fix(sqlness): redact all volatile text (#4583)

Commit Message:

 Add SQLNESS replacements for RoundRobinBatch and region patterns

.env.example

@@ -14,10 +14,11 @@ GT_AZBLOB_CONTAINER=AZBLOB container
 GT_AZBLOB_ACCOUNT_NAME=AZBLOB account name
 GT_AZBLOB_ACCOUNT_KEY=AZBLOB account key
 GT_AZBLOB_ENDPOINT=AZBLOB endpoint
-# Settings for gcs test 
-GT_GCS_BUCKET = GCS bucket 
+# Settings for gcs test
+GT_GCS_BUCKET = GCS bucket
 GT_GCS_SCOPE  = GCS scope
-GT_GCS_CREDENTIAL_PATH = GCS credential path 
+GT_GCS_CREDENTIAL_PATH = GCS credential path
+GT_GCS_CREDENTIAL = GCS credential
 GT_GCS_ENDPOINT = GCS end point
 # Settings for kafka wal test
 GT_KAFKA_ENDPOINTS = localhost:9092

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

@@ -62,12 +62,13 @@ runs:
     # Get proper backtraces in mac Sonoma. Currently there's an issue with the new
     # linker that prevents backtraces from getting printed correctly.
     #
-    # <https://github.com/rust-lang/rust/issues/113783> 
+    # <https://github.com/rust-lang/rust/issues/113783>
     - name: Run integration tests
       if: ${{ inputs.disable-run-tests == 'false' }}
       shell: bash
-      env: 
+      env:
         CARGO_BUILD_RUSTFLAGS: "-Clink-arg=-Wl,-ld_classic"
+        SQLNESS_OPTS: "--preserve-state"
       run: |
         make test sqlness-test
 
@@ -81,7 +82,7 @@ runs:
 
     - name: Build greptime binary
       shell: bash
-      env: 
+      env:
         CARGO_BUILD_RUSTFLAGS: "-Clink-arg=-Wl,-ld_classic"
       run: |
         make build \

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

@@ -40,7 +40,7 @@ runs:
     - name: Install Python
       uses: actions/setup-python@v5
       with:
-        python-version: '3.10'
+        python-version: "3.10"
 
     - name: Install PyArrow Package
       shell: pwsh
@@ -62,13 +62,14 @@ runs:
       env:
         RUSTUP_WINDOWS_PATH_ADD_BIN: 1 # Workaround for https://github.com/nextest-rs/nextest/issues/1493
         RUST_BACKTRACE: 1
+        SQLNESS_OPTS: "--preserve-state"
 
     - name: Upload sqlness logs
       if: ${{ failure() }} # Only upload logs when the integration tests failed.
       uses: actions/upload-artifact@v4
       with:
         name: sqlness-logs
-        path: /tmp/greptime-*.log
+        path: C:\Users\RUNNER~1\AppData\Local\Temp\sqlness*
         retention-days: 3
 
     - name: Build greptime binary

.github/actions/setup-greptimedb-cluster/with-disk.yaml

@@ -1,18 +1,13 @@
 meta:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
 datanode:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
+    compact_rt_size = 2
 frontend:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4

.github/actions/setup-greptimedb-cluster/with-minio-and-cache.yaml

@@ -1,29 +1,24 @@
 meta:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
-    
+    global_rt_size = 4
+
     [datanode]
     [datanode.client]
     timeout = "60s"
 datanode:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
+    compact_rt_size = 2
     
     [storage]
     cache_path = "/data/greptimedb/s3cache"
     cache_capacity = "256MB"
 frontend:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
 
     [meta_client]
     ddl_timeout = "60s"

.github/actions/setup-greptimedb-cluster/with-minio.yaml

@@ -1,25 +1,20 @@
 meta:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
     
     [datanode]
     [datanode.client]
     timeout = "60s"
 datanode:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
+    compact_rt_size = 2
 frontend:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
 
     [meta_client]
     ddl_timeout = "60s"

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

@@ -1,9 +1,7 @@
 meta:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
     
     [wal]
     provider = "kafka"
@@ -15,22 +13,19 @@ meta:
     [datanode.client]
     timeout = "60s"
 datanode:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
+    compact_rt_size = 2
 
     [wal]
     provider = "kafka"
     broker_endpoints = ["kafka.kafka-cluster.svc.cluster.local:9092"]
     linger = "2ms"
 frontend:
-  config: |-
+  configData: |-
     [runtime]
-    read_rt_size = 8
-    write_rt_size = 8
-    bg_rt_size = 8
+    global_rt_size = 4
 
     [meta_client]
     ddl_timeout = "60s"
@@ -43,3 +38,8 @@ objectStorage:
   credentials:
     accessKeyId: rootuser
     secretAccessKey: rootpass123
+remoteWal:
+   enabled: true
+   kafka:
+     brokerEndpoints: 
+      - "kafka.kafka-cluster.svc.cluster.local:9092"

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

@@ -0,0 +1,30 @@
+name: Setup PostgreSQL
+description: Deploy PostgreSQL on Kubernetes
+inputs:
+  postgres-replicas:
+    default: 1
+    description: "Number of PostgreSQL replicas"
+  namespace:
+    default: "postgres-namespace"
+  postgres-version:
+    default: "14.2"
+    description: "PostgreSQL version"
+  storage-size:
+    default: "1Gi"
+    description: "Storage size for PostgreSQL"
+
+runs:
+  using: composite
+  steps:
+  - name: Install PostgreSQL
+    shell: bash
+    run: |
+      helm upgrade \
+        --install postgresql oci://registry-1.docker.io/bitnamicharts/postgresql \
+        --set replicaCount=${{ inputs.postgres-replicas }} \
+        --set image.tag=${{ inputs.postgres-version }} \
+        --set persistence.size=${{ inputs.storage-size }} \
+        --set postgresql.username=greptimedb \
+        --set postgresql.password=admin \
+        --create-namespace \
+        -n ${{ inputs.namespace }}

.github/workflows/develop.yml

@@ -141,9 +141,22 @@ jobs:
     runs-on: ubuntu-latest
     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" ]
     steps:
+      - name: Remove unused software
+        run: |
+          echo "Disk space before:"
+          df -h
+          [[ -d /usr/share/dotnet ]] && sudo rm -rf /usr/share/dotnet
+          [[ -d /usr/local/lib/android ]] && sudo rm -rf /usr/local/lib/android
+          [[ -d /opt/ghc ]] && sudo rm -rf /opt/ghc
+          [[ -d /opt/hostedtoolcache/CodeQL ]] && sudo rm -rf /opt/hostedtoolcache/CodeQL
+          sudo docker image prune --all --force
+          sudo docker builder prune -a
+          echo "Disk space after:"
+          df -h
       - uses: actions/checkout@v4
       - uses: arduino/setup-protoc@v3
         with:
@@ -192,6 +205,18 @@ jobs:
       matrix:
         target: [ "unstable_fuzz_create_table_standalone" ]
     steps:
+      - name: Remove unused software
+        run: |
+          echo "Disk space before:"
+          df -h
+          [[ -d /usr/share/dotnet ]] && sudo rm -rf /usr/share/dotnet
+          [[ -d /usr/local/lib/android ]] && sudo rm -rf /usr/local/lib/android
+          [[ -d /opt/ghc ]] && sudo rm -rf /opt/ghc
+          [[ -d /opt/hostedtoolcache/CodeQL ]] && sudo rm -rf /opt/hostedtoolcache/CodeQL
+          sudo docker image prune --all --force
+          sudo docker builder prune -a
+          echo "Disk space after:"
+          df -h
       - uses: actions/checkout@v4
       - uses: arduino/setup-protoc@v3
         with:
@@ -284,24 +309,24 @@ jobs:
     strategy:
       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: 
-          - name: "Disk"
-            minio: false
-            kafka: false
-            values: "with-disk.yaml"
-          - name: "Minio"
-            minio: true
-            kafka: false
-            values: "with-minio.yaml"
-          - name: "Minio with Cache"
-            minio: true
-            kafka: false
-            values: "with-minio-and-cache.yaml"
+        mode:
           - name: "Remote WAL"
             minio: true
             kafka: true
             values: "with-remote-wal.yaml"
     steps:
+      - name: Remove unused software
+        run: |
+          echo "Disk space before:"
+          df -h
+          [[ -d /usr/share/dotnet ]] && sudo rm -rf /usr/share/dotnet
+          [[ -d /usr/local/lib/android ]] && sudo rm -rf /usr/local/lib/android
+          [[ -d /opt/ghc ]] && sudo rm -rf /opt/ghc
+          [[ -d /opt/hostedtoolcache/CodeQL ]] && sudo rm -rf /opt/hostedtoolcache/CodeQL
+          sudo docker image prune --all --force
+          sudo docker builder prune -a
+          echo "Disk space after:"
+          df -h
       - uses: actions/checkout@v4
       - name: Setup Kind
         uses: ./.github/actions/setup-kind
@@ -313,6 +338,8 @@ jobs:
         uses: ./.github/actions/setup-kafka-cluster
       - name: Setup Etcd cluser
         uses: ./.github/actions/setup-etcd-cluster
+      - name: Setup Postgres cluser
+        uses: ./.github/actions/setup-postgres-cluster
       # Prepares for fuzz tests
       - uses: arduino/setup-protoc@v3
         with:
@@ -426,6 +453,18 @@ jobs:
             kafka: true
             values: "with-remote-wal.yaml"
     steps:
+      - name: Remove unused software
+        run: |
+          echo "Disk space before:"
+          df -h
+          [[ -d /usr/share/dotnet ]] && sudo rm -rf /usr/share/dotnet
+          [[ -d /usr/local/lib/android ]] && sudo rm -rf /usr/local/lib/android
+          [[ -d /opt/ghc ]] && sudo rm -rf /opt/ghc
+          [[ -d /opt/hostedtoolcache/CodeQL ]] && sudo rm -rf /opt/hostedtoolcache/CodeQL
+          sudo docker image prune --all --force
+          sudo docker builder prune -a
+          echo "Disk space after:"
+          df -h
       - uses: actions/checkout@v4
       - name: Setup Kind
         uses: ./.github/actions/setup-kind
@@ -439,6 +478,8 @@ jobs:
         uses: ./.github/actions/setup-kafka-cluster
       - name: Setup Etcd cluser
         uses: ./.github/actions/setup-etcd-cluster
+      - name: Setup Postgres cluser
+        uses: ./.github/actions/setup-postgres-cluster
       # Prepares for fuzz tests
       - uses: arduino/setup-protoc@v3
         with:
@@ -556,6 +597,10 @@ jobs:
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
+      - if: matrix.mode.kafka
+        name: Setup kafka server
+        working-directory: tests-integration/fixtures/kafka
+        run: docker compose -f docker-compose-standalone.yml up -d --wait
       - name: Download pre-built binaries
         uses: actions/download-artifact@v4
         with:
@@ -563,10 +608,6 @@ jobs:
           path: .
       - name: Unzip binaries
         run: tar -xvf ./bins.tar.gz
-      - if: matrix.mode.kafka
-        name: Setup kafka server
-        working-directory: tests-integration/fixtures/kafka
-        run: docker compose -f docker-compose-standalone.yml up -d --wait
       - name: Run sqlness
         run: RUST_BACKTRACE=1 ./bins/sqlness-runner ${{ matrix.mode.opts }} -c ./tests/cases --bins-dir ./bins --preserve-state
       - name: Upload sqlness logs
@@ -665,6 +706,9 @@ jobs:
       - name: Setup minio
         working-directory: tests-integration/fixtures/minio
         run: docker compose -f docker-compose-standalone.yml up -d --wait
+      - name: Setup postgres server
+        working-directory: tests-integration/fixtures/postgres
+        run: docker compose -f docker-compose-standalone.yml up -d --wait
       - name: Run nextest cases
         run: cargo llvm-cov nextest --workspace --lcov --output-path lcov.info -F pyo3_backend -F dashboard
         env:
@@ -681,7 +725,9 @@ jobs:
           GT_MINIO_REGION: us-west-2
           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_KAFKA_ENDPOINTS: 127.0.0.1:9092
+          GT_KAFKA_SASL_ENDPOINTS: 127.0.0.1:9093
           UNITTEST_LOG_DIR: "__unittest_logs"
       - name: Codecov upload
         uses: codecov/codecov-action@v4

.github/workflows/nightly-ci.yml

@@ -33,6 +33,13 @@ jobs:
           aws-region: ${{ vars.AWS_CI_TEST_BUCKET_REGION }}
           aws-access-key-id: ${{ secrets.AWS_CI_TEST_ACCESS_KEY_ID }}
           aws-secret-access-key: ${{ secrets.AWS_CI_TEST_SECRET_ACCESS_KEY }}
+      - name: Upload sqlness logs
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: sqlness-logs-kind
+          path: /tmp/kind/
+          retention-days: 3
 
   sqlness-windows:
     name: Sqlness tests on Windows
@@ -51,13 +58,15 @@ jobs:
       - name: Rust Cache
         uses: Swatinem/rust-cache@v2
       - name: Run sqlness
-        run: cargo sqlness
+        run: make sqlness-test
+        env:
+          SQLNESS_OPTS: "--preserve-state"
       - name: Upload sqlness logs
-        if: always()
+        if: failure()
         uses: actions/upload-artifact@v4
         with:
           name: sqlness-logs
-          path: /tmp/greptime-*.log
+          path: C:\Users\RUNNER~1\AppData\Local\Temp\sqlness*
           retention-days: 3
 
   test-on-windows:
@@ -109,11 +118,7 @@ jobs:
 
   check-status:
     name: Check status
-    needs: [
-      sqlness-test,
-      sqlness-windows,
-      test-on-windows,
-    ]
+    needs: [sqlness-test, sqlness-windows, test-on-windows]
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
     runs-on: ubuntu-20.04
     outputs:
@@ -127,9 +132,7 @@ jobs:
   notification:
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' && always() }} # Not requiring successful dependent jobs, always run.
     name: Send notification to Greptime team
-    needs: [
-      check-status
-    ]
+    needs: [check-status]
     runs-on: ubuntu-20.04
     env:
       SLACK_WEBHOOK_URL: ${{ secrets.SLACK_WEBHOOK_URL_DEVELOP_CHANNEL }}

CONTRIBUTING.md

@@ -14,7 +14,7 @@ Follow our [README](https://github.com/GreptimeTeam/greptimedb#readme) to get th
 
 It can feel intimidating to contribute to a complex project, but it can also be exciting and fun. These general notes will help everyone participate in this communal activity.
 
-- Follow the [Code of Conduct](https://github.com/GreptimeTeam/greptimedb/blob/main/CODE_OF_CONDUCT.md)
+- Follow the [Code of Conduct](https://github.com/GreptimeTeam/.github/blob/main/.github/CODE_OF_CONDUCT.md)
 - Small changes make huge differences. We will happily accept a PR making a single character change if it helps move forward. Don't wait to have everything working.
 - Check the closed issues before opening your issue.
 - Try to follow the existing style of the code.
@@ -30,7 +30,7 @@ Pull requests are great, but we accept all kinds of other help if you like. Such
 
 ## Code of Conduct
 
-Also, there are things that we are not looking for because they don't match the goals of the product or benefit the community. Please read [Code of Conduct](https://github.com/GreptimeTeam/greptimedb/blob/main/CODE_OF_CONDUCT.md); we hope everyone can keep good manners and become an honored member.
+Also, there are things that we are not looking for because they don't match the goals of the product or benefit the community. Please read [Code of Conduct](https://github.com/GreptimeTeam/.github/blob/main/.github/CODE_OF_CONDUCT.md); we hope everyone can keep good manners and become an honored member.
 
 ## License
 

Cargo.toml

@@ -64,7 +64,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "0.9.0"
+version = "0.9.2"
 edition = "2021"
 license = "Apache-2.0"
 
@@ -104,27 +104,27 @@ clap = { version = "4.4", features = ["derive"] }
 config = "0.13.0"
 crossbeam-utils = "0.8"
 dashmap = "5.4"
-datafusion = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "d7bda5c9b762426e81f144296deadc87e5f4a0b8" }
-datafusion-common = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "d7bda5c9b762426e81f144296deadc87e5f4a0b8" }
-datafusion-expr = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "d7bda5c9b762426e81f144296deadc87e5f4a0b8" }
-datafusion-functions = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "d7bda5c9b762426e81f144296deadc87e5f4a0b8" }
-datafusion-optimizer = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "d7bda5c9b762426e81f144296deadc87e5f4a0b8" }
-datafusion-physical-expr = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "d7bda5c9b762426e81f144296deadc87e5f4a0b8" }
-datafusion-physical-plan = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "d7bda5c9b762426e81f144296deadc87e5f4a0b8" }
-datafusion-sql = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "d7bda5c9b762426e81f144296deadc87e5f4a0b8" }
-datafusion-substrait = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "d7bda5c9b762426e81f144296deadc87e5f4a0b8" }
+datafusion = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "7823ef2f63663907edab46af0d51359900f608d6" }
+datafusion-common = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "7823ef2f63663907edab46af0d51359900f608d6" }
+datafusion-expr = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "7823ef2f63663907edab46af0d51359900f608d6" }
+datafusion-functions = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "7823ef2f63663907edab46af0d51359900f608d6" }
+datafusion-optimizer = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "7823ef2f63663907edab46af0d51359900f608d6" }
+datafusion-physical-expr = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "7823ef2f63663907edab46af0d51359900f608d6" }
+datafusion-physical-plan = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "7823ef2f63663907edab46af0d51359900f608d6" }
+datafusion-sql = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "7823ef2f63663907edab46af0d51359900f608d6" }
+datafusion-substrait = { git = "https://github.com/waynexia/arrow-datafusion.git", rev = "7823ef2f63663907edab46af0d51359900f608d6" }
 derive_builder = "0.12"
 dotenv = "0.15"
 etcd-client = { version = "0.13" }
 fst = "0.4.7"
 futures = "0.3"
 futures-util = "0.3"
-greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "5c801650435d464891114502539b701c77a1b914" }
+greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "c437b55725b7f5224fe9d46db21072b4a682ee4b" }
 humantime = "2.1"
 humantime-serde = "1.1"
 itertools = "0.10"
 lazy_static = "1.4"
-meter-core = { git = "https://github.com/GreptimeTeam/greptime-meter.git", rev = "80b72716dcde47ec4161478416a5c6c21343364d" }
+meter-core = { git = "https://github.com/GreptimeTeam/greptime-meter.git", rev = "80eb97c24c88af4dd9a86f8bbaf50e741d4eb8cd" }
 mockall = "0.11.4"
 moka = "0.12"
 notify = "6.1"
@@ -151,14 +151,19 @@ reqwest = { version = "0.12", default-features = false, features = [
     "stream",
     "multipart",
 ] }
-rskafka = "0.5"
+# SCRAM-SHA-512 requires https://github.com/dequbed/rsasl/pull/48, https://github.com/influxdata/rskafka/pull/247 
+rskafka = { git = "https://github.com/WenyXu/rskafka.git", rev = "940c6030012c5b746fad819fb72e3325b26e39de", features = [
+    "transport-tls",
+] }
 rstest = "0.21"
 rstest_reuse = "0.7"
 rust_decimal = "1.33"
+rustc-hash = "2.0"
 schemars = "0.8"
 serde = { version = "1.0", features = ["derive"] }
 serde_json = { version = "1.0", features = ["float_roundtrip"] }
 serde_with = "3"
+shadow-rs = "0.31"
 smallvec = { version = "1", features = ["serde"] }
 snafu = "0.8"
 sysinfo = "0.30"
@@ -169,6 +174,7 @@ sqlparser = { git = "https://github.com/GreptimeTeam/sqlparser-rs.git", rev = "5
 strum = { version = "0.25", features = ["derive"] }
 tempfile = "3"
 tokio = { version = "1.36", features = ["full"] }
+tokio-postgres = "0.7"
 tokio-stream = { version = "0.1" }
 tokio-util = { version = "0.7", features = ["io-util", "compat"] }
 toml = "0.8.8"
@@ -183,7 +189,7 @@ auth = { path = "src/auth" }
 cache = { path = "src/cache" }
 catalog = { path = "src/catalog" }
 client = { path = "src/client" }
-cmd = { path = "src/cmd" }
+cmd = { path = "src/cmd", default-features = false }
 common-base = { path = "src/common/base" }
 common-catalog = { path = "src/common/catalog" }
 common-config = { path = "src/common/config" }
@@ -213,7 +219,7 @@ datanode = { path = "src/datanode" }
 datatypes = { path = "src/datatypes" }
 file-engine = { path = "src/file-engine" }
 flow = { path = "src/flow" }
-frontend = { path = "src/frontend" }
+frontend = { path = "src/frontend", default-features = false }
 index = { path = "src/index" }
 log-store = { path = "src/log-store" }
 meta-client = { path = "src/meta-client" }
@@ -238,7 +244,7 @@ table = { path = "src/table" }
 
 [workspace.dependencies.meter-macros]
 git = "https://github.com/GreptimeTeam/greptime-meter.git"
-rev = "80b72716dcde47ec4161478416a5c6c21343364d"
+rev = "80eb97c24c88af4dd9a86f8bbaf50e741d4eb8cd"
 
 [profile.release]
 debug = 1

Makefile

@@ -15,6 +15,7 @@ RUST_TOOLCHAIN ?= $(shell cat rust-toolchain.toml | grep channel | cut -d'"' -f2
 CARGO_REGISTRY_CACHE ?= ${HOME}/.cargo/registry
 ARCH := $(shell uname -m | sed 's/x86_64/amd64/' | sed 's/aarch64/arm64/')
 OUTPUT_DIR := $(shell if [ "$(RELEASE)" = "true" ]; then echo "release"; elif [ ! -z "$(CARGO_PROFILE)" ]; then echo "$(CARGO_PROFILE)" ; else echo "debug"; fi)
+SQLNESS_OPTS ?=
 
 # The arguments for running integration tests.
 ETCD_VERSION ?= v3.5.9
@@ -161,7 +162,7 @@ nextest: ## Install nextest tools.
 
 .PHONY: sqlness-test
 sqlness-test: ## Run sqlness test.
-	cargo sqlness
+	cargo sqlness ${SQLNESS_OPTS}
 
 # Run fuzz test ${FUZZ_TARGET}.
 RUNS ?= 1
@@ -172,7 +173,7 @@ fuzz:
 
 .PHONY: fuzz-ls
 fuzz-ls:
-	cargo fuzz list --fuzz-dir tests-fuzz 
+	cargo fuzz list --fuzz-dir tests-fuzz
 
 .PHONY: check
 check: ## Cargo check all the targets.

README.md

@@ -6,7 +6,7 @@
   </picture>
 </p>
 
-<h2 align="center">Unified Time Series Database for Metrics, Events, and Logs</h2>
+<h2 align="center">Unified Time Series Database for Metrics, Logs, and Events</h2>
 
 <div align="center">
 <h3 align="center">
@@ -50,7 +50,7 @@
 
 ## Introduction
 
-**GreptimeDB** is an open-source unified time-series database for **Metrics**, **Events**, and **Logs** (also **Traces** in plan). You can gain real-time insights from Edge to Cloud at any scale.
+**GreptimeDB** is an open-source unified time-series database for **Metrics**, **Logs**, and **Events** (also **Traces** in plan). You can gain real-time insights from Edge to Cloud at any scale.
 
 ## Why GreptimeDB
 
@@ -58,7 +58,7 @@ Our core developers have been building time-series data platforms for years. Bas
 
 * **Unified all kinds of time series**
 
-  GreptimeDB treats all time series as contextual events with timestamp, and thus unifies the processing of metrics and events. It supports analyzing metrics and events with SQL and PromQL, and doing streaming with continuous aggregation.
+  GreptimeDB treats all time series as contextual events with timestamp, and thus unifies the processing of metrics, logs, and events. It supports analyzing metrics, logs, and events with SQL and PromQL, and doing streaming with continuous aggregation.
 
 * **Cloud-Edge collaboration**
 
@@ -104,10 +104,10 @@ Read more about [Installation](https://docs.greptime.com/getting-started/install
 
 ## Getting Started
 
-* [Quickstart](https://docs.greptime.com/getting-started/quick-start/overview)
-* [Write Data](https://docs.greptime.com/user-guide/clients/overview)
-* [Query Data](https://docs.greptime.com/user-guide/query-data/overview)
-* [Operations](https://docs.greptime.com/user-guide/operations/overview)
+* [Quickstart](https://docs.greptime.com/getting-started/quick-start)
+* [User Guide](https://docs.greptime.com/user-guide/overview)
+* [Demos](https://github.com/GreptimeTeam/demo-scene)
+* [FAQ](https://docs.greptime.com/faq-and-others/faq)
 
 ## Build
 
@@ -150,9 +150,10 @@ Our official Grafana dashboard is available at [grafana](grafana/README.md) dire
 
 ## Project Status
 
-The current version has not yet reached General Availability version standards.
-In line with our Greptime 2024 Roadmap, we plan to achieve a production-level
-version with the update to v1.0 in August. [[Join Force]](https://github.com/GreptimeTeam/greptimedb/issues/3412)
+The current version has not yet reached the standards for General Availability. 
+According to our Greptime 2024 Roadmap, we aim to achieve a production-level version with the release of v1.0 by the end of 2024. [Join Us](https://github.com/GreptimeTeam/greptimedb/issues/3412)
+
+We welcome you to test and use GreptimeDB. Some users have already adopted it in their production environments. If you're interested in trying it out, please use the latest stable release available.
 
 ## Community
 

config/config.md

@@ -16,9 +16,8 @@
 | `enable_telemetry` | Bool | `true` | Enable telemetry to collect anonymous usage data. |
 | `default_timezone` | String | `None` | The default timezone of the server. |
 | `runtime` | -- | -- | The runtime options. |
-| `runtime.read_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
-| `runtime.write_rt_size` | Integer | `8` | The number of threads to execute the runtime for global write operations. |
-| `runtime.bg_rt_size` | Integer | `4` | The number of threads to execute the runtime for global background operations. |
+| `runtime.global_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
+| `runtime.compact_rt_size` | Integer | `4` | The number of threads to execute the runtime for global write operations. |
 | `http` | -- | -- | The HTTP server options. |
 | `http.addr` | String | `127.0.0.1:4000` | The address to bind the HTTP server. |
 | `http.timeout` | String | `30s` | HTTP request timeout. Set to 0 to disable timeout. |
@@ -68,6 +67,11 @@
 | `wal.prefill_log_files` | Bool | `false` | Whether to pre-create log files on start up.<br/>**It's only used when the provider is `raft_engine`**. |
 | `wal.sync_period` | String | `10s` | Duration for fsyncing log files.<br/>**It's only used when the provider is `raft_engine`**. |
 | `wal.broker_endpoints` | Array | -- | The Kafka broker endpoints.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.num_topics` | Integer | `64` | Number of topics to be created upon start.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.selector_type` | String | `round_robin` | Topic selector type.<br/>Available selector types:<br/>- `round_robin` (default)<br/>**It's only used when the provider is `kafka`**. |
+| `wal.topic_name_prefix` | String | `greptimedb_wal_topic` | A Kafka topic is constructed by concatenating `topic_name_prefix` and `topic_id`.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.replication_factor` | Integer | `1` | Expected number of replicas of each partition.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.create_topic_timeout` | String | `30s` | Above which a topic creation operation will be cancelled.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.max_batch_bytes` | String | `1MB` | The max size of a single producer batch.<br/>Warning: Kafka has a default limit of 1MB per message in a topic.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.consumer_wait_timeout` | String | `100ms` | The consumer wait timeout.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.backoff_init` | String | `500ms` | The initial backoff delay.<br/>**It's only used when the provider is `kafka`**. |
@@ -94,6 +98,7 @@
 | `storage.account_key` | String | `None` | The account key of the azure account.<br/>**It's only used when the storage type is `Azblob`**. |
 | `storage.scope` | String | `None` | The scope of the google cloud storage.<br/>**It's only used when the storage type is `Gcs`**. |
 | `storage.credential_path` | String | `None` | The credential path of the google cloud storage.<br/>**It's only used when the storage type is `Gcs`**. |
+| `storage.credential` | String | `None` | The credential of the google cloud storage.<br/>**It's only used when the storage type is `Gcs`**. |
 | `storage.container` | String | `None` | The container of the azure account.<br/>**It's only used when the storage type is `Azblob`**. |
 | `storage.sas_token` | String | `None` | The sas token of the azure account.<br/>**It's only used when the storage type is `Azblob`**. |
 | `storage.endpoint` | String | `None` | The endpoint of the S3 service.<br/>**It's only used when the storage type is `S3`, `Oss`, `Gcs` and `Azblob`**. |
@@ -129,6 +134,8 @@
 | `region_engine.mito.inverted_index.apply_on_query` | String | `auto` | Whether to apply the index on query<br/>- `auto`: automatically (default)<br/>- `disable`: never |
 | `region_engine.mito.inverted_index.mem_threshold_on_create` | String | `auto` | Memory threshold for performing an external sort during index creation.<br/>- `auto`: automatically determine the threshold based on the system memory size (default)<br/>- `unlimited`: no memory limit<br/>- `[size]` e.g. `64MB`: fixed memory threshold |
 | `region_engine.mito.inverted_index.intermediate_path` | String | `""` | Deprecated, use `region_engine.mito.index.aux_path` instead. |
+| `region_engine.mito.inverted_index.metadata_cache_size` | String | `64MiB` | Cache size for inverted index metadata. |
+| `region_engine.mito.inverted_index.content_cache_size` | String | `128MiB` | Cache size for inverted index content. |
 | `region_engine.mito.fulltext_index` | -- | -- | The options for full-text index in Mito engine. |
 | `region_engine.mito.fulltext_index.create_on_flush` | String | `auto` | Whether to create the index on flush.<br/>- `auto`: automatically (default)<br/>- `disable`: never |
 | `region_engine.mito.fulltext_index.create_on_compaction` | String | `auto` | Whether to create the index on compaction.<br/>- `auto`: automatically (default)<br/>- `disable`: never |
@@ -144,7 +151,7 @@
 | `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
 | `logging.level` | String | `None` | The log level. Can be `info`/`debug`/`warn`/`error`. |
 | `logging.enable_otlp_tracing` | Bool | `false` | Enable OTLP tracing. |
-| `logging.otlp_endpoint` | String | `None` | The OTLP tracing endpoint. |
+| `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.tracing_sample_ratio` | -- | -- | The percentage of tracing will be sampled and exported.<br/>Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.<br/>ratio > 1 are treated as 1. Fractions < 0 are treated as 0 |
 | `logging.tracing_sample_ratio.default_ratio` | Float | `1.0` | -- |
@@ -166,12 +173,10 @@
 
 | Key | Type | Default | Descriptions |
 | --- | -----| ------- | ----------- |
-| `mode` | String | `standalone` | The running mode of the datanode. It can be `standalone` or `distributed`. |
 | `default_timezone` | String | `None` | The default timezone of the server. |
 | `runtime` | -- | -- | The runtime options. |
-| `runtime.read_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
-| `runtime.write_rt_size` | Integer | `8` | The number of threads to execute the runtime for global write operations. |
-| `runtime.bg_rt_size` | Integer | `4` | The number of threads to execute the runtime for global background operations. |
+| `runtime.global_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
+| `runtime.compact_rt_size` | Integer | `4` | The number of threads to execute the runtime for global write operations. |
 | `heartbeat` | -- | -- | The heartbeat options. |
 | `heartbeat.interval` | String | `18s` | Interval for sending heartbeat messages to the metasrv. |
 | `heartbeat.retry_interval` | String | `3s` | Interval for retrying to send heartbeat messages to the metasrv. |
@@ -231,7 +236,7 @@
 | `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
 | `logging.level` | String | `None` | The log level. Can be `info`/`debug`/`warn`/`error`. |
 | `logging.enable_otlp_tracing` | Bool | `false` | Enable OTLP tracing. |
-| `logging.otlp_endpoint` | String | `None` | The OTLP tracing endpoint. |
+| `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.tracing_sample_ratio` | -- | -- | The percentage of tracing will be sampled and exported.<br/>Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.<br/>ratio > 1 are treated as 1. Fractions < 0 are treated as 0 |
 | `logging.tracing_sample_ratio.default_ratio` | Float | `1.0` | -- |
@@ -255,29 +260,28 @@
 | `bind_addr` | String | `127.0.0.1:3002` | The bind address of metasrv. |
 | `server_addr` | String | `127.0.0.1:3002` | The communication server address for frontend and datanode to connect to metasrv,  "127.0.0.1:3002" by default for localhost. |
 | `store_addr` | String | `127.0.0.1:2379` | Etcd server address. |
-| `selector` | String | `lease_based` | Datanode selector type.<br/>- `lease_based` (default value).<br/>- `load_based`<br/>For details, please see "https://docs.greptime.com/developer-guide/metasrv/selector". |
+| `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_telemetry` | Bool | `true` | Whether to enable greptimedb telemetry. |
 | `store_key_prefix` | String | `""` | If it's not empty, the metasrv will store all data with this key prefix. |
 | `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). |
 | `runtime` | -- | -- | The runtime options. |
-| `runtime.read_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
-| `runtime.write_rt_size` | Integer | `8` | The number of threads to execute the runtime for global write operations. |
-| `runtime.bg_rt_size` | Integer | `4` | The number of threads to execute the runtime for global background operations. |
+| `runtime.global_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
+| `runtime.compact_rt_size` | Integer | `4` | The number of threads to execute the runtime for global write operations. |
 | `procedure` | -- | -- | Procedure storage options. |
 | `procedure.max_retry_times` | Integer | `12` | Procedure max retry time. |
 | `procedure.retry_delay` | String | `500ms` | Initial retry delay of procedures, increases exponentially |
 | `procedure.max_metadata_value_size` | String | `1500KiB` | Auto split large value<br/>GreptimeDB procedure uses etcd as the default metadata storage backend.<br/>The etcd the maximum size of any request is 1.5 MiB<br/>1500KiB = 1536KiB (1.5MiB) - 36KiB (reserved size of key)<br/>Comments out the `max_metadata_value_size`, for don't split large value (no limit). |
 | `failure_detector` | -- | -- | -- |
-| `failure_detector.threshold` | Float | `8.0` | -- |
-| `failure_detector.min_std_deviation` | String | `100ms` | -- |
-| `failure_detector.acceptable_heartbeat_pause` | String | `10000ms` | -- |
-| `failure_detector.first_heartbeat_estimate` | String | `1000ms` | -- |
+| `failure_detector.threshold` | Float | `8.0` | The threshold value used by the failure detector to determine failure conditions. |
+| `failure_detector.min_std_deviation` | String | `100ms` | The minimum standard deviation of the heartbeat intervals, used to calculate acceptable variations. |
+| `failure_detector.acceptable_heartbeat_pause` | String | `10000ms` | The acceptable pause duration between heartbeats, used to determine if a heartbeat interval is acceptable. |
+| `failure_detector.first_heartbeat_estimate` | String | `1000ms` | The initial estimate of the heartbeat interval used by the failure detector. |
 | `datanode` | -- | -- | Datanode options. |
 | `datanode.client` | -- | -- | Datanode client options. |
-| `datanode.client.timeout` | String | `10s` | -- |
-| `datanode.client.connect_timeout` | String | `10s` | -- |
-| `datanode.client.tcp_nodelay` | Bool | `true` | -- |
+| `datanode.client.timeout` | String | `10s` | Operation timeout. |
+| `datanode.client.connect_timeout` | String | `10s` | Connect server timeout. |
+| `datanode.client.tcp_nodelay` | Bool | `true` | `TCP_NODELAY` option for accepted connections. |
 | `wal` | -- | -- | -- |
 | `wal.provider` | String | `raft_engine` | -- |
 | `wal.broker_endpoints` | Array | -- | The broker endpoints of the Kafka cluster. |
@@ -294,7 +298,7 @@
 | `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
 | `logging.level` | String | `None` | The log level. Can be `info`/`debug`/`warn`/`error`. |
 | `logging.enable_otlp_tracing` | Bool | `false` | Enable OTLP tracing. |
-| `logging.otlp_endpoint` | String | `None` | The OTLP tracing endpoint. |
+| `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.tracing_sample_ratio` | -- | -- | The percentage of tracing will be sampled and exported.<br/>Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.<br/>ratio > 1 are treated as 1. Fractions < 0 are treated as 0 |
 | `logging.tracing_sample_ratio.default_ratio` | Float | `1.0` | -- |
@@ -337,9 +341,8 @@
 | `grpc.tls.key_path` | String | `None` | Private key file path. |
 | `grpc.tls.watch` | Bool | `false` | Watch for Certificate and key file change and auto reload.<br/>For now, gRPC tls config does not support auto reload. |
 | `runtime` | -- | -- | The runtime options. |
-| `runtime.read_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
-| `runtime.write_rt_size` | Integer | `8` | The number of threads to execute the runtime for global write operations. |
-| `runtime.bg_rt_size` | Integer | `4` | The number of threads to execute the runtime for global background operations. |
+| `runtime.global_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
+| `runtime.compact_rt_size` | Integer | `4` | The number of threads to execute the runtime for global write operations. |
 | `heartbeat` | -- | -- | The heartbeat options. |
 | `heartbeat.interval` | String | `3s` | Interval for sending heartbeat messages to the metasrv. |
 | `heartbeat.retry_interval` | String | `3s` | Interval for retrying to send heartbeat messages to the metasrv. |
@@ -371,6 +374,8 @@
 | `wal.backoff_max` | String | `10s` | The maximum backoff delay.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.backoff_base` | Integer | `2` | The exponential backoff rate, i.e. next backoff = base * current backoff.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.backoff_deadline` | String | `5mins` | The deadline of retries.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.create_index` | Bool | `true` | Whether to enable WAL index creation.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.dump_index_interval` | String | `60s` | The interval for dumping WAL indexes.<br/>**It's only used when the provider is `kafka`**. |
 | `storage` | -- | -- | The data storage options. |
 | `storage.data_home` | String | `/tmp/greptimedb/` | The working home directory. |
 | `storage.type` | String | `File` | The storage type used to store the data.<br/>- `File`: the data is stored in the local file system.<br/>- `S3`: the data is stored in the S3 object storage.<br/>- `Gcs`: the data is stored in the Google Cloud Storage.<br/>- `Azblob`: the data is stored in the Azure Blob Storage.<br/>- `Oss`: the data is stored in the Aliyun OSS. |
@@ -385,6 +390,7 @@
 | `storage.account_key` | String | `None` | The account key of the azure account.<br/>**It's only used when the storage type is `Azblob`**. |
 | `storage.scope` | String | `None` | The scope of the google cloud storage.<br/>**It's only used when the storage type is `Gcs`**. |
 | `storage.credential_path` | String | `None` | The credential path of the google cloud storage.<br/>**It's only used when the storage type is `Gcs`**. |
+| `storage.credential` | String | `None` | The credential of the google cloud storage.<br/>**It's only used when the storage type is `Gcs`**. |
 | `storage.container` | String | `None` | The container of the azure account.<br/>**It's only used when the storage type is `Azblob`**. |
 | `storage.sas_token` | String | `None` | The sas token of the azure account.<br/>**It's only used when the storage type is `Azblob`**. |
 | `storage.endpoint` | String | `None` | The endpoint of the S3 service.<br/>**It's only used when the storage type is `S3`, `Oss`, `Gcs` and `Azblob`**. |
@@ -435,7 +441,7 @@
 | `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
 | `logging.level` | String | `None` | The log level. Can be `info`/`debug`/`warn`/`error`. |
 | `logging.enable_otlp_tracing` | Bool | `false` | Enable OTLP tracing. |
-| `logging.otlp_endpoint` | String | `None` | The OTLP tracing endpoint. |
+| `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.tracing_sample_ratio` | -- | -- | The percentage of tracing will be sampled and exported.<br/>Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.<br/>ratio > 1 are treated as 1. Fractions < 0 are treated as 0 |
 | `logging.tracing_sample_ratio.default_ratio` | Float | `1.0` | -- |
@@ -480,7 +486,7 @@
 | `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
 | `logging.level` | String | `None` | The log level. Can be `info`/`debug`/`warn`/`error`. |
 | `logging.enable_otlp_tracing` | Bool | `false` | Enable OTLP tracing. |
-| `logging.otlp_endpoint` | String | `None` | The OTLP tracing endpoint. |
+| `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.tracing_sample_ratio` | -- | -- | The percentage of tracing will be sampled and exported.<br/>Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.<br/>ratio > 1 are treated as 1. Fractions < 0 are treated as 0 |
 | `logging.tracing_sample_ratio.default_ratio` | Float | `1.0` | -- |

config/datanode.example.toml

@@ -73,11 +73,9 @@ watch = false
 ## The runtime options.
 [runtime]
 ## The number of threads to execute the runtime for global read operations.
-read_rt_size = 8
+global_rt_size = 8
 ## The number of threads to execute the runtime for global write operations.
-write_rt_size = 8
-## The number of threads to execute the runtime for global background operations.
-bg_rt_size = 4
+compact_rt_size = 4
 
 ## The heartbeat options.
 [heartbeat]
@@ -189,6 +187,32 @@ backoff_base = 2
 ## **It's only used when the provider is `kafka`**.
 backoff_deadline = "5mins"
 
+## Whether to enable WAL index creation.
+## **It's only used when the provider is `kafka`**.
+create_index = true
+
+## The interval for dumping WAL indexes.
+## **It's only used when the provider is `kafka`**.
+dump_index_interval = "60s"
+
+# The Kafka SASL configuration.
+# **It's only used when the provider is `kafka`**.
+# Available SASL mechanisms:
+# - `PLAIN`
+# - `SCRAM-SHA-256`
+# - `SCRAM-SHA-512`
+# [wal.sasl]
+# type = "SCRAM-SHA-512"
+# username = "user_kafka"
+# password = "secret"
+
+# The Kafka TLS configuration.
+# **It's only used when the provider is `kafka`**.
+# [wal.tls]
+# server_ca_cert_path = "/path/to/server_cert"
+# client_cert_path = "/path/to/client_cert"
+# client_key_path = "/path/to/key"
+
 # Example of using S3 as the storage.
 # [storage]
 # type = "S3"
@@ -225,6 +249,7 @@ backoff_deadline = "5mins"
 # root = "data"
 # scope = "test"
 # credential_path = "123456"
+# credential = "base64-credential"
 # endpoint = "https://storage.googleapis.com"
 
 ## The data storage options.
@@ -296,6 +321,11 @@ scope = "test"
 ## +toml2docs:none-default
 credential_path = "test"
 
+## The credential of the google cloud storage.
+## **It's only used when the storage type is `Gcs`**.
+## +toml2docs:none-default
+credential= "base64-credential"
+
 ## The container of the azure account.
 ## **It's only used when the storage type is `Azblob`**.
 ## +toml2docs:none-default
@@ -495,8 +525,7 @@ level = "info"
 enable_otlp_tracing = false
 
 ## The OTLP tracing endpoint.
-## +toml2docs:none-default
-otlp_endpoint = ""
+otlp_endpoint = "http://localhost:4317"
 
 ## Whether to append logs to stdout.
 append_stdout = true

config/flownode.example.toml

@@ -70,8 +70,7 @@ level = "info"
 enable_otlp_tracing = false
 
 ## The OTLP tracing endpoint.
-## +toml2docs:none-default
-otlp_endpoint = ""
+otlp_endpoint = "http://localhost:4317"
 
 ## Whether to append logs to stdout.
 append_stdout = true

config/frontend.example.toml

@@ -1,6 +1,3 @@
-## The running mode of the datanode. It can be `standalone` or `distributed`.
-mode = "standalone"
-
 ## The default timezone of the server.
 ## +toml2docs:none-default
 default_timezone = "UTC"
@@ -8,11 +5,9 @@ default_timezone = "UTC"
 ## The runtime options.
 [runtime]
 ## The number of threads to execute the runtime for global read operations.
-read_rt_size = 8
+global_rt_size = 8
 ## The number of threads to execute the runtime for global write operations.
-write_rt_size = 8
-## The number of threads to execute the runtime for global background operations.
-bg_rt_size = 4
+compact_rt_size = 4
 
 ## The heartbeat options.
 [heartbeat]
@@ -182,8 +177,7 @@ level = "info"
 enable_otlp_tracing = false
 
 ## The OTLP tracing endpoint.
-## +toml2docs:none-default
-otlp_endpoint = ""
+otlp_endpoint = "http://localhost:4317"
 
 ## Whether to append logs to stdout.
 append_stdout = true

config/metasrv.example.toml

@@ -11,10 +11,11 @@ server_addr = "127.0.0.1:3002"
 store_addr = "127.0.0.1:2379"
 
 ## Datanode selector type.
-## - `lease_based` (default value).
+## - `round_robin` (default value)
+## - `lease_based`
 ## - `load_based`
 ## For details, please see "https://docs.greptime.com/developer-guide/metasrv/selector".
-selector = "lease_based"
+selector = "round_robin"
 
 ## Store data in memory.
 use_memory_store = false
@@ -34,11 +35,9 @@ enable_region_failover = false
 ## The runtime options.
 [runtime]
 ## The number of threads to execute the runtime for global read operations.
-read_rt_size = 8
+global_rt_size = 8
 ## The number of threads to execute the runtime for global write operations.
-write_rt_size = 8
-## The number of threads to execute the runtime for global background operations.
-bg_rt_size = 4
+compact_rt_size = 4
 
 ## Procedure storage options.
 [procedure]
@@ -58,17 +57,32 @@ max_metadata_value_size = "1500KiB"
 
 # Failure detectors options.
 [failure_detector]
+
+## The threshold value used by the failure detector to determine failure conditions.
 threshold = 8.0
+
+## The minimum standard deviation of the heartbeat intervals, used to calculate acceptable variations.
 min_std_deviation = "100ms"
+
+## The acceptable pause duration between heartbeats, used to determine if a heartbeat interval is acceptable.
 acceptable_heartbeat_pause = "10000ms"
+
+## The initial estimate of the heartbeat interval used by the failure detector.
 first_heartbeat_estimate = "1000ms"
 
 ## Datanode options.
 [datanode]
+
 ## Datanode client options.
 [datanode.client]
+
+## Operation timeout.
 timeout = "10s"
+
+## Connect server timeout.
 connect_timeout = "10s"
+
+## `TCP_NODELAY` option for accepted connections.
 tcp_nodelay = true
 
 [wal]
@@ -110,6 +124,24 @@ backoff_base = 2
 ## Stop reconnecting if the total wait time reaches the deadline. If this config is missing, the reconnecting won't terminate.
 backoff_deadline = "5mins"
 
+# The Kafka SASL configuration.
+# **It's only used when the provider is `kafka`**.
+# Available SASL mechanisms:
+# - `PLAIN`
+# - `SCRAM-SHA-256`
+# - `SCRAM-SHA-512`
+# [wal.sasl]
+# type = "SCRAM-SHA-512"
+# username = "user_kafka"
+# password = "secret"
+
+# The Kafka TLS configuration.
+# **It's only used when the provider is `kafka`**.
+# [wal.tls]
+# server_ca_cert_path = "/path/to/server_cert"
+# client_cert_path = "/path/to/client_cert"
+# client_key_path = "/path/to/key"
+
 ## The logging options.
 [logging]
 ## The directory to store the log files.
@@ -123,8 +155,7 @@ level = "info"
 enable_otlp_tracing = false
 
 ## The OTLP tracing endpoint.
-## +toml2docs:none-default
-otlp_endpoint = ""
+otlp_endpoint = "http://localhost:4317"
 
 ## Whether to append logs to stdout.
 append_stdout = true

config/standalone.example.toml

@@ -11,11 +11,9 @@ default_timezone = "UTC"
 ## The runtime options.
 [runtime]
 ## The number of threads to execute the runtime for global read operations.
-read_rt_size = 8
+global_rt_size = 8
 ## The number of threads to execute the runtime for global write operations.
-write_rt_size = 8
-## The number of threads to execute the runtime for global background operations.
-bg_rt_size = 4
+compact_rt_size = 4
 
 ## The HTTP server options.
 [http]
@@ -173,6 +171,28 @@ sync_period = "10s"
 ## **It's only used when the provider is `kafka`**.
 broker_endpoints = ["127.0.0.1:9092"]
 
+## Number of topics to be created upon start.
+## **It's only used when the provider is `kafka`**.
+num_topics = 64
+
+## Topic selector type.
+## Available selector types:
+## - `round_robin` (default)
+## **It's only used when the provider is `kafka`**.
+selector_type = "round_robin"
+
+## A Kafka topic is constructed by concatenating `topic_name_prefix` and `topic_id`.
+## **It's only used when the provider is `kafka`**.
+topic_name_prefix = "greptimedb_wal_topic"
+
+## Expected number of replicas of each partition.
+## **It's only used when the provider is `kafka`**.
+replication_factor = 1
+
+## Above which a topic creation operation will be cancelled.
+## **It's only used when the provider is `kafka`**.
+create_topic_timeout = "30s"
+
 ## The max size of a single producer batch.
 ## Warning: Kafka has a default limit of 1MB per message in a topic.
 ## **It's only used when the provider is `kafka`**.
@@ -198,6 +218,24 @@ backoff_base = 2
 ## **It's only used when the provider is `kafka`**.
 backoff_deadline = "5mins"
 
+# The Kafka SASL configuration.
+# **It's only used when the provider is `kafka`**.
+# Available SASL mechanisms:
+# - `PLAIN`
+# - `SCRAM-SHA-256`
+# - `SCRAM-SHA-512`
+# [wal.sasl]
+# type = "SCRAM-SHA-512"
+# username = "user_kafka"
+# password = "secret"
+
+# The Kafka TLS configuration.
+# **It's only used when the provider is `kafka`**.
+# [wal.tls]
+# server_ca_cert_path = "/path/to/server_cert"
+# client_cert_path = "/path/to/client_cert"
+# client_key_path = "/path/to/key"
+
 ## Metadata storage options.
 [metadata_store]
 ## Kv file size in bytes.
@@ -248,6 +286,7 @@ retry_delay = "500ms"
 # root = "data"
 # scope = "test"
 # credential_path = "123456"
+# credential = "base64-credential"
 # endpoint = "https://storage.googleapis.com"
 
 ## The data storage options.
@@ -319,6 +358,11 @@ scope = "test"
 ## +toml2docs:none-default
 credential_path = "test"
 
+## The credential of the google cloud storage.
+## **It's only used when the storage type is `Gcs`**.
+## +toml2docs:none-default
+credential = "base64-credential"
+
 ## The container of the azure account.
 ## **It's only used when the storage type is `Azblob`**.
 ## +toml2docs:none-default
@@ -459,6 +503,12 @@ mem_threshold_on_create = "auto"
 ## Deprecated, use `region_engine.mito.index.aux_path` instead.
 intermediate_path = ""
 
+## Cache size for inverted index metadata.
+metadata_cache_size = "64MiB"
+
+## Cache size for inverted index content.
+content_cache_size = "128MiB"
+
 ## The options for full-text index in Mito engine.
 [region_engine.mito.fulltext_index]
 
@@ -518,8 +568,7 @@ level = "info"
 enable_otlp_tracing = false
 
 ## The OTLP tracing endpoint.
-## +toml2docs:none-default
-otlp_endpoint = ""
+otlp_endpoint = "http://localhost:4317"
 
 ## Whether to append logs to stdout.
 append_stdout = true

docker/docker-compose/cluster-with-etcd.yaml

@@ -1,9 +1,9 @@
 x-custom:
   etcd_initial_cluster_token: &etcd_initial_cluster_token "--initial-cluster-token=etcd-cluster"
   etcd_common_settings: &etcd_common_settings
-    image: quay.io/coreos/etcd:v3.5.10
+    image: "${ETCD_REGISTRY:-quay.io}/${ETCD_NAMESPACE:-coreos}/etcd:${ETCD_VERSION:-v3.5.10}"
     entrypoint: /usr/local/bin/etcd
-  greptimedb_image: &greptimedb_image docker.io/greptimedb/greptimedb:latest
+  greptimedb_image: &greptimedb_image "${GREPTIMEDB_REGISTRY:-docker.io}/${GREPTIMEDB_NAMESPACE:-greptime}/greptimedb:${GREPTIMEDB_VERSION:-latest}"
 
 services:
   etcd0:

docs/how-to/how-to-write-fuzz-tests.md

@@ -105,7 +105,7 @@ use tests_fuzz::utils::{init_greptime_connections, Connections};
 
 fuzz_target!(|input: FuzzInput| {
     common_telemetry::init_default_ut_logging();
-    common_runtime::block_on_write(async {
+    common_runtime::block_on_global(async {
         let Connections { mysql } = init_greptime_connections().await;
             let mut rng = ChaChaRng::seed_from_u64(input.seed);
             let columns = rng.gen_range(2..30);

grafana/README.md

@@ -25,7 +25,7 @@ Please ensure the following configuration before importing the dashboard into Gr
 
 __1. Prometheus scrape config__
 
-Assign `greptime_pod` label to each host target. We use this label to identify each node instance.
+Configure Prometheus to scrape the cluster.
 
 ```yml
 # example config
@@ -34,27 +34,15 @@ Assign `greptime_pod` label to each host target. We use this label to identify e
 scrape_configs:
   - job_name: metasrv
     static_configs:
-    - targets: ['<ip>:<port>']
-      labels:
-        greptime_pod: metasrv
+    - targets: ['<metasrv-ip>:<port>']
 
   - job_name: datanode
     static_configs:
-    - targets: ['<ip>:<port>']
-      labels:
-        greptime_pod: datanode1
-    - targets: ['<ip>:<port>']
-      labels:
-        greptime_pod: datanode2
-    - targets: ['<ip>:<port>']
-      labels:
-        greptime_pod: datanode3
+    - targets: ['<datanode0-ip>:<port>', '<datanode1-ip>:<port>', '<datanode2-ip>:<port>']
 
   - job_name: frontend
     static_configs:
-    - targets: ['<ip>:<port>']
-      labels:
-        greptime_pod: frontend
+    - targets: ['<frontend-ip>:<port>']
 ```
 
 __2. Grafana config__
@@ -63,4 +51,4 @@ Create a Prometheus data source in Grafana before using this dashboard. We use `
 
 ### Usage
 
-Use `datasource` or `greptime_pod` on the upper-left corner to filter data from certain node.
+Use `datasource` or `instance` on the upper-left corner to filter data from certain node.

scripts/install.sh

@@ -1,62 +1,72 @@
-#!/bin/sh
+#!/usr/bin/env bash
 
 set -ue
 
 OS_TYPE=
 ARCH_TYPE=
+
+# Set the GitHub token to avoid GitHub API rate limit.
+# You can run with `GITHUB_TOKEN`:
+#  GITHUB_TOKEN=<your_token> ./scripts/install.sh
+GITHUB_TOKEN=${GITHUB_TOKEN:-}
+
 VERSION=${1:-latest}
 GITHUB_ORG=GreptimeTeam
 GITHUB_REPO=greptimedb
 BIN=greptime
 
-get_os_type() {
-    os_type="$(uname -s)"
+function get_os_type() {
+  os_type="$(uname -s)"
 
-    case "$os_type" in
+  case "$os_type" in
     Darwin)
-        OS_TYPE=darwin
-        ;;
+      OS_TYPE=darwin
+      ;;
     Linux)
-        OS_TYPE=linux
-        ;;
+      OS_TYPE=linux
+      ;;
     *)
-        echo "Error: Unknown OS type: $os_type"
-        exit 1
-    esac
+      echo "Error: Unknown OS type: $os_type"
+      exit 1
+  esac
 }
 
-get_arch_type() {
-    arch_type="$(uname -m)"
+function get_arch_type() {
+  arch_type="$(uname -m)"
 
-    case "$arch_type" in
+  case "$arch_type" in
     arm64)
-        ARCH_TYPE=arm64
-        ;;
+      ARCH_TYPE=arm64
+      ;;
     aarch64)
-        ARCH_TYPE=arm64
-        ;;
+      ARCH_TYPE=arm64
+      ;;
     x86_64)
-        ARCH_TYPE=amd64
-        ;;
+      ARCH_TYPE=amd64
+      ;;
     amd64)
-        ARCH_TYPE=amd64
-        ;;
+      ARCH_TYPE=amd64
+      ;;
     *)
-        echo "Error: Unknown CPU type: $arch_type"
-        exit 1
-    esac
+      echo "Error: Unknown CPU type: $arch_type"
+      exit 1
+  esac
 }
 
-get_os_type
-get_arch_type
-
-if [ -n "${OS_TYPE}" ] && [ -n "${ARCH_TYPE}" ]; then
-    # Use the latest nightly version.
+function download_artifact() {
+  if [ -n "${OS_TYPE}" ] && [ -n "${ARCH_TYPE}" ]; then
+    # Use the latest stable released version.
+    # GitHub API reference: https://docs.github.com/en/rest/releases/releases?apiVersion=2022-11-28#get-the-latest-release.
     if [ "${VERSION}" = "latest" ]; then
-        VERSION=$(curl -s -XGET "https://api.github.com/repos/${GITHUB_ORG}/${GITHUB_REPO}/releases" | grep tag_name | grep nightly | cut -d: -f 2 | sed 's/.*"\(.*\)".*/\1/' | uniq | sort -r | head -n 1)
-        if [ -z "${VERSION}" ]; then
-            echo "Failed to get the latest version."
-            exit 1
+      # To avoid other tools dependency, we choose to use `curl` to get the version metadata and parsed by `sed`.
+      VERSION=$(curl -sL \
+        -H "Accept: application/vnd.github+json" \
+        -H "X-GitHub-Api-Version: 2022-11-28" \
+        ${GITHUB_TOKEN:+-H "Authorization: Bearer $GITHUB_TOKEN"} \
+        "https://api.github.com/repos/${GITHUB_ORG}/${GITHUB_REPO}/releases/latest" | sed -n 's/.*"tag_name": "\([^"]*\)".*/\1/p')
+      if [ -z "${VERSION}" ]; then
+        echo "Failed to get the latest stable released version."
+        exit 1
         fi
     fi
 
@@ -73,4 +83,9 @@ if [ -n "${OS_TYPE}" ] && [ -n "${ARCH_TYPE}" ]; then
       rm -r "${PACKAGE_NAME%.tar.gz}" && \
       echo "Run './${BIN} --help' to get started"
     fi
-fi
+  fi
+}
+
+get_os_type
+get_arch_type
+download_artifact

src/auth/src/error.rs

@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use common_error::ext::ErrorExt;
+use common_error::ext::{BoxedError, ErrorExt};
 use common_error::status_code::StatusCode;
 use common_macro::stack_trace_debug;
 use snafu::{Location, Snafu};
@@ -38,6 +38,14 @@ pub enum Error {
         location: Location,
     },
 
+    #[snafu(display("Authentication source failure"))]
+    AuthBackend {
+        #[snafu(implicit)]
+        location: Location,
+        #[snafu(source)]
+        source: BoxedError,
+    },
+
     #[snafu(display("User not found, username: {}", username))]
     UserNotFound { username: String },
 
@@ -81,6 +89,7 @@ impl ErrorExt for Error {
             Error::FileWatch { .. } => StatusCode::InvalidArguments,
             Error::InternalState { .. } => StatusCode::Unexpected,
             Error::Io { .. } => StatusCode::StorageUnavailable,
+            Error::AuthBackend { .. } => StatusCode::Internal,
 
             Error::UserNotFound { .. } => StatusCode::UserNotFound,
             Error::UnsupportedPasswordType { .. } => StatusCode::UnsupportedPasswordType,

src/catalog/Cargo.toml

@@ -40,6 +40,7 @@ moka = { workspace = true, features = ["future", "sync"] }
 partition.workspace = true
 paste = "1.0"
 prometheus.workspace = true
+rustc-hash.workspace = true
 serde_json.workspace = true
 session.workspace = true
 snafu.workspace = true

src/catalog/src/system_schema.rs

@@ -15,6 +15,8 @@
 pub mod information_schema;
 mod memory_table;
 pub mod pg_catalog;
+mod predicate;
+mod utils;
 
 use std::collections::HashMap;
 use std::sync::Arc;

src/catalog/src/system_schema/information_schema.rs

@@ -18,14 +18,12 @@ pub mod flows;
 mod information_memory_table;
 pub mod key_column_usage;
 mod partitions;
-mod predicate;
 mod region_peers;
 mod runtime_metrics;
 pub mod schemata;
 mod table_constraints;
 mod table_names;
 pub mod tables;
-pub(crate) mod utils;
 mod views;
 
 use std::collections::HashMap;
@@ -37,7 +35,6 @@ use common_recordbatch::SendableRecordBatchStream;
 use datatypes::schema::SchemaRef;
 use lazy_static::lazy_static;
 use paste::paste;
-pub(crate) use predicate::Predicates;
 use store_api::storage::{ScanRequest, TableId};
 use table::metadata::TableType;
 use table::TableRef;
@@ -58,6 +55,7 @@ use crate::system_schema::information_schema::schemata::InformationSchemaSchemat
 use crate::system_schema::information_schema::table_constraints::InformationSchemaTableConstraints;
 use crate::system_schema::information_schema::tables::InformationSchemaTables;
 use crate::system_schema::memory_table::MemoryTable;
+pub(crate) use crate::system_schema::predicate::Predicates;
 use crate::system_schema::SystemSchemaProvider;
 use crate::CatalogManager;
 

src/catalog/src/system_schema/information_schema/cluster_info.rs

@@ -41,7 +41,8 @@ use store_api::storage::{ScanRequest, TableId};
 
 use super::CLUSTER_INFO;
 use crate::error::{CreateRecordBatchSnafu, InternalSnafu, ListNodesSnafu, Result};
-use crate::system_schema::information_schema::{utils, InformationTable, Predicates};
+use crate::system_schema::information_schema::{InformationTable, Predicates};
+use crate::system_schema::utils;
 use crate::CatalogManager;
 
 const PEER_ID: &str = "peer_id";

src/catalog/src/system_schema/information_schema/information_memory_table.rs

@@ -19,7 +19,7 @@ use datatypes::schema::{Schema, SchemaRef};
 use datatypes::vectors::{Int64Vector, StringVector, VectorRef};
 
 use super::table_names::*;
-use crate::system_schema::memory_table::tables::{
+use crate::system_schema::utils::tables::{
     bigint_column, datetime_column, string_column, string_columns,
 };
 

src/catalog/src/system_schema/information_schema/schemata.rs

@@ -36,7 +36,8 @@ use crate::error::{
     CreateRecordBatchSnafu, InternalSnafu, Result, TableMetadataManagerSnafu,
     UpgradeWeakCatalogManagerRefSnafu,
 };
-use crate::system_schema::information_schema::{utils, InformationTable, Predicates};
+use crate::system_schema::information_schema::{InformationTable, Predicates};
+use crate::system_schema::utils;
 use crate::CatalogManager;
 
 pub const CATALOG_NAME: &str = "catalog_name";

src/catalog/src/system_schema/information_schema/tables.rs

@@ -298,7 +298,7 @@ impl InformationSchemaTablesBuilder {
         self.data_free.push(Some(0));
         self.auto_increment.push(Some(0));
         self.row_format.push(Some("Fixed"));
-        self.table_collation.push(None);
+        self.table_collation.push(Some("utf8_bin"));
         self.update_time.push(None);
         self.check_time.push(None);
 

src/catalog/src/system_schema/information_schema/views.rs

@@ -26,7 +26,7 @@ use datafusion::physical_plan::SendableRecordBatchStream as DfSendableRecordBatc
 use datatypes::prelude::{ConcreteDataType, ScalarVectorBuilder, VectorRef};
 use datatypes::schema::{ColumnSchema, Schema, SchemaRef};
 use datatypes::value::Value;
-use datatypes::vectors::{BooleanVectorBuilder, StringVectorBuilder};
+use datatypes::vectors::StringVectorBuilder;
 use futures::TryStreamExt;
 use snafu::{OptionExt, ResultExt};
 use store_api::storage::{ScanRequest, TableId};
@@ -76,7 +76,7 @@ impl InformationSchemaViews {
             ColumnSchema::new(TABLE_NAME, ConcreteDataType::string_datatype(), false),
             ColumnSchema::new(VIEW_DEFINITION, ConcreteDataType::string_datatype(), false),
             ColumnSchema::new(CHECK_OPTION, ConcreteDataType::string_datatype(), true),
-            ColumnSchema::new(IS_UPDATABLE, ConcreteDataType::boolean_datatype(), true),
+            ColumnSchema::new(IS_UPDATABLE, ConcreteDataType::string_datatype(), true),
             ColumnSchema::new(DEFINER, ConcreteDataType::string_datatype(), true),
             ColumnSchema::new(SECURITY_TYPE, ConcreteDataType::string_datatype(), true),
             ColumnSchema::new(
@@ -148,7 +148,7 @@ struct InformationSchemaViewsBuilder {
     table_names: StringVectorBuilder,
     view_definitions: StringVectorBuilder,
     check_options: StringVectorBuilder,
-    is_updatable: BooleanVectorBuilder,
+    is_updatable: StringVectorBuilder,
     definer: StringVectorBuilder,
     security_type: StringVectorBuilder,
     character_set_client: StringVectorBuilder,
@@ -170,7 +170,7 @@ impl InformationSchemaViewsBuilder {
             table_names: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             view_definitions: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             check_options: StringVectorBuilder::with_capacity(INIT_CAPACITY),
-            is_updatable: BooleanVectorBuilder::with_capacity(INIT_CAPACITY),
+            is_updatable: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             definer: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             security_type: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             character_set_client: StringVectorBuilder::with_capacity(INIT_CAPACITY),
@@ -241,7 +241,8 @@ impl InformationSchemaViewsBuilder {
         self.table_names.push(Some(table_name));
         self.view_definitions.push(Some(definition));
         self.check_options.push(None);
-        self.is_updatable.push(Some(true));
+        // View is not updatable, statements such UPDATE , DELETE , and INSERT are illegal and are rejected.
+        self.is_updatable.push(Some("NO"));
         self.definer.push(None);
         self.security_type.push(None);
         self.character_set_client.push(Some("utf8"));

src/catalog/src/system_schema/memory_table.rs

@@ -13,7 +13,6 @@
 // limitations under the License.
 
 mod table_columns;
-pub mod tables;
 
 use std::sync::Arc;
 

src/catalog/src/system_schema/pg_catalog.rs

@@ -13,6 +13,8 @@
 // limitations under the License.
 
 mod pg_catalog_memory_table;
+mod pg_class;
+mod pg_namespace;
 mod table_names;
 
 use std::collections::HashMap;
@@ -23,11 +25,14 @@ use datatypes::schema::ColumnSchema;
 use lazy_static::lazy_static;
 use paste::paste;
 use pg_catalog_memory_table::get_schema_columns;
+use pg_class::PGClass;
+use pg_namespace::PGNamespace;
 use table::TableRef;
 pub use table_names::*;
 
-use super::memory_table::tables::u32_column;
+use self::pg_namespace::oid_map::{PGNamespaceOidMap, PGNamespaceOidMapRef};
 use super::memory_table::MemoryTable;
+use super::utils::tables::u32_column;
 use super::{SystemSchemaProvider, SystemSchemaProviderInner, SystemTableRef};
 use crate::CatalogManager;
 
@@ -46,8 +51,11 @@ fn oid_column() -> ColumnSchema {
 /// [`PGCatalogProvider`] is the provider for a schema named `pg_catalog`, it is not a catalog.
 pub struct PGCatalogProvider {
     catalog_name: String,
-    _catalog_manager: Weak<dyn CatalogManager>,
+    catalog_manager: Weak<dyn CatalogManager>,
     tables: HashMap<String, TableRef>,
+
+    // Workaround to store mapping of schema_name to a numeric id
+    namespace_oid_map: PGNamespaceOidMapRef,
 }
 
 impl SystemSchemaProvider for PGCatalogProvider {
@@ -79,8 +87,9 @@ impl PGCatalogProvider {
     pub fn new(catalog_name: String, catalog_manager: Weak<dyn CatalogManager>) -> Self {
         let mut provider = Self {
             catalog_name,
-            _catalog_manager: catalog_manager,
+            catalog_manager,
             tables: HashMap::new(),
+            namespace_oid_map: Arc::new(PGNamespaceOidMap::new()),
         };
         provider.build_tables();
         provider
@@ -90,9 +99,19 @@ impl PGCatalogProvider {
         // SECURITY NOTE:
         // Must follow the same security rules as [`InformationSchemaProvider::build_tables`].
         let mut tables = HashMap::new();
+        // TODO(J0HN50N133): modeling the table_name as a enum type to get rid of expect/unwrap here
+        // It's safe to unwrap here because we are sure that the constants have been handle correctly inside system_table.
         for name in MEMORY_TABLES.iter() {
             tables.insert(name.to_string(), self.build_table(name).expect(name));
         }
+        tables.insert(
+            PG_NAMESPACE.to_string(),
+            self.build_table(PG_NAMESPACE).expect(PG_NAMESPACE),
+        );
+        tables.insert(
+            PG_CLASS.to_string(),
+            self.build_table(PG_CLASS).expect(PG_NAMESPACE),
+        );
         self.tables = tables;
     }
 }
@@ -105,6 +124,16 @@ impl SystemSchemaProviderInner for PGCatalogProvider {
     fn system_table(&self, name: &str) -> Option<SystemTableRef> {
         match name {
             table_names::PG_TYPE => setup_memory_table!(PG_TYPE),
+            table_names::PG_NAMESPACE => Some(Arc::new(PGNamespace::new(
+                self.catalog_name.clone(),
+                self.catalog_manager.clone(),
+                self.namespace_oid_map.clone(),
+            ))),
+            table_names::PG_CLASS => Some(Arc::new(PGClass::new(
+                self.catalog_name.clone(),
+                self.catalog_manager.clone(),
+                self.namespace_oid_map.clone(),
+            ))),
             _ => None,
         }
     }

src/catalog/src/system_schema/pg_catalog/pg_catalog_memory_table.rs

@@ -20,7 +20,7 @@ use datatypes::vectors::{Int16Vector, StringVector, UInt32Vector, VectorRef};
 use super::oid_column;
 use super::table_names::PG_TYPE;
 use crate::memory_table_cols;
-use crate::system_schema::memory_table::tables::{i16_column, string_column};
+use crate::system_schema::utils::tables::{i16_column, string_column};
 
 fn pg_type_schema_columns() -> (Vec<ColumnSchema>, Vec<VectorRef>) {
     // TODO(j0hn50n133): acquire this information from `DataType` instead of hardcoding it to avoid regression.

src/catalog/src/system_schema/pg_catalog/pg_class.rs

@@ -0,0 +1,263 @@
+// 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, Weak};
+
+use arrow_schema::SchemaRef as ArrowSchemaRef;
+use common_catalog::consts::PG_CATALOG_PG_CLASS_TABLE_ID;
+use common_error::ext::BoxedError;
+use common_recordbatch::adapter::RecordBatchStreamAdapter;
+use common_recordbatch::{DfSendableRecordBatchStream, RecordBatch};
+use datafusion::execution::TaskContext;
+use datafusion::physical_plan::stream::RecordBatchStreamAdapter as DfRecordBatchStreamAdapter;
+use datafusion::physical_plan::streaming::PartitionStream as DfPartitionStream;
+use datatypes::scalars::ScalarVectorBuilder;
+use datatypes::schema::{Schema, SchemaRef};
+use datatypes::value::Value;
+use datatypes::vectors::{StringVectorBuilder, UInt32VectorBuilder, VectorRef};
+use futures::TryStreamExt;
+use snafu::{OptionExt, ResultExt};
+use store_api::storage::ScanRequest;
+use table::metadata::TableType;
+
+use super::pg_namespace::oid_map::PGNamespaceOidMapRef;
+use super::{OID_COLUMN_NAME, PG_CLASS};
+use crate::error::{
+    CreateRecordBatchSnafu, InternalSnafu, Result, UpgradeWeakCatalogManagerRefSnafu,
+};
+use crate::information_schema::Predicates;
+use crate::system_schema::utils::tables::{string_column, u32_column};
+use crate::system_schema::SystemTable;
+use crate::CatalogManager;
+
+// === column name ===
+pub const RELNAME: &str = "relname";
+pub const RELNAMESPACE: &str = "relnamespace";
+pub const RELKIND: &str = "relkind";
+pub const RELOWNER: &str = "relowner";
+
+// === enum value of relkind ===
+pub const RELKIND_TABLE: &str = "r";
+pub const RELKIND_VIEW: &str = "v";
+
+/// The initial capacity of the vector builders.
+const INIT_CAPACITY: usize = 42;
+/// The dummy owner id for the namespace.
+const DUMMY_OWNER_ID: u32 = 0;
+
+/// The `pg_catalog.pg_class` table implementation.
+pub(super) struct PGClass {
+    schema: SchemaRef,
+    catalog_name: String,
+    catalog_manager: Weak<dyn CatalogManager>,
+
+    // Workaround to convert schema_name to a numeric id
+    namespace_oid_map: PGNamespaceOidMapRef,
+}
+
+impl PGClass {
+    pub(super) fn new(
+        catalog_name: String,
+        catalog_manager: Weak<dyn CatalogManager>,
+        namespace_oid_map: PGNamespaceOidMapRef,
+    ) -> Self {
+        Self {
+            schema: Self::schema(),
+            catalog_name,
+            catalog_manager,
+            namespace_oid_map,
+        }
+    }
+
+    fn schema() -> SchemaRef {
+        Arc::new(Schema::new(vec![
+            u32_column(OID_COLUMN_NAME),
+            string_column(RELNAME),
+            u32_column(RELNAMESPACE),
+            string_column(RELKIND),
+            u32_column(RELOWNER),
+        ]))
+    }
+
+    fn builder(&self) -> PGClassBuilder {
+        PGClassBuilder::new(
+            self.schema.clone(),
+            self.catalog_name.clone(),
+            self.catalog_manager.clone(),
+            self.namespace_oid_map.clone(),
+        )
+    }
+}
+
+impl SystemTable for PGClass {
+    fn table_id(&self) -> table::metadata::TableId {
+        PG_CATALOG_PG_CLASS_TABLE_ID
+    }
+
+    fn table_name(&self) -> &'static str {
+        PG_CLASS
+    }
+
+    fn schema(&self) -> SchemaRef {
+        self.schema.clone()
+    }
+
+    fn to_stream(
+        &self,
+        request: ScanRequest,
+    ) -> Result<common_recordbatch::SendableRecordBatchStream> {
+        let schema = self.schema.arrow_schema().clone();
+        let mut builder = self.builder();
+        let stream = Box::pin(DfRecordBatchStreamAdapter::new(
+            schema,
+            futures::stream::once(async move {
+                builder
+                    .make_class(Some(request))
+                    .await
+                    .map(|x| x.into_df_record_batch())
+                    .map_err(Into::into)
+            }),
+        ));
+        Ok(Box::pin(
+            RecordBatchStreamAdapter::try_new(stream)
+                .map_err(BoxedError::new)
+                .context(InternalSnafu)?,
+        ))
+    }
+}
+
+impl DfPartitionStream for PGClass {
+    fn schema(&self) -> &ArrowSchemaRef {
+        self.schema.arrow_schema()
+    }
+
+    fn execute(&self, _: Arc<TaskContext>) -> DfSendableRecordBatchStream {
+        let schema = self.schema.arrow_schema().clone();
+        let mut builder = self.builder();
+        Box::pin(DfRecordBatchStreamAdapter::new(
+            schema,
+            futures::stream::once(async move {
+                builder
+                    .make_class(None)
+                    .await
+                    .map(|x| x.into_df_record_batch())
+                    .map_err(Into::into)
+            }),
+        ))
+    }
+}
+
+/// Builds the `pg_catalog.pg_class` table row by row
+/// TODO(J0HN50N133): `relowner` is always the [`DUMMY_OWNER_ID`] cuz we don't have user.
+/// Once we have user system, make it the actual owner of the table.
+struct PGClassBuilder {
+    schema: SchemaRef,
+    catalog_name: String,
+    catalog_manager: Weak<dyn CatalogManager>,
+    namespace_oid_map: PGNamespaceOidMapRef,
+
+    oid: UInt32VectorBuilder,
+    relname: StringVectorBuilder,
+    relnamespace: UInt32VectorBuilder,
+    relkind: StringVectorBuilder,
+    relowner: UInt32VectorBuilder,
+}
+
+impl PGClassBuilder {
+    fn new(
+        schema: SchemaRef,
+        catalog_name: String,
+        catalog_manager: Weak<dyn CatalogManager>,
+        namespace_oid_map: PGNamespaceOidMapRef,
+    ) -> Self {
+        Self {
+            schema,
+            catalog_name,
+            catalog_manager,
+            namespace_oid_map,
+
+            oid: UInt32VectorBuilder::with_capacity(INIT_CAPACITY),
+            relname: StringVectorBuilder::with_capacity(INIT_CAPACITY),
+            relnamespace: UInt32VectorBuilder::with_capacity(INIT_CAPACITY),
+            relkind: StringVectorBuilder::with_capacity(INIT_CAPACITY),
+            relowner: UInt32VectorBuilder::with_capacity(INIT_CAPACITY),
+        }
+    }
+
+    async fn make_class(&mut self, request: Option<ScanRequest>) -> Result<RecordBatch> {
+        let catalog_name = self.catalog_name.clone();
+        let catalog_manager = self
+            .catalog_manager
+            .upgrade()
+            .context(UpgradeWeakCatalogManagerRefSnafu)?;
+        let predicates = Predicates::from_scan_request(&request);
+        for schema_name in catalog_manager.schema_names(&catalog_name).await? {
+            let mut stream = catalog_manager.tables(&catalog_name, &schema_name);
+            while let Some(table) = stream.try_next().await? {
+                let table_info = table.table_info();
+                self.add_class(
+                    &predicates,
+                    table_info.table_id(),
+                    &schema_name,
+                    &table_info.name,
+                    if table_info.table_type == TableType::View {
+                        RELKIND_VIEW
+                    } else {
+                        RELKIND_TABLE
+                    },
+                );
+            }
+        }
+        self.finish()
+    }
+
+    fn add_class(
+        &mut self,
+        predicates: &Predicates,
+        oid: u32,
+        schema: &str,
+        table: &str,
+        kind: &str,
+    ) {
+        let namespace_oid = self.namespace_oid_map.get_oid(schema);
+        let row = [
+            (OID_COLUMN_NAME, &Value::from(oid)),
+            (RELNAMESPACE, &Value::from(schema)),
+            (RELNAME, &Value::from(table)),
+            (RELKIND, &Value::from(kind)),
+            (RELOWNER, &Value::from(DUMMY_OWNER_ID)),
+        ];
+
+        if !predicates.eval(&row) {
+            return;
+        }
+
+        self.oid.push(Some(oid));
+        self.relnamespace.push(Some(namespace_oid));
+        self.relname.push(Some(table));
+        self.relkind.push(Some(kind));
+        self.relowner.push(Some(DUMMY_OWNER_ID));
+    }
+
+    fn finish(&mut self) -> Result<RecordBatch> {
+        let columns: Vec<VectorRef> = vec![
+            Arc::new(self.oid.finish()),
+            Arc::new(self.relname.finish()),
+            Arc::new(self.relnamespace.finish()),
+            Arc::new(self.relkind.finish()),
+            Arc::new(self.relowner.finish()),
+        ];
+        RecordBatch::new(self.schema.clone(), columns).context(CreateRecordBatchSnafu)
+    }
+}

src/catalog/src/system_schema/pg_catalog/pg_namespace.rs

@@ -0,0 +1,206 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+pub(super) mod oid_map;
+
+use std::sync::{Arc, Weak};
+
+use arrow_schema::SchemaRef as ArrowSchemaRef;
+use common_catalog::consts::PG_CATALOG_PG_NAMESPACE_TABLE_ID;
+use common_error::ext::BoxedError;
+use common_recordbatch::adapter::RecordBatchStreamAdapter;
+use common_recordbatch::{DfSendableRecordBatchStream, RecordBatch, SendableRecordBatchStream};
+use datafusion::execution::TaskContext;
+use datafusion::physical_plan::stream::RecordBatchStreamAdapter as DfRecordBatchStreamAdapter;
+use datafusion::physical_plan::streaming::PartitionStream as DfPartitionStream;
+use datatypes::scalars::ScalarVectorBuilder;
+use datatypes::schema::{Schema, SchemaRef};
+use datatypes::value::Value;
+use datatypes::vectors::{StringVectorBuilder, UInt32VectorBuilder, VectorRef};
+use snafu::{OptionExt, ResultExt};
+use store_api::storage::ScanRequest;
+
+use super::{PGNamespaceOidMapRef, OID_COLUMN_NAME, PG_NAMESPACE};
+use crate::error::{
+    CreateRecordBatchSnafu, InternalSnafu, Result, UpgradeWeakCatalogManagerRefSnafu,
+};
+use crate::information_schema::Predicates;
+use crate::system_schema::utils::tables::{string_column, u32_column};
+use crate::system_schema::SystemTable;
+use crate::CatalogManager;
+
+/// The `pg_catalog.pg_namespace` table implementation.
+/// namespace is a schema in greptime
+
+const NSPNAME: &str = "nspname";
+const INIT_CAPACITY: usize = 42;
+
+pub(super) struct PGNamespace {
+    schema: SchemaRef,
+    catalog_name: String,
+    catalog_manager: Weak<dyn CatalogManager>,
+
+    // Workaround to convert schema_name to a numeric id
+    oid_map: PGNamespaceOidMapRef,
+}
+
+impl PGNamespace {
+    pub(super) fn new(
+        catalog_name: String,
+        catalog_manager: Weak<dyn CatalogManager>,
+        oid_map: PGNamespaceOidMapRef,
+    ) -> Self {
+        Self {
+            schema: Self::schema(),
+            catalog_name,
+            catalog_manager,
+            oid_map,
+        }
+    }
+
+    fn schema() -> SchemaRef {
+        Arc::new(Schema::new(vec![
+            // TODO(J0HN50N133): we do not have a numeric schema id, use schema name as a workaround. Use a proper schema id once we have it.
+            u32_column(OID_COLUMN_NAME),
+            string_column(NSPNAME),
+        ]))
+    }
+
+    fn builder(&self) -> PGNamespaceBuilder {
+        PGNamespaceBuilder::new(
+            self.schema.clone(),
+            self.catalog_name.clone(),
+            self.catalog_manager.clone(),
+            self.oid_map.clone(),
+        )
+    }
+}
+
+impl SystemTable for PGNamespace {
+    fn schema(&self) -> SchemaRef {
+        self.schema.clone()
+    }
+
+    fn table_id(&self) -> table::metadata::TableId {
+        PG_CATALOG_PG_NAMESPACE_TABLE_ID
+    }
+
+    fn table_name(&self) -> &'static str {
+        PG_NAMESPACE
+    }
+
+    fn to_stream(&self, request: ScanRequest) -> Result<SendableRecordBatchStream> {
+        let schema = self.schema.arrow_schema().clone();
+        let mut builder = self.builder();
+        let stream = Box::pin(DfRecordBatchStreamAdapter::new(
+            schema,
+            futures::stream::once(async move {
+                builder
+                    .make_namespace(Some(request))
+                    .await
+                    .map(|x| x.into_df_record_batch())
+                    .map_err(Into::into)
+            }),
+        ));
+        Ok(Box::pin(
+            RecordBatchStreamAdapter::try_new(stream)
+                .map_err(BoxedError::new)
+                .context(InternalSnafu)?,
+        ))
+    }
+}
+
+impl DfPartitionStream for PGNamespace {
+    fn schema(&self) -> &ArrowSchemaRef {
+        self.schema.arrow_schema()
+    }
+
+    fn execute(&self, _: Arc<TaskContext>) -> DfSendableRecordBatchStream {
+        let schema = self.schema.arrow_schema().clone();
+        let mut builder = self.builder();
+        Box::pin(DfRecordBatchStreamAdapter::new(
+            schema,
+            futures::stream::once(async move {
+                builder
+                    .make_namespace(None)
+                    .await
+                    .map(|x| x.into_df_record_batch())
+                    .map_err(Into::into)
+            }),
+        ))
+    }
+}
+
+/// Builds the `pg_catalog.pg_namespace` table row by row
+/// `oid` use schema name as a workaround since we don't have numeric schema id.
+/// `nspname` is the schema name.
+struct PGNamespaceBuilder {
+    schema: SchemaRef,
+    catalog_name: String,
+    catalog_manager: Weak<dyn CatalogManager>,
+    namespace_oid_map: PGNamespaceOidMapRef,
+
+    oid: UInt32VectorBuilder,
+    nspname: StringVectorBuilder,
+}
+
+impl PGNamespaceBuilder {
+    fn new(
+        schema: SchemaRef,
+        catalog_name: String,
+        catalog_manager: Weak<dyn CatalogManager>,
+        namespace_oid_map: PGNamespaceOidMapRef,
+    ) -> Self {
+        Self {
+            schema,
+            catalog_name,
+            catalog_manager,
+            namespace_oid_map,
+            oid: UInt32VectorBuilder::with_capacity(INIT_CAPACITY),
+            nspname: StringVectorBuilder::with_capacity(INIT_CAPACITY),
+        }
+    }
+
+    /// Construct the `pg_catalog.pg_namespace` virtual table
+    async fn make_namespace(&mut self, request: Option<ScanRequest>) -> Result<RecordBatch> {
+        let catalog_name = self.catalog_name.clone();
+        let catalog_manager = self
+            .catalog_manager
+            .upgrade()
+            .context(UpgradeWeakCatalogManagerRefSnafu)?;
+        let predicates = Predicates::from_scan_request(&request);
+        for schema_name in catalog_manager.schema_names(&catalog_name).await? {
+            self.add_namespace(&predicates, &schema_name);
+        }
+        self.finish()
+    }
+    fn finish(&mut self) -> Result<RecordBatch> {
+        let columns: Vec<VectorRef> =
+            vec![Arc::new(self.oid.finish()), Arc::new(self.nspname.finish())];
+        RecordBatch::new(self.schema.clone(), columns).context(CreateRecordBatchSnafu)
+    }
+
+    fn add_namespace(&mut self, predicates: &Predicates, schema_name: &str) {
+        let oid = self.namespace_oid_map.get_oid(schema_name);
+        let row = [
+            (OID_COLUMN_NAME, &Value::from(oid)),
+            (NSPNAME, &Value::from(schema_name)),
+        ];
+        if !predicates.eval(&row) {
+            return;
+        }
+        self.oid.push(Some(oid));
+        self.nspname.push(Some(schema_name));
+    }
+}

src/catalog/src/system_schema/pg_catalog/pg_namespace/oid_map.rs

@@ -0,0 +1,100 @@
+// 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::hash::BuildHasher;
+use std::sync::Arc;
+
+use dashmap::DashMap;
+use rustc_hash::FxSeededState;
+
+pub type PGNamespaceOidMapRef = Arc<PGNamespaceOidMap>;
+// Workaround to convert schema_name to a numeric id,
+// remove this when we have numeric schema id in greptime
+pub struct PGNamespaceOidMap {
+    oid_map: DashMap<String, u32>,
+
+    // Rust use SipHasher by default, which provides resistance against DOS attacks.
+    // This will produce different hash value between each greptime instance. This will
+    // cause the sqlness test fail. We need a deterministic hash here to provide
+    // same oid for the same schema name with best effort and DOS attacks aren't concern here.
+    hasher: FxSeededState,
+}
+
+impl PGNamespaceOidMap {
+    pub fn new() -> Self {
+        Self {
+            oid_map: DashMap::new(),
+            hasher: FxSeededState::with_seed(0), // PLEASE DO NOT MODIFY THIS SEED VALUE!!!
+        }
+    }
+
+    fn oid_is_used(&self, oid: u32) -> bool {
+        self.oid_map.iter().any(|e| *e.value() == oid)
+    }
+
+    pub fn get_oid(&self, schema_name: &str) -> u32 {
+        if let Some(oid) = self.oid_map.get(schema_name) {
+            *oid
+        } else {
+            let mut oid = self.hasher.hash_one(schema_name) as u32;
+            while self.oid_is_used(oid) {
+                oid = self.hasher.hash_one(oid) as u32;
+            }
+            self.oid_map.insert(schema_name.to_string(), oid);
+            oid
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    use super::*;
+
+    #[test]
+    fn oid_is_stable() {
+        let oid_map_1 = PGNamespaceOidMap::new();
+        let oid_map_2 = PGNamespaceOidMap::new();
+
+        let schema = "schema";
+        let oid = oid_map_1.get_oid(schema);
+
+        // oid keep stable in the same instance
+        assert_eq!(oid, oid_map_1.get_oid(schema));
+
+        // oid keep stable between different instances
+        assert_eq!(oid, oid_map_2.get_oid(schema));
+    }
+
+    #[test]
+    fn oid_collision() {
+        let oid_map = PGNamespaceOidMap::new();
+
+        let key1 = "3178510";
+        let key2 = "4215648";
+
+        // have collision
+        assert_eq!(
+            oid_map.hasher.hash_one(key1) as u32,
+            oid_map.hasher.hash_one(key2) as u32
+        );
+
+        // insert them into oid_map
+        let oid1 = oid_map.get_oid(key1);
+        let oid2 = oid_map.get_oid(key2);
+
+        // they should have different id
+        assert_ne!(oid1, oid2);
+    }
+}

src/catalog/src/system_schema/predicate.rs

@@ -25,7 +25,7 @@ type ColumnName = String;
 /// we only support these simple predicates currently.
 /// TODO(dennis): supports more predicate types.
 #[derive(Clone, PartialEq, Eq, Debug)]
-enum Predicate {
+pub(crate) enum Predicate {
     Eq(ColumnName, Value),
     Like(ColumnName, String, bool),
     NotEq(ColumnName, Value),

src/catalog/src/system_schema/utils.rs

@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+pub mod tables;
+
 use std::sync::{Arc, Weak};
 
 use common_config::Mode;

src/client/src/database.rs

@@ -33,9 +33,12 @@ use common_telemetry::tracing_context::W3cTrace;
 use futures_util::StreamExt;
 use prost::Message;
 use snafu::{ensure, ResultExt};
+use tonic::metadata::AsciiMetadataKey;
 use tonic::transport::Channel;
 
-use crate::error::{ConvertFlightDataSnafu, Error, IllegalFlightMessagesSnafu, ServerSnafu};
+use crate::error::{
+    ConvertFlightDataSnafu, Error, IllegalFlightMessagesSnafu, InvalidAsciiSnafu, ServerSnafu,
+};
 use crate::{from_grpc_response, Client, Result};
 
 #[derive(Clone, Debug, Default)]
@@ -130,6 +133,36 @@ impl Database {
         self.handle(Request::Inserts(requests)).await
     }
 
+    pub async fn insert_with_hints(
+        &self,
+        requests: InsertRequests,
+        hints: &[(&str, &str)],
+    ) -> Result<u32> {
+        let mut client = make_database_client(&self.client)?.inner;
+        let request = self.to_rpc_request(Request::Inserts(requests));
+
+        let mut request = tonic::Request::new(request);
+        let metadata = request.metadata_mut();
+        for (key, value) in hints {
+            let key = AsciiMetadataKey::from_bytes(format!("x-greptime-hint-{}", key).as_bytes())
+                .map_err(|_| {
+                InvalidAsciiSnafu {
+                    value: key.to_string(),
+                }
+                .build()
+            })?;
+            let value = value.parse().map_err(|_| {
+                InvalidAsciiSnafu {
+                    value: value.to_string(),
+                }
+                .build()
+            })?;
+            metadata.insert(key, value);
+        }
+        let response = client.handle(request).await?.into_inner();
+        from_grpc_response(response)
+    }
+
     async fn handle(&self, request: Request) -> Result<u32> {
         let mut client = make_database_client(&self.client)?.inner;
         let request = self.to_rpc_request(request);

src/client/src/error.rs

@@ -122,6 +122,13 @@ pub enum Error {
         #[snafu(implicit)]
         location: Location,
     },
+
+    #[snafu(display("Failed to parse ascii string: {}", value))]
+    InvalidAscii {
+        value: String,
+        #[snafu(implicit)]
+        location: Location,
+    },
 }
 
 pub type Result<T> = std::result::Result<T, Error>;
@@ -143,6 +150,8 @@ impl ErrorExt for Error {
             | Error::ConvertFlightData { source, .. }
             | Error::CreateTlsChannel { source, .. } => source.status_code(),
             Error::IllegalGrpcClientState { .. } => StatusCode::Unexpected,
+
+            Error::InvalidAscii { .. } => StatusCode::InvalidArguments,
         }
     }
 

src/client/src/flow.rs

@@ -16,7 +16,7 @@ use api::v1::flow::{FlowRequest, FlowResponse};
 use api::v1::region::InsertRequests;
 use common_error::ext::BoxedError;
 use common_meta::node_manager::Flownode;
-use snafu::{location, Location, ResultExt};
+use snafu::{location, ResultExt};
 
 use crate::error::Result;
 use crate::Client;

src/client/src/region.rs

@@ -33,7 +33,7 @@ use common_telemetry::error;
 use common_telemetry::tracing_context::TracingContext;
 use prost::Message;
 use query::query_engine::DefaultSerializer;
-use snafu::{location, Location, OptionExt, ResultExt};
+use snafu::{location, OptionExt, ResultExt};
 use substrait::{DFLogicalSubstraitConvertor, SubstraitPlan};
 use tokio_stream::StreamExt;
 

src/cmd/Cargo.toml

@@ -10,7 +10,9 @@ name = "greptime"
 path = "src/bin/greptime.rs"
 
 [features]
+default = ["python"]
 tokio-console = ["common-telemetry/tokio-console"]
+python = ["frontend/python"]
 
 [lints]
 workspace = true
@@ -47,7 +49,7 @@ either = "1.8"
 etcd-client.workspace = true
 file-engine.workspace = true
 flow.workspace = true
-frontend.workspace = true
+frontend = { workspace = true, default-features = false }
 futures.workspace = true
 human-panic = "1.2.2"
 lazy_static.workspace = true

src/cmd/src/bin/greptime.rs

@@ -62,8 +62,37 @@ enum SubCommand {
 #[global_allocator]
 static ALLOC: tikv_jemallocator::Jemalloc = tikv_jemallocator::Jemalloc;
 
+#[cfg(debug_assertions)]
+fn main() -> Result<()> {
+    use snafu::ResultExt;
+    // Set the stack size to 8MB for the thread so it wouldn't overflow on large stack usage in debug mode
+    // see https://github.com/GreptimeTeam/greptimedb/pull/4317
+    // and https://github.com/rust-lang/rust/issues/34283
+    std::thread::Builder::new()
+        .name("main_spawn".to_string())
+        .stack_size(8 * 1024 * 1024)
+        .spawn(|| {
+            {
+                tokio::runtime::Builder::new_multi_thread()
+                    .thread_stack_size(8 * 1024 * 1024)
+                    .enable_all()
+                    .build()
+                    .expect("Failed building the Runtime")
+                    .block_on(main_body())
+            }
+        })
+        .context(cmd::error::SpawnThreadSnafu)?
+        .join()
+        .expect("Couldn't join on the associated thread")
+}
+
+#[cfg(not(debug_assertions))]
 #[tokio::main]
 async fn main() -> Result<()> {
+    main_body().await
+}
+
+async fn main_body() -> Result<()> {
     setup_human_panic();
     start(Command::parse()).await
 }

src/cmd/src/cli/export.rs

@@ -21,11 +21,12 @@ use base64::engine::general_purpose;
 use base64::Engine;
 use clap::{Parser, ValueEnum};
 use client::DEFAULT_SCHEMA_NAME;
-use common_telemetry::{debug, error, info, warn};
+use common_catalog::consts::DEFAULT_CATALOG_NAME;
+use common_telemetry::{debug, error, info};
 use serde_json::Value;
 use servers::http::greptime_result_v1::GreptimedbV1Response;
 use servers::http::GreptimeQueryOutput;
-use snafu::{OptionExt, ResultExt};
+use snafu::ResultExt;
 use tokio::fs::File;
 use tokio::io::{AsyncWriteExt, BufWriter};
 use tokio::sync::Semaphore;
@@ -34,22 +35,20 @@ use tracing_appender::non_blocking::WorkerGuard;
 
 use crate::cli::{Instance, Tool};
 use crate::error::{
-    EmptyResultSnafu, Error, FileIoSnafu, HttpQuerySqlSnafu, InvalidDatabaseNameSnafu, Result,
-    SerdeJsonSnafu,
+    EmptyResultSnafu, Error, FileIoSnafu, HttpQuerySqlSnafu, Result, SerdeJsonSnafu,
 };
 
 type TableReference = (String, String, String);
 
 #[derive(Debug, Default, Clone, ValueEnum)]
 enum ExportTarget {
-    /// Corresponding to `SHOW CREATE TABLE`
+    /// Export all table schemas, corresponding to `SHOW CREATE TABLE`.
+    Schema,
+    /// Export all table data, corresponding to `COPY DATABASE TO`.
+    Data,
+    /// Export all table schemas and data at once.
     #[default]
-    CreateTable,
-    /// Corresponding to `EXPORT TABLE`
-    #[deprecated(note = "Please use `DatabaseData` instead.")]
-    TableData,
-    /// Corresponding to `EXPORT DATABASE`
-    DatabaseData,
+    All,
 }
 
 #[derive(Debug, Default, Parser)]
@@ -75,7 +74,7 @@ pub struct ExportCommand {
     max_retry: usize,
 
     /// Things to export
-    #[clap(long, short = 't', value_enum)]
+    #[clap(long, short = 't', value_enum, default_value = "all")]
     target: ExportTarget,
 
     /// A half-open time range: [start_time, end_time).
@@ -178,7 +177,7 @@ impl Export {
         if let Some(schema) = &self.schema {
             Ok(vec![(self.catalog.clone(), schema.clone())])
         } else {
-            let result = self.sql("show databases").await?;
+            let result = self.sql("SHOW DATABASES").await?;
             let Some(records) = result else {
                 EmptyResultSnafu.fail()?
             };
@@ -205,9 +204,11 @@ impl Export {
     ) -> Result<(Vec<TableReference>, Vec<TableReference>)> {
         // Puts all metric table first
         let sql = format!(
-            "select table_catalog, table_schema, table_name from \
-            information_schema.columns where column_name = '__tsid' \
-            and table_catalog = \'{catalog}\' and table_schema = \'{schema}\'"
+            "SELECT table_catalog, table_schema, table_name \
+            FROM information_schema.columns \
+            WHERE column_name = '__tsid' \
+                and table_catalog = \'{catalog}\' \
+                and table_schema = \'{schema}\'"
         );
         let result = self.sql(&sql).await?;
         let Some(records) = result else {
@@ -227,9 +228,11 @@ impl Export {
 
         // TODO: SQL injection hurts
         let sql = format!(
-            "select table_catalog, table_schema, table_name from \
-            information_schema.tables where table_type = \'BASE TABLE\' \
-            and table_catalog = \'{catalog}\' and table_schema = \'{schema}\'",
+            "SELECT table_catalog, table_schema, table_name \
+            FROM information_schema.tables \
+            WHERE table_type = \'BASE TABLE\' \
+                and table_catalog = \'{catalog}\' \
+                and table_schema = \'{schema}\'",
         );
         let result = self.sql(&sql).await?;
         let Some(records) = result else {
@@ -266,7 +269,7 @@ impl Export {
 
     async fn show_create_table(&self, catalog: &str, schema: &str, table: &str) -> Result<String> {
         let sql = format!(
-            r#"show create table "{}"."{}"."{}""#,
+            r#"SHOW CREATE TABLE "{}"."{}"."{}""#,
             catalog, schema, table
         );
         let result = self.sql(&sql).await?;
@@ -341,99 +344,6 @@ impl Export {
         Ok(())
     }
 
-    async fn export_table_data(&self) -> Result<()> {
-        let timer = Instant::now();
-        let semaphore = Arc::new(Semaphore::new(self.parallelism));
-        let db_names = self.iter_db_names().await?;
-        let db_count = db_names.len();
-        let mut tasks = Vec::with_capacity(db_names.len());
-        for (catalog, schema) in db_names {
-            let semaphore_moved = semaphore.clone();
-            tasks.push(async move {
-                let _permit = semaphore_moved.acquire().await.unwrap();
-                tokio::fs::create_dir_all(&self.output_dir)
-                    .await
-                    .context(FileIoSnafu)?;
-                let output_dir = Path::new(&self.output_dir).join(format!("{catalog}-{schema}/"));
-                // Ignores metric physical tables
-                let (metrics_tables, table_list) = self.get_table_list(&catalog, &schema).await?;
-                for (_, _, table_name) in metrics_tables {
-                    warn!("Ignores metric physical table: {table_name}");
-                }
-                for (catalog_name, schema_name, table_name) in table_list {
-                    // copy table to
-                    let sql = format!(
-                        r#"Copy "{}"."{}"."{}" TO '{}{}.parquet' WITH (format='parquet');"#,
-                        catalog_name,
-                        schema_name,
-                        table_name,
-                        output_dir.to_str().unwrap(),
-                        table_name,
-                    );
-                    info!("Executing sql: {sql}");
-                    self.sql(&sql).await?;
-                }
-                info!("Finished exporting {catalog}.{schema} data");
-
-                // export copy from sql
-                let dir_filenames = match output_dir.read_dir() {
-                    Ok(dir) => dir,
-                    Err(_) => {
-                        warn!("empty database {catalog}.{schema}");
-                        return Ok(());
-                    }
-                };
-
-                let copy_from_file =
-                    Path::new(&self.output_dir).join(format!("{catalog}-{schema}_copy_from.sql"));
-                let mut writer =
-                    BufWriter::new(File::create(copy_from_file).await.context(FileIoSnafu)?);
-
-                for table_file in dir_filenames {
-                    let table_file = table_file.unwrap();
-                    let table_name = table_file
-                        .file_name()
-                        .into_string()
-                        .unwrap()
-                        .replace(".parquet", "");
-
-                    writer
-                        .write(
-                            format!(
-                                "copy {} from '{}' with (format='parquet');\n",
-                                table_name,
-                                table_file.path().to_str().unwrap()
-                            )
-                            .as_bytes(),
-                        )
-                        .await
-                        .context(FileIoSnafu)?;
-                }
-                writer.flush().await.context(FileIoSnafu)?;
-
-                info!("finished exporting {catalog}.{schema} copy_from.sql");
-
-                Ok::<(), Error>(())
-            });
-        }
-
-        let success = futures::future::join_all(tasks)
-            .await
-            .into_iter()
-            .filter(|r| match r {
-                Ok(_) => true,
-                Err(e) => {
-                    error!(e; "export job failed");
-                    false
-                }
-            })
-            .count();
-        let elapsed = timer.elapsed();
-        info!("Success {success}/{db_count} jobs, costs: {:?}", elapsed);
-
-        Ok(())
-    }
-
     async fn export_database_data(&self) -> Result<()> {
         let timer = Instant::now();
         let semaphore = Arc::new(Semaphore::new(self.parallelism));
@@ -530,20 +440,23 @@ impl Export {
 impl Tool for Export {
     async fn do_work(&self) -> Result<()> {
         match self.target {
-            ExportTarget::CreateTable => self.export_create_table().await,
-            ExportTarget::TableData => self.export_table_data().await,
-            ExportTarget::DatabaseData => self.export_database_data().await,
+            ExportTarget::Schema => self.export_create_table().await,
+            ExportTarget::Data => self.export_database_data().await,
+            ExportTarget::All => {
+                self.export_create_table().await?;
+                self.export_database_data().await
+            }
         }
     }
 }
 
 /// Split at `-`.
 fn split_database(database: &str) -> Result<(String, Option<String>)> {
-    let (catalog, schema) = database
-        .split_once('-')
-        .with_context(|| InvalidDatabaseNameSnafu {
-            database: database.to_string(),
-        })?;
+    let (catalog, schema) = match database.split_once('-') {
+        Some((catalog, schema)) => (catalog, schema),
+        None => (DEFAULT_CATALOG_NAME, database),
+    };
+
     if schema == "*" {
         Ok((catalog.to_string(), None))
     } else {
@@ -558,10 +471,26 @@ mod tests {
     use common_catalog::consts::{DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
     use common_telemetry::logging::LoggingOptions;
 
+    use crate::cli::export::split_database;
     use crate::error::Result as CmdResult;
     use crate::options::GlobalOptions;
     use crate::{cli, standalone, App};
 
+    #[test]
+    fn test_split_database() {
+        let result = split_database("catalog-schema").unwrap();
+        assert_eq!(result, ("catalog".to_string(), Some("schema".to_string())));
+
+        let result = split_database("schema").unwrap();
+        assert_eq!(result, ("greptime".to_string(), Some("schema".to_string())));
+
+        let result = split_database("catalog-*").unwrap();
+        assert_eq!(result, ("catalog".to_string(), None));
+
+        let result = split_database("*").unwrap();
+        assert_eq!(result, ("greptime".to_string(), None));
+    }
+
     #[tokio::test(flavor = "multi_thread")]
     async fn test_export_create_table_with_quoted_names() -> CmdResult<()> {
         let output_dir = tempfile::tempdir().unwrap();
@@ -603,7 +532,7 @@ mod tests {
             "--output-dir",
             &*output_dir.path().to_string_lossy(),
             "--target",
-            "create-table",
+            "schema",
         ]);
         let mut cli_app = cli.build(LoggingOptions::default()).await?;
         cli_app.start().await?;

src/cmd/src/cli/repl.rs

@@ -34,7 +34,6 @@ use common_telemetry::debug;
 use either::Either;
 use meta_client::client::MetaClientBuilder;
 use query::datafusion::DatafusionQueryEngine;
-use query::logical_optimizer::LogicalOptimizer;
 use query::parser::QueryLanguageParser;
 use query::plan::LogicalPlan;
 use query::query_engine::{DefaultSerializer, QueryEngineState};
@@ -289,6 +288,7 @@ async fn create_query_engine(meta_addr: &str) -> Result<DatafusionQueryEngine> {
         None,
         None,
         None,
+        None,
         false,
         plugins.clone(),
     ));

src/cmd/src/error.rs

@@ -298,13 +298,6 @@ pub enum Error {
         error: std::io::Error,
     },
 
-    #[snafu(display("Invalid database name: {}", database))]
-    InvalidDatabaseName {
-        #[snafu(implicit)]
-        location: Location,
-        database: String,
-    },
-
     #[snafu(display("Failed to create directory {}", dir))]
     CreateDir {
         dir: String,
@@ -312,6 +305,12 @@ pub enum Error {
         error: std::io::Error,
     },
 
+    #[snafu(display("Failed to spawn thread"))]
+    SpawnThread {
+        #[snafu(source)]
+        error: std::io::Error,
+    },
+
     #[snafu(display("Other error"))]
     Other {
         source: BoxedError,
@@ -384,8 +383,7 @@ impl ErrorExt for Error {
             | Error::ConnectEtcd { .. }
             | Error::NotDataFromOutput { .. }
             | Error::CreateDir { .. }
-            | Error::EmptyResult { .. }
-            | Error::InvalidDatabaseName { .. } => StatusCode::InvalidArguments,
+            | Error::EmptyResult { .. } => StatusCode::InvalidArguments,
 
             Error::StartProcedureManager { source, .. }
             | Error::StopProcedureManager { source, .. } => source.status_code(),
@@ -403,7 +401,9 @@ impl ErrorExt for Error {
             Error::SubstraitEncodeLogicalPlan { source, .. } => source.status_code(),
             Error::StartCatalogManager { source, .. } => source.status_code(),
 
-            Error::SerdeJson { .. } | Error::FileIo { .. } => StatusCode::Unexpected,
+            Error::SerdeJson { .. } | Error::FileIo { .. } | Error::SpawnThread { .. } => {
+                StatusCode::Unexpected
+            }
 
             Error::Other { source, .. } => source.status_code(),
 

src/cmd/src/flownode.rs

@@ -24,6 +24,7 @@ use common_grpc::channel_manager::ChannelConfig;
 use common_meta::cache::{CacheRegistryBuilder, LayeredCacheRegistryBuilder};
 use common_meta::heartbeat::handler::parse_mailbox_message::ParseMailboxMessageHandler;
 use common_meta::heartbeat::handler::HandlerGroupExecutor;
+use common_meta::key::flow::FlowMetadataManager;
 use common_meta::key::TableMetadataManager;
 use common_telemetry::info;
 use common_telemetry::logging::TracingOptions;
@@ -296,11 +297,13 @@ impl StartCommand {
             Arc::new(executor),
         );
 
+        let flow_metadata_manager = Arc::new(FlowMetadataManager::new(cached_meta_backend.clone()));
         let flownode_builder = FlownodeBuilder::new(
             opts,
             Plugins::new(),
             table_metadata_manager,
             catalog_manager.clone(),
+            flow_metadata_manager,
         )
         .with_heartbeat_task(heartbeat_task);
 

src/cmd/src/frontend.rs

@@ -243,7 +243,6 @@ impl StartCommand {
                 .get_or_insert_with(MetaClientOptions::default)
                 .metasrv_addrs
                 .clone_from(metasrv_addrs);
-            opts.mode = Mode::Distributed;
         }
 
         if let Some(user_provider) = &self.user_provider {
@@ -317,7 +316,7 @@ impl StartCommand {
         );
 
         let catalog_manager = KvBackendCatalogManager::new(
-            opts.mode,
+            Mode::Distributed,
             Some(meta_client.clone()),
             cached_meta_backend.clone(),
             layered_cache_registry.clone(),
@@ -448,7 +447,6 @@ mod tests {
 
         let fe_opts = command.load_options(&Default::default()).unwrap().component;
 
-        assert_eq!(Mode::Distributed, fe_opts.mode);
         assert_eq!("127.0.0.1:4000".to_string(), fe_opts.http.addr);
         assert_eq!(Duration::from_secs(30), fe_opts.http.timeout);
 

src/cmd/src/lib.rs

@@ -30,7 +30,7 @@ pub mod standalone;
 
 lazy_static::lazy_static! {
     static ref APP_VERSION: prometheus::IntGaugeVec =
-        prometheus::register_int_gauge_vec!("greptime_app_version", "app version", &["short_version", "version"]).unwrap();
+        prometheus::register_int_gauge_vec!("greptime_app_version", "app version", &["version", "short_version"]).unwrap();
 }
 
 #[async_trait]
@@ -74,16 +74,16 @@ pub trait App: Send {
 }
 
 /// Log the versions of the application, and the arguments passed to the cli.
-/// `version_string` should be the same as the output of cli "--version";
-/// and the `app_version` is the short version of the codes, often consist of git branch and commit.
-pub fn log_versions(version_string: &str, app_version: &str) {
+/// `version` should be the same as the output of cli "--version";
+/// and the `short_version` is the short version of the codes, often consist of git branch and commit.
+pub fn log_versions(version: &str, short_version: &str) {
     // Report app version as gauge.
     APP_VERSION
-        .with_label_values(&[env!("CARGO_PKG_VERSION"), app_version])
+        .with_label_values(&[env!("CARGO_PKG_VERSION"), short_version])
         .inc();
 
     // Log version and argument flags.
-    info!("GreptimeDB version: {}", version_string);
+    info!("GreptimeDB version: {}", version);
 
     log_env_flags();
 }

src/cmd/src/metasrv.rs

@@ -22,6 +22,7 @@ use common_telemetry::info;
 use common_telemetry::logging::TracingOptions;
 use common_version::{short_version, version};
 use meta_srv::bootstrap::MetasrvInstance;
+use meta_srv::metasrv::BackendImpl;
 use snafu::ResultExt;
 use tracing_appender::non_blocking::WorkerGuard;
 
@@ -137,6 +138,9 @@ struct StartCommand {
     /// The max operations per txn
     #[clap(long)]
     max_txn_ops: Option<usize>,
+    /// The database backend.
+    #[clap(long, value_enum)]
+    backend: Option<BackendImpl>,
 }
 
 impl StartCommand {
@@ -219,6 +223,12 @@ impl StartCommand {
             opts.max_txn_ops = max_txn_ops;
         }
 
+        if let Some(backend) = &self.backend {
+            opts.backend.clone_from(backend);
+        } else {
+            opts.backend = BackendImpl::default()
+        }
+
         // Disable dashboard in metasrv.
         opts.http.disable_dashboard = true;
 

src/cmd/src/standalone.rs

@@ -182,7 +182,6 @@ impl StandaloneOptions {
     pub fn frontend_options(&self) -> FrontendOptions {
         let cloned_opts = self.clone();
         FrontendOptions {
-            mode: cloned_opts.mode,
             default_timezone: cloned_opts.default_timezone,
             http: cloned_opts.http,
             grpc: cloned_opts.grpc,
@@ -477,11 +476,13 @@ impl StartCommand {
             .await
             .context(StartDatanodeSnafu)?;
 
+        let flow_metadata_manager = Arc::new(FlowMetadataManager::new(kv_backend.clone()));
         let flow_builder = FlownodeBuilder::new(
             Default::default(),
             plugins.clone(),
             table_metadata_manager.clone(),
             catalog_manager.clone(),
+            flow_metadata_manager.clone(),
         );
         let flownode = Arc::new(
             flow_builder
@@ -512,7 +513,6 @@ impl StartCommand {
             opts.wal.into(),
             kv_backend.clone(),
         ));
-        let flow_metadata_manager = Arc::new(FlowMetadataManager::new(kv_backend.clone()));
         let table_meta_allocator = Arc::new(TableMetadataAllocator::new(
             table_id_sequence,
             wal_options_allocator.clone(),
@@ -636,7 +636,6 @@ mod tests {
     use common_test_util::temp_dir::create_named_temp_file;
     use common_wal::config::DatanodeWalConfig;
     use datanode::config::{FileConfig, GcsConfig};
-    use servers::Mode;
 
     use super::*;
     use crate::options::GlobalOptions;
@@ -736,7 +735,6 @@ mod tests {
         let fe_opts = options.frontend_options();
         let dn_opts = options.datanode_options();
         let logging_opts = options.logging;
-        assert_eq!(Mode::Standalone, fe_opts.mode);
         assert_eq!("127.0.0.1:4000".to_string(), fe_opts.http.addr);
         assert_eq!(Duration::from_secs(33), fe_opts.http.timeout);
         assert_eq!(ReadableSize::mb(128), fe_opts.http.body_limit);

src/cmd/tests/load_config_test.rs

@@ -22,7 +22,7 @@ use common_grpc::channel_manager::{
     DEFAULT_MAX_GRPC_RECV_MESSAGE_SIZE, DEFAULT_MAX_GRPC_SEND_MESSAGE_SIZE,
 };
 use common_runtime::global::RuntimeOptions;
-use common_telemetry::logging::LoggingOptions;
+use common_telemetry::logging::{LoggingOptions, DEFAULT_OTLP_ENDPOINT};
 use common_wal::config::raft_engine::RaftEngineConfig;
 use common_wal::config::DatanodeWalConfig;
 use datanode::config::{DatanodeOptions, RegionEngineConfig, StorageConfig};
@@ -46,9 +46,8 @@ fn test_load_datanode_example_config() {
 
     let expected = GreptimeOptions::<DatanodeOptions> {
         runtime: RuntimeOptions {
-            read_rt_size: 8,
-            write_rt_size: 8,
-            bg_rt_size: 4,
+            global_rt_size: 8,
+            compact_rt_size: 4,
         },
         component: DatanodeOptions {
             node_id: Some(42),
@@ -89,7 +88,7 @@ fn test_load_datanode_example_config() {
             ],
             logging: LoggingOptions {
                 level: Some("info".to_string()),
-                otlp_endpoint: Some("".to_string()),
+                otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                 tracing_sample_ratio: Some(Default::default()),
                 ..Default::default()
             },
@@ -119,9 +118,8 @@ fn test_load_frontend_example_config() {
             .unwrap();
     let expected = GreptimeOptions::<FrontendOptions> {
         runtime: RuntimeOptions {
-            read_rt_size: 8,
-            write_rt_size: 8,
-            bg_rt_size: 4,
+            global_rt_size: 8,
+            compact_rt_size: 4,
         },
         component: FrontendOptions {
             default_timezone: Some("UTC".to_string()),
@@ -138,7 +136,7 @@ fn test_load_frontend_example_config() {
             }),
             logging: LoggingOptions {
                 level: Some("info".to_string()),
-                otlp_endpoint: Some("".to_string()),
+                otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                 tracing_sample_ratio: Some(Default::default()),
                 ..Default::default()
             },
@@ -167,17 +165,16 @@ fn test_load_metasrv_example_config() {
             .unwrap();
     let expected = GreptimeOptions::<MetasrvOptions> {
         runtime: RuntimeOptions {
-            read_rt_size: 8,
-            write_rt_size: 8,
-            bg_rt_size: 4,
+            global_rt_size: 8,
+            compact_rt_size: 4,
         },
         component: MetasrvOptions {
-            selector: SelectorType::LeaseBased,
+            selector: SelectorType::default(),
             data_home: "/tmp/metasrv/".to_string(),
             logging: LoggingOptions {
                 dir: "/tmp/greptimedb/logs".to_string(),
                 level: Some("info".to_string()),
-                otlp_endpoint: Some("".to_string()),
+                otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                 tracing_sample_ratio: Some(Default::default()),
                 ..Default::default()
             },
@@ -200,9 +197,8 @@ fn test_load_standalone_example_config() {
             .unwrap();
     let expected = GreptimeOptions::<StandaloneOptions> {
         runtime: RuntimeOptions {
-            read_rt_size: 8,
-            write_rt_size: 8,
-            bg_rt_size: 4,
+            global_rt_size: 8,
+            compact_rt_size: 4,
         },
         component: StandaloneOptions {
             default_timezone: Some("UTC".to_string()),
@@ -232,7 +228,7 @@ fn test_load_standalone_example_config() {
             },
             logging: LoggingOptions {
                 level: Some("info".to_string()),
-                otlp_endpoint: Some("".to_string()),
+                otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                 tracing_sample_ratio: Some(Default::default()),
                 ..Default::default()
             },

src/common/catalog/src/consts.rs

@@ -103,6 +103,7 @@ pub const INFORMATION_SCHEMA_FLOW_TABLE_ID: u32 = 33;
 /// ----- Begin of pg_catalog tables -----
 pub const PG_CATALOG_PG_CLASS_TABLE_ID: u32 = 256;
 pub const PG_CATALOG_PG_TYPE_TABLE_ID: u32 = 257;
+pub const PG_CATALOG_PG_NAMESPACE_TABLE_ID: u32 = 258;
 
 /// ----- End of pg_catalog tables -----
 pub const MITO_ENGINE: &str = "mito";

src/common/datasource/src/file_format/csv.rs

@@ -185,7 +185,7 @@ impl FileFormat for CsvFormat {
         let schema_infer_max_record = self.schema_infer_max_record;
         let has_header = self.has_header;
 
-        common_runtime::spawn_blocking_read(move || {
+        common_runtime::spawn_blocking_global(move || {
             let reader = SyncIoBridge::new(decoded);
 
             let (schema, _records_read) =

src/common/datasource/src/file_format/json.rs

@@ -101,7 +101,7 @@ impl FileFormat for JsonFormat {
 
         let schema_infer_max_record = self.schema_infer_max_record;
 
-        common_runtime::spawn_blocking_read(move || {
+        common_runtime::spawn_blocking_global(move || {
             let mut reader = BufReader::new(SyncIoBridge::new(decoded));
 
             let iter = ValueIter::new(&mut reader, schema_infer_max_record);

src/common/datasource/src/object_store/fs.rs

@@ -19,9 +19,8 @@ use snafu::ResultExt;
 use crate::error::{BuildBackendSnafu, Result};
 
 pub fn build_fs_backend(root: &str) -> Result<ObjectStore> {
-    let mut builder = Fs::default();
-    let _ = builder.root(root);
-    let object_store = ObjectStore::new(builder)
+    let builder = Fs::default();
+    let object_store = ObjectStore::new(builder.root(root))
         .context(BuildBackendSnafu)?
         .layer(
             object_store::layers::LoggingLayer::default()

src/common/datasource/src/object_store/s3.rs

@@ -44,28 +44,26 @@ pub fn build_s3_backend(
     path: &str,
     connection: &HashMap<String, String>,
 ) -> Result<ObjectStore> {
-    let mut builder = S3::default();
-
-    let _ = builder.root(path).bucket(host);
+    let mut builder = S3::default().root(path).bucket(host);
 
     if let Some(endpoint) = connection.get(ENDPOINT) {
-        let _ = builder.endpoint(endpoint);
+        builder = builder.endpoint(endpoint);
     }
 
     if let Some(region) = connection.get(REGION) {
-        let _ = builder.region(region);
+        builder = builder.region(region);
     }
 
     if let Some(key_id) = connection.get(ACCESS_KEY_ID) {
-        let _ = builder.access_key_id(key_id);
+        builder = builder.access_key_id(key_id);
     }
 
     if let Some(key) = connection.get(SECRET_ACCESS_KEY) {
-        let _ = builder.secret_access_key(key);
+        builder = builder.secret_access_key(key);
     }
 
     if let Some(session_token) = connection.get(SESSION_TOKEN) {
-        let _ = builder.security_token(session_token);
+        builder = builder.session_token(session_token);
     }
 
     if let Some(enable_str) = connection.get(ENABLE_VIRTUAL_HOST_STYLE) {
@@ -79,7 +77,7 @@ pub fn build_s3_backend(
             .build()
         })?;
         if enable {
-            let _ = builder.enable_virtual_host_style();
+            builder = builder.enable_virtual_host_style();
         }
     }
 

src/common/datasource/src/test_util.rs

@@ -47,19 +47,15 @@ pub fn format_schema(schema: Schema) -> Vec<String> {
 }
 
 pub fn test_store(root: &str) -> ObjectStore {
-    let mut builder = Fs::default();
-    let _ = builder.root(root);
-
-    ObjectStore::new(builder).unwrap().finish()
+    let builder = Fs::default();
+    ObjectStore::new(builder.root(root)).unwrap().finish()
 }
 
 pub fn test_tmp_store(root: &str) -> (ObjectStore, TempDir) {
     let dir = create_temp_dir(root);
 
-    let mut builder = Fs::default();
-    let _ = builder.root("/");
-
-    (ObjectStore::new(builder).unwrap().finish(), dir)
+    let builder = Fs::default();
+    (ObjectStore::new(builder.root("/")).unwrap().finish(), dir)
 }
 
 pub fn test_basic_schema() -> SchemaRef {

src/common/error/src/ext.rs

@@ -53,6 +53,20 @@ pub trait ErrorExt: StackError {
             }
         }
     }
+
+    /// Find out root level error for nested error
+    fn root_cause(&self) -> Option<&dyn std::error::Error>
+    where
+        Self: Sized,
+    {
+        let error = self.last();
+        if let Some(external_error) = error.source() {
+            let external_root = external_error.sources().last().unwrap();
+            Some(external_root)
+        } else {
+            None
+        }
+    }
 }
 
 pub trait StackError: std::error::Error {

src/common/error/src/status_code.rs

@@ -36,7 +36,7 @@ pub enum StatusCode {
     InvalidArguments = 1004,
     /// The task is cancelled.
     Cancelled = 1005,
-    /// Illegal state, can be exposed to users
+    /// Illegal state, can be exposed to users.
     IllegalState = 1006,
     // ====== End of common status code ================
 
@@ -55,33 +55,34 @@ pub enum StatusCode {
     // ====== Begin of catalog related status code =====
     /// Table already exists.
     TableAlreadyExists = 4000,
-    /// Table not found
+    /// Table not found.
     TableNotFound = 4001,
-    /// Table column not found
+    /// Table column not found.
     TableColumnNotFound = 4002,
-    /// Table column already exists
+    /// Table column already exists.
     TableColumnExists = 4003,
-    /// Database not found
+    /// Database not found.
     DatabaseNotFound = 4004,
-    /// Region not found
+    /// Region not found.
     RegionNotFound = 4005,
-    /// Region already exists
+    /// Region already exists.
     RegionAlreadyExists = 4006,
+    /// Region is read-only in current state.
     RegionReadonly = 4007,
     /// Region is not in a proper state to handle specific request.
     RegionNotReady = 4008,
-    /// Region is temporarily in busy state
+    /// Region is temporarily in busy state.
     RegionBusy = 4009,
-    /// Table is temporarily unable to handle the request
+    /// Table is temporarily unable to handle the request.
     TableUnavailable = 4010,
-    /// Database not found
+    /// Database already exists.
     DatabaseAlreadyExists = 4011,
     // ====== End of catalog related status code =======
 
     // ====== Begin of storage related status code =====
-    /// Storage is temporarily unable to handle the request
+    /// Storage is temporarily unable to handle the request.
     StorageUnavailable = 5000,
-    /// Request is outdated, e.g., version mismatch
+    /// Request is outdated, e.g., version mismatch.
     RequestOutdated = 5001,
     // ====== End of storage related status code =======
 
@@ -89,24 +90,24 @@ pub enum StatusCode {
     /// Runtime resources exhausted, like creating threads failed.
     RuntimeResourcesExhausted = 6000,
 
-    /// Rate limit exceeded
+    /// Rate limit exceeded.
     RateLimited = 6001,
     // ====== End of server related status code =======
 
     // ====== Begin of auth related status code =====
-    /// User not exist
+    /// User not exist.
     UserNotFound = 7000,
-    /// Unsupported password type
+    /// Unsupported password type.
     UnsupportedPasswordType = 7001,
-    /// Username and password does not match
+    /// Username and password does not match.
     UserPasswordMismatch = 7002,
-    /// Not found http authorization header
+    /// Not found http authorization header.
     AuthHeaderNotFound = 7003,
-    /// Invalid http authorization header
+    /// Invalid http authorization header.
     InvalidAuthHeader = 7004,
-    /// Illegal request to connect catalog-schema
+    /// Illegal request to connect catalog-schema.
     AccessDenied = 7005,
-    /// User is not authorized to perform the operation
+    /// User is not authorized to perform the operation.
     PermissionDenied = 7006,
     // ====== End of auth related status code =====
 

src/common/function/Cargo.toml

@@ -31,6 +31,7 @@ serde.workspace = true
 serde_json.workspace = true
 session.workspace = true
 snafu.workspace = true
+sql.workspace = true
 statrs = "0.16"
 store-api.workspace = true
 table.workspace = true

src/common/function/src/flush_flow.rs

@@ -0,0 +1,164 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use common_error::ext::BoxedError;
+use common_macro::admin_fn;
+use common_query::error::{
+    ExecuteSnafu, InvalidFuncArgsSnafu, MissingFlowServiceHandlerSnafu, Result,
+    UnsupportedInputDataTypeSnafu,
+};
+use common_query::prelude::Signature;
+use datafusion::logical_expr::Volatility;
+use datatypes::value::{Value, ValueRef};
+use session::context::QueryContextRef;
+use snafu::{ensure, ResultExt};
+use sql::parser::ParserContext;
+use store_api::storage::ConcreteDataType;
+
+use crate::handlers::FlowServiceHandlerRef;
+
+fn flush_signature() -> Signature {
+    Signature::uniform(
+        1,
+        vec![ConcreteDataType::string_datatype()],
+        Volatility::Immutable,
+    )
+}
+
+#[admin_fn(
+    name = FlushFlowFunction,
+    display_name = flush_flow,
+    sig_fn = flush_signature,
+    ret = uint64
+)]
+pub(crate) async fn flush_flow(
+    flow_service_handler: &FlowServiceHandlerRef,
+    query_ctx: &QueryContextRef,
+    params: &[ValueRef<'_>],
+) -> Result<Value> {
+    let (catalog_name, flow_name) = parse_flush_flow(params, query_ctx)?;
+
+    let res = flow_service_handler
+        .flush(&catalog_name, &flow_name, query_ctx.clone())
+        .await?;
+    let affected_rows = res.affected_rows;
+
+    Ok(Value::from(affected_rows))
+}
+
+fn parse_flush_flow(
+    params: &[ValueRef<'_>],
+    query_ctx: &QueryContextRef,
+) -> Result<(String, String)> {
+    ensure!(
+        params.len() == 1,
+        InvalidFuncArgsSnafu {
+            err_msg: format!(
+                "The length of the args is not correct, expect 1, have: {}",
+                params.len()
+            ),
+        }
+    );
+
+    let ValueRef::String(flow_name) = params[0] else {
+        return UnsupportedInputDataTypeSnafu {
+            function: "flush_flow",
+            datatypes: params.iter().map(|v| v.data_type()).collect::<Vec<_>>(),
+        }
+        .fail();
+    };
+    let obj_name = ParserContext::parse_table_name(flow_name, query_ctx.sql_dialect())
+        .map_err(BoxedError::new)
+        .context(ExecuteSnafu)?;
+
+    let (catalog_name, flow_name) = match &obj_name.0[..] {
+        [flow_name] => (
+            query_ctx.current_catalog().to_string(),
+            flow_name.value.clone(),
+        ),
+        [catalog, flow_name] => (catalog.value.clone(), flow_name.value.clone()),
+        _ => {
+            return InvalidFuncArgsSnafu {
+                err_msg: format!(
+                    "expect flow name to be <catalog>.<flow-name> or <flow-name>, actual: {}",
+                    obj_name
+                ),
+            }
+            .fail()
+        }
+    };
+    Ok((catalog_name, flow_name))
+}
+
+#[cfg(test)]
+mod test {
+    use std::sync::Arc;
+
+    use datatypes::scalars::ScalarVector;
+    use datatypes::vectors::StringVector;
+    use session::context::QueryContext;
+
+    use super::*;
+    use crate::function::{Function, FunctionContext};
+
+    #[test]
+    fn test_flush_flow_metadata() {
+        let f = FlushFlowFunction;
+        assert_eq!("flush_flow", f.name());
+        assert_eq!(
+            ConcreteDataType::uint64_datatype(),
+            f.return_type(&[]).unwrap()
+        );
+        assert_eq!(
+            f.signature(),
+            Signature::uniform(
+                1,
+                vec![ConcreteDataType::string_datatype()],
+                Volatility::Immutable,
+            )
+        );
+    }
+
+    #[test]
+    fn test_missing_flow_service() {
+        let f = FlushFlowFunction;
+
+        let args = vec!["flow_name"];
+        let args = args
+            .into_iter()
+            .map(|arg| Arc::new(StringVector::from_slice(&[arg])) as _)
+            .collect::<Vec<_>>();
+
+        let result = f.eval(FunctionContext::default(), &args).unwrap_err();
+        assert_eq!(
+            "Missing FlowServiceHandler, not expected",
+            result.to_string()
+        );
+    }
+
+    #[test]
+    fn test_parse_flow_args() {
+        let testcases = [
+            ("flow_name", ("greptime", "flow_name")),
+            ("catalog.flow_name", ("catalog", "flow_name")),
+        ];
+        for (input, expected) in testcases.iter() {
+            let args = vec![*input];
+            let args = args.into_iter().map(ValueRef::String).collect::<Vec<_>>();
+
+            let result = parse_flush_flow(&args, &QueryContext::arc()).unwrap();
+            assert_eq!(*expected, (result.0.as_str(), result.1.as_str()));
+        }
+    }
+}

src/common/function/src/handlers.rs

@@ -65,6 +65,19 @@ pub trait ProcedureServiceHandler: Send + Sync {
     async fn query_procedure_state(&self, pid: &str) -> Result<ProcedureStateResponse>;
 }
 
+/// This flow service handler is only use for flush flow for now.
+#[async_trait]
+pub trait FlowServiceHandler: Send + Sync {
+    async fn flush(
+        &self,
+        catalog: &str,
+        flow: &str,
+        ctx: QueryContextRef,
+    ) -> Result<api::v1::flow::FlowResponse>;
+}
+
 pub type TableMutationHandlerRef = Arc<dyn TableMutationHandler>;
 
 pub type ProcedureServiceHandlerRef = Arc<dyn ProcedureServiceHandler>;
+
+pub type FlowServiceHandlerRef = Arc<dyn FlowServiceHandler>;

src/common/function/src/lib.rs

@@ -15,6 +15,7 @@
 #![feature(let_chains)]
 #![feature(try_blocks)]
 
+mod flush_flow;
 mod macros;
 pub mod scalars;
 mod system;

src/common/function/src/scalars/expression/unary.rs

@@ -19,8 +19,6 @@ use snafu::ResultExt;
 
 use crate::scalars::expression::ctx::EvalContext;
 
-/// TODO: remove the allow_unused when it's used.
-#[allow(unused)]
 pub fn scalar_unary_op<L: Scalar, O: Scalar, F>(
     l: &VectorRef,
     f: F,

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

@@ -14,11 +14,9 @@
 
 use std::sync::Arc;
 mod greatest;
-mod to_timezone;
 mod to_unixtime;
 
 use greatest::GreatestFunction;
-use to_timezone::ToTimezoneFunction;
 use to_unixtime::ToUnixtimeFunction;
 
 use crate::function_registry::FunctionRegistry;
@@ -27,7 +25,6 @@ pub(crate) struct TimestampFunction;
 
 impl TimestampFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(ToTimezoneFunction));
         registry.register(Arc::new(ToUnixtimeFunction));
         registry.register(Arc::new(GreatestFunction));
     }

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

@@ -1,313 +0,0 @@
-// Copyright 2023 Greptime Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-use std::fmt;
-use std::sync::Arc;
-
-use common_query::error::{InvalidFuncArgsSnafu, Result, UnsupportedInputDataTypeSnafu};
-use common_query::prelude::Signature;
-use common_time::{Timestamp, Timezone};
-use datatypes::data_type::ConcreteDataType;
-use datatypes::prelude::VectorRef;
-use datatypes::types::TimestampType;
-use datatypes::value::Value;
-use datatypes::vectors::{
-    Int64Vector, StringVector, TimestampMicrosecondVector, TimestampMillisecondVector,
-    TimestampNanosecondVector, TimestampSecondVector, Vector,
-};
-use snafu::{ensure, OptionExt};
-
-use crate::function::{Function, FunctionContext};
-use crate::helper;
-
-#[derive(Clone, Debug, Default)]
-pub struct ToTimezoneFunction;
-
-const NAME: &str = "to_timezone";
-
-fn convert_to_timezone(arg: &str) -> Option<Timezone> {
-    Timezone::from_tz_string(arg).ok()
-}
-
-fn convert_to_timestamp(arg: &Value) -> Option<Timestamp> {
-    match arg {
-        Value::Timestamp(ts) => Some(*ts),
-        Value::Int64(i) => Some(Timestamp::new_millisecond(*i)),
-        _ => None,
-    }
-}
-
-impl fmt::Display for ToTimezoneFunction {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-        write!(f, "TO_TIMEZONE")
-    }
-}
-
-impl Function for ToTimezoneFunction {
-    fn name(&self) -> &str {
-        NAME
-    }
-
-    fn return_type(&self, input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
-        // type checked by signature - MUST BE timestamp
-        Ok(input_types[0].clone())
-    }
-
-    fn signature(&self) -> Signature {
-        helper::one_of_sigs2(
-            vec![
-                ConcreteDataType::int32_datatype(),
-                ConcreteDataType::int64_datatype(),
-                ConcreteDataType::timestamp_second_datatype(),
-                ConcreteDataType::timestamp_millisecond_datatype(),
-                ConcreteDataType::timestamp_microsecond_datatype(),
-                ConcreteDataType::timestamp_nanosecond_datatype(),
-            ],
-            vec![ConcreteDataType::string_datatype()],
-        )
-    }
-
-    fn eval(&self, _ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 2,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect exactly 2, have: {}",
-                    columns.len()
-                ),
-            }
-        );
-
-        let array = columns[0].to_arrow_array();
-        let times = match columns[0].data_type() {
-            ConcreteDataType::Int64(_) | ConcreteDataType::Int32(_) => {
-                let vector = Int64Vector::try_from_arrow_array(array).unwrap();
-                (0..vector.len())
-                    .map(|i| convert_to_timestamp(&vector.get(i)))
-                    .collect::<Vec<_>>()
-            }
-            ConcreteDataType::Timestamp(ts) => match ts {
-                TimestampType::Second(_) => {
-                    let vector = TimestampSecondVector::try_from_arrow_array(array).unwrap();
-                    (0..vector.len())
-                        .map(|i| convert_to_timestamp(&vector.get(i)))
-                        .collect::<Vec<_>>()
-                }
-                TimestampType::Millisecond(_) => {
-                    let vector = TimestampMillisecondVector::try_from_arrow_array(array).unwrap();
-                    (0..vector.len())
-                        .map(|i| convert_to_timestamp(&vector.get(i)))
-                        .collect::<Vec<_>>()
-                }
-                TimestampType::Microsecond(_) => {
-                    let vector = TimestampMicrosecondVector::try_from_arrow_array(array).unwrap();
-                    (0..vector.len())
-                        .map(|i| convert_to_timestamp(&vector.get(i)))
-                        .collect::<Vec<_>>()
-                }
-                TimestampType::Nanosecond(_) => {
-                    let vector = TimestampNanosecondVector::try_from_arrow_array(array).unwrap();
-                    (0..vector.len())
-                        .map(|i| convert_to_timestamp(&vector.get(i)))
-                        .collect::<Vec<_>>()
-                }
-            },
-            _ => UnsupportedInputDataTypeSnafu {
-                function: NAME,
-                datatypes: columns.iter().map(|c| c.data_type()).collect::<Vec<_>>(),
-            }
-            .fail()?,
-        };
-
-        let tzs = {
-            let array = columns[1].to_arrow_array();
-            let vector = StringVector::try_from_arrow_array(&array)
-                .ok()
-                .with_context(|| UnsupportedInputDataTypeSnafu {
-                    function: NAME,
-                    datatypes: columns.iter().map(|c| c.data_type()).collect::<Vec<_>>(),
-                })?;
-            (0..vector.len())
-                .map(|i| convert_to_timezone(&vector.get(i).to_string()))
-                .collect::<Vec<_>>()
-        };
-
-        let result = times
-            .iter()
-            .zip(tzs.iter())
-            .map(|(time, tz)| match (time, tz) {
-                (Some(time), _) => Some(time.to_timezone_aware_string(tz.as_ref())),
-                _ => None,
-            })
-            .collect::<Vec<Option<String>>>();
-        Ok(Arc::new(StringVector::from(result)))
-    }
-}
-
-#[cfg(test)]
-mod tests {
-
-    use datatypes::scalars::ScalarVector;
-    use datatypes::timestamp::{
-        TimestampMicrosecond, TimestampMillisecond, TimestampNanosecond, TimestampSecond,
-    };
-    use datatypes::vectors::{Int64Vector, StringVector};
-
-    use super::*;
-
-    #[test]
-    fn test_timestamp_to_timezone() {
-        let f = ToTimezoneFunction;
-        assert_eq!("to_timezone", f.name());
-
-        let results = vec![
-            Some("1969-12-31 19:00:01"),
-            None,
-            Some("1970-01-01 03:00:01"),
-            None,
-        ];
-        let times: Vec<Option<TimestampSecond>> = vec![
-            Some(TimestampSecond::new(1)),
-            None,
-            Some(TimestampSecond::new(1)),
-            None,
-        ];
-        let ts_vector: TimestampSecondVector =
-            TimestampSecondVector::from_owned_iterator(times.into_iter());
-        let tzs = vec![Some("America/New_York"), None, Some("Europe/Moscow"), None];
-        let args: Vec<VectorRef> = vec![
-            Arc::new(ts_vector),
-            Arc::new(StringVector::from(tzs.clone())),
-        ];
-        let vector = f.eval(FunctionContext::default(), &args).unwrap();
-        assert_eq!(4, vector.len());
-        let expect_times: VectorRef = Arc::new(StringVector::from(results));
-        assert_eq!(expect_times, vector);
-
-        let results = vec![
-            Some("1969-12-31 19:00:00.001"),
-            None,
-            Some("1970-01-01 03:00:00.001"),
-            None,
-        ];
-        let times: Vec<Option<TimestampMillisecond>> = vec![
-            Some(TimestampMillisecond::new(1)),
-            None,
-            Some(TimestampMillisecond::new(1)),
-            None,
-        ];
-        let ts_vector: TimestampMillisecondVector =
-            TimestampMillisecondVector::from_owned_iterator(times.into_iter());
-        let args: Vec<VectorRef> = vec![
-            Arc::new(ts_vector),
-            Arc::new(StringVector::from(tzs.clone())),
-        ];
-        let vector = f.eval(FunctionContext::default(), &args).unwrap();
-        assert_eq!(4, vector.len());
-        let expect_times: VectorRef = Arc::new(StringVector::from(results));
-        assert_eq!(expect_times, vector);
-
-        let results = vec![
-            Some("1969-12-31 19:00:00.000001"),
-            None,
-            Some("1970-01-01 03:00:00.000001"),
-            None,
-        ];
-        let times: Vec<Option<TimestampMicrosecond>> = vec![
-            Some(TimestampMicrosecond::new(1)),
-            None,
-            Some(TimestampMicrosecond::new(1)),
-            None,
-        ];
-        let ts_vector: TimestampMicrosecondVector =
-            TimestampMicrosecondVector::from_owned_iterator(times.into_iter());
-
-        let args: Vec<VectorRef> = vec![
-            Arc::new(ts_vector),
-            Arc::new(StringVector::from(tzs.clone())),
-        ];
-        let vector = f.eval(FunctionContext::default(), &args).unwrap();
-        assert_eq!(4, vector.len());
-        let expect_times: VectorRef = Arc::new(StringVector::from(results));
-        assert_eq!(expect_times, vector);
-
-        let results = vec![
-            Some("1969-12-31 19:00:00.000000001"),
-            None,
-            Some("1970-01-01 03:00:00.000000001"),
-            None,
-        ];
-        let times: Vec<Option<TimestampNanosecond>> = vec![
-            Some(TimestampNanosecond::new(1)),
-            None,
-            Some(TimestampNanosecond::new(1)),
-            None,
-        ];
-        let ts_vector: TimestampNanosecondVector =
-            TimestampNanosecondVector::from_owned_iterator(times.into_iter());
-
-        let args: Vec<VectorRef> = vec![
-            Arc::new(ts_vector),
-            Arc::new(StringVector::from(tzs.clone())),
-        ];
-        let vector = f.eval(FunctionContext::default(), &args).unwrap();
-        assert_eq!(4, vector.len());
-        let expect_times: VectorRef = Arc::new(StringVector::from(results));
-        assert_eq!(expect_times, vector);
-    }
-
-    #[test]
-    fn test_numerical_to_timezone() {
-        let f = ToTimezoneFunction;
-        let results = vec![
-            Some("1969-12-31 19:00:00.001"),
-            None,
-            Some("1970-01-01 03:00:00.001"),
-            None,
-            Some("2024-03-26 23:01:50"),
-            None,
-            Some("2024-03-27 06:02:00"),
-            None,
-        ];
-        let times: Vec<Option<i64>> = vec![
-            Some(1),
-            None,
-            Some(1),
-            None,
-            Some(1711508510000),
-            None,
-            Some(1711508520000),
-            None,
-        ];
-        let ts_vector: Int64Vector = Int64Vector::from_owned_iterator(times.into_iter());
-        let tzs = vec![
-            Some("America/New_York"),
-            None,
-            Some("Europe/Moscow"),
-            None,
-            Some("America/New_York"),
-            None,
-            Some("Europe/Moscow"),
-            None,
-        ];
-        let args: Vec<VectorRef> = vec![
-            Arc::new(ts_vector),
-            Arc::new(StringVector::from(tzs.clone())),
-        ];
-        let vector = f.eval(FunctionContext::default(), &args).unwrap();
-        assert_eq!(8, vector.len());
-        let expect_times: VectorRef = Arc::new(StringVector::from(results));
-        assert_eq!(expect_times, vector);
-    }
-}

src/common/function/src/state.rs

@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use crate::handlers::{ProcedureServiceHandlerRef, TableMutationHandlerRef};
+use crate::handlers::{FlowServiceHandlerRef, ProcedureServiceHandlerRef, TableMutationHandlerRef};
 
 /// Shared state for SQL functions.
 /// The handlers in state may be `None` in cli command-line or test cases.
@@ -22,6 +22,8 @@ pub struct FunctionState {
     pub table_mutation_handler: Option<TableMutationHandlerRef>,
     // The procedure service handler
     pub procedure_service_handler: Option<ProcedureServiceHandlerRef>,
+    // The flownode handler
+    pub flow_service_handler: Option<FlowServiceHandlerRef>,
 }
 
 impl FunctionState {
@@ -42,9 +44,10 @@ impl FunctionState {
             CompactTableRequest, DeleteRequest, FlushTableRequest, InsertRequest,
         };
 
-        use crate::handlers::{ProcedureServiceHandler, TableMutationHandler};
+        use crate::handlers::{FlowServiceHandler, ProcedureServiceHandler, TableMutationHandler};
         struct MockProcedureServiceHandler;
         struct MockTableMutationHandler;
+        struct MockFlowServiceHandler;
         const ROWS: usize = 42;
 
         #[async_trait]
@@ -116,9 +119,22 @@ impl FunctionState {
             }
         }
 
+        #[async_trait]
+        impl FlowServiceHandler for MockFlowServiceHandler {
+            async fn flush(
+                &self,
+                _catalog: &str,
+                _flow: &str,
+                _ctx: QueryContextRef,
+            ) -> Result<api::v1::flow::FlowResponse> {
+                todo!()
+            }
+        }
+
         Self {
             table_mutation_handler: Some(Arc::new(MockTableMutationHandler)),
             procedure_service_handler: Some(Arc::new(MockProcedureServiceHandler)),
+            flow_service_handler: Some(Arc::new(MockFlowServiceHandler)),
         }
     }
 }

src/common/function/src/system.rs

@@ -14,6 +14,7 @@
 
 mod build;
 mod database;
+mod pg_catalog;
 mod procedure_state;
 mod timezone;
 mod version;
@@ -22,6 +23,7 @@ use std::sync::Arc;
 
 use build::BuildFunction;
 use database::DatabaseFunction;
+use pg_catalog::PGCatalogFunction;
 use procedure_state::ProcedureStateFunction;
 use timezone::TimezoneFunction;
 use version::VersionFunction;
@@ -37,5 +39,6 @@ impl SystemFunction {
         registry.register(Arc::new(DatabaseFunction));
         registry.register(Arc::new(TimezoneFunction));
         registry.register(Arc::new(ProcedureStateFunction));
+        PGCatalogFunction::register(registry);
     }
 }

src/common/function/src/system/pg_catalog.rs

@@ -12,22 +12,28 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+mod pg_get_userbyid;
+mod table_is_visible;
+
 use std::sync::Arc;
 
-use datafusion::physical_plan::ExecutionPlan;
+use pg_get_userbyid::PGGetUserByIdFunction;
+use table_is_visible::PGTableIsVisibleFunction;
 
-use crate::error::Result;
-use crate::plan::LogicalPlan;
-use crate::query_engine::QueryEngineContext;
+use crate::function_registry::FunctionRegistry;
 
-/// Physical query planner that converts a `LogicalPlan` to an
-/// `ExecutionPlan` suitable for execution.
-#[async_trait::async_trait]
-pub trait PhysicalPlanner {
-    /// Create a physical plan from a logical plan
-    async fn create_physical_plan(
-        &self,
-        ctx: &mut QueryEngineContext,
-        logical_plan: &LogicalPlan,
-    ) -> Result<Arc<dyn ExecutionPlan>>;
+#[macro_export]
+macro_rules! pg_catalog_func_fullname {
+    ($name:literal) => {
+        concat!("pg_catalog.", $name)
+    };
+}
+
+pub(super) struct PGCatalogFunction;
+
+impl PGCatalogFunction {
+    pub fn register(registry: &FunctionRegistry) {
+        registry.register(Arc::new(PGTableIsVisibleFunction));
+        registry.register(Arc::new(PGGetUserByIdFunction));
+    }
 }

src/common/function/src/system/pg_catalog/pg_get_userbyid.rs

@@ -0,0 +1,72 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::fmt::{self};
+use std::sync::Arc;
+
+use common_query::error::Result;
+use common_query::prelude::{Signature, Volatility};
+use datatypes::prelude::{ConcreteDataType, DataType, VectorRef};
+use datatypes::types::LogicalPrimitiveType;
+use datatypes::with_match_primitive_type_id;
+use num_traits::AsPrimitive;
+
+use crate::function::{Function, FunctionContext};
+use crate::scalars::expression::{scalar_unary_op, EvalContext};
+
+#[derive(Clone, Debug, Default)]
+pub struct PGGetUserByIdFunction;
+
+const NAME: &str = crate::pg_catalog_func_fullname!("pg_get_userbyid");
+
+impl fmt::Display for PGGetUserByIdFunction {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, crate::pg_catalog_func_fullname!("PG_GET_USERBYID"))
+    }
+}
+
+impl Function for PGGetUserByIdFunction {
+    fn name(&self) -> &str {
+        NAME
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::string_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        Signature::uniform(
+            1,
+            vec![ConcreteDataType::uint32_datatype()],
+            Volatility::Immutable,
+        )
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        with_match_primitive_type_id!(columns[0].data_type().logical_type_id(), |$T| {
+            let col = scalar_unary_op::<<$T as LogicalPrimitiveType>::Native, String, _>(&columns[0], pg_get_user_by_id, &mut EvalContext::default())?;
+            Ok(Arc::new(col))
+        }, {
+            unreachable!()
+        })
+    }
+}
+
+fn pg_get_user_by_id<I>(table_oid: Option<I>, _ctx: &mut EvalContext) -> Option<String>
+where
+    I: AsPrimitive<u32>,
+{
+    // TODO(J0HN50N133): We lack way to get the user_info by a numeric value. Once we have it, we can implement this function.
+    table_oid.map(|_| "".to_string())
+}

src/common/function/src/system/pg_catalog/table_is_visible.rs

@@ -0,0 +1,72 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::fmt::{self};
+use std::sync::Arc;
+
+use common_query::error::Result;
+use common_query::prelude::{Signature, Volatility};
+use datatypes::prelude::{ConcreteDataType, DataType, VectorRef};
+use datatypes::types::LogicalPrimitiveType;
+use datatypes::with_match_primitive_type_id;
+use num_traits::AsPrimitive;
+
+use crate::function::{Function, FunctionContext};
+use crate::scalars::expression::{scalar_unary_op, EvalContext};
+
+#[derive(Clone, Debug, Default)]
+pub struct PGTableIsVisibleFunction;
+
+const NAME: &str = crate::pg_catalog_func_fullname!("pg_table_is_visible");
+
+impl fmt::Display for PGTableIsVisibleFunction {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, crate::pg_catalog_func_fullname!("PG_TABLE_IS_VISIBLE"))
+    }
+}
+
+impl Function for PGTableIsVisibleFunction {
+    fn name(&self) -> &str {
+        NAME
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::boolean_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        Signature::uniform(
+            1,
+            vec![ConcreteDataType::uint32_datatype()],
+            Volatility::Immutable,
+        )
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        with_match_primitive_type_id!(columns[0].data_type().logical_type_id(), |$T| {
+            let col = scalar_unary_op::<<$T as LogicalPrimitiveType>::Native, bool, _>(&columns[0], pg_table_is_visible, &mut EvalContext::default())?;
+            Ok(Arc::new(col))
+        }, {
+            unreachable!()
+        })
+    }
+}
+
+fn pg_table_is_visible<I>(table_oid: Option<I>, _ctx: &mut EvalContext) -> Option<bool>
+where
+    I: AsPrimitive<u32>,
+{
+    // There is no table visibility in greptime, so we always return true
+    table_oid.map(|_| true)
+}

src/common/function/src/system/procedure_state.rs

@@ -12,26 +12,19 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::fmt;
-
 use api::v1::meta::ProcedureStatus;
 use common_macro::admin_fn;
 use common_meta::rpc::procedure::ProcedureStateResponse;
-use common_query::error::Error::ThreadJoin;
 use common_query::error::{
     InvalidFuncArgsSnafu, MissingProcedureServiceHandlerSnafu, Result,
     UnsupportedInputDataTypeSnafu,
 };
 use common_query::prelude::{Signature, Volatility};
-use common_telemetry::error;
 use datatypes::prelude::*;
-use datatypes::vectors::VectorRef;
 use serde::Serialize;
 use session::context::QueryContextRef;
-use snafu::{ensure, Location, OptionExt};
+use snafu::ensure;
 
-use crate::ensure_greptime;
-use crate::function::{Function, FunctionContext};
 use crate::handlers::ProcedureServiceHandlerRef;
 
 #[derive(Serialize)]
@@ -103,6 +96,7 @@ mod tests {
     use datatypes::vectors::StringVector;
 
     use super::*;
+    use crate::function::{Function, FunctionContext};
 
     #[test]
     fn test_procedure_state_misc() {

src/common/function/src/table.rs

@@ -22,6 +22,7 @@ use flush_compact_region::{CompactRegionFunction, FlushRegionFunction};
 use flush_compact_table::{CompactTableFunction, FlushTableFunction};
 use migrate_region::MigrateRegionFunction;
 
+use crate::flush_flow::FlushFlowFunction;
 use crate::function_registry::FunctionRegistry;
 
 /// Table functions
@@ -35,5 +36,6 @@ impl TableFunction {
         registry.register(Arc::new(CompactRegionFunction));
         registry.register(Arc::new(FlushTableFunction));
         registry.register(Arc::new(CompactTableFunction));
+        registry.register(Arc::new(FlushFlowFunction));
     }
 }

src/common/function/src/table/flush_compact_region.rs

@@ -12,23 +12,16 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::fmt;
-
 use common_macro::admin_fn;
-use common_query::error::Error::ThreadJoin;
 use common_query::error::{
     InvalidFuncArgsSnafu, MissingTableMutationHandlerSnafu, Result, UnsupportedInputDataTypeSnafu,
 };
 use common_query::prelude::{Signature, Volatility};
-use common_telemetry::error;
 use datatypes::prelude::*;
-use datatypes::vectors::VectorRef;
 use session::context::QueryContextRef;
-use snafu::{ensure, Location, OptionExt};
+use snafu::ensure;
 use store_api::storage::RegionId;
 
-use crate::ensure_greptime;
-use crate::function::{Function, FunctionContext};
 use crate::handlers::TableMutationHandlerRef;
 use crate::helper::cast_u64;
 
@@ -84,6 +77,7 @@ mod tests {
     use datatypes::vectors::UInt64Vector;
 
     use super::*;
+    use crate::function::{Function, FunctionContext};
 
     macro_rules! define_region_function_test {
         ($name: ident, $func: ident) => {

src/common/function/src/table/flush_compact_table.rs

@@ -12,32 +12,29 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::fmt;
 use std::str::FromStr;
 
 use api::v1::region::{compact_request, StrictWindow};
 use common_error::ext::BoxedError;
 use common_macro::admin_fn;
-use common_query::error::Error::ThreadJoin;
 use common_query::error::{
     InvalidFuncArgsSnafu, MissingTableMutationHandlerSnafu, Result, TableMutationSnafu,
     UnsupportedInputDataTypeSnafu,
 };
 use common_query::prelude::{Signature, Volatility};
-use common_telemetry::{error, info};
+use common_telemetry::info;
 use datatypes::prelude::*;
-use datatypes::vectors::VectorRef;
 use session::context::QueryContextRef;
 use session::table_name::table_name_to_full_name;
-use snafu::{ensure, Location, OptionExt, ResultExt};
+use snafu::{ensure, ResultExt};
 use table::requests::{CompactTableRequest, FlushTableRequest};
 
-use crate::ensure_greptime;
-use crate::function::{Function, FunctionContext};
 use crate::handlers::TableMutationHandlerRef;
 
 /// Compact type: strict window.
 const COMPACT_TYPE_STRICT_WINDOW: &str = "strict_window";
+/// Compact type: strict window (short name).
+const COMPACT_TYPE_STRICT_WINDOW_SHORT: &str = "swcs";
 
 #[admin_fn(
     name = FlushTableFunction,
@@ -173,8 +170,12 @@ fn parse_compact_params(
     })
 }
 
+/// Parses compaction strategy type. For `strict_window` or `swcs` strict window compaction is chose,
+/// otherwise choose regular (TWCS) compaction.
 fn parse_compact_type(type_str: &str, option: Option<&str>) -> Result<compact_request::Options> {
-    if type_str.eq_ignore_ascii_case(COMPACT_TYPE_STRICT_WINDOW) {
+    if type_str.eq_ignore_ascii_case(COMPACT_TYPE_STRICT_WINDOW)
+        | type_str.eq_ignore_ascii_case(COMPACT_TYPE_STRICT_WINDOW_SHORT)
+    {
         let window_seconds = option
             .map(|v| {
                 i64::from_str(v).map_err(|_| {
@@ -209,6 +210,7 @@ mod tests {
     use session::context::QueryContext;
 
     use super::*;
+    use crate::function::{Function, FunctionContext};
 
     macro_rules! define_table_function_test {
         ($name: ident, $func: ident) => {
@@ -354,6 +356,17 @@ mod tests {
                     compact_options: Options::Regular(Default::default()),
                 },
             ),
+            (
+                &["table", "swcs", "120"],
+                CompactTableRequest {
+                    catalog_name: DEFAULT_CATALOG_NAME.to_string(),
+                    schema_name: DEFAULT_SCHEMA_NAME.to_string(),
+                    table_name: "table".to_string(),
+                    compact_options: Options::StrictWindow(StrictWindow {
+                        window_seconds: 120,
+                    }),
+                },
+            ),
         ]);
 
         assert!(parse_compact_params(

src/common/function/src/table/migrate_region.rs

@@ -12,24 +12,16 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::fmt::{self};
 use std::time::Duration;
 
 use common_macro::admin_fn;
 use common_meta::rpc::procedure::MigrateRegionRequest;
-use common_query::error::Error::ThreadJoin;
 use common_query::error::{InvalidFuncArgsSnafu, MissingProcedureServiceHandlerSnafu, Result};
 use common_query::prelude::{Signature, TypeSignature, Volatility};
-use common_telemetry::error;
-use datatypes::data_type::DataType;
 use datatypes::prelude::ConcreteDataType;
 use datatypes::value::{Value, ValueRef};
-use datatypes::vectors::VectorRef;
 use session::context::QueryContextRef;
-use snafu::{Location, OptionExt};
 
-use crate::ensure_greptime;
-use crate::function::{Function, FunctionContext};
 use crate::handlers::ProcedureServiceHandlerRef;
 use crate::helper::cast_u64;
 
@@ -128,9 +120,10 @@ mod tests {
     use std::sync::Arc;
 
     use common_query::prelude::TypeSignature;
-    use datatypes::vectors::{StringVector, UInt64Vector};
+    use datatypes::vectors::{StringVector, UInt64Vector, VectorRef};
 
     use super::*;
+    use crate::function::{Function, FunctionContext};
 
     #[test]
     fn test_migrate_region_misc() {

src/common/greptimedb-telemetry/src/lib.rs

@@ -72,7 +72,7 @@ impl GreptimeDBTelemetryTask {
         match self {
             GreptimeDBTelemetryTask::Enable((task, _)) => {
                 print_anonymous_usage_data_disclaimer();
-                task.start(common_runtime::bg_runtime())
+                task.start(common_runtime::global_runtime())
             }
             GreptimeDBTelemetryTask::Disable => Ok(()),
         }

src/common/grpc/src/channel_manager.rs

@@ -225,7 +225,7 @@ impl ChannelManager {
         }
 
         let pool = self.pool.clone();
-        let _handle = common_runtime::spawn_bg(async {
+        let _handle = common_runtime::spawn_global(async {
             recycle_channel_in_loop(pool, RECYCLE_CHANNEL_INTERVAL_SECS).await;
         });
         info!(

src/common/macro/src/admin_fn.rs

@@ -153,6 +153,7 @@ fn build_struct(
     let ret = Ident::new(&format!("{ret}_datatype"), ret.span());
     let uppcase_display_name = display_name.to_uppercase();
     // Get the handler name in function state by the argument ident
+    // TODO(discord9): consider simple depend injection if more handlers are needed
     let (handler, snafu_type) = match handler_type.to_string().as_str() {
         "ProcedureServiceHandlerRef" => (
             Ident::new("procedure_service_handler", handler_type.span()),
@@ -163,6 +164,11 @@ fn build_struct(
             Ident::new("table_mutation_handler", handler_type.span()),
             Ident::new("MissingTableMutationHandlerSnafu", handler_type.span()),
         ),
+
+        "FlowServiceHandlerRef" => (
+            Ident::new("flow_service_handler", handler_type.span()),
+            Ident::new("MissingFlowServiceHandlerSnafu", handler_type.span()),
+        ),
         handler => ok!(error!(
             handler_type.span(),
             format!("Unknown handler type: {handler}")
@@ -174,29 +180,29 @@ fn build_struct(
         #[derive(Debug)]
         #vis struct #name;
 
-        impl fmt::Display for #name {
-            fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        impl std::fmt::Display for #name {
+            fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
                 write!(f, #uppcase_display_name)
             }
         }
 
 
-        impl Function for #name {
+        impl crate::function::Function for #name {
             fn name(&self) -> &'static str {
                 #display_name
             }
 
-            fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
-                Ok(ConcreteDataType::#ret())
+            fn return_type(&self, _input_types: &[store_api::storage::ConcreteDataType]) -> common_query::error::Result<store_api::storage::ConcreteDataType> {
+                Ok(store_api::storage::ConcreteDataType::#ret())
             }
 
             fn signature(&self) -> Signature {
                 #sig_fn()
             }
 
-            fn eval(&self, func_ctx: FunctionContext, columns: &[VectorRef]) ->  Result<VectorRef> {
+            fn eval(&self, func_ctx: crate::function::FunctionContext, columns: &[datatypes::vectors::VectorRef]) ->  common_query::error::Result<datatypes::vectors::VectorRef> {
                 // Ensure under the `greptime` catalog for security
-                ensure_greptime!(func_ctx);
+                crate::ensure_greptime!(func_ctx);
 
                 let columns_num = columns.len();
                 let rows_num = if columns.is_empty() {
@@ -208,6 +214,9 @@ fn build_struct(
 
                 // TODO(dennis): DataFusion doesn't support async UDF currently
                 std::thread::spawn(move || {
+                    use snafu::OptionExt;
+                    use datatypes::data_type::DataType;
+
                     let query_ctx = &func_ctx.query_ctx;
                     let handler = func_ctx
                         .state
@@ -215,11 +224,11 @@ fn build_struct(
                         .as_ref()
                         .context(#snafu_type)?;
 
-                    let mut builder = ConcreteDataType::#ret()
+                    let mut builder = store_api::storage::ConcreteDataType::#ret()
                         .create_mutable_vector(rows_num);
 
                     if columns_num == 0 {
-                        let result = common_runtime::block_on_read(async move {
+                        let result = common_runtime::block_on_global(async move {
                             #fn_name(handler, query_ctx, &[]).await
                         })?;
 
@@ -230,7 +239,7 @@ fn build_struct(
                                 .map(|vector| vector.get_ref(i))
                                 .collect();
 
-                            let result = common_runtime::block_on_read(async move {
+                            let result = common_runtime::block_on_global(async move {
                                 #fn_name(handler, query_ctx, &args).await
                             })?;
 
@@ -242,9 +251,9 @@ fn build_struct(
                 })
                     .join()
                     .map_err(|e| {
-                        error!(e; "Join thread error");
-                        ThreadJoin {
-                            location: Location::default(),
+                        common_telemetry::error!(e; "Join thread error");
+                        common_query::error::Error::ThreadJoin {
+                            location: snafu::Location::default(),
                         }
                     })?
 

src/common/macro/src/lib.rs

@@ -73,7 +73,7 @@ pub fn range_fn(args: TokenStream, input: TokenStream) -> TokenStream {
 
 /// Attribute macro to convert a normal function to SQL administration function. The annotated function
 /// should accept:
-///    - `&ProcedureServiceHandlerRef` or `&TableMutationHandlerRef` as the first argument,
+///    - `&ProcedureServiceHandlerRef` or `&TableMutationHandlerRef` or `FlowServiceHandlerRef` as the first argument,
 ///    - `&QueryContextRef` as the second argument, and
 ///    - `&[ValueRef<'_>]` as the third argument which is SQL function input values in each row.
 /// Return type must be `common_query::error::Result<Value>`.
@@ -85,6 +85,8 @@ pub fn range_fn(args: TokenStream, input: TokenStream) -> TokenStream {
 /// - `ret`: The return type of the generated SQL function, it will be transformed into `ConcreteDataType::{ret}_datatype()` result.
 /// - `display_name`: The display name of the generated SQL function.
 /// - `sig_fn`: the function to returns `Signature` of generated `Function`.
+///
+/// Note that this macro should only be used in `common-function` crate for now
 #[proc_macro_attribute]
 pub fn admin_fn(args: TokenStream, input: TokenStream) -> TokenStream {
     process_admin_fn(args, input)

src/common/meta/Cargo.toml

@@ -6,6 +6,7 @@ license.workspace = true
 
 [features]
 testing = []
+pg_kvbackend = ["dep:tokio-postgres"]
 
 [lints]
 workspace = true
@@ -56,6 +57,7 @@ store-api.workspace = true
 strum.workspace = true
 table.workspace = true
 tokio.workspace = true
+tokio-postgres = { workspace = true, optional = true }
 tonic.workspace = true
 typetag = "0.2"
 

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

@@ -227,7 +227,7 @@ impl Procedure for DropTableProcedure {
     }
 
     fn rollback_supported(&self) -> bool {
-        !matches!(self.data.state, DropTableState::Prepare)
+        !matches!(self.data.state, DropTableState::Prepare) && self.data.allow_rollback
     }
 
     async fn rollback(&mut self, _: &ProcedureContext) -> ProcedureResult<()> {
@@ -256,6 +256,8 @@ pub struct DropTableData {
     pub task: DropTableTask,
     pub physical_region_routes: Vec<RegionRoute>,
     pub physical_table_id: Option<TableId>,
+    #[serde(default)]
+    pub allow_rollback: bool,
 }
 
 impl DropTableData {
@@ -266,6 +268,7 @@ impl DropTableData {
             task,
             physical_region_routes: vec![],
             physical_table_id: None,
+            allow_rollback: false,
         }
     }
 

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

@@ -12,8 +12,12 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use common_catalog::format_full_table_name;
+use snafu::OptionExt;
+use store_api::metric_engine_consts::METRIC_ENGINE_NAME;
+
 use crate::ddl::drop_table::DropTableProcedure;
-use crate::error::Result;
+use crate::error::{self, Result};
 
 impl DropTableProcedure {
     /// Fetches the table info and physical table route.
@@ -29,6 +33,23 @@ impl DropTableProcedure {
         self.data.physical_region_routes = physical_table_route_value.region_routes;
         self.data.physical_table_id = Some(physical_table_id);
 
+        if physical_table_id == self.data.table_id() {
+            let table_info_value = self
+                .context
+                .table_metadata_manager
+                .table_info_manager()
+                .get(task.table_id)
+                .await?
+                .with_context(|| error::TableInfoNotFoundSnafu {
+                    table: format_full_table_name(&task.catalog, &task.schema, &task.table),
+                })?
+                .into_inner();
+
+            let engine = table_info_value.table_info.meta.engine;
+            // rollback only if dropping the metric physical table fails
+            self.data.allow_rollback = engine.as_str() == METRIC_ENGINE_NAME
+        }
+
         Ok(())
     }
 }

src/common/meta/src/ddl/test_util.rs

@@ -23,6 +23,7 @@ use std::collections::HashMap;
 use api::v1::meta::Partition;
 use api::v1::{ColumnDataType, SemanticType};
 use common_procedure::Status;
+use store_api::metric_engine_consts::{LOGICAL_TABLE_METADATA_KEY, METRIC_ENGINE_NAME};
 use table::metadata::{RawTableInfo, TableId};
 
 use crate::ddl::create_logical_tables::CreateLogicalTablesProcedure;
@@ -130,6 +131,11 @@ pub fn test_create_logical_table_task(name: &str) -> CreateTableTask {
         .time_index("ts")
         .primary_keys(["host".into()])
         .table_name(name)
+        .engine(METRIC_ENGINE_NAME)
+        .table_options(HashMap::from([(
+            LOGICAL_TABLE_METADATA_KEY.to_string(),
+            "phy".to_string(),
+        )]))
         .build()
         .unwrap()
         .into();
@@ -166,6 +172,7 @@ pub fn test_create_physical_table_task(name: &str) -> CreateTableTask {
         .time_index("ts")
         .primary_keys(["value".into()])
         .table_name(name)
+        .engine(METRIC_ENGINE_NAME)
         .build()
         .unwrap()
         .into();

src/common/meta/src/ddl/test_util/create_table.rs

@@ -127,7 +127,7 @@ pub fn build_raw_table_info_from_expr(expr: &CreateTableExpr) -> RawTableInfo {
             engine: expr.engine.to_string(),
             next_column_id: expr.column_defs.len() as u32,
             region_numbers: vec![],
-            options: TableOptions::default(),
+            options: TableOptions::try_from_iter(&expr.table_options).unwrap(),
             created_on: DateTime::default(),
             partition_key_indices: vec![],
         },

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

@@ -91,6 +91,7 @@ async fn test_on_prepare_table() {
     // Drop if exists
     let mut procedure = DropTableProcedure::new(cluster_id, task, ddl_context.clone());
     procedure.on_prepare().await.unwrap();
+    assert!(!procedure.rollback_supported());
 
     let task = new_drop_table_task(table_name, table_id, false);
     // Drop table
@@ -224,9 +225,12 @@ async fn test_on_rollback() {
         let task = new_drop_table_task("phy_table", physical_table_id, false);
         let mut procedure = DropTableProcedure::new(cluster_id, task, ddl_context.clone());
         procedure.on_prepare().await.unwrap();
+        assert!(procedure.rollback_supported());
         procedure.on_delete_metadata().await.unwrap();
+        assert!(procedure.rollback_supported());
         procedure.rollback(&ctx).await.unwrap();
         // Rollback again
+        assert!(procedure.rollback_supported());
         procedure.rollback(&ctx).await.unwrap();
         let kvs = kv_backend.dump();
         assert_eq!(kvs, expected_kvs);
@@ -236,12 +240,7 @@ async fn test_on_rollback() {
     let task = new_drop_table_task("foo", table_ids[0], false);
     let mut procedure = DropTableProcedure::new(cluster_id, task, ddl_context.clone());
     procedure.on_prepare().await.unwrap();
-    procedure.on_delete_metadata().await.unwrap();
-    procedure.rollback(&ctx).await.unwrap();
-    // Rollback again
-    procedure.rollback(&ctx).await.unwrap();
-    let kvs = kv_backend.dump();
-    assert_eq!(kvs, expected_kvs);
+    assert!(!procedure.rollback_supported());
 }
 
 fn new_drop_table_task(table_name: &str, table_id: TableId, drop_if_exists: bool) -> DropTableTask {

src/common/meta/src/ddl/utils.rs

@@ -15,7 +15,7 @@
 use common_catalog::consts::METRIC_ENGINE;
 use common_error::ext::BoxedError;
 use common_procedure::error::Error as ProcedureError;
-use snafu::{ensure, location, Location, OptionExt};
+use snafu::{ensure, location, OptionExt};
 use store_api::metric_engine_consts::LOGICAL_TABLE_METADATA_KEY;
 use table::metadata::TableId;
 

src/common/meta/src/error.rs

@@ -499,6 +499,13 @@ pub enum Error {
         error: rskafka::client::error::Error,
     },
 
+    #[snafu(display("Failed to create TLS Config"))]
+    TlsConfig {
+        #[snafu(implicit)]
+        location: Location,
+        source: common_wal::error::Error,
+    },
+
     #[snafu(display("Failed to resolve Kafka broker endpoint."))]
     ResolveKafkaEndpoint { source: common_wal::error::Error },
 
@@ -636,6 +643,15 @@ pub enum Error {
         location: Location,
     },
 
+    #[snafu(display("Failed to parse {} from str to utf8", name))]
+    StrFromUtf8 {
+        name: String,
+        #[snafu(source)]
+        error: std::str::Utf8Error,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
     #[snafu(display("Value not exists"))]
     ValueNotExist {
         #[snafu(implicit)]
@@ -644,6 +660,24 @@ pub enum Error {
 
     #[snafu(display("Failed to get cache"))]
     GetCache { source: Arc<Error> },
+
+    #[cfg(feature = "pg_kvbackend")]
+    #[snafu(display("Failed to execute via Postgres"))]
+    PostgresExecution {
+        #[snafu(source)]
+        error: tokio_postgres::Error,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
+    #[cfg(feature = "pg_kvbackend")]
+    #[snafu(display("Failed to connect to Postgres"))]
+    ConnectPostgres {
+        #[snafu(source)]
+        error: tokio_postgres::Error,
+        #[snafu(implicit)]
+        location: Location,
+    },
 }
 
 pub type Result<T> = std::result::Result<T, Error>;
@@ -697,7 +731,8 @@ impl ErrorExt for Error {
             | UnexpectedLogicalRouteTable { .. }
             | ProcedureOutput { .. }
             | FromUtf8 { .. }
-            | MetadataCorruption { .. } => StatusCode::Unexpected,
+            | MetadataCorruption { .. }
+            | StrFromUtf8 { .. } => StatusCode::Unexpected,
 
             SendMessage { .. } | GetKvCache { .. } | CacheNotGet { .. } | RenameTable { .. } => {
                 StatusCode::Internal
@@ -714,7 +749,8 @@ impl ErrorExt for Error {
             | AlterLogicalTablesInvalidArguments { .. }
             | CreateLogicalTablesInvalidArguments { .. }
             | MismatchPrefix { .. }
-            | DelimiterNotFound { .. } => StatusCode::InvalidArguments,
+            | DelimiterNotFound { .. }
+            | TlsConfig { .. } => StatusCode::InvalidArguments,
 
             FlowNotFound { .. } => StatusCode::FlowNotFound,
             FlowRouteNotFound { .. } => StatusCode::Unexpected,
@@ -741,6 +777,11 @@ impl ErrorExt for Error {
             | ParseNum { .. }
             | InvalidRole { .. }
             | EmptyDdlTasks { .. } => StatusCode::InvalidArguments,
+
+            #[cfg(feature = "pg_kvbackend")]
+            PostgresExecution { .. } => StatusCode::Internal,
+            #[cfg(feature = "pg_kvbackend")]
+            ConnectPostgres { .. } => StatusCode::Internal,
         }
     }
 

src/common/meta/src/kv_backend.rs

@@ -31,6 +31,8 @@ use crate::rpc::KeyValue;
 pub mod chroot;
 pub mod etcd;
 pub mod memory;
+#[cfg(feature = "pg_kvbackend")]
+pub mod postgres;
 pub mod test;
 pub mod txn;
 

src/common/meta/src/kv_backend/postgres.rs

@@ -0,0 +1,626 @@
+// 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::any::Any;
+use std::borrow::Cow;
+use std::sync::Arc;
+
+use snafu::ResultExt;
+use tokio_postgres::types::ToSql;
+use tokio_postgres::{Client, NoTls};
+
+use super::{KvBackend, TxnService};
+use crate::error::{ConnectPostgresSnafu, Error, PostgresExecutionSnafu, Result, StrFromUtf8Snafu};
+use crate::kv_backend::txn::{Txn as KvTxn, TxnResponse as KvTxnResponse};
+use crate::kv_backend::KvBackendRef;
+use crate::rpc::store::{
+    BatchDeleteRequest, BatchDeleteResponse, BatchGetRequest, BatchGetResponse, BatchPutRequest,
+    BatchPutResponse, CompareAndPutRequest, CompareAndPutResponse, DeleteRangeRequest,
+    DeleteRangeResponse, PutRequest, PutResponse, RangeRequest, RangeResponse,
+};
+use crate::rpc::KeyValue;
+
+/// Posgres backend store for metasrv
+pub struct PgStore {
+    // TODO: Consider using sqlx crate.
+    client: Client,
+}
+
+const EMPTY: &[u8] = &[0];
+
+// TODO: allow users to configure metadata table name.
+const METADKV_CREATION: &str =
+    "CREATE TABLE IF NOT EXISTS greptime_metakv(k varchar PRIMARY KEY, v varchar)";
+
+const FULL_TABLE_SCAN: &str = "SELECT k, v FROM greptime_metakv $1 ORDER BY K";
+
+const POINT_GET: &str = "SELECT k, v FROM greptime_metakv WHERE k = $1";
+
+const PREFIX_SCAN: &str = "SELECT k, v FROM greptime_metakv WHERE k LIKE $1 ORDER BY K";
+
+const RANGE_SCAN_LEFT_BOUNDED: &str = "SELECT k, v FROM greptime_metakv WHERE k >= $1 ORDER BY K";
+
+const RANGE_SCAN_FULL_RANGE: &str =
+    "SELECT k, v FROM greptime_metakv WHERE k >= $1 AND K < $2 ORDER BY K";
+
+const FULL_TABLE_DELETE: &str = "DELETE FROM greptime_metakv RETURNING k,v";
+
+const POINT_DELETE: &str = "DELETE FROM greptime_metakv WHERE K = $1 RETURNING k,v;";
+
+const PREFIX_DELETE: &str = "DELETE FROM greptime_metakv WHERE k LIKE $1 RETURNING k,v;";
+
+const RANGE_DELETE_LEFT_BOUNDED: &str = "DELETE FROM greptime_metakv WHERE k >= $1 RETURNING k,v;";
+
+const RANGE_DELETE_FULL_RANGE: &str =
+    "DELETE FROM greptime_metakv WHERE k >= $1 AND K < $2 RETURNING k,v;";
+
+const CAS: &str = r#"
+WITH prev AS (
+    SELECT k,v FROM greptime_metakv WHERE k = $1 AND v = $2
+), update AS (
+UPDATE greptime_metakv
+SET k=$1,
+v=$2
+WHERE 
+    k=$1 AND v=$3
+)
+
+SELECT k, v FROM prev;
+"#;
+
+const PUT_IF_NOT_EXISTS: &str = r#"    
+WITH prev AS (
+    select k,v from greptime_metakv where k = $1
+), insert AS (
+    INSERT INTO greptime_metakv
+    VALUES ($1, $2)
+    ON CONFLICT (k) DO NOTHING
+)
+
+SELECT k, v FROM prev;"#;
+
+impl PgStore {
+    /// Create pgstore impl of KvBackendRef from url.
+    pub async fn with_url(url: &str) -> Result<KvBackendRef> {
+        // TODO: support tls.
+        let (client, conn) = tokio_postgres::connect(url, NoTls)
+            .await
+            .context(ConnectPostgresSnafu)?;
+        tokio::spawn(async move { conn.await.context(ConnectPostgresSnafu) });
+        Self::with_pg_client(client).await
+    }
+
+    /// Create pgstore impl of KvBackendRef from tokio-postgres client.
+    pub async fn with_pg_client(client: Client) -> Result<KvBackendRef> {
+        // This step ensures the postgres metadata backend is ready to use.
+        // We check if greptime_metakv table exists, and we will create a new table
+        // if it does not exist.
+        client
+            .execute(METADKV_CREATION, &[])
+            .await
+            .context(PostgresExecutionSnafu)?;
+        Ok(Arc::new(Self { client }))
+    }
+
+    async fn put_if_not_exists(&self, key: &str, value: &str) -> Result<bool> {
+        let res = self
+            .client
+            .query(PUT_IF_NOT_EXISTS, &[&key, &value])
+            .await
+            .context(PostgresExecutionSnafu)?;
+        Ok(res.is_empty())
+    }
+}
+
+fn select_range_template(req: &RangeRequest) -> &str {
+    if req.range_end.is_empty() {
+        return POINT_GET;
+    }
+    if req.key == EMPTY && req.range_end == EMPTY {
+        FULL_TABLE_SCAN
+    } else if req.range_end == EMPTY {
+        RANGE_SCAN_LEFT_BOUNDED
+    } else if is_prefix_range(&req.key, &req.range_end) {
+        PREFIX_SCAN
+    } else {
+        RANGE_SCAN_FULL_RANGE
+    }
+}
+
+fn select_range_delete_template(req: &DeleteRangeRequest) -> &str {
+    if req.range_end.is_empty() {
+        return POINT_DELETE;
+    }
+    if req.key == EMPTY && req.range_end == EMPTY {
+        FULL_TABLE_DELETE
+    } else if req.range_end == EMPTY {
+        RANGE_DELETE_LEFT_BOUNDED
+    } else if is_prefix_range(&req.key, &req.range_end) {
+        PREFIX_DELETE
+    } else {
+        RANGE_DELETE_FULL_RANGE
+    }
+}
+
+// Generate dynamic parameterized sql for batch get.
+fn generate_batch_get_query(key_len: usize) -> String {
+    let in_placeholders: Vec<String> = (1..=key_len).map(|i| format!("${}", i)).collect();
+    let in_clause = in_placeholders.join(", ");
+    format!(
+        "SELECT k, v FROM greptime_metakv WHERE k in ({});",
+        in_clause
+    )
+}
+
+// Generate dynamic parameterized sql for batch delete.
+fn generate_batch_delete_query(key_len: usize) -> String {
+    let in_placeholders: Vec<String> = (1..=key_len).map(|i| format!("${}", i)).collect();
+    let in_clause = in_placeholders.join(", ");
+    format!(
+        "DELETE FROM greptime_metakv WHERE k in ({}) RETURNING k, v;",
+        in_clause
+    )
+}
+
+// Generate dynamic parameterized sql for batch upsert.
+fn generate_batch_upsert_query(kv_len: usize) -> String {
+    let in_placeholders: Vec<String> = (1..=kv_len).map(|i| format!("${}", i)).collect();
+    let in_clause = in_placeholders.join(", ");
+    let mut param_index = kv_len + 1;
+    let mut values_placeholders = Vec::new();
+    for _ in 0..kv_len {
+        values_placeholders.push(format!("(${0}, ${1})", param_index, param_index + 1));
+        param_index += 2;
+    }
+    let values_clause = values_placeholders.join(", ");
+
+    format!(
+        r#"
+    WITH prev AS (
+        SELECT k,v FROM greptime_metakv WHERE k IN ({in_clause})
+    ), update AS (
+    INSERT INTO greptime_metakv (k, v) VALUES
+        {values_clause}
+    ON CONFLICT (
+        k
+    ) DO UPDATE SET
+        v = excluded.v
+    )
+
+    SELECT k, v FROM prev;
+    "#
+    )
+}
+
+//  Trim null byte at the end and convert bytes to string.
+fn process_bytes<'a>(data: &'a [u8], name: &str) -> Result<&'a str> {
+    let mut len = data.len();
+    // remove trailing null bytes to avoid error in postgres encoding.
+    while len > 0 && data[len - 1] == 0 {
+        len -= 1;
+    }
+    let res = std::str::from_utf8(&data[0..len]).context(StrFromUtf8Snafu { name })?;
+    Ok(res)
+}
+
+#[async_trait::async_trait]
+impl KvBackend for PgStore {
+    fn name(&self) -> &str {
+        "Postgres"
+    }
+
+    fn as_any(&self) -> &dyn Any {
+        self
+    }
+
+    async fn range(&self, req: RangeRequest) -> Result<RangeResponse> {
+        let mut params = vec![];
+        let template = select_range_template(&req);
+        if req.key != EMPTY {
+            let key = process_bytes(&req.key, "rangeKey")?;
+            if template == PREFIX_SCAN {
+                let prefix = format!("{key}%");
+                params.push(Cow::Owned(prefix))
+            } else {
+                params.push(Cow::Borrowed(key))
+            }
+        }
+        if template == RANGE_SCAN_FULL_RANGE && req.range_end != EMPTY {
+            let range_end = process_bytes(&req.range_end, "rangeEnd")?;
+            params.push(Cow::Borrowed(range_end));
+        }
+        let limit = req.limit as usize;
+        let limit_cause = match limit > 0 {
+            true => format!(" LIMIT {};", limit + 1),
+            false => ";".to_string(),
+        };
+        let template = format!("{}{}", template, limit_cause);
+        let params: Vec<&(dyn ToSql + Sync)> = params
+            .iter()
+            .map(|x| match x {
+                Cow::Borrowed(borrowed) => borrowed as &(dyn ToSql + Sync),
+                Cow::Owned(owned) => owned as &(dyn ToSql + Sync),
+            })
+            .collect();
+        let res = self
+            .client
+            .query(&template, &params)
+            .await
+            .context(PostgresExecutionSnafu)?;
+        let kvs: Vec<KeyValue> = res
+            .into_iter()
+            .map(|r| {
+                let key: String = r.get(0);
+                if req.keys_only {
+                    return KeyValue {
+                        key: key.into_bytes(),
+                        value: vec![],
+                    };
+                }
+                let value: String = r.get(1);
+                KeyValue {
+                    key: key.into_bytes(),
+                    value: value.into_bytes(),
+                }
+            })
+            .collect();
+        if limit == 0 || limit > kvs.len() {
+            return Ok(RangeResponse { kvs, more: false });
+        }
+        let (filtered_kvs, _) = kvs.split_at(limit);
+        Ok(RangeResponse {
+            kvs: filtered_kvs.to_vec(),
+            more: kvs.len() > limit,
+        })
+    }
+
+    async fn put(&self, req: PutRequest) -> Result<PutResponse> {
+        let kv = KeyValue {
+            key: req.key,
+            value: req.value,
+        };
+        let mut res = self
+            .batch_put(BatchPutRequest {
+                kvs: vec![kv],
+                prev_kv: req.prev_kv,
+            })
+            .await?;
+
+        if !res.prev_kvs.is_empty() {
+            return Ok(PutResponse {
+                prev_kv: Some(res.prev_kvs.remove(0)),
+            });
+        }
+        Ok(PutResponse { prev_kv: None })
+    }
+
+    async fn batch_put(&self, req: BatchPutRequest) -> Result<BatchPutResponse> {
+        let mut in_params = Vec::with_capacity(req.kvs.len());
+        let mut values_params = Vec::with_capacity(req.kvs.len() * 2);
+
+        for kv in &req.kvs {
+            let processed_key = process_bytes(&kv.key, "BatchPutRequestKey")?;
+            in_params.push(processed_key);
+
+            let processed_value = process_bytes(&kv.value, "BatchPutRequestValue")?;
+            values_params.push(processed_key);
+            values_params.push(processed_value);
+        }
+        in_params.extend(values_params);
+        let params: Vec<&(dyn ToSql + Sync)> =
+            in_params.iter().map(|x| x as &(dyn ToSql + Sync)).collect();
+
+        let query = generate_batch_upsert_query(req.kvs.len());
+        let res = self
+            .client
+            .query(&query, &params)
+            .await
+            .context(PostgresExecutionSnafu)?;
+        if req.prev_kv {
+            let kvs: Vec<KeyValue> = res
+                .into_iter()
+                .map(|r| {
+                    let key: String = r.get(0);
+                    let value: String = r.get(1);
+                    KeyValue {
+                        key: key.into_bytes(),
+                        value: value.into_bytes(),
+                    }
+                })
+                .collect();
+            if !kvs.is_empty() {
+                return Ok(BatchPutResponse { prev_kvs: kvs });
+            }
+        }
+        Ok(BatchPutResponse { prev_kvs: vec![] })
+    }
+
+    async fn batch_get(&self, req: BatchGetRequest) -> Result<BatchGetResponse> {
+        if req.keys.is_empty() {
+            return Ok(BatchGetResponse { kvs: vec![] });
+        }
+        let query = generate_batch_get_query(req.keys.len());
+        let value_params = req
+            .keys
+            .iter()
+            .map(|k| process_bytes(k, "BatchGetRequestKey"))
+            .collect::<Result<Vec<&str>>>()?;
+        let params: Vec<&(dyn ToSql + Sync)> = value_params
+            .iter()
+            .map(|x| x as &(dyn ToSql + Sync))
+            .collect();
+        let res = self
+            .client
+            .query(&query, &params)
+            .await
+            .context(PostgresExecutionSnafu)?;
+        let kvs: Vec<KeyValue> = res
+            .into_iter()
+            .map(|r| {
+                let key: String = r.get(0);
+                let value: String = r.get(1);
+                KeyValue {
+                    key: key.into_bytes(),
+                    value: value.into_bytes(),
+                }
+            })
+            .collect();
+        Ok(BatchGetResponse { kvs })
+    }
+
+    async fn delete_range(&self, req: DeleteRangeRequest) -> Result<DeleteRangeResponse> {
+        let mut params = vec![];
+        let template = select_range_delete_template(&req);
+        if req.key != EMPTY {
+            let key = process_bytes(&req.key, "deleteRangeKey")?;
+            if template == PREFIX_DELETE {
+                let prefix = format!("{key}%");
+                params.push(Cow::Owned(prefix));
+            } else {
+                params.push(Cow::Borrowed(key));
+            }
+        }
+        if template == RANGE_DELETE_FULL_RANGE && req.range_end != EMPTY {
+            let range_end = process_bytes(&req.range_end, "deleteRangeEnd")?;
+            params.push(Cow::Borrowed(range_end));
+        }
+        let params: Vec<&(dyn ToSql + Sync)> = params
+            .iter()
+            .map(|x| match x {
+                Cow::Borrowed(borrowed) => borrowed as &(dyn ToSql + Sync),
+                Cow::Owned(owned) => owned as &(dyn ToSql + Sync),
+            })
+            .collect();
+
+        let res = self
+            .client
+            .query(template, &params)
+            .await
+            .context(PostgresExecutionSnafu)?;
+        let deleted = res.len() as i64;
+        if !req.prev_kv {
+            return Ok({
+                DeleteRangeResponse {
+                    deleted,
+                    prev_kvs: vec![],
+                }
+            });
+        }
+        let kvs: Vec<KeyValue> = res
+            .into_iter()
+            .map(|r| {
+                let key: String = r.get(0);
+                let value: String = r.get(1);
+                KeyValue {
+                    key: key.into_bytes(),
+                    value: value.into_bytes(),
+                }
+            })
+            .collect();
+        Ok(DeleteRangeResponse {
+            deleted,
+            prev_kvs: kvs,
+        })
+    }
+
+    async fn batch_delete(&self, req: BatchDeleteRequest) -> Result<BatchDeleteResponse> {
+        if req.keys.is_empty() {
+            return Ok(BatchDeleteResponse { prev_kvs: vec![] });
+        }
+        let query = generate_batch_delete_query(req.keys.len());
+        let value_params = req
+            .keys
+            .iter()
+            .map(|k| process_bytes(k, "BatchDeleteRequestKey"))
+            .collect::<Result<Vec<&str>>>()?;
+        let params: Vec<&(dyn ToSql + Sync)> = value_params
+            .iter()
+            .map(|x| x as &(dyn ToSql + Sync))
+            .collect();
+        let res = self
+            .client
+            .query(&query, &params)
+            .await
+            .context(PostgresExecutionSnafu)?;
+        if !req.prev_kv {
+            return Ok(BatchDeleteResponse { prev_kvs: vec![] });
+        }
+        let kvs: Vec<KeyValue> = res
+            .into_iter()
+            .map(|r| {
+                let key: String = r.get(0);
+                let value: String = r.get(1);
+                KeyValue {
+                    key: key.into_bytes(),
+                    value: value.into_bytes(),
+                }
+            })
+            .collect();
+        Ok(BatchDeleteResponse { prev_kvs: kvs })
+    }
+
+    async fn compare_and_put(&self, req: CompareAndPutRequest) -> Result<CompareAndPutResponse> {
+        let key = process_bytes(&req.key, "CASKey")?;
+        let value = process_bytes(&req.value, "CASValue")?;
+        if req.expect.is_empty() {
+            let put_res = self.put_if_not_exists(key, value).await?;
+            return Ok(CompareAndPutResponse {
+                success: put_res,
+                prev_kv: None,
+            });
+        }
+        let expect = process_bytes(&req.expect, "CASExpect")?;
+
+        let res = self
+            .client
+            .query(CAS, &[&key, &value, &expect])
+            .await
+            .context(PostgresExecutionSnafu)?;
+        match res.is_empty() {
+            true => Ok(CompareAndPutResponse {
+                success: false,
+                prev_kv: None,
+            }),
+            false => {
+                let mut kvs: Vec<KeyValue> = res
+                    .into_iter()
+                    .map(|r| {
+                        let key: String = r.get(0);
+                        let value: String = r.get(1);
+                        KeyValue {
+                            key: key.into_bytes(),
+                            value: value.into_bytes(),
+                        }
+                    })
+                    .collect();
+                Ok(CompareAndPutResponse {
+                    success: true,
+                    prev_kv: Some(kvs.remove(0)),
+                })
+            }
+        }
+    }
+}
+
+#[async_trait::async_trait]
+impl TxnService for PgStore {
+    type Error = Error;
+
+    async fn txn(&self, _txn: KvTxn) -> Result<KvTxnResponse> {
+        // TODO: implement txn for pg kv backend.
+        unimplemented!()
+    }
+
+    fn max_txn_ops(&self) -> usize {
+        unreachable!("postgres backend does not support max_txn_ops!")
+    }
+}
+
+fn is_prefix_range(start: &[u8], end: &[u8]) -> bool {
+    if start.len() != end.len() {
+        return false;
+    }
+    let l = start.len();
+    let same_prefix = start[0..l - 1] == end[0..l - 1];
+    if let (Some(rhs), Some(lhs)) = (start.last(), end.last()) {
+        return same_prefix && (*rhs + 1) == *lhs;
+    }
+    false
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::kv_backend::test::{
+        prepare_kv_with_prefix, test_kv_batch_delete_with_prefix, test_kv_batch_get_with_prefix,
+        test_kv_compare_and_put_with_prefix, test_kv_delete_range_with_prefix,
+        test_kv_put_with_prefix, test_kv_range_2_with_prefix, test_kv_range_with_prefix,
+        unprepare_kv,
+    };
+
+    async fn build_pg_kv_backend() -> Option<PgStore> {
+        let endpoints = std::env::var("GT_POSTGRES_ENDPOINTS").unwrap_or_default();
+        if endpoints.is_empty() {
+            return None;
+        }
+
+        let (client, connection) = tokio_postgres::connect(&endpoints, NoTls).await.unwrap();
+        tokio::spawn(connection);
+        let _ = client.execute(METADKV_CREATION, &[]).await;
+        Some(PgStore { client })
+    }
+
+    #[tokio::test]
+    async fn test_put() {
+        if let Some(kv_backend) = build_pg_kv_backend().await {
+            let prefix = b"put/";
+            prepare_kv_with_prefix(&kv_backend, prefix.to_vec()).await;
+            test_kv_put_with_prefix(&kv_backend, prefix.to_vec()).await;
+            unprepare_kv(&kv_backend, prefix).await;
+        }
+    }
+
+    #[tokio::test]
+    async fn test_range() {
+        if let Some(kv_backend) = build_pg_kv_backend().await {
+            let prefix = b"range/";
+            prepare_kv_with_prefix(&kv_backend, prefix.to_vec()).await;
+            test_kv_range_with_prefix(&kv_backend, prefix.to_vec()).await;
+            unprepare_kv(&kv_backend, prefix).await;
+        }
+    }
+
+    #[tokio::test]
+    async fn test_range_2() {
+        if let Some(kv_backend) = build_pg_kv_backend().await {
+            test_kv_range_2_with_prefix(kv_backend, b"range2/".to_vec()).await;
+        }
+    }
+
+    #[tokio::test]
+    async fn test_batch_get() {
+        if let Some(kv_backend) = build_pg_kv_backend().await {
+            let prefix = b"batchGet/";
+            prepare_kv_with_prefix(&kv_backend, prefix.to_vec()).await;
+            test_kv_batch_get_with_prefix(&kv_backend, prefix.to_vec()).await;
+            unprepare_kv(&kv_backend, prefix).await;
+        }
+    }
+
+    #[tokio::test(flavor = "multi_thread")]
+    async fn test_compare_and_put() {
+        if let Some(kv_backend) = build_pg_kv_backend().await {
+            let kv_backend = Arc::new(kv_backend);
+            test_kv_compare_and_put_with_prefix(kv_backend, b"compareAndPut/".to_vec()).await;
+        }
+    }
+
+    #[tokio::test]
+    async fn test_delete_range() {
+        if let Some(kv_backend) = build_pg_kv_backend().await {
+            let prefix = b"deleteRange/";
+            prepare_kv_with_prefix(&kv_backend, prefix.to_vec()).await;
+            test_kv_delete_range_with_prefix(kv_backend, prefix.to_vec()).await;
+        }
+    }
+
+    #[tokio::test]
+    async fn test_batch_delete() {
+        if let Some(kv_backend) = build_pg_kv_backend().await {
+            let prefix = b"batchDelete/";
+            prepare_kv_with_prefix(&kv_backend, prefix.to_vec()).await;
+            test_kv_batch_delete_with_prefix(kv_backend, prefix.to_vec()).await;
+        }
+    }
+}

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

@@ -1079,6 +1079,7 @@ pub struct QueryContext {
     current_schema: String,
     timezone: String,
     extensions: HashMap<String, String>,
+    channel: u8,
 }
 
 impl From<QueryContextRef> for QueryContext {
@@ -1088,6 +1089,7 @@ impl From<QueryContextRef> for QueryContext {
             current_schema: query_context.current_schema().to_string(),
             timezone: query_context.timezone().to_string(),
             extensions: query_context.extensions(),
+            channel: query_context.channel() as u8,
         }
     }
 }
@@ -1099,6 +1101,7 @@ impl From<QueryContext> for PbQueryContext {
             current_schema,
             timezone,
             extensions,
+            channel,
         }: QueryContext,
     ) -> Self {
         PbQueryContext {
@@ -1106,6 +1109,7 @@ impl From<QueryContext> for PbQueryContext {
             current_schema,
             timezone,
             extensions,
+            channel: channel as u32,
         }
     }
 }