            }
+
+            let hll_bytes = hll_opt.unwrap();
+
+            // Deserialize the HyperLogLogPlus from its bincode representation
+            let mut hll: HllStateType = match bincode::deserialize(hll_bytes) {
+                Ok(h) => h,
+                Err(e) => {
+                    common_telemetry::trace!("Failed to deserialize HyperLogLogPlus: {}", e);
+                    builder.push_null();
+                    continue;
+                }
+            };
+
+            builder.push(Some(hll.count().round() as u64));
+        }
+
+        Ok(builder.to_vector())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use datatypes::vectors::BinaryVector;
+
+    use super::*;
+    use crate::utils::FixedRandomState;
+
+    #[test]
+    fn test_hll_count_function() {
+        let function = HllCalcFunction;
+        assert_eq!("hll_count", function.name());
+        assert_eq!(
+            ConcreteDataType::uint64_datatype(),
+            function
+                .return_type(&[ConcreteDataType::uint64_datatype()])
+                .unwrap()
+        );
+
+        // Create a test HLL
+        let mut hll = HllStateType::new(14, FixedRandomState::new()).unwrap();
+        for i in 1..=10 {
+            hll.insert(&i.to_string());
+        }
+
+        let serialized_bytes = bincode::serialize(&hll).unwrap();
+        let args: Vec<VectorRef> = vec![Arc::new(BinaryVector::from(vec![Some(serialized_bytes)]))];
+
+        let result = function.eval(FunctionContext::default(), &args).unwrap();
+        assert_eq!(result.len(), 1);
+
+        // Test cardinality estimate
+        if let datatypes::value::Value::UInt64(v) = result.get(0) {
+            assert_eq!(v, 10);
+        } else {
+            panic!("Expected uint64 value");
+        }
+    }
+
+    #[test]
+    fn test_hll_count_function_errors() {
+        let function = HllCalcFunction;
+
+        // Test with invalid number of arguments
+        let args: Vec<VectorRef> = vec![];
+        let result = function.eval(FunctionContext::default(), &args);
+        assert!(result.is_err());
+        assert!(result
+            .unwrap_err()
+            .to_string()
+            .contains("hll_count expects 1 argument"));
+
+        // Test with invalid binary data
+        let args: Vec<VectorRef> = vec![Arc::new(BinaryVector::from(vec![Some(vec![1, 2, 3])]))]; // Invalid binary data
+        let result = function.eval(FunctionContext::default(), &args).unwrap();
+        assert_eq!(result.len(), 1);
+        assert!(matches!(result.get(0), datatypes::value::Value::Null));
+    }
+}

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

@@ -0,0 +1,211 @@
+// 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.
+
+//! Implementation of the scalar function `uddsketch_calc`.
+
+use std::fmt;
+use std::fmt::Display;
+use std::sync::Arc;
+
+use common_query::error::{DowncastVectorSnafu, InvalidFuncArgsSnafu, Result};
+use common_query::prelude::{Signature, Volatility};
+use datatypes::data_type::ConcreteDataType;
+use datatypes::prelude::Vector;
+use datatypes::scalars::{ScalarVector, ScalarVectorBuilder};
+use datatypes::vectors::{BinaryVector, Float64VectorBuilder, MutableVector, VectorRef};
+use snafu::OptionExt;
+use uddsketch::UDDSketch;
+
+use crate::function::{Function, FunctionContext};
+use crate::function_registry::FunctionRegistry;
+
+const NAME: &str = "uddsketch_calc";
+
+/// UddSketchCalcFunction implements the scalar function `uddsketch_calc`.
+///
+/// It accepts two arguments:
+/// 1. A percentile (as f64) for which to compute the estimated quantile (e.g. 0.95 for p95).
+/// 2. The serialized UDDSketch state, as produced by the aggregator (binary).
+///
+/// For each row, it deserializes the sketch and returns the computed quantile value.
+#[derive(Debug, Default)]
+pub struct UddSketchCalcFunction;
+
+impl UddSketchCalcFunction {
+    pub fn register(registry: &FunctionRegistry) {
+        registry.register(Arc::new(UddSketchCalcFunction));
+    }
+}
+
+impl Display for UddSketchCalcFunction {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "{}", NAME.to_ascii_uppercase())
+    }
+}
+
+impl Function for UddSketchCalcFunction {
+    fn name(&self) -> &str {
+        NAME
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::float64_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        // First argument: percentile (float64)
+        // Second argument: UDDSketch state (binary)
+        Signature::exact(
+            vec![
+                ConcreteDataType::float64_datatype(),
+                ConcreteDataType::binary_datatype(),
+            ],
+            Volatility::Immutable,
+        )
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        if columns.len() != 2 {
+            return InvalidFuncArgsSnafu {
+                err_msg: format!("uddsketch_calc expects 2 arguments, got {}", columns.len()),
+            }
+            .fail();
+        }
+
+        let perc_vec = &columns[0];
+        let sketch_vec = columns[1]
+            .as_any()
+            .downcast_ref::<BinaryVector>()
+            .with_context(|| DowncastVectorSnafu {
+                err_msg: format!("expect BinaryVector, got {}", columns[1].vector_type_name()),
+            })?;
+        let len = sketch_vec.len();
+        let mut builder = Float64VectorBuilder::with_capacity(len);
+
+        for i in 0..len {
+            let perc_opt = perc_vec.get(i).as_f64_lossy();
+            let sketch_opt = sketch_vec.get_data(i);
+
+            if sketch_opt.is_none() || perc_opt.is_none() {
+                builder.push_null();
+                continue;
+            }
+
+            let sketch_bytes = sketch_opt.unwrap();
+            let perc = perc_opt.unwrap();
+
+            // Deserialize the UDDSketch from its bincode representation
+            let sketch: UDDSketch = match bincode::deserialize(sketch_bytes) {
+                Ok(s) => s,
+                Err(e) => {
+                    common_telemetry::trace!("Failed to deserialize UDDSketch: {}", e);
+                    builder.push_null();
+                    continue;
+                }
+            };
+
+            // Compute the estimated quantile from the sketch
+            let result = sketch.estimate_quantile(perc);
+            builder.push(Some(result));
+        }
+
+        Ok(builder.to_vector())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::Arc;
+
+    use datatypes::vectors::{BinaryVector, Float64Vector};
+
+    use super::*;
+
+    #[test]
+    fn test_uddsketch_calc_function() {
+        let function = UddSketchCalcFunction;
+        assert_eq!("uddsketch_calc", function.name());
+        assert_eq!(
+            ConcreteDataType::float64_datatype(),
+            function
+                .return_type(&[ConcreteDataType::float64_datatype()])
+                .unwrap()
+        );
+
+        // Create a test sketch
+        let mut sketch = UDDSketch::new(128, 0.01);
+        sketch.add_value(10.0);
+        sketch.add_value(20.0);
+        sketch.add_value(30.0);
+        sketch.add_value(40.0);
+        sketch.add_value(50.0);
+        sketch.add_value(60.0);
+        sketch.add_value(70.0);
+        sketch.add_value(80.0);
+        sketch.add_value(90.0);
+        sketch.add_value(100.0);
+
+        // Get expected values directly from the sketch
+        let expected_p50 = sketch.estimate_quantile(0.5);
+        let expected_p90 = sketch.estimate_quantile(0.9);
+        let expected_p95 = sketch.estimate_quantile(0.95);
+
+        let serialized = bincode::serialize(&sketch).unwrap();
+        let percentiles = vec![0.5, 0.9, 0.95];
+
+        let args: Vec<VectorRef> = vec![
+            Arc::new(Float64Vector::from_vec(percentiles.clone())),
+            Arc::new(BinaryVector::from(vec![Some(serialized.clone()); 3])),
+        ];
+
+        let result = function.eval(FunctionContext::default(), &args).unwrap();
+        assert_eq!(result.len(), 3);
+
+        // Test median (p50)
+        assert!(
+            matches!(result.get(0), datatypes::value::Value::Float64(v) if (v - expected_p50).abs() < 1e-10)
+        );
+        // Test p90
+        assert!(
+            matches!(result.get(1), datatypes::value::Value::Float64(v) if (v - expected_p90).abs() < 1e-10)
+        );
+        // Test p95
+        assert!(
+            matches!(result.get(2), datatypes::value::Value::Float64(v) if (v - expected_p95).abs() < 1e-10)
+        );
+    }
+
+    #[test]
+    fn test_uddsketch_calc_function_errors() {
+        let function = UddSketchCalcFunction;
+
+        // Test with invalid number of arguments
+        let args: Vec<VectorRef> = vec![Arc::new(Float64Vector::from_vec(vec![0.95]))];
+        let result = function.eval(FunctionContext::default(), &args);
+        assert!(result.is_err());
+        assert!(result
+            .unwrap_err()
+            .to_string()
+            .contains("uddsketch_calc expects 2 arguments"));
+
+        // Test with invalid binary data
+        let args: Vec<VectorRef> = vec![
+            Arc::new(Float64Vector::from_vec(vec![0.95])),
+            Arc::new(BinaryVector::from(vec![Some(vec![1, 2, 3])])), // Invalid binary data
+        ];
+        let result = function.eval(FunctionContext::default(), &args).unwrap();
+        assert_eq!(result.len(), 1);
+        assert!(matches!(result.get(0), datatypes::value::Value::Null));
+    }
+}

src/common/function/src/utils.rs

@@ -12,6 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::hash::BuildHasher;
+
+use ahash::RandomState;
+use serde::{Deserialize, Serialize};
+
 /// Escapes special characters in the provided pattern string for `LIKE`.
 ///
 /// Specifically, it prefixes the backslash (`\`), percent (`%`), and underscore (`_`)
@@ -32,6 +37,71 @@ pub fn escape_like_pattern(pattern: &str) -> String {
         })
         .collect::<String>()
 }
+
+/// A random state with fixed seeds.
+///
+/// This is used to ensure that the hash values are consistent across
+/// different processes, and easy to serialize and deserialize.
+#[derive(Debug)]
+pub struct FixedRandomState {
+    state: RandomState,
+}
+
+impl FixedRandomState {
+    // some random seeds
+    const RANDOM_SEED_0: u64 = 0x517cc1b727220a95;
+    const RANDOM_SEED_1: u64 = 0x428a2f98d728ae22;
+    const RANDOM_SEED_2: u64 = 0x7137449123ef65cd;
+    const RANDOM_SEED_3: u64 = 0xb5c0fbcfec4d3b2f;
+
+    pub fn new() -> Self {
+        Self {
+            state: ahash::RandomState::with_seeds(
+                Self::RANDOM_SEED_0,
+                Self::RANDOM_SEED_1,
+                Self::RANDOM_SEED_2,
+                Self::RANDOM_SEED_3,
+            ),
+        }
+    }
+}
+
+impl Default for FixedRandomState {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl BuildHasher for FixedRandomState {
+    type Hasher = ahash::AHasher;
+
+    fn build_hasher(&self) -> Self::Hasher {
+        self.state.build_hasher()
+    }
+
+    fn hash_one<T: std::hash::Hash>(&self, x: T) -> u64 {
+        self.state.hash_one(x)
+    }
+}
+
+impl Serialize for FixedRandomState {
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: serde::Serializer,
+    {
+        serializer.serialize_unit()
+    }
+}
+
+impl<'de> Deserialize<'de> for FixedRandomState {
+    fn deserialize<D>(_deserializer: D) -> Result<Self, D::Error>
+    where
+        D: serde::Deserializer<'de>,
+    {
+        Ok(Self::new())
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

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

@@ -16,7 +16,6 @@ use std::collections::HashMap;
 use std::sync::Arc;
 
 use futures::future::BoxFuture;
-use futures::TryStreamExt;
 use moka::future::Cache;
 use moka::ops::compute::Op;
 use table::metadata::TableId;
@@ -54,9 +53,13 @@ fn init_factory(table_flow_manager: TableFlowManagerRef) -> Initializer<TableId,
         Box::pin(async move {
             table_flow_manager
                 .flows(table_id)
-                .map_ok(|(key, value)| (key.flownode_id(), value.peer))
-                .try_collect::<HashMap<_, _>>()
                 .await
+                .map(|flows| {
+                    flows
+                        .into_iter()
+                        .map(|(key, value)| (key.flownode_id(), value.peer))
+                        .collect::<HashMap<_, _>>()
+                })
                 // We must cache the `HashSet` even if it's empty,
                 // to avoid future requests to the remote storage next time;
                 // If the value is added to the remote storage,

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> {
+impl From<&NodeInfoKey> for Vec<u8> {
-    fn from(key: NodeInfoKey) -> Self {
+    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

@@ -15,6 +15,7 @@
 mod metadata;
 
 use std::collections::BTreeMap;
+use std::fmt;
 
 use api::v1::flow::flow_request::Body as PbFlowRequest;
 use api::v1::flow::{CreateRequest, FlowRequest, FlowRequestHeader};
@@ -28,7 +29,6 @@ use common_procedure::{
 use common_telemetry::info;
 use common_telemetry::tracing_context::TracingContext;
 use futures::future::join_all;
-use futures::TryStreamExt;
 use itertools::Itertools;
 use serde::{Deserialize, Serialize};
 use snafu::{ensure, ResultExt};
@@ -77,6 +77,7 @@ impl CreateFlowProcedure {
                 query_context,
                 state: CreateFlowState::Prepare,
                 prev_flow_info_value: None,
+                flow_type: None,
             },
         }
     }
@@ -104,7 +105,7 @@ impl CreateFlowProcedure {
         if create_if_not_exists && or_replace {
             // this is forbidden because not clear what does that mean exactly
             return error::UnsupportedSnafu {
-                operation: "Create flow with both `IF NOT EXISTS` and `OR REPLACE`".to_string(),
+                operation: "Create flow with both `IF NOT EXISTS` and `OR REPLACE`",
             }
             .fail();
         }
@@ -129,9 +130,10 @@ impl CreateFlowProcedure {
                 .flow_metadata_manager
                 .flow_route_manager()
                 .routes(flow_id)
-                .map_ok(|(_, value)| value.peer)
+                .await?
-                .try_collect::<Vec<_>>()
+                .into_iter()
-                .await?;
+                .map(|(_, value)| value.peer)
+                .collect::<Vec<_>>();
             self.data.flow_id = Some(flow_id);
             self.data.peers = peers;
             info!("Replacing flow, flow_id: {}", flow_id);
@@ -175,6 +177,8 @@ impl CreateFlowProcedure {
             self.allocate_flow_id().await?;
         }
         self.data.state = CreateFlowState::CreateFlows;
+        // determine flow type
+        self.data.flow_type = Some(determine_flow_type(&self.data.task));
 
         Ok(Status::executing(true))
     }
@@ -309,6 +313,11 @@ impl Procedure for CreateFlowProcedure {
     }
 }
 
+pub fn determine_flow_type(_flow_task: &CreateFlowTask) -> FlowType {
+    // TODO(discord9): determine flow type
+    FlowType::RecordingRule
+}
+
 /// The state of [CreateFlowProcedure].
 #[derive(Debug, Clone, Serialize, Deserialize, AsRefStr, PartialEq)]
 pub enum CreateFlowState {
@@ -322,6 +331,35 @@ pub enum CreateFlowState {
     CreateMetadata,
 }
 
+/// The type of flow.
+#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
+pub enum FlowType {
+    /// The flow is a recording rule task.
+    RecordingRule,
+    /// The flow is a streaming task.
+    Streaming,
+}
+
+impl FlowType {
+    pub const RECORDING_RULE: &str = "recording_rule";
+    pub const STREAMING: &str = "streaming";
+}
+
+impl Default for FlowType {
+    fn default() -> Self {
+        Self::RecordingRule
+    }
+}
+
+impl fmt::Display for FlowType {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            FlowType::RecordingRule => write!(f, "{}", FlowType::RECORDING_RULE),
+            FlowType::Streaming => write!(f, "{}", FlowType::STREAMING),
+        }
+    }
+}
+
 /// The serializable data.
 #[derive(Debug, Serialize, Deserialize)]
 pub struct CreateFlowData {
@@ -335,6 +373,7 @@ pub struct CreateFlowData {
     /// For verify if prev value is consistent when need to update flow metadata.
     /// only set when `or_replace` is true.
     pub(crate) prev_flow_info_value: Option<DeserializedValueWithBytes<FlowInfoValue>>,
+    pub(crate) flow_type: Option<FlowType>,
 }
 
 impl From<&CreateFlowData> for CreateRequest {
@@ -342,7 +381,7 @@ impl From<&CreateFlowData> for CreateRequest {
         let flow_id = value.flow_id.unwrap();
         let source_table_ids = &value.source_table_ids;
 
-        CreateRequest {
+        let mut req = CreateRequest {
             flow_id: Some(api::v1::FlowId { id: flow_id }),
             source_table_ids: source_table_ids
                 .iter()
@@ -356,7 +395,11 @@ impl From<&CreateFlowData> for CreateRequest {
             comment: value.task.comment.clone(),
             sql: value.task.sql.clone(),
             flow_options: value.task.flow_options.clone(),
-        }
+        };
+
+        let flow_type = value.flow_type.unwrap_or_default().to_string();
+        req.flow_options.insert("flow_type".to_string(), flow_type);
+        req
     }
 }
 
@@ -369,7 +412,7 @@ impl From<&CreateFlowData> for (FlowInfoValue, Vec<(FlowPartitionId, FlowRouteVa
             expire_after,
             comment,
             sql,
-            flow_options: options,
+            flow_options: mut options,
             ..
         } = value.task.clone();
 
@@ -386,19 +429,21 @@ impl From<&CreateFlowData> for (FlowInfoValue, Vec<(FlowPartitionId, FlowRouteVa
             .map(|(idx, peer)| (idx as u32, FlowRouteValue { peer: peer.clone() }))
             .collect::<Vec<_>>();
 
-        (
+        let flow_type = value.flow_type.unwrap_or_default().to_string();
-            FlowInfoValue {
+        options.insert("flow_type".to_string(), flow_type);
-                source_table_ids: value.source_table_ids.clone(),
+
-                sink_table_name,
+        let flow_info = FlowInfoValue {
-                flownode_ids,
+            source_table_ids: value.source_table_ids.clone(),
-                catalog_name,
+            sink_table_name,
-                flow_name,
+            flownode_ids,
-                raw_sql: sql,
+            catalog_name,
-                expire_after,
+            flow_name,
-                comment,
+            raw_sql: sql,
-                options,
+            expire_after,
-            },
+            comment,
-            flow_routes,
+            options,
-        )
+        };
+
+        (flow_info, flow_routes)
     }
 }

src/common/meta/src/ddl/drop_database/executor.rs

@@ -128,7 +128,7 @@ impl State for DropDatabaseExecutor {
             .await?;
         executor.invalidate_table_cache(ddl_ctx).await?;
         executor
-            .on_drop_regions(ddl_ctx, &self.physical_region_routes)
+            .on_drop_regions(ddl_ctx, &self.physical_region_routes, true)
             .await?;
         info!("Table: {}({}) is dropped", self.table_name, self.table_id);
 

src/common/meta/src/ddl/drop_flow/metadata.rs

@@ -13,7 +13,6 @@
 // limitations under the License.
 
 use common_catalog::format_full_flow_name;
-use futures::TryStreamExt;
 use snafu::{ensure, OptionExt};
 
 use crate::ddl::drop_flow::DropFlowProcedure;
@@ -39,9 +38,10 @@ impl DropFlowProcedure {
             .flow_metadata_manager
             .flow_route_manager()
             .routes(self.data.task.flow_id)
-            .map_ok(|(_, value)| value)
+            .await?
-            .try_collect::<Vec<_>>()
+            .into_iter()
-            .await?;
+            .map(|(_, value)| value)
+            .collect::<Vec<_>>();
         ensure!(
             !flow_route_values.is_empty(),
             error::FlowRouteNotFoundSnafu {

src/common/meta/src/ddl/drop_table.rs

@@ -156,7 +156,7 @@ impl DropTableProcedure {
 
     pub async fn on_datanode_drop_regions(&mut self) -> Result<Status> {
         self.executor
-            .on_drop_regions(&self.context, &self.data.physical_region_routes)
+            .on_drop_regions(&self.context, &self.data.physical_region_routes, false)
             .await?;
         self.data.state = DropTableState::DeleteTombstone;
         Ok(Status::executing(true))

src/common/meta/src/ddl/drop_table/executor.rs

@@ -214,6 +214,7 @@ impl DropTableExecutor {
         &self,
         ctx: &DdlContext,
         region_routes: &[RegionRoute],
+        fast_path: bool,
     ) -> Result<()> {
         let leaders = find_leaders(region_routes);
         let mut drop_region_tasks = Vec::with_capacity(leaders.len());
@@ -236,6 +237,7 @@ impl DropTableExecutor {
                     }),
                     body: Some(region_request::Body::Drop(PbDropRegionRequest {
                         region_id: region_id.as_u64(),
+                        fast_path,
                     })),
                 };
                 let datanode = datanode.clone();

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

@@ -16,9 +16,9 @@ pub mod flow_info;
 pub(crate) mod flow_name;
 pub(crate) mod flow_route;
 pub mod flow_state;
+mod flownode_addr_helper;
 pub(crate) mod flownode_flow;
 pub(crate) mod table_flow;
-
 use std::ops::Deref;
 use std::sync::Arc;
 
@@ -506,7 +506,6 @@ mod tests {
         let routes = flow_metadata_manager
             .flow_route_manager()
             .routes(flow_id)
-            .try_collect::<Vec<_>>()
             .await
             .unwrap();
         assert_eq!(
@@ -538,7 +537,6 @@ mod tests {
             let nodes = flow_metadata_manager
                 .table_flow_manager()
                 .flows(table_id)
-                .try_collect::<Vec<_>>()
                 .await
                 .unwrap();
             assert_eq!(
@@ -727,7 +725,6 @@ mod tests {
         let routes = flow_metadata_manager
             .flow_route_manager()
             .routes(flow_id)
-            .try_collect::<Vec<_>>()
             .await
             .unwrap();
         assert_eq!(
@@ -759,7 +756,6 @@ mod tests {
             let nodes = flow_metadata_manager
                 .table_flow_manager()
                 .flows(table_id)
-                .try_collect::<Vec<_>>()
                 .await
                 .unwrap();
             assert_eq!(

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

@@ -12,14 +12,15 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use futures::stream::BoxStream;
+use futures::TryStreamExt;
 use lazy_static::lazy_static;
 use regex::Regex;
 use serde::{Deserialize, Serialize};
 use snafu::OptionExt;
 
 use crate::error::{self, Result};
-use crate::key::flow::FlowScoped;
+use crate::key::flow::{flownode_addr_helper, FlowScoped};
+use crate::key::node_address::NodeAddressKey;
 use crate::key::{BytesAdapter, FlowId, FlowPartitionId, MetadataKey, MetadataValue};
 use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
@@ -167,10 +168,7 @@ impl FlowRouteManager {
     }
 
     /// Retrieves all [FlowRouteValue]s of the specified `flow_id`.
-    pub fn routes(
+    pub async fn routes(&self, flow_id: FlowId) -> Result<Vec<(FlowRouteKey, FlowRouteValue)>> {
-        &self,
-        flow_id: FlowId,
-    ) -> BoxStream<'static, Result<(FlowRouteKey, FlowRouteValue)>> {
         let start_key = FlowRouteKey::range_start_key(flow_id);
         let req = RangeRequest::new().with_prefix(start_key);
         let stream = PaginationStream::new(
@@ -181,7 +179,9 @@ impl FlowRouteManager {
         )
         .into_stream();
 
-        Box::pin(stream)
+        let mut res = stream.try_collect::<Vec<_>>().await?;
+        self.remap_flow_route_addresses(&mut res).await?;
+        Ok(res)
     }
 
     /// Builds a create flow routes transaction.
@@ -203,6 +203,28 @@ impl FlowRouteManager {
 
         Ok(Txn::new().and_then(txns))
     }
+
+    async fn remap_flow_route_addresses(
+        &self,
+        flow_routes: &mut [(FlowRouteKey, FlowRouteValue)],
+    ) -> Result<()> {
+        let keys = flow_routes
+            .iter()
+            .map(|(_, value)| NodeAddressKey::with_flownode(value.peer.id))
+            .collect();
+        let flow_node_addrs =
+            flownode_addr_helper::get_flownode_addresses(&self.kv_backend, keys).await?;
+        for (_, flow_route_value) in flow_routes.iter_mut() {
+            let flownode_id = flow_route_value.peer.id;
+            // If an id lacks a corresponding address in the `flow_node_addrs`,
+            // it means the old address in `table_flow_value` is still valid,
+            // which is expected.
+            if let Some(node_addr) = flow_node_addrs.get(&flownode_id) {
+                flow_route_value.peer.addr = node_addr.peer.addr.clone();
+            }
+        }
+        Ok(())
+    }
 }
 
 #[cfg(test)]

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

@@ -0,0 +1,47 @@
+// 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::HashMap;
+
+use crate::error::Result;
+use crate::key::node_address::{NodeAddressKey, NodeAddressValue};
+use crate::key::{MetadataKey, MetadataValue};
+use crate::kv_backend::KvBackendRef;
+use crate::rpc::store::BatchGetRequest;
+
+/// Get the addresses of the flownodes.
+/// The result is a map: node_id -> NodeAddressValue
+pub(crate) async fn get_flownode_addresses(
+    kv_backend: &KvBackendRef,
+    keys: Vec<NodeAddressKey>,
+) -> Result<HashMap<u64, NodeAddressValue>> {
+    if keys.is_empty() {
+        return Ok(HashMap::default());
+    }
+
+    let req = BatchGetRequest {
+        keys: keys.into_iter().map(|k| k.to_bytes()).collect(),
+    };
+    kv_backend
+        .batch_get(req)
+        .await?
+        .kvs
+        .into_iter()
+        .map(|kv| {
+            let key = NodeAddressKey::from_bytes(&kv.key)?;
+            let value = NodeAddressValue::try_from_raw_value(&kv.value)?;
+            Ok((key.node_id, value))
+        })
+        .collect()
+}

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

@@ -14,7 +14,7 @@
 
 use std::sync::Arc;
 
-use futures::stream::BoxStream;
+use futures::TryStreamExt;
 use lazy_static::lazy_static;
 use regex::Regex;
 use serde::{Deserialize, Serialize};
@@ -22,7 +22,8 @@ use snafu::OptionExt;
 use table::metadata::TableId;
 
 use crate::error::{self, Result};
-use crate::key::flow::FlowScoped;
+use crate::key::flow::{flownode_addr_helper, FlowScoped};
+use crate::key::node_address::NodeAddressKey;
 use crate::key::{BytesAdapter, FlowId, FlowPartitionId, MetadataKey, MetadataValue};
 use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
@@ -196,10 +197,7 @@ impl TableFlowManager {
     /// Retrieves all [TableFlowKey]s of the specified `table_id`.
     ///
     /// TODO(discord9): add cache for it since range request does not support cache.
-    pub fn flows(
+    pub async fn flows(&self, table_id: TableId) -> Result<Vec<(TableFlowKey, TableFlowValue)>> {
-        &self,
-        table_id: TableId,
-    ) -> BoxStream<'static, Result<(TableFlowKey, TableFlowValue)>> {
         let start_key = TableFlowKey::range_start_key(table_id);
         let req = RangeRequest::new().with_prefix(start_key);
         let stream = PaginationStream::new(
@@ -210,7 +208,9 @@ impl TableFlowManager {
         )
         .into_stream();
 
-        Box::pin(stream)
+        let mut res = stream.try_collect::<Vec<_>>().await?;
+        self.remap_table_flow_addresses(&mut res).await?;
+        Ok(res)
     }
 
     /// Builds a create table flow transaction.
@@ -238,6 +238,28 @@ impl TableFlowManager {
 
         Ok(Txn::new().and_then(txns))
     }
+
+    async fn remap_table_flow_addresses(
+        &self,
+        table_flows: &mut [(TableFlowKey, TableFlowValue)],
+    ) -> Result<()> {
+        let keys = table_flows
+            .iter()
+            .map(|(_, value)| NodeAddressKey::with_flownode(value.peer.id))
+            .collect::<Vec<_>>();
+        let flownode_addrs =
+            flownode_addr_helper::get_flownode_addresses(&self.kv_backend, keys).await?;
+        for (_, table_flow_value) in table_flows.iter_mut() {
+            let flownode_id = table_flow_value.peer.id;
+            // If an id lacks a corresponding address in the `flow_node_addrs`,
+            // it means the old address in `table_flow_value` is still valid,
+            // which is expected.
+            if let Some(flownode_addr) = flownode_addrs.get(&flownode_id) {
+                table_flow_value.peer.addr = flownode_addr.peer.addr.clone();
+            }
+        }
+        Ok(())
+    }
 }
 
 #[cfg(test)]

src/common/meta/src/key/node_address.rs

@@ -39,6 +39,10 @@ impl NodeAddressKey {
     pub fn with_datanode(node_id: u64) -> Self {
         Self::new(Role::Datanode, node_id)
     }
+
+    pub fn with_flownode(node_id: u64) -> Self {
+        Self::new(Role::Flownode, node_id)
+    }
 }
 
 #[derive(Debug, PartialEq, Serialize, Deserialize, 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/region_server.rs

@@ -1218,7 +1218,10 @@ mod tests {
         );
 
         let response = mock_region_server
-            .handle_request(region_id, RegionRequest::Drop(RegionDropRequest {}))
+            .handle_request(
+                region_id,
+                RegionRequest::Drop(RegionDropRequest { fast_path: false }),
+            )
             .await
             .unwrap();
         assert_eq!(response.affected_rows, 0);
@@ -1310,7 +1313,10 @@ mod tests {
             .insert(region_id, RegionEngineWithStatus::Ready(engine.clone()));
 
         mock_region_server
-            .handle_request(region_id, RegionRequest::Drop(RegionDropRequest {}))
+            .handle_request(
+                region_id,
+                RegionRequest::Drop(RegionDropRequest { fast_path: false }),
+            )
             .await
             .unwrap_err();
 

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/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 {

src/flow/src/heartbeat.rs

@@ -60,12 +60,12 @@ async fn query_flow_state(
 #[derive(Clone)]
 pub struct HeartbeatTask {
     node_id: u64,
+    node_epoch: u64,
     peer_addr: String,
     meta_client: Arc<MetaClient>,
     report_interval: Duration,
     retry_interval: Duration,
     resp_handler_executor: HeartbeatResponseHandlerExecutorRef,
-    start_time_ms: u64,
     running: Arc<AtomicBool>,
     query_stat_size: Option<SizeReportSender>,
 }
@@ -83,12 +83,12 @@ impl HeartbeatTask {
     ) -> Self {
         Self {
             node_id: opts.node_id.unwrap_or(0),
+            node_epoch: common_time::util::current_time_millis() as u64,
             peer_addr: addrs::resolve_addr(&opts.grpc.bind_addr, Some(&opts.grpc.server_addr)),
             meta_client,
             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,
             running: Arc::new(AtomicBool::new(false)),
             query_stat_size: None,
         }
@@ -103,6 +103,11 @@ impl HeartbeatTask {
             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
@@ -181,7 +186,7 @@ impl HeartbeatTask {
         mut outgoing_rx: mpsc::Receiver<OutgoingMessage>,
     ) {
         let report_interval = self.report_interval;
-        let start_time_ms = self.start_time_ms;
+        let node_epoch = self.node_epoch;
         let self_peer = Some(Peer {
             id: self.node_id,
             addr: self.peer_addr.clone(),
@@ -198,7 +203,8 @@ impl HeartbeatTask {
 
             let heartbeat_request = HeartbeatRequest {
                 peer: self_peer,
-                info: Self::build_node_info(start_time_ms),
+                node_epoch,
+                info: Self::build_node_info(node_epoch),
                 ..Default::default()
             };
 
@@ -208,6 +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
                         }
@@ -230,6 +237,8 @@ impl HeartbeatTask {
                 // set the timeout to half of the report interval so that it wouldn't delay heartbeat if something went horribly wrong
                 latest_report = query_flow_state(&query_stat_size, report_interval / 2).await;
             }
+
+            info!("flownode heartbeat task stopped.");
         });
     }
 
@@ -247,7 +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;
@@ -273,7 +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/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,
+    report_interval: Duration,
-    retry_interval: u64,
+    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,
+            report_interval: heartbeat_opts.interval,
-            retry_interval: heartbeat_opts.retry_interval.as_millis() as u64,
+            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
+                        capture_self.start_with_retry(retry_interval).await;
-                            .start_with_retry(Duration::from_millis(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/frontend/src/instance.rs

@@ -237,6 +237,13 @@ impl Instance {
 
         let output = match stmt {
             Statement::Query(_) | Statement::Explain(_) | Statement::Delete(_) => {
+                // TODO: remove this when format is supported in datafusion
+                if let Statement::Explain(explain) = &stmt {
+                    if let Some(format) = explain.format() {
+                        query_ctx.set_explain_format(format.to_string());
+                    }
+                }
+
                 let stmt = QueryStatement::Sql(stmt);
                 let plan = self
                     .statement_executor

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/index/src/fulltext_index/tests.rs

@@ -25,7 +25,7 @@ use crate::fulltext_index::create::{FulltextIndexCreator, TantivyFulltextIndexCr
 use crate::fulltext_index::search::{FulltextIndexSearcher, RowId, TantivyFulltextIndexSearcher};
 use crate::fulltext_index::{Analyzer, Config};
 
-async fn new_bounded_stager(prefix: &str) -> (TempDir, Arc<BoundedStager>) {
+async fn new_bounded_stager(prefix: &str) -> (TempDir, Arc<BoundedStager<String>>) {
     let staging_dir = create_temp_dir(prefix);
     let path = staging_dir.path().to_path_buf();
     (
@@ -68,13 +68,13 @@ async fn test_search(
     let file_accessor = Arc::new(MockFileAccessor::new(prefix));
     let puffin_manager = FsPuffinManager::new(stager, file_accessor);
 
-    let file_name = "fulltext_index";
+    let file_name = "fulltext_index".to_string();
-    let blob_key = "fulltext_index";
+    let blob_key = "fulltext_index".to_string();
-    let mut writer = puffin_manager.writer(file_name).await.unwrap();
+    let mut writer = puffin_manager.writer(&file_name).await.unwrap();
-    create_index(prefix, &mut writer, blob_key, texts, config).await;
+    create_index(prefix, &mut writer, &blob_key, texts, config).await;
 
-    let reader = puffin_manager.reader(file_name).await.unwrap();
+    let reader = puffin_manager.reader(&file_name).await.unwrap();
-    let index_dir = reader.dir(blob_key).await.unwrap();
+    let index_dir = reader.dir(&blob_key).await.unwrap();
     let searcher = TantivyFulltextIndexSearcher::new(index_dir.path()).unwrap();
     let results = searcher.search(query).await.unwrap();
 

src/meta-client/src/client/heartbeat.rs

@@ -198,13 +198,13 @@ impl Inner {
             }
         );
 
-        let leader = self
+        let leader_addr = self
             .ask_leader
             .as_ref()
             .unwrap()
             .get_leader()
             .context(error::NoLeaderSnafu)?;
-        let mut leader = self.make_client(leader)?;
+        let mut leader = self.make_client(&leader_addr)?;
 
         let (sender, receiver) = mpsc::channel::<HeartbeatRequest>(128);
 
@@ -236,7 +236,11 @@ impl Inner {
             .await
             .map_err(error::Error::from)?
             .context(error::CreateHeartbeatStreamSnafu)?;
-        info!("Success to create heartbeat stream to server: {:#?}", res);
+
+        info!(
+            "Success to create heartbeat stream to server: {}, response: {:#?}",
+            leader_addr, res
+        );
 
         Ok((
             HeartbeatSender::new(self.id, self.role, sender),

src/meta-srv/src/handler.rs

@@ -44,6 +44,7 @@ use mailbox_handler::MailboxHandler;
 use on_leader_start_handler::OnLeaderStartHandler;
 use publish_heartbeat_handler::PublishHeartbeatHandler;
 use region_lease_handler::RegionLeaseHandler;
+use remap_flow_peer_handler::RemapFlowPeerHandler;
 use response_header_handler::ResponseHeaderHandler;
 use snafu::{OptionExt, ResultExt};
 use store_api::storage::RegionId;
@@ -71,6 +72,7 @@ pub mod mailbox_handler;
 pub mod on_leader_start_handler;
 pub mod publish_heartbeat_handler;
 pub mod region_lease_handler;
+pub mod remap_flow_peer_handler;
 pub mod response_header_handler;
 
 #[async_trait::async_trait]
@@ -573,6 +575,7 @@ impl HeartbeatHandlerGroupBuilder {
             self.add_handler_last(publish_heartbeat_handler);
         }
         self.add_handler_last(CollectStatsHandler::new(self.flush_stats_factor));
+        self.add_handler_last(RemapFlowPeerHandler::default());
 
         if let Some(flow_state_handler) = self.flow_state_handler.take() {
             self.add_handler_last(flow_state_handler);
@@ -853,7 +856,7 @@ mod tests {
             .unwrap();
 
         let handlers = group.handlers;
-        assert_eq!(12, handlers.len());
+        assert_eq!(13, handlers.len());
 
         let names = [
             "ResponseHeaderHandler",
@@ -868,6 +871,7 @@ mod tests {
             "MailboxHandler",
             "FilterInactiveRegionStatsHandler",
             "CollectStatsHandler",
+            "RemapFlowPeerHandler",
         ];
 
         for (handler, name) in handlers.iter().zip(names.into_iter()) {
@@ -888,7 +892,7 @@ mod tests {
 
         let group = builder.build().unwrap();
         let handlers = group.handlers;
-        assert_eq!(13, handlers.len());
+        assert_eq!(14, handlers.len());
 
         let names = [
             "ResponseHeaderHandler",
@@ -904,6 +908,7 @@ mod tests {
             "CollectStatsHandler",
             "FilterInactiveRegionStatsHandler",
             "CollectStatsHandler",
+            "RemapFlowPeerHandler",
         ];
 
         for (handler, name) in handlers.iter().zip(names.into_iter()) {
@@ -921,7 +926,7 @@ mod tests {
 
         let group = builder.build().unwrap();
         let handlers = group.handlers;
-        assert_eq!(13, handlers.len());
+        assert_eq!(14, handlers.len());
 
         let names = [
             "CollectStatsHandler",
@@ -937,6 +942,7 @@ mod tests {
             "MailboxHandler",
             "FilterInactiveRegionStatsHandler",
             "CollectStatsHandler",
+            "RemapFlowPeerHandler",
         ];
 
         for (handler, name) in handlers.iter().zip(names.into_iter()) {
@@ -954,7 +960,7 @@ mod tests {
 
         let group = builder.build().unwrap();
         let handlers = group.handlers;
-        assert_eq!(13, handlers.len());
+        assert_eq!(14, handlers.len());
 
         let names = [
             "ResponseHeaderHandler",
@@ -970,6 +976,7 @@ mod tests {
             "CollectStatsHandler",
             "FilterInactiveRegionStatsHandler",
             "CollectStatsHandler",
+            "RemapFlowPeerHandler",
         ];
 
         for (handler, name) in handlers.iter().zip(names.into_iter()) {
@@ -987,7 +994,7 @@ mod tests {
 
         let group = builder.build().unwrap();
         let handlers = group.handlers;
-        assert_eq!(13, handlers.len());
+        assert_eq!(14, handlers.len());
 
         let names = [
             "ResponseHeaderHandler",
@@ -1003,6 +1010,7 @@ mod tests {
             "FilterInactiveRegionStatsHandler",
             "CollectStatsHandler",
             "ResponseHeaderHandler",
+            "RemapFlowPeerHandler",
         ];
 
         for (handler, name) in handlers.iter().zip(names.into_iter()) {
@@ -1020,7 +1028,7 @@ mod tests {
 
         let group = builder.build().unwrap();
         let handlers = group.handlers;
-        assert_eq!(12, handlers.len());
+        assert_eq!(13, handlers.len());
 
         let names = [
             "ResponseHeaderHandler",
@@ -1035,6 +1043,7 @@ mod tests {
             "CollectStatsHandler",
             "FilterInactiveRegionStatsHandler",
             "CollectStatsHandler",
+            "RemapFlowPeerHandler",
         ];
 
         for (handler, name) in handlers.iter().zip(names.into_iter()) {
@@ -1052,7 +1061,7 @@ mod tests {
 
         let group = builder.build().unwrap();
         let handlers = group.handlers;
-        assert_eq!(12, handlers.len());
+        assert_eq!(13, handlers.len());
 
         let names = [
             "ResponseHeaderHandler",
@@ -1067,6 +1076,7 @@ mod tests {
             "MailboxHandler",
             "FilterInactiveRegionStatsHandler",
             "ResponseHeaderHandler",
+            "RemapFlowPeerHandler",
         ];
 
         for (handler, name) in handlers.iter().zip(names.into_iter()) {
@@ -1084,7 +1094,7 @@ mod tests {
 
         let group = builder.build().unwrap();
         let handlers = group.handlers;
-        assert_eq!(12, handlers.len());
+        assert_eq!(13, handlers.len());
 
         let names = [
             "CollectStatsHandler",
@@ -1099,6 +1109,7 @@ mod tests {
             "MailboxHandler",
             "FilterInactiveRegionStatsHandler",
             "CollectStatsHandler",
+            "RemapFlowPeerHandler",
         ];
 
         for (handler, name) in handlers.iter().zip(names.into_iter()) {

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

@@ -23,8 +23,8 @@ pub struct CheckLeaderHandler;
 
 #[async_trait::async_trait]
 impl HeartbeatHandler for CheckLeaderHandler {
-    fn is_acceptable(&self, role: Role) -> bool {
+    fn is_acceptable(&self, _role: Role) -> bool {
-        role == Role::Datanode
+        true
     }
 
     async fn handle(

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/handler/collect_stats_handler.rs

@@ -21,7 +21,7 @@ use common_meta::key::node_address::{NodeAddressKey, NodeAddressValue};
 use common_meta::key::{MetadataKey, MetadataValue};
 use common_meta::peer::Peer;
 use common_meta::rpc::store::PutRequest;
-use common_telemetry::{error, warn};
+use common_telemetry::{error, info, warn};
 use dashmap::DashMap;
 use snafu::ResultExt;
 
@@ -185,6 +185,10 @@ async fn rewrite_node_address(ctx: &mut Context, stat: &Stat) {
 
         match ctx.leader_cached_kv_backend.put(put).await {
             Ok(_) => {
+                info!(
+                    "Successfully updated datanode `NodeAddressValue`: {:?}",
+                    peer
+                );
                 // broadcast invalidating cache
                 let cache_idents = stat
                     .table_ids()
@@ -200,11 +204,14 @@ async fn rewrite_node_address(ctx: &mut Context, stat: &Stat) {
                 }
             }
             Err(e) => {
-                error!(e; "Failed to update NodeAddressValue: {:?}", peer);
+                error!(e; "Failed to update datanode `NodeAddressValue`: {:?}", peer);
             }
         }
     } else {
-        warn!("Failed to serialize NodeAddressValue: {:?}", peer);
+        warn!(
+            "Failed to serialize datanode `NodeAddressValue`: {:?}",
+            peer
+        );
     }
 }
 

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

@@ -0,0 +1,92 @@
+// 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 api::v1::meta::{HeartbeatRequest, Peer, Role};
+use common_meta::key::node_address::{NodeAddressKey, NodeAddressValue};
+use common_meta::key::{MetadataKey, MetadataValue};
+use common_meta::rpc::store::PutRequest;
+use common_telemetry::{error, info, warn};
+use dashmap::DashMap;
+
+use crate::handler::{HandleControl, HeartbeatAccumulator, HeartbeatHandler};
+use crate::metasrv::Context;
+use crate::Result;
+
+#[derive(Debug, Default)]
+pub struct RemapFlowPeerHandler {
+    /// flow_node_id -> epoch
+    epoch_cache: DashMap<u64, u64>,
+}
+
+#[async_trait::async_trait]
+impl HeartbeatHandler for RemapFlowPeerHandler {
+    fn is_acceptable(&self, role: Role) -> bool {
+        role == Role::Flownode
+    }
+
+    async fn handle(
+        &self,
+        req: &HeartbeatRequest,
+        ctx: &mut Context,
+        _acc: &mut HeartbeatAccumulator,
+    ) -> Result<HandleControl> {
+        let Some(peer) = req.peer.as_ref() else {
+            return Ok(HandleControl::Continue);
+        };
+
+        let current_epoch = req.node_epoch;
+        let flow_node_id = peer.id;
+
+        let refresh = if let Some(mut epoch) = self.epoch_cache.get_mut(&flow_node_id) {
+            if current_epoch > *epoch.value() {
+                *epoch.value_mut() = current_epoch;
+                true
+            } else {
+                false
+            }
+        } else {
+            self.epoch_cache.insert(flow_node_id, current_epoch);
+            true
+        };
+
+        if refresh {
+            rewrite_node_address(ctx, peer).await;
+        }
+
+        Ok(HandleControl::Continue)
+    }
+}
+
+async fn rewrite_node_address(ctx: &mut Context, peer: &Peer) {
+    let key = NodeAddressKey::with_flownode(peer.id).to_bytes();
+    if let Ok(value) = NodeAddressValue::new(peer.clone().into()).try_as_raw_value() {
+        let put = PutRequest {
+            key,
+            value,
+            prev_kv: false,
+        };
+
+        match ctx.leader_cached_kv_backend.put(put).await {
+            Ok(_) => {
+                info!("Successfully updated flow `NodeAddressValue`: {:?}", peer);
+                // TODO(discord): broadcast invalidating cache to all frontends
+            }
+            Err(e) => {
+                error!(e; "Failed to update flow `NodeAddressValue`: {:?}", peer);
+            }
+        }
+    } else {
+        warn!("Failed to serialize flow `NodeAddressValue`: {:?}", peer);
+    }
+}

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::{self, Result};
-use crate::error::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/meta-srv/src/service/procedure.rs

@@ -17,13 +17,15 @@ use std::time::Duration;
 
 use api::v1::meta::{
     procedure_service_server, DdlTaskRequest as PbDdlTaskRequest,
-    DdlTaskResponse as PbDdlTaskResponse, MigrateRegionRequest, MigrateRegionResponse,
+    DdlTaskResponse as PbDdlTaskResponse, Error, MigrateRegionRequest, MigrateRegionResponse,
     ProcedureDetailRequest, ProcedureDetailResponse, ProcedureStateResponse, QueryProcedureRequest,
+    ResponseHeader,
 };
 use common_meta::ddl::ExecutorContext;
 use common_meta::rpc::ddl::{DdlTask, SubmitDdlTaskRequest};
 use common_meta::rpc::procedure;
-use snafu::{ensure, OptionExt, ResultExt};
+use common_telemetry::warn;
+use snafu::{OptionExt, ResultExt};
 use tonic::{Request, Response};
 
 use super::GrpcResult;
@@ -37,6 +39,16 @@ impl procedure_service_server::ProcedureService for Metasrv {
         &self,
         request: Request<QueryProcedureRequest>,
     ) -> GrpcResult<ProcedureStateResponse> {
+        if !self.is_leader() {
+            let resp = ProcedureStateResponse {
+                header: Some(ResponseHeader::failed(0, Error::is_not_leader())),
+                ..Default::default()
+            };
+
+            warn!("The current meta is not leader, but a `query procedure state` request have reached the meta. Detail: {:?}.", request);
+            return Ok(Response::new(resp));
+        }
+
         let QueryProcedureRequest { header, pid, .. } = request.into_inner();
         let _header = header.context(error::MissingRequestHeaderSnafu)?;
         let pid = pid.context(error::MissingRequiredParameterSnafu { param: "pid" })?;
@@ -57,6 +69,16 @@ impl procedure_service_server::ProcedureService for Metasrv {
     }
 
     async fn ddl(&self, request: Request<PbDdlTaskRequest>) -> GrpcResult<PbDdlTaskResponse> {
+        if !self.is_leader() {
+            let resp = PbDdlTaskResponse {
+                header: Some(ResponseHeader::failed(0, Error::is_not_leader())),
+                ..Default::default()
+            };
+
+            warn!("The current meta is not leader, but a `ddl` request have reached the meta. Detail: {:?}.", request);
+            return Ok(Response::new(resp));
+        }
+
         let PbDdlTaskRequest {
             header,
             query_context,
@@ -99,12 +121,15 @@ impl procedure_service_server::ProcedureService for Metasrv {
         &self,
         request: Request<MigrateRegionRequest>,
     ) -> GrpcResult<MigrateRegionResponse> {
-        ensure!(
+        if !self.is_leader() {
-            self.meta_peer_client().is_leader(),
+            let resp = MigrateRegionResponse {
-            error::UnexpectedSnafu {
+                header: Some(ResponseHeader::failed(0, Error::is_not_leader())),
-                violated: "Trying to submit a region migration procedure to non-leader meta server"
+                ..Default::default()
-            }
+            };
-        );
+
+            warn!("The current meta is not leader, but a `migrate` request have reached the meta. Detail: {:?}.", request);
+            return Ok(Response::new(resp));
+        }
 
         let MigrateRegionRequest {
             header,
@@ -150,6 +175,16 @@ impl procedure_service_server::ProcedureService for Metasrv {
         &self,
         request: Request<ProcedureDetailRequest>,
     ) -> GrpcResult<ProcedureDetailResponse> {
+        if !self.is_leader() {
+            let resp = ProcedureDetailResponse {
+                header: Some(ResponseHeader::failed(0, Error::is_not_leader())),
+                ..Default::default()
+            };
+
+            warn!("The current meta is not leader, but a `procedure details` request have reached the meta. Detail: {:?}.", request);
+            return Ok(Response::new(resp));
+        }
+
         let ProcedureDetailRequest { header } = request.into_inner();
         let _header = header.context(error::MissingRequestHeaderSnafu)?;
         let metas = self

src/metric-engine/src/data_region.rs

@@ -142,6 +142,7 @@ impl DataRegion {
                 c.column_id = new_column_id_start + delta as u32;
                 c.column_schema.set_nullable();
                 match index_options {
+                    IndexOptions::None => {}
                     IndexOptions::Inverted => {
                         c.column_schema.set_inverted_index(true);
                     }

src/metric-engine/src/engine/create.rs

@@ -21,7 +21,7 @@ use api::v1::SemanticType;
 use common_telemetry::info;
 use common_time::{Timestamp, FOREVER};
 use datatypes::data_type::ConcreteDataType;
-use datatypes::schema::ColumnSchema;
+use datatypes::schema::{ColumnSchema, SkippingIndexOptions};
 use datatypes::value::Value;
 use mito2::engine::MITO_ENGINE_NAME;
 use object_store::util::join_dir;
@@ -55,6 +55,8 @@ use crate::error::{
 use crate::metrics::PHYSICAL_REGION_COUNT;
 use crate::utils::{self, to_data_region_id, to_metadata_region_id};
 
+const DEFAULT_TABLE_ID_SKIPPING_INDEX_GRANULARITY: u32 = 1024;
+
 impl MetricEngineInner {
     pub async fn create_regions(
         &self,
@@ -440,6 +442,7 @@ impl MetricEngineInner {
     ///
     /// Return `[table_id_col, tsid_col]`
     fn internal_column_metadata() -> [ColumnMetadata; 2] {
+        // Safety: BloomFilter is a valid skipping index type
         let metric_name_col = ColumnMetadata {
             column_id: ReservedColumnId::table_id(),
             semantic_type: SemanticType::Tag,
@@ -448,7 +451,11 @@ impl MetricEngineInner {
                 ConcreteDataType::uint32_datatype(),
                 false,
             )
-            .with_inverted_index(true),
+            .with_skipping_options(SkippingIndexOptions {
+                granularity: DEFAULT_TABLE_ID_SKIPPING_INDEX_GRANULARITY,
+                index_type: datatypes::schema::SkippingIndexType::BloomFilter,
+            })
+            .unwrap(),
         };
         let tsid_col = ColumnMetadata {
             column_id: ReservedColumnId::tsid(),

src/metric-engine/src/engine/drop.rs

@@ -30,9 +30,10 @@ impl MetricEngineInner {
     pub async fn drop_region(
         &self,
         region_id: RegionId,
-        _req: RegionDropRequest,
+        req: RegionDropRequest,
     ) -> Result<AffectedRows> {
         let data_region_id = utils::to_data_region_id(region_id);
+        let fast_path = req.fast_path;
 
         // enclose the guard in a block to prevent the guard from polluting the async context
         let (is_physical_region, is_physical_region_busy) = {
@@ -52,7 +53,7 @@ impl MetricEngineInner {
 
         if is_physical_region {
             // check if there is no logical region relates to this physical region
-            if is_physical_region_busy {
+            if is_physical_region_busy && !fast_path {
                 // reject if there is any present logical region
                 return Err(PhysicalRegionBusySnafu {
                     region_id: data_region_id,
@@ -60,9 +61,21 @@ impl MetricEngineInner {
                 .build());
             }
 
-            self.drop_physical_region(data_region_id).await
+            return self.drop_physical_region(data_region_id).await;
+        }
+
+        if fast_path {
+            // for fast path, we don't delete the metadata in the metadata region.
+            // it only remove the logical region from the engine state.
+            //
+            // The drop database procedure will ensure the metadata region and data region are dropped eventually.
+            self.state
+                .write()
+                .unwrap()
+                .remove_logical_region(region_id)?;
+
+            Ok(0)
         } else {
-            // cannot merge these two `if` otherwise the stupid type checker will complain
             let metadata_region_id = self
                 .state
                 .read()
@@ -87,13 +100,16 @@ impl MetricEngineInner {
         // Since the physical regions are going to be dropped, we don't need to
         // update the contents in metadata region.
         self.mito
-            .handle_request(data_region_id, RegionRequest::Drop(RegionDropRequest {}))
+            .handle_request(
+                data_region_id,
+                RegionRequest::Drop(RegionDropRequest { fast_path: false }),
+            )
             .await
             .with_context(|_| CloseMitoRegionSnafu { region_id })?;
         self.mito
             .handle_request(
                 metadata_region_id,
-                RegionRequest::Drop(RegionDropRequest {}),
+                RegionRequest::Drop(RegionDropRequest { fast_path: false }),
             )
             .await
             .with_context(|_| CloseMitoRegionSnafu { region_id })?;

src/metric-engine/src/engine/options.rs

@@ -40,6 +40,7 @@ pub struct PhysicalRegionOptions {
 #[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
 pub enum IndexOptions {
     #[default]
+    None,
     Inverted,
     Skipping {
         granularity: u32,

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();
@@ -146,11 +171,14 @@ impl AccessLayer {
         } else {
             // Write cache is disabled.
             let store = self.object_store.clone();
+            let path_provider = RegionFilePathFactory::new(self.region_dir.clone());
             let indexer_builder = IndexerBuilderImpl {
                 op_type: request.op_type,
                 metadata: request.metadata.clone(),
                 row_group_size: write_opts.row_group_size,
-                puffin_manager: self.puffin_manager_factory.build(store),
+                puffin_manager: self
+                    .puffin_manager_factory
+                    .build(store, path_provider.clone()),
                 intermediate_manager: self.intermediate_manager.clone(),
                 index_options: request.index_options,
                 inverted_index_config: request.inverted_index_config,
@@ -161,14 +189,19 @@ impl AccessLayer {
                 self.object_store.clone(),
                 request.metadata,
                 indexer_builder,
-                RegionFilePathFactory {
+                path_provider,
-                    region_dir: self.region_dir.clone(),
-                },
             )
             .await;
-            writer
+            let sst_info = writer
-                .write_all(request.source, request.max_sequence, write_opts)
+                .write_all(request.source, request.max_sequence, write_opts, metrics)
-                .await?
+                .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.
@@ -188,28 +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 struct SstWriteRequest {
-    pub(crate) op_type: OperationType,
+    pub op_type: OperationType,
-    pub(crate) metadata: RegionMetadataRef,
+    pub metadata: RegionMetadataRef,
-    pub(crate) source: Source,
+    pub source: Source,
-    pub(crate) cache_manager: CacheManagerRef,
+    pub cache_manager: CacheManagerRef,
     #[allow(dead_code)]
-    pub(crate) storage: Option<String>,
+    pub storage: Option<String>,
-    pub(crate) max_sequence: Option<SequenceNumber>,
+    pub max_sequence: Option<SequenceNumber>,
 
     /// Configs for index
-    pub(crate) index_options: IndexOptions,
+    pub index_options: IndexOptions,
-    pub(crate) inverted_index_config: InvertedIndexConfig,
+    pub inverted_index_config: InvertedIndexConfig,
-    pub(crate) fulltext_index_config: FulltextIndexConfig,
+    pub fulltext_index_config: FulltextIndexConfig,
-    pub(crate) bloom_filter_index_config: BloomFilterConfig,
+    pub bloom_filter_index_config: BloomFilterConfig,
 }
 
 pub(crate) async fn new_fs_cache_store(root: &str) -> Result<ObjectStore> {
@@ -248,8 +306,18 @@ pub trait FilePathProvider: Send + Sync {
 /// Path provider that builds paths in local write cache.
 #[derive(Clone)]
 pub(crate) struct WriteCachePathProvider {
-    pub(crate) region_id: RegionId,
+    region_id: RegionId,
-    pub(crate) file_cache: FileCacheRef,
+    file_cache: FileCacheRef,
+}
+
+impl WriteCachePathProvider {
+    /// Creates a new `WriteCachePathProvider` instance.
+    pub fn new(region_id: RegionId, file_cache: FileCacheRef) -> Self {
+        Self {
+            region_id,
+            file_cache,
+        }
+    }
 }
 
 impl FilePathProvider for WriteCachePathProvider {
@@ -267,7 +335,14 @@ impl FilePathProvider for WriteCachePathProvider {
 /// Path provider that builds paths in region storage path.
 #[derive(Clone, Debug)]
 pub(crate) struct RegionFilePathFactory {
-    pub(crate) region_dir: String,
+    region_dir: String,
+}
+
+impl RegionFilePathFactory {
+    /// Creates a new `RegionFilePathFactory` instance.
+    pub fn new(region_dir: String) -> Self {
+        Self { region_dir }
+    }
 }
 
 impl FilePathProvider for RegionFilePathFactory {

src/mito2/src/cache/file_cache.rs

@@ -187,9 +187,12 @@ impl FileCache {
     }
 
     /// Removes a file from the cache explicitly.
+    /// It always tries to remove the file from the local store because we may not have the file
+    /// in the memory index if upload is failed.
     pub(crate) async fn remove(&self, key: IndexKey) {
         let file_path = self.cache_file_path(key);
         self.memory_index.remove(&key).await;
+        // Always delete the file from the local store.
         if let Err(e) = self.local_store.delete(&file_path).await {
             warn!(e; "Failed to delete a cached file {}", file_path);
         }

src/mito2/src/cache/write_cache.rs

@@ -22,6 +22,7 @@ use common_telemetry::{debug, info};
 use futures::AsyncWriteExt;
 use object_store::ObjectStore;
 use snafu::ResultExt;
+use store_api::storage::RegionId;
 
 use crate::access_layer::{
     new_fs_cache_store, FilePathProvider, RegionFilePathFactory, SstInfoArray, SstWriteRequest,
@@ -39,6 +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.
 ///
@@ -114,15 +116,14 @@ impl WriteCache {
         let region_id = write_request.metadata.region_id;
 
         let store = self.file_cache.local_store();
-        let path_provider = WriteCachePathProvider {
+        let path_provider = WriteCachePathProvider::new(region_id, self.file_cache.clone());
-            file_cache: self.file_cache.clone(),
-            region_id,
-        };
         let indexer = IndexerBuilderImpl {
             op_type: write_request.op_type,
             metadata: write_request.metadata.clone(),
             row_group_size: write_opts.row_group_size,
-            puffin_manager: self.puffin_manager_factory.build(store),
+            puffin_manager: self
+                .puffin_manager_factory
+                .build(store, path_provider.clone()),
             intermediate_manager: self.intermediate_manager.clone(),
             index_options: write_request.index_options,
             inverted_index_config: write_request.inverted_index_config,
@@ -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();
@@ -150,24 +156,41 @@ impl WriteCache {
             return Ok(sst_info);
         }
 
+        let mut upload_tracker = UploadTracker::new(region_id);
+        let mut err = None;
         let remote_store = &upload_request.remote_store;
         for sst in &sst_info {
             let parquet_key = IndexKey::new(region_id, sst.file_id, FileType::Parquet);
             let parquet_path = upload_request
                 .dest_path_provider
                 .build_sst_file_path(sst.file_id);
-            self.upload(parquet_key, &parquet_path, remote_store)
+            if let Err(e) = self.upload(parquet_key, &parquet_path, remote_store).await {
-                .await?;
+                err = Some(e);
+                break;
+            }
+            upload_tracker.push_uploaded_file(parquet_path);
 
             if sst.index_metadata.file_size > 0 {
                 let puffin_key = IndexKey::new(region_id, sst.file_id, FileType::Puffin);
-                let puffin_path = &upload_request
+                let puffin_path = upload_request
                     .dest_path_provider
                     .build_index_file_path(sst.file_id);
-                self.upload(puffin_key, puffin_path, remote_store).await?;
+                if let Err(e) = self.upload(puffin_key, &puffin_path, remote_store).await {
+                    err = Some(e);
+                    break;
+                }
+                upload_tracker.push_uploaded_file(puffin_path);
             }
         }
 
+        if let Some(err) = err {
+            // Cleans files on failure.
+            upload_tracker
+                .clean(&sst_info, &self.file_cache, remote_store)
+                .await;
+            return Err(err);
+        }
+
         Ok(sst_info)
     }
 
@@ -333,6 +356,61 @@ pub struct SstUploadRequest {
     pub remote_store: ObjectStore,
 }
 
+/// A structs to track files to upload and clean them if upload failed.
+struct UploadTracker {
+    /// Id of the region to track.
+    region_id: RegionId,
+    /// Paths of files uploaded successfully.
+    files_uploaded: Vec<String>,
+}
+
+impl UploadTracker {
+    /// Creates a new instance of `UploadTracker` for a given region.
+    fn new(region_id: RegionId) -> Self {
+        Self {
+            region_id,
+            files_uploaded: Vec::new(),
+        }
+    }
+
+    /// Add a file path to the list of uploaded files.
+    fn push_uploaded_file(&mut self, path: String) {
+        self.files_uploaded.push(path);
+    }
+
+    /// Cleans uploaded files and files in the file cache at best effort.
+    async fn clean(
+        &self,
+        sst_info: &SstInfoArray,
+        file_cache: &FileCacheRef,
+        remote_store: &ObjectStore,
+    ) {
+        common_telemetry::info!(
+            "Start cleaning files on upload failure, region: {}, num_ssts: {}",
+            self.region_id,
+            sst_info.len()
+        );
+
+        // Cleans files in the file cache first.
+        for sst in sst_info {
+            let parquet_key = IndexKey::new(self.region_id, sst.file_id, FileType::Parquet);
+            file_cache.remove(parquet_key).await;
+
+            if sst.index_metadata.file_size > 0 {
+                let puffin_key = IndexKey::new(self.region_id, sst.file_id, FileType::Puffin);
+                file_cache.remove(puffin_key).await;
+            }
+        }
+
+        // Cleans uploaded files.
+        for file_path in &self.files_uploaded {
+            if let Err(e) = remote_store.delete(file_path).await {
+                common_telemetry::error!(e; "Failed to delete file {}", file_path);
+            }
+        }
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use common_test_util::temp_dir::create_temp_dir;
@@ -355,9 +433,7 @@ mod tests {
         // and now just use local file system to mock.
         let mut env = TestEnv::new();
         let mock_store = env.init_object_store_manager();
-        let path_provider = RegionFilePathFactory {
+        let path_provider = RegionFilePathFactory::new("test".to_string());
-            region_dir: "test".to_string(),
-        };
 
         let local_dir = create_temp_dir("");
         let local_store = new_fs_store(local_dir.path().to_str().unwrap());
@@ -488,9 +564,7 @@ mod tests {
             ..Default::default()
         };
         let upload_request = SstUploadRequest {
-            dest_path_provider: RegionFilePathFactory {
+            dest_path_provider: RegionFilePathFactory::new(data_home.clone()),
-                region_dir: data_home.clone(),
-            },
             remote_store: mock_store.clone(),
         };
 

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/engine/drop_test.rs

@@ -56,7 +56,10 @@ async fn test_engine_drop_region() {
 
     // It's okay to drop a region doesn't exist.
     engine
-        .handle_request(region_id, RegionRequest::Drop(RegionDropRequest {}))
+        .handle_request(
+            region_id,
+            RegionRequest::Drop(RegionDropRequest { fast_path: false }),
+        )
         .await
         .unwrap_err();
 
@@ -86,7 +89,10 @@ async fn test_engine_drop_region() {
 
     // drop the created region.
     engine
-        .handle_request(region_id, RegionRequest::Drop(RegionDropRequest {}))
+        .handle_request(
+            region_id,
+            RegionRequest::Drop(RegionDropRequest { fast_path: false }),
+        )
         .await
         .unwrap();
     assert!(!engine.is_region_exists(region_id));
@@ -192,7 +198,10 @@ async fn test_engine_drop_region_for_custom_store() {
 
     // Drop the custom region.
     engine
-        .handle_request(custom_region_id, RegionRequest::Drop(RegionDropRequest {}))
+        .handle_request(
+            custom_region_id,
+            RegionRequest::Drop(RegionDropRequest { fast_path: false }),
+        )
         .await
         .unwrap();
     assert!(!engine.is_region_exists(custom_region_id));

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/read/scan_region.rs

@@ -32,7 +32,6 @@ use tokio::sync::{mpsc, Semaphore};
 use tokio_stream::wrappers::ReceiverStream;
 
 use crate::access_layer::AccessLayerRef;
-use crate::cache::file_cache::FileCacheRef;
 use crate::cache::CacheStrategy;
 use crate::config::DEFAULT_SCAN_CHANNEL_SIZE;
 use crate::error::Result;
@@ -427,12 +426,7 @@ impl ScanRegion {
             return None;
         }
 
-        let file_cache = || -> Option<FileCacheRef> {
+        let file_cache = self.cache_strategy.write_cache().map(|w| w.file_cache());
-            let write_cache = self.cache_strategy.write_cache()?;
-            let file_cache = write_cache.file_cache();
-            Some(file_cache)
-        }();
-
         let inverted_index_cache = self.cache_strategy.inverted_index_cache().cloned();
 
         let puffin_metadata_cache = self.cache_strategy.puffin_metadata_cache().cloned();
@@ -467,14 +461,8 @@ impl ScanRegion {
             return None;
         }
 
-        let file_cache = || -> Option<FileCacheRef> {
+        let file_cache = self.cache_strategy.write_cache().map(|w| w.file_cache());
-            let write_cache = self.cache_strategy.write_cache()?;
-            let file_cache = write_cache.file_cache();
-            Some(file_cache)
-        }();
-
         let bloom_filter_index_cache = self.cache_strategy.bloom_filter_index_cache().cloned();
-
         let puffin_metadata_cache = self.cache_strategy.puffin_metadata_cache().cloned();
 
         BloomFilterIndexApplierBuilder::new(
@@ -499,12 +487,18 @@ impl ScanRegion {
             return None;
         }
 
+        let file_cache = self.cache_strategy.write_cache().map(|w| w.file_cache());
+        let puffin_metadata_cache = self.cache_strategy.puffin_metadata_cache().cloned();
+
         FulltextIndexApplierBuilder::new(
             self.access_layer.region_dir().to_string(),
+            self.version.metadata.region_id,
             self.access_layer.object_store().clone(),
             self.access_layer.puffin_manager_factory().clone(),
             self.version.metadata.as_ref(),
         )
+        .with_file_cache(file_cache)
+        .with_puffin_metadata_cache(puffin_metadata_cache)
         .build(&self.request.filters)
         .inspect_err(|err| warn!(err; "Failed to build fulltext index applier"))
         .ok()

src/mito2/src/request.rs

@@ -35,8 +35,8 @@ use store_api::metadata::{ColumnMetadata, RegionMetadata, RegionMetadataRef};
 use store_api::region_engine::{SetRegionRoleStateResponse, SettableRegionRoleState};
 use store_api::region_request::{
     AffectedRows, RegionAlterRequest, RegionCatchupRequest, RegionCloseRequest,
-    RegionCompactRequest, RegionCreateRequest, RegionDropRequest, RegionFlushRequest,
+    RegionCompactRequest, RegionCreateRequest, RegionFlushRequest, RegionOpenRequest,
-    RegionOpenRequest, RegionRequest, RegionTruncateRequest,
+    RegionRequest, RegionTruncateRequest,
 };
 use store_api::storage::{RegionId, SequenceNumber};
 use tokio::sync::oneshot::{self, Receiver, Sender};
@@ -624,10 +624,10 @@ impl WorkerRequest {
                 sender: sender.into(),
                 request: DdlRequest::Create(v),
             }),
-            RegionRequest::Drop(v) => WorkerRequest::Ddl(SenderDdlRequest {
+            RegionRequest::Drop(_) => WorkerRequest::Ddl(SenderDdlRequest {
                 region_id,
                 sender: sender.into(),
-                request: DdlRequest::Drop(v),
+                request: DdlRequest::Drop,
             }),
             RegionRequest::Open(v) => WorkerRequest::Ddl(SenderDdlRequest {
                 region_id,
@@ -690,7 +690,7 @@ impl WorkerRequest {
 #[derive(Debug)]
 pub(crate) enum DdlRequest {
     Create(RegionCreateRequest),
-    Drop(RegionDropRequest),
+    Drop,
     Open((RegionOpenRequest, Option<WalEntryReceiver>)),
     Close(RegionCloseRequest),
     Alter(RegionAlterRequest),

src/mito2/src/sst/file.rs

@@ -174,31 +174,8 @@ impl FileMeta {
             .contains(&IndexType::BloomFilterIndex)
     }
 
-    /// Returns the size of the inverted index file
+    pub fn index_file_size(&self) -> u64 {
-    pub fn inverted_index_size(&self) -> Option<u64> {
+        self.index_file_size
-        if self.available_indexes.len() == 1 && self.inverted_index_available() {
-            Some(self.index_file_size)
-        } else {
-            None
-        }
-    }
-
-    /// Returns the size of the fulltext index file
-    pub fn fulltext_index_size(&self) -> Option<u64> {
-        if self.available_indexes.len() == 1 && self.fulltext_index_available() {
-            Some(self.index_file_size)
-        } else {
-            None
-        }
-    }
-
-    /// Returns the size of the bloom filter index file
-    pub fn bloom_filter_index_size(&self) -> Option<u64> {
-        if self.available_indexes.len() == 1 && self.bloom_filter_index_available() {
-            Some(self.index_file_size)
-        } else {
-            None
-        }
     }
 }