fix: last non null iter not init (#4687 )

chore: bump version v0.9.3 (#4684 )
chore(dockerfile): remove mysql and postgresql clients in greptimedb image (#4685 )
2025-12-25 07:30:02 +00:00 · 2024-09-06 04:13:23 +00:00 · 2024-09-06 02:31:41 +00:00 · 2024-09-05 16:00:53 +00:00 · 2024-09-05 06:15:38 +00:00 · 2024-09-05 06:13:22 +00:00
175 changed files with 4985 additions and 2451 deletions
--- a/.github/actions/build-macos-artifacts/action.yml
+++ b/.github/actions/build-macos-artifacts/action.yml
@@ -42,7 +42,7 @@ runs:
    - name: Install rust toolchain
      uses: actions-rust-lang/setup-rust-toolchain@v1
      with:
-        targets: ${{ inputs.arch }}
+        target: ${{ inputs.arch }}

    - name: Start etcd # For integration tests.
      if: ${{ inputs.disable-run-tests == 'false' }}
--- a/.github/actions/build-windows-artifacts/action.yml
+++ b/.github/actions/build-windows-artifacts/action.yml
@@ -27,7 +27,7 @@ runs:
    - name: Install rust toolchain
      uses: actions-rust-lang/setup-rust-toolchain@v1
      with:
-        targets: ${{ inputs.arch }}
+        target: ${{ inputs.arch }}
        components: llvm-tools-preview

    - name: Rust Cache
--- a/.github/actions/setup-etcd-cluster/action.yml
+++ b/.github/actions/setup-etcd-cluster/action.yml
@@ -18,6 +18,8 @@ runs:
        --set replicaCount=${{ inputs.etcd-replicas }} \
        --set resources.requests.cpu=50m \
        --set resources.requests.memory=128Mi \
+        --set resources.limits.cpu=1000m \
+        --set resources.limits.memory=2Gi \
        --set auth.rbac.create=false \
        --set auth.rbac.token.enabled=false \
        --set persistence.size=2Gi \
--- a/.github/actions/start-runner/action.yml
+++ b/.github/actions/start-runner/action.yml
@@ -38,7 +38,7 @@ runs:
  steps:
    - name: Configure AWS credentials
      if: startsWith(inputs.runner, 'ec2')
-      uses: aws-actions/configure-aws-credentials@v2
+      uses: aws-actions/configure-aws-credentials@v4
      with:
        aws-access-key-id: ${{ inputs.aws-access-key-id }}
        aws-secret-access-key: ${{ inputs.aws-secret-access-key }}
--- a/.github/actions/stop-runner/action.yml
+++ b/.github/actions/stop-runner/action.yml
@@ -25,7 +25,7 @@ runs:
  steps:
    - name: Configure AWS credentials
      if: ${{ inputs.label && inputs.ec2-instance-id }}
-      uses: aws-actions/configure-aws-credentials@v2
+      uses: aws-actions/configure-aws-credentials@v4
      with:
        aws-access-key-id: ${{ inputs.aws-access-key-id }}
        aws-secret-access-key: ${{ inputs.aws-secret-access-key }}
--- a/.github/workflows/develop.yml
+++ b/.github/workflows/develop.yml
@@ -616,8 +616,8 @@ jobs:
        with:
          # Shares across multiple jobs
          shared-key: "check-rust-fmt"
-      - name: Run cargo fmt
-        run: cargo fmt --all -- --check
+      - name: Check format
+        run: make fmt-check

  clippy:
    name: Clippy
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -33,6 +33,7 @@ on:
        description: The runner uses to build linux-arm64 artifacts
        default: ec2-c6g.4xlarge-arm64
        options:
+          - ubuntu-2204-32-cores-arm
          - ec2-c6g.xlarge-arm64 # 4C8G
          - ec2-c6g.2xlarge-arm64 # 8C16G
          - ec2-c6g.4xlarge-arm64 # 16C32G
@@ -98,16 +99,6 @@ permissions:
  contents: write # Allows the action to create a release.

 jobs:
-  check-builder-rust-version:
-    name: Check rust version in builder
-    runs-on: ubuntu-20.04
-    steps:
-      - uses: actions/checkout@v4
-      - name: Check Rust toolchain version
-        shell: bash
-        run: |
-          ./scripts/check-builder-rust-version.sh
-
  allocate-runners:
    name: Allocate runners
    if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
@@ -132,6 +123,11 @@ jobs:
        with:
          fetch-depth: 0

+      - name: Check Rust toolchain version
+        shell: bash
+        run: |
+          ./scripts/check-builder-rust-version.sh
+
      # The create-version will create a global variable named 'version' in the global workflows.
      # - If it's a tag push release, the version is the tag name(${{ github.ref_name }});
      # - If it's a scheduled release, the version is '${{ env.NEXT_RELEASE_VERSION }}-nightly-$buildTime', like v0.2.0-nigthly-20230313;
@@ -256,7 +252,8 @@ jobs:
          cargo-profile: ${{ env.CARGO_PROFILE }}
          features: ${{ matrix.features }}
          version: ${{ needs.allocate-runners.outputs.version }}
-          disable-run-tests: ${{ env.DISABLE_RUN_TESTS }}
+          # We decide to disable the integration tests on macOS because it's unnecessary and time-consuming.
+          disable-run-tests: true
          artifacts-dir: ${{ matrix.artifacts-dir-prefix }}-${{ needs.allocate-runners.outputs.version }}

      - name: Set build macos result
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -214,7 +214,7 @@ checksum = "d301b3b94cb4b2f23d7917810addbbaff90738e0ca2be692bd027e70d7e0330c"

 [[package]]
 name = "api"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "common-base",
 "common-decimal",
@@ -762,7 +762,7 @@ dependencies = [

 [[package]]
 name = "auth"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "async-trait",
@@ -1286,7 +1286,7 @@ dependencies = [

 [[package]]
 name = "cache"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "catalog",
 "common-error",
@@ -1294,7 +1294,7 @@ dependencies = [
 "common-meta",
 "moka",
 "snafu 0.8.4",
- "substrait 0.9.2",
+ "substrait 0.9.3",
 ]

 [[package]]
@@ -1321,7 +1321,7 @@ checksum = "37b2a672a2cb129a2e41c10b1224bb368f9f37a2b16b612598138befd7b37eb5"

 [[package]]
 name = "catalog"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arrow",
@@ -1647,7 +1647,7 @@ checksum = "4b82cf0babdbd58558212896d1a4272303a57bdb245c2bf1147185fb45640e70"

 [[package]]
 name = "client"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arc-swap",
@@ -1677,7 +1677,7 @@ dependencies = [
 "serde_json",
 "snafu 0.8.4",
 "substrait 0.37.3",
- "substrait 0.9.2",
+ "substrait 0.9.3",
 "tokio",
 "tokio-stream",
 "tonic 0.11.0",
@@ -1707,7 +1707,7 @@ dependencies = [

 [[package]]
 name = "cmd"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-trait",
 "auth",
@@ -1763,7 +1763,7 @@ dependencies = [
 "session",
 "snafu 0.8.4",
 "store-api",
- "substrait 0.9.2",
+ "substrait 0.9.3",
 "table",
 "temp-env",
 "tempfile",
@@ -1809,7 +1809,7 @@ checksum = "55b672471b4e9f9e95499ea597ff64941a309b2cdbffcc46f2cc5e2d971fd335"

 [[package]]
 name = "common-base"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "anymap",
 "bitvec",
@@ -1825,7 +1825,7 @@ dependencies = [

 [[package]]
 name = "common-catalog"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "chrono",
 "common-error",
@@ -1836,7 +1836,7 @@ dependencies = [

 [[package]]
 name = "common-config"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "common-base",
 "common-error",
@@ -1859,7 +1859,7 @@ dependencies = [

 [[package]]
 name = "common-datasource"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "arrow",
 "arrow-schema",
@@ -1896,7 +1896,7 @@ dependencies = [

 [[package]]
 name = "common-decimal"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "bigdecimal",
 "common-error",
@@ -1909,7 +1909,7 @@ dependencies = [

 [[package]]
 name = "common-error"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "snafu 0.8.4",
 "strum 0.25.0",
@@ -1918,7 +1918,7 @@ dependencies = [

 [[package]]
 name = "common-frontend"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "async-trait",
@@ -1933,7 +1933,7 @@ dependencies = [

 [[package]]
 name = "common-function"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arc-swap",
@@ -1970,7 +1970,7 @@ dependencies = [

 [[package]]
 name = "common-greptimedb-telemetry"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-trait",
 "common-runtime",
@@ -1987,7 +1987,7 @@ dependencies = [

 [[package]]
 name = "common-grpc"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arrow-flight",
@@ -2013,7 +2013,7 @@ dependencies = [

 [[package]]
 name = "common-grpc-expr"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "common-base",
@@ -2031,7 +2031,7 @@ dependencies = [

 [[package]]
 name = "common-macro"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "arc-swap",
 "common-query",
@@ -2045,7 +2045,7 @@ dependencies = [

 [[package]]
 name = "common-mem-prof"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "common-error",
 "common-macro",
@@ -2058,7 +2058,7 @@ dependencies = [

 [[package]]
 name = "common-meta"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "anymap2",
 "api",
@@ -2114,11 +2114,11 @@ dependencies = [

 [[package]]
 name = "common-plugins"
-version = "0.9.2"
+version = "0.9.3"

 [[package]]
 name = "common-procedure"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-stream",
 "async-trait",
@@ -2144,7 +2144,7 @@ dependencies = [

 [[package]]
 name = "common-procedure-test"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-trait",
 "common-procedure",
@@ -2152,7 +2152,7 @@ dependencies = [

 [[package]]
 name = "common-query"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "async-trait",
@@ -2178,7 +2178,7 @@ dependencies = [

 [[package]]
 name = "common-recordbatch"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "arc-swap",
 "common-error",
@@ -2197,7 +2197,7 @@ dependencies = [

 [[package]]
 name = "common-runtime"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-trait",
 "common-error",
@@ -2219,7 +2219,7 @@ dependencies = [

 [[package]]
 name = "common-telemetry"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "atty",
 "backtrace",
@@ -2246,7 +2246,7 @@ dependencies = [

 [[package]]
 name = "common-test-util"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "client",
 "common-query",
@@ -2258,7 +2258,7 @@ dependencies = [

 [[package]]
 name = "common-time"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "arrow",
 "chrono",
@@ -2274,7 +2274,7 @@ dependencies = [

 [[package]]
 name = "common-version"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "build-data",
 "const_format",
@@ -2285,7 +2285,7 @@ dependencies = [

 [[package]]
 name = "common-wal"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "common-base",
 "common-error",
@@ -3093,7 +3093,7 @@ dependencies = [

 [[package]]
 name = "datanode"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arrow-flight",
@@ -3142,7 +3142,7 @@ dependencies = [
 "session",
 "snafu 0.8.4",
 "store-api",
- "substrait 0.9.2",
+ "substrait 0.9.3",
 "table",
 "tokio",
 "toml 0.8.14",
@@ -3151,7 +3151,7 @@ dependencies = [

 [[package]]
 name = "datatypes"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -3721,7 +3721,7 @@ dependencies = [

 [[package]]
 name = "file-engine"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "async-trait",
@@ -3823,7 +3823,7 @@ checksum = "28a80e3145d8ad11ba0995949bbcf48b9df2be62772b3d351ef017dff6ecb853"

 [[package]]
 name = "flow"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arrow",
@@ -3880,7 +3880,7 @@ dependencies = [
 "snafu 0.8.4",
 "store-api",
 "strum 0.25.0",
- "substrait 0.9.2",
+ "substrait 0.9.3",
 "table",
 "tokio",
 "tonic 0.11.0",
@@ -3927,7 +3927,7 @@ checksum = "6c2141d6d6c8512188a7891b4b01590a45f6dac67afb4f255c4124dbb86d4eaa"

 [[package]]
 name = "frontend"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arc-swap",
@@ -5078,7 +5078,7 @@ dependencies = [

 [[package]]
 name = "index"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-trait",
 "asynchronous-codec",
@@ -5858,7 +5858,7 @@ checksum = "90ed8c1e510134f979dbc4f070f87d4313098b704861a105fe34231c70a3901c"

 [[package]]
 name = "log-store"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-stream",
 "async-trait",
@@ -6170,7 +6170,7 @@ dependencies = [

 [[package]]
 name = "meta-client"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "async-trait",
@@ -6196,7 +6196,7 @@ dependencies = [

 [[package]]
 name = "meta-srv"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "async-trait",
@@ -6274,7 +6274,7 @@ dependencies = [

 [[package]]
 name = "metric-engine"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "aquamarine",
@@ -6365,7 +6365,7 @@ dependencies = [

 [[package]]
 name = "mito2"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "aquamarine",
@@ -7012,7 +7012,7 @@ dependencies = [

 [[package]]
 name = "object-store"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "anyhow",
 "bytes",
@@ -7259,7 +7259,7 @@ dependencies = [

 [[package]]
 name = "operator"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "async-trait",
@@ -7304,7 +7304,7 @@ dependencies = [
 "sql",
 "sqlparser 0.45.0 (git+https://github.com/GreptimeTeam/sqlparser-rs.git?rev=54a267ac89c09b11c0c88934690530807185d3e7)",
 "store-api",
- "substrait 0.9.2",
+ "substrait 0.9.3",
 "table",
 "tokio",
 "tokio-util",
@@ -7554,7 +7554,7 @@ dependencies = [

 [[package]]
 name = "partition"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "async-trait",
@@ -7843,7 +7843,7 @@ checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"

 [[package]]
 name = "pipeline"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "ahash 0.8.11",
 "api",
@@ -8004,7 +8004,7 @@ dependencies = [

 [[package]]
 name = "plugins"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "auth",
 "common-base",
@@ -8273,7 +8273,7 @@ dependencies = [

 [[package]]
 name = "promql"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "ahash 0.8.11",
 "async-trait",
@@ -8508,7 +8508,7 @@ dependencies = [

 [[package]]
 name = "puffin"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-compression 0.4.11",
 "async-trait",
@@ -8630,7 +8630,7 @@ dependencies = [

 [[package]]
 name = "query"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "ahash 0.8.11",
 "api",
@@ -8693,7 +8693,7 @@ dependencies = [
 "stats-cli",
 "store-api",
 "streaming-stats",
- "substrait 0.9.2",
+ "substrait 0.9.3",
 "table",
 "tokio",
 "tokio-stream",
@@ -10055,7 +10055,7 @@ checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"

 [[package]]
 name = "script"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arc-swap",
@@ -10349,7 +10349,7 @@ dependencies = [

 [[package]]
 name = "servers"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "aide",
 "api",
@@ -10455,7 +10455,7 @@ dependencies = [

 [[package]]
 name = "session"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arc-swap",
@@ -10756,7 +10756,7 @@ dependencies = [

 [[package]]
 name = "sql"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "chrono",
@@ -10816,7 +10816,7 @@ dependencies = [

 [[package]]
 name = "sqlness-runner"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-trait",
 "clap 4.5.7",
@@ -11033,7 +11033,7 @@ dependencies = [

 [[package]]
 name = "store-api"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "aquamarine",
@@ -11202,7 +11202,7 @@ dependencies = [

 [[package]]
 name = "substrait"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "async-trait",
 "bytes",
@@ -11403,7 +11403,7 @@ dependencies = [

 [[package]]
 name = "table"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "async-trait",
@@ -11668,7 +11668,7 @@ checksum = "3369f5ac52d5eb6ab48c6b4ffdc8efbcad6b89c765749064ba298f2c68a16a76"

 [[package]]
 name = "tests-fuzz"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "arbitrary",
 "async-trait",
@@ -11710,7 +11710,7 @@ dependencies = [

 [[package]]
 name = "tests-integration"
-version = "0.9.2"
+version = "0.9.3"
 dependencies = [
 "api",
 "arrow-flight",
@@ -11770,7 +11770,7 @@ dependencies = [
 "sql",
 "sqlx",
 "store-api",
- "substrait 0.9.2",
+ "substrait 0.9.3",
 "table",
 "tempfile",
 "time",
@@ -12489,6 +12489,16 @@ dependencies = [
 "web-time 0.2.4",
 ]

+[[package]]
+name = "tracing-serde"
+version = "0.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bc6b213177105856957181934e4920de57730fc69bf42c37ee5bb664d406d9e1"
+dependencies = [
+ "serde",
+ "tracing-core",
+]
+
 [[package]]
 name = "tracing-subscriber"
 version = "0.3.18"
@@ -12499,12 +12509,15 @@ dependencies = [
 "nu-ansi-term",
 "once_cell",
 "regex",
+ "serde",
+ "serde_json",
 "sharded-slab",
 "smallvec",
 "thread_local",
 "tracing",
 "tracing-core",
 "tracing-log 0.2.0",
+ "tracing-serde",
 ]

 [[package]]
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -64,7 +64,7 @@ members = [
 resolver = "2"

 [workspace.package]
-version = "0.9.2"
+version = "0.9.3"
 edition = "2021"
 license = "Apache-2.0"

--- a/1
+++ b/1
@@ -191,6 +191,7 @@ fix-clippy: ## Fix clippy violations.
 .PHONY: fmt-check
 fmt-check: ## Check code format.
 	cargo fmt --all -- --check
+	python3 scripts/check-snafu.py

 .PHONY: start-etcd
 start-etcd: ## Start single node etcd for testing purpose.
--- a/config/config.md
+++ b/config/config.md
@@ -15,6 +15,8 @@
 | `mode` | String | `standalone` | The running mode of the datanode. It can be `standalone` or `distributed`. |
 | `enable_telemetry` | Bool | `true` | Enable telemetry to collect anonymous usage data. |
 | `default_timezone` | String | `None` | The default timezone of the server. |
+| `init_regions_in_background` | Bool | `false` | Initialize all regions in the background during the startup.<br/>By default, it provides services after all regions have been initialized. |
+| `init_regions_parallelism` | Integer | `16` | Parallelism of initializing regions. |
 | `runtime` | -- | -- | The runtime options. |
 | `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. |
@@ -127,6 +129,7 @@
 | `region_engine.mito.scan_parallelism` | Integer | `0` | Parallelism to scan a region (default: 1/4 of cpu cores).<br/>- `0`: using the default value (1/4 of cpu cores).<br/>- `1`: scan in current thread.<br/>- `n`: scan in parallelism n. |
 | `region_engine.mito.parallel_scan_channel_size` | Integer | `32` | Capacity of the channel to send data from parallel scan tasks to the main task. |
 | `region_engine.mito.allow_stale_entries` | Bool | `false` | Whether to allow stale WAL entries read during replay. |
+| `region_engine.mito.min_compaction_interval` | String | `0m` | Minimum time interval between two compactions.<br/>To align with the old behavior, the default value is 0 (no restrictions). |
 | `region_engine.mito.index` | -- | -- | The options for index in Mito engine. |
 | `region_engine.mito.index.aux_path` | String | `""` | Auxiliary directory path for the index in filesystem, used to store intermediate files for<br/>creating the index and staging files for searching the index, defaults to `{data_home}/index_intermediate`.<br/>The default name for this directory is `index_intermediate` for backward compatibility.<br/><br/>This path contains two subdirectories:<br/>- `__intm`: for storing intermediate files used during creating index.<br/>- `staging`: for storing staging files used during searching index. |
 | `region_engine.mito.index.staging_size` | String | `2GB` | The max capacity of the staging directory. |
@@ -150,11 +153,12 @@
 | `region_engine.mito.memtable.fork_dictionary_bytes` | String | `1GiB` | Max dictionary bytes.<br/>Only available for `partition_tree` memtable. |
 | `region_engine.file` | -- | -- | Enable the file engine. |
 | `logging` | -- | -- | The logging options. |
-| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
+| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. If set to empty, logs will not be written to 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 | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
+| `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
 | `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` | -- |
 | `export_metrics` | -- | -- | The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
@@ -235,11 +239,12 @@
 | `datanode.client.connect_timeout` | String | `10s` | -- |
 | `datanode.client.tcp_nodelay` | Bool | `true` | -- |
 | `logging` | -- | -- | The logging options. |
-| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
+| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. If set to empty, logs will not be written to 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 | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
+| `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
 | `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` | -- |
 | `export_metrics` | -- | -- | The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
@@ -299,11 +304,12 @@
 | `wal.backoff_base` | Integer | `2` | Exponential backoff rate, i.e. next backoff = base * current backoff. |
 | `wal.backoff_deadline` | String | `5mins` | Stop reconnecting if the total wait time reaches the deadline. If this config is missing, the reconnecting won't terminate. |
 | `logging` | -- | -- | The logging options. |
-| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
+| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. If set to empty, logs will not be written to 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 | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
+| `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
 | `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` | -- |
 | `export_metrics` | -- | -- | The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
@@ -426,6 +432,7 @@
 | `region_engine.mito.scan_parallelism` | Integer | `0` | Parallelism to scan a region (default: 1/4 of cpu cores).<br/>- `0`: using the default value (1/4 of cpu cores).<br/>- `1`: scan in current thread.<br/>- `n`: scan in parallelism n. |
 | `region_engine.mito.parallel_scan_channel_size` | Integer | `32` | Capacity of the channel to send data from parallel scan tasks to the main task. |
 | `region_engine.mito.allow_stale_entries` | Bool | `false` | Whether to allow stale WAL entries read during replay. |
+| `region_engine.mito.min_compaction_interval` | String | `0m` | Minimum time interval between two compactions.<br/>To align with the old behavior, the default value is 0 (no restrictions). |
 | `region_engine.mito.index` | -- | -- | The options for index in Mito engine. |
 | `region_engine.mito.index.aux_path` | String | `""` | Auxiliary directory path for the index in filesystem, used to store intermediate files for<br/>creating the index and staging files for searching the index, defaults to `{data_home}/index_intermediate`.<br/>The default name for this directory is `index_intermediate` for backward compatibility.<br/><br/>This path contains two subdirectories:<br/>- `__intm`: for storing intermediate files used during creating index.<br/>- `staging`: for storing staging files used during searching index. |
 | `region_engine.mito.index.staging_size` | String | `2GB` | The max capacity of the staging directory. |
@@ -447,11 +454,12 @@
 | `region_engine.mito.memtable.fork_dictionary_bytes` | String | `1GiB` | Max dictionary bytes.<br/>Only available for `partition_tree` memtable. |
 | `region_engine.file` | -- | -- | Enable the file engine. |
 | `logging` | -- | -- | The logging options. |
-| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
+| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. If set to empty, logs will not be written to 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 | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
+| `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
 | `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` | -- |
 | `export_metrics` | -- | -- | The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
@@ -492,11 +500,12 @@
 | `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. |
 | `logging` | -- | -- | The logging options. |
-| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. |
+| `logging.dir` | String | `/tmp/greptimedb/logs` | The directory to store the log files. If set to empty, logs will not be written to 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 | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
+| `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
 | `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` | -- |
 | `tracing` | -- | -- | The tracing options. Only effect when compiled with `tokio-console` feature. |
--- a/config/datanode.example.toml
+++ b/config/datanode.example.toml
@@ -336,7 +336,7 @@ 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"
+credential = "base64-credential"

 ## The container of the azure account.
 ## **It's only used when the storage type is `Azblob`**.
@@ -360,9 +360,23 @@ region = "us-west-2"

 # Custom storage options
 # [[storage.providers]]
+# name = "S3"
 # type = "S3"
+# bucket = "greptimedb"
+# root = "data"
+# access_key_id = "test"
+# secret_access_key = "123456"
+# endpoint = "https://s3.amazonaws.com"
+# region = "us-west-2"
 # [[storage.providers]]
+# name = "Gcs"
 # type = "Gcs"
+# bucket = "greptimedb"
+# root = "data"
+# scope = "test"
+# credential_path = "123456"
+# credential = "base64-credential"
+# endpoint = "https://storage.googleapis.com"

 ## The region engine options. You can configure multiple region engines.
 [[region_engine]]
@@ -441,6 +455,10 @@ parallel_scan_channel_size = 32
 ## Whether to allow stale WAL entries read during replay.
 allow_stale_entries = false

+## Minimum time interval between two compactions.
+## To align with the old behavior, the default value is 0 (no restrictions).
+min_compaction_interval = "0m"
+
 ## The options for index in Mito engine.
 [region_engine.mito.index]

@@ -531,7 +549,7 @@ fork_dictionary_bytes = "1GiB"

 ## The logging options.
 [logging]
-## The directory to store the log files.
+## The directory to store the log files. If set to empty, logs will not be written to files.
 dir = "/tmp/greptimedb/logs"

 ## The log level. Can be `info`/`debug`/`warn`/`error`.
@@ -547,6 +565,9 @@ otlp_endpoint = "http://localhost:4317"
 ## Whether to append logs to stdout.
 append_stdout = true

+## The log format. Can be `text`/`json`.
+log_format = "text"
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
--- a/config/flownode.example.toml
+++ b/config/flownode.example.toml
@@ -59,7 +59,7 @@ retry_interval = "3s"

 ## The logging options.
 [logging]
-## The directory to store the log files.
+## The directory to store the log files. If set to empty, logs will not be written to files.
 dir = "/tmp/greptimedb/logs"

 ## The log level. Can be `info`/`debug`/`warn`/`error`.
@@ -75,6 +75,9 @@ otlp_endpoint = "http://localhost:4317"
 ## Whether to append logs to stdout.
 append_stdout = true

+## The log format. Can be `text`/`json`.
+log_format = "text"
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
--- a/config/frontend.example.toml
+++ b/config/frontend.example.toml
@@ -166,7 +166,7 @@ tcp_nodelay = true

 ## The logging options.
 [logging]
-## The directory to store the log files.
+## The directory to store the log files. If set to empty, logs will not be written to files.
 dir = "/tmp/greptimedb/logs"

 ## The log level. Can be `info`/`debug`/`warn`/`error`.
@@ -182,6 +182,9 @@ otlp_endpoint = "http://localhost:4317"
 ## Whether to append logs to stdout.
 append_stdout = true

+## The log format. Can be `text`/`json`.
+log_format = "text"
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
--- a/config/metasrv.example.toml
+++ b/config/metasrv.example.toml
@@ -153,7 +153,7 @@ backoff_deadline = "5mins"

 ## The logging options.
 [logging]
-## The directory to store the log files.
+## The directory to store the log files. If set to empty, logs will not be written to files.
 dir = "/tmp/greptimedb/logs"

 ## The log level. Can be `info`/`debug`/`warn`/`error`.
@@ -169,6 +169,9 @@ otlp_endpoint = "http://localhost:4317"
 ## Whether to append logs to stdout.
 append_stdout = true

+## The log format. Can be `text`/`json`.
+log_format = "text"
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
--- a/config/standalone.example.toml
+++ b/config/standalone.example.toml
@@ -8,6 +8,13 @@ enable_telemetry = true
 ## +toml2docs:none-default
 default_timezone = "UTC"

+## Initialize all regions in the background during the startup.
+## By default, it provides services after all regions have been initialized.
+init_regions_in_background = false
+
+## Parallelism of initializing regions.
+init_regions_parallelism = 16
+
 ## The runtime options.
 [runtime]
 ## The number of threads to execute the runtime for global read operations.
@@ -391,9 +398,23 @@ region = "us-west-2"

 # Custom storage options
 # [[storage.providers]]
+# name = "S3"
 # type = "S3"
+# bucket = "greptimedb"
+# root = "data"
+# access_key_id = "test"
+# secret_access_key = "123456"
+# endpoint = "https://s3.amazonaws.com"
+# region = "us-west-2"
 # [[storage.providers]]
+# name = "Gcs"
 # type = "Gcs"
+# bucket = "greptimedb"
+# root = "data"
+# scope = "test"
+# credential_path = "123456"
+# credential = "base64-credential"
+# endpoint = "https://storage.googleapis.com"

 ## The region engine options. You can configure multiple region engines.
 [[region_engine]]
@@ -472,6 +493,10 @@ parallel_scan_channel_size = 32
 ## Whether to allow stale WAL entries read during replay.
 allow_stale_entries = false

+## Minimum time interval between two compactions.
+## To align with the old behavior, the default value is 0 (no restrictions).
+min_compaction_interval = "0m"
+
 ## The options for index in Mito engine.
 [region_engine.mito.index]

@@ -568,7 +593,7 @@ fork_dictionary_bytes = "1GiB"

 ## The logging options.
 [logging]
-## The directory to store the log files.
+## The directory to store the log files. If set to empty, logs will not be written to files.
 dir = "/tmp/greptimedb/logs"

 ## The log level. Can be `info`/`debug`/`warn`/`error`.
@@ -584,6 +609,9 @@ otlp_endpoint = "http://localhost:4317"
 ## Whether to append logs to stdout.
 append_stdout = true

+## The log format. Can be `text`/`json`.
+log_format = "text"
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
--- a/scripts/check-snafu.py
+++ b/scripts/check-snafu.py
@@ -0,0 +1,69 @@
+# Copyright 2023 Greptime Team
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import re
+
+
+def find_rust_files(directory):
+    error_files = []
+    other_rust_files = []
+    for root, _, files in os.walk(directory):
+        for file in files:
+            if file == "error.rs":
+                error_files.append(os.path.join(root, file))
+            elif file.endswith(".rs"):
+                other_rust_files.append(os.path.join(root, file))
+    return error_files, other_rust_files
+
+
+def extract_branch_names(file_content):
+    pattern = re.compile(r"#\[snafu\(display\([^\)]*\)\)\]\s*(\w+)\s*\{")
+    return pattern.findall(file_content)
+
+
+def check_snafu_in_files(branch_name, rust_files):
+    branch_name_snafu = f"{branch_name}Snafu"
+    for rust_file in rust_files:
+        with open(rust_file, "r") as file:
+            content = file.read()
+            if branch_name_snafu in content:
+                return True
+    return False
+
+
+def main():
+    error_files, other_rust_files = find_rust_files(".")
+    branch_names = []
+
+    for error_file in error_files:
+        with open(error_file, "r") as file:
+            content = file.read()
+            branch_names.extend(extract_branch_names(content))
+
+    unused_snafu = [
+        branch_name
+        for branch_name in branch_names
+        if not check_snafu_in_files(branch_name, other_rust_files)
+    ]
+
+    for name in unused_snafu:
+        print(name)
+
+    if unused_snafu:
+        raise SystemExit(1)
+
+
+if __name__ == "__main__":
+    main()
--- a/src/auth/src/tests.rs
+++ b/src/auth/src/tests.rs
@@ -13,9 +13,11 @@
 // limitations under the License.

 use common_base::secrets::ExposeSecret;
+use common_error::ext::BoxedError;
+use snafu::{OptionExt, ResultExt};

 use crate::error::{
-    AccessDeniedSnafu, Result, UnsupportedPasswordTypeSnafu, UserNotFoundSnafu,
+    AccessDeniedSnafu, AuthBackendSnafu, Result, UnsupportedPasswordTypeSnafu, UserNotFoundSnafu,
    UserPasswordMismatchSnafu,
 };
 use crate::user_info::DefaultUserInfo;
@@ -49,6 +51,19 @@ impl MockUserProvider {
        info.schema.clone_into(&mut self.schema);
        info.username.clone_into(&mut self.username);
    }
+
+    // this is a deliberate function to ref AuthBackendSnafu
+    // so that it won't get deleted in the future
+    pub fn ref_auth_backend_snafu(&self) -> Result<()> {
+        let none_option = None;
+
+        none_option
+            .context(UserNotFoundSnafu {
+                username: "no_user".to_string(),
+            })
+            .map_err(BoxedError::new)
+            .context(AuthBackendSnafu)
+    }
 }

 #[async_trait::async_trait]
--- a/src/auth/tests/mod.rs
+++ b/src/auth/tests/mod.rs
@@ -18,6 +18,7 @@ use std::sync::Arc;

 use api::v1::greptime_request::Request;
 use auth::error::Error::InternalState;
+use auth::error::InternalStateSnafu;
 use auth::{PermissionChecker, PermissionCheckerRef, PermissionReq, PermissionResp, UserInfoRef};
 use sql::statements::show::{ShowDatabases, ShowKind};
 use sql::statements::statement::Statement;
@@ -33,9 +34,10 @@ impl PermissionChecker for DummyPermissionChecker {
        match req {
            PermissionReq::GrpcRequest(_) => Ok(PermissionResp::Allow),
            PermissionReq::SqlStatement(_) => Ok(PermissionResp::Reject),
-            _ => Err(InternalState {
+            _ => InternalStateSnafu {
                msg: "testing".to_string(),
-            }),
+            }
+            .fail(),
        }
    }
 }
--- a/src/catalog/src/error.rs
+++ b/src/catalog/src/error.rs
@@ -97,13 +97,6 @@ pub enum Error {
        source: table::error::Error,
    },

-    #[snafu(display("System catalog is not valid: {}", msg))]
-    SystemCatalog {
-        msg: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Cannot find catalog by name: {}", catalog_name))]
    CatalogNotFound {
        catalog_name: String,
@@ -186,13 +179,6 @@ pub enum Error {
        source: common_query::error::Error,
    },

-    #[snafu(display("Failed to perform metasrv operation"))]
-    Metasrv {
-        #[snafu(implicit)]
-        location: Location,
-        source: meta_client::error::Error,
-    },
-
    #[snafu(display("Invalid table info in catalog"))]
    InvalidTableInfoInCatalog {
        #[snafu(implicit)]
@@ -288,8 +274,6 @@ impl ErrorExt for Error {

            Error::FlowInfoNotFound { .. } => StatusCode::FlowNotFound,

-            Error::SystemCatalog { .. } => StatusCode::StorageUnavailable,
-
            Error::UpgradeWeakCatalogManagerRef { .. } => StatusCode::Internal,

            Error::CreateRecordBatch { source, .. } => source.status_code(),
@@ -303,7 +287,6 @@ impl ErrorExt for Error {

            Error::CreateTable { source, .. } => source.status_code(),

-            Error::Metasrv { source, .. } => source.status_code(),
            Error::DecodePlan { source, .. } => source.status_code(),
            Error::InvalidTableInfoInCatalog { source, .. } => source.status_code(),

@@ -338,27 +321,6 @@ mod tests {

    use super::*;

-    #[test]
-    pub fn test_error_status_code() {
-        assert_eq!(
-            StatusCode::TableAlreadyExists,
-            Error::TableExists {
-                table: "some_table".to_string(),
-                location: Location::generate(),
-            }
-            .status_code()
-        );
-
-        assert_eq!(
-            StatusCode::StorageUnavailable,
-            Error::SystemCatalog {
-                msg: String::default(),
-                location: Location::generate(),
-            }
-            .status_code()
-        );
-    }
-
    #[test]
    pub fn test_errors_to_datafusion_error() {
        let e: DataFusionError = Error::TableExists {
--- a/src/catalog/src/kvbackend/client.rs
+++ b/src/catalog/src/kvbackend/client.rs
@@ -20,8 +20,8 @@ use std::time::Duration;

 use common_error::ext::BoxedError;
 use common_meta::cache_invalidator::KvCacheInvalidator;
-use common_meta::error::Error::{CacheNotGet, GetKvCache};
-use common_meta::error::{CacheNotGetSnafu, Error, ExternalSnafu, Result};
+use common_meta::error::Error::CacheNotGet;
+use common_meta::error::{CacheNotGetSnafu, Error, ExternalSnafu, GetKvCacheSnafu, Result};
 use common_meta::kv_backend::{KvBackend, KvBackendRef, TxnService};
 use common_meta::rpc::store::{
    BatchDeleteRequest, BatchDeleteResponse, BatchGetRequest, BatchGetResponse, BatchPutRequest,
@@ -282,8 +282,11 @@ impl KvBackend for CachedMetaKvBackend {
                _ => Err(e),
            },
        }
-        .map_err(|e| GetKvCache {
-            err_msg: e.to_string(),
+        .map_err(|e| {
+            GetKvCacheSnafu {
+                err_msg: e.to_string(),
+            }
+            .build()
        });

        // "cache.invalidate_key" and "cache.try_get_with_by_ref" are not mutually exclusive. So we need
--- a/src/catalog/src/kvbackend/manager.rs
+++ b/src/catalog/src/kvbackend/manager.rs
@@ -313,7 +313,7 @@ struct SystemCatalog {
    catalog_cache: Cache<String, Arc<InformationSchemaProvider>>,
    pg_catalog_cache: Cache<String, Arc<PGCatalogProvider>>,

-    // system_schema_provier for default catalog
+    // system_schema_provider for default catalog
    information_schema_provider: Arc<InformationSchemaProvider>,
    pg_catalog_provider: Arc<PGCatalogProvider>,
    backend: KvBackendRef,
--- a/src/client/src/database.rs
+++ b/src/client/src/database.rs
@@ -37,7 +37,8 @@ use tonic::metadata::AsciiMetadataKey;
 use tonic::transport::Channel;

 use crate::error::{
-    ConvertFlightDataSnafu, Error, IllegalFlightMessagesSnafu, InvalidAsciiSnafu, ServerSnafu,
+    ConvertFlightDataSnafu, Error, FlightGetSnafu, IllegalFlightMessagesSnafu, InvalidAsciiSnafu,
+    ServerSnafu,
 };
 use crate::{from_grpc_response, Client, Result};

@@ -225,16 +226,18 @@ impl Database {

        let mut client = self.client.make_flight_client()?;

-        let response = client.mut_inner().do_get(request).await.map_err(|e| {
+        let response = client.mut_inner().do_get(request).await.or_else(|e| {
            let tonic_code = e.code();
            let e: Error = e.into();
            let code = e.status_code();
            let msg = e.to_string();
-            let error = Error::FlightGet {
-                tonic_code,
-                addr: client.addr().to_string(),
-                source: BoxedError::new(ServerSnafu { code, msg }.build()),
-            };
+            let error =
+                Err(BoxedError::new(ServerSnafu { code, msg }.build())).with_context(|_| {
+                    FlightGetSnafu {
+                        addr: client.addr().to_string(),
+                        tonic_code,
+                    }
+                });
            error!(
                "Failed to do Flight get, addr: {}, code: {}, source: {:?}",
                client.addr(),
--- a/src/client/src/error.rs
+++ b/src/client/src/error.rs
@@ -39,13 +39,6 @@ pub enum Error {
        source: BoxedError,
    },

-    #[snafu(display("Failure occurs during handling request"))]
-    HandleRequest {
-        #[snafu(implicit)]
-        location: Location,
-        source: BoxedError,
-    },
-
    #[snafu(display("Failed to convert FlightData"))]
    ConvertFlightData {
        #[snafu(implicit)]
@@ -116,13 +109,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Failed to send request with streaming: {}", err_msg))]
-    ClientStreaming {
-        err_msg: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Failed to parse ascii string: {}", value))]
    InvalidAscii {
        value: String,
@@ -138,12 +124,10 @@ impl ErrorExt for Error {
        match self {
            Error::IllegalFlightMessages { .. }
            | Error::MissingField { .. }
-            | Error::IllegalDatabaseResponse { .. }
-            | Error::ClientStreaming { .. } => StatusCode::Internal,
+            | Error::IllegalDatabaseResponse { .. } => StatusCode::Internal,

            Error::Server { code, .. } => *code,
            Error::FlightGet { source, .. }
-            | Error::HandleRequest { source, .. }
            | Error::RegionServer { source, .. }
            | Error::FlowServer { source, .. } => source.status_code(),
            Error::CreateChannel { source, .. }
--- a/src/client/src/flow.rs
+++ b/src/client/src/flow.rs
@@ -16,9 +16,9 @@ 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, ResultExt};
+use snafu::ResultExt;

-use crate::error::Result;
+use crate::error::{FlowServerSnafu, Result};
 use crate::Client;

 #[derive(Debug)]
@@ -57,15 +57,10 @@ impl FlowRequester {
        let response = client
            .handle_create_remove(request)
            .await
-            .map_err(|e| {
+            .or_else(|e| {
                let code = e.code();
                let err: crate::error::Error = e.into();
-                crate::error::Error::FlowServer {
-                    addr,
-                    code,
-                    source: BoxedError::new(err),
-                    location: location!(),
-                }
+                Err(BoxedError::new(err)).context(FlowServerSnafu { addr, code })
            })?
            .into_inner();
        Ok(response)
@@ -88,15 +83,10 @@ impl FlowRequester {
        let response = client
            .handle_mirror_request(requests)
            .await
-            .map_err(|e| {
+            .or_else(|e| {
                let code = e.code();
                let err: crate::error::Error = e.into();
-                crate::error::Error::FlowServer {
-                    addr,
-                    code,
-                    source: BoxedError::new(err),
-                    location: location!(),
-                }
+                Err(BoxedError::new(err)).context(FlowServerSnafu { addr, code })
            })?
            .into_inner();
        Ok(response)
--- a/src/client/src/region.rs
+++ b/src/client/src/region.rs
@@ -38,8 +38,8 @@ use substrait::{DFLogicalSubstraitConvertor, SubstraitPlan};
 use tokio_stream::StreamExt;

 use crate::error::{
-    self, ConvertFlightDataSnafu, IllegalDatabaseResponseSnafu, IllegalFlightMessagesSnafu,
-    MissingFieldSnafu, Result, ServerSnafu,
+    self, ConvertFlightDataSnafu, FlightGetSnafu, IllegalDatabaseResponseSnafu,
+    IllegalFlightMessagesSnafu, MissingFieldSnafu, Result, ServerSnafu,
 };
 use crate::{metrics, Client, Error};

@@ -103,11 +103,14 @@ impl RegionRequester {
                let e: error::Error = e.into();
                let code = e.status_code();
                let msg = e.to_string();
-                let error = Error::FlightGet {
-                    tonic_code,
-                    addr: flight_client.addr().to_string(),
-                    source: BoxedError::new(ServerSnafu { code, msg }.build()),
-                };
+                let error = ServerSnafu { code, msg }
+                    .fail::<()>()
+                    .map_err(BoxedError::new)
+                    .with_context(|_| FlightGetSnafu {
+                        tonic_code,
+                        addr: flight_client.addr().to_string(),
+                    })
+                    .unwrap_err();
                error!(
                    e; "Failed to do Flight get, addr: {}, code: {}",
                    flight_client.addr(),
--- a/src/cmd/src/cli.rs
+++ b/src/cmd/src/cli.rs
@@ -21,6 +21,8 @@ mod export;
 mod helper;

 // Wait for https://github.com/GreptimeTeam/greptimedb/issues/2373
+mod database;
+mod import;
 #[allow(unused)]
 mod repl;

@@ -32,6 +34,7 @@ pub use repl::Repl;
 use tracing_appender::non_blocking::WorkerGuard;

 use self::export::ExportCommand;
+use crate::cli::import::ImportCommand;
 use crate::error::Result;
 use crate::options::GlobalOptions;
 use crate::App;
@@ -114,6 +117,7 @@ enum SubCommand {
    // Attach(AttachCommand),
    Bench(BenchTableMetadataCommand),
    Export(ExportCommand),
+    Import(ImportCommand),
 }

 impl SubCommand {
@@ -122,6 +126,7 @@ impl SubCommand {
            // SubCommand::Attach(cmd) => cmd.build().await,
            SubCommand::Bench(cmd) => cmd.build(guard).await,
            SubCommand::Export(cmd) => cmd.build(guard).await,
+            SubCommand::Import(cmd) => cmd.build(guard).await,
        }
    }
 }
--- a/src/cmd/src/cli/database.rs
+++ b/src/cmd/src/cli/database.rs
@@ -0,0 +1,119 @@
+// 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 base64::engine::general_purpose;
+use base64::Engine;
+use common_catalog::consts::{DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
+use serde_json::Value;
+use servers::http::greptime_result_v1::GreptimedbV1Response;
+use servers::http::GreptimeQueryOutput;
+use snafu::ResultExt;
+
+use crate::error::{HttpQuerySqlSnafu, Result, SerdeJsonSnafu};
+
+pub(crate) struct DatabaseClient {
+    addr: String,
+    catalog: String,
+    auth_header: Option<String>,
+}
+
+impl DatabaseClient {
+    pub fn new(addr: String, catalog: String, auth_basic: Option<String>) -> Self {
+        let auth_header = if let Some(basic) = auth_basic {
+            let encoded = general_purpose::STANDARD.encode(basic);
+            Some(format!("basic {}", encoded))
+        } else {
+            None
+        };
+
+        Self {
+            addr,
+            catalog,
+            auth_header,
+        }
+    }
+
+    pub async fn sql_in_public(&self, sql: &str) -> Result<Option<Vec<Vec<Value>>>> {
+        self.sql(sql, DEFAULT_SCHEMA_NAME).await
+    }
+
+    /// Execute sql query.
+    pub async fn sql(&self, sql: &str, schema: &str) -> Result<Option<Vec<Vec<Value>>>> {
+        let url = format!("http://{}/v1/sql", self.addr);
+        let params = [
+            ("db", format!("{}-{}", self.catalog, schema)),
+            ("sql", sql.to_string()),
+        ];
+        let mut request = reqwest::Client::new()
+            .post(&url)
+            .form(&params)
+            .header("Content-Type", "application/x-www-form-urlencoded");
+        if let Some(ref auth) = self.auth_header {
+            request = request.header("Authorization", auth);
+        }
+
+        let response = request.send().await.with_context(|_| HttpQuerySqlSnafu {
+            reason: format!("bad url: {}", url),
+        })?;
+        let response = response
+            .error_for_status()
+            .with_context(|_| HttpQuerySqlSnafu {
+                reason: format!("query failed: {}", sql),
+            })?;
+
+        let text = response.text().await.with_context(|_| HttpQuerySqlSnafu {
+            reason: "cannot get response text".to_string(),
+        })?;
+
+        let body = serde_json::from_str::<GreptimedbV1Response>(&text).context(SerdeJsonSnafu)?;
+        Ok(body.output().first().and_then(|output| match output {
+            GreptimeQueryOutput::Records(records) => Some(records.rows().clone()),
+            GreptimeQueryOutput::AffectedRows(_) => None,
+        }))
+    }
+}
+
+/// Split at `-`.
+pub(crate) fn split_database(database: &str) -> Result<(String, Option<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 {
+        Ok((catalog.to_string(), Some(schema.to_string())))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[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));
+    }
+}
--- a/src/cmd/src/cli/export.rs
+++ b/src/cmd/src/cli/export.rs
@@ -13,30 +13,23 @@
 // limitations under the License.

 use std::collections::HashSet;
-use std::path::Path;
+use std::path::PathBuf;
 use std::sync::Arc;

 use async_trait::async_trait;
-use base64::engine::general_purpose;
-use base64::Engine;
 use clap::{Parser, ValueEnum};
-use client::DEFAULT_SCHEMA_NAME;
-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::ResultExt;
+use snafu::{OptionExt, ResultExt};
 use tokio::fs::File;
 use tokio::io::{AsyncWriteExt, BufWriter};
 use tokio::sync::Semaphore;
 use tokio::time::Instant;
 use tracing_appender::non_blocking::WorkerGuard;

-use crate::cli::{Instance, Tool};
-use crate::error::{
-    EmptyResultSnafu, Error, FileIoSnafu, HttpQuerySqlSnafu, Result, SerdeJsonSnafu,
-};
+use crate::cli::database::DatabaseClient;
+use crate::cli::{database, Instance, Tool};
+use crate::error::{EmptyResultSnafu, Error, FileIoSnafu, Result, SchemaNotFoundSnafu};

 type TableReference = (String, String, String);

@@ -94,26 +87,21 @@ pub struct ExportCommand {

 impl ExportCommand {
    pub async fn build(&self, guard: Vec<WorkerGuard>) -> Result<Instance> {
-        let (catalog, schema) = split_database(&self.database)?;
+        let (catalog, schema) = database::split_database(&self.database)?;

-        let auth_header = if let Some(basic) = &self.auth_basic {
-            let encoded = general_purpose::STANDARD.encode(basic);
-            Some(format!("basic {}", encoded))
-        } else {
-            None
-        };
+        let database_client =
+            DatabaseClient::new(self.addr.clone(), catalog.clone(), self.auth_basic.clone());

        Ok(Instance::new(
            Box::new(Export {
-                addr: self.addr.clone(),
                catalog,
                schema,
+                database_client,
                output_dir: self.output_dir.clone(),
                parallelism: self.export_jobs,
                target: self.target.clone(),
                start_time: self.start_time.clone(),
                end_time: self.end_time.clone(),
-                auth_header,
            }),
            guard,
        ))
@@ -121,78 +109,59 @@ impl ExportCommand {
 }

 pub struct Export {
-    addr: String,
    catalog: String,
    schema: Option<String>,
+    database_client: DatabaseClient,
    output_dir: String,
    parallelism: usize,
    target: ExportTarget,
    start_time: Option<String>,
    end_time: Option<String>,
-    auth_header: Option<String>,
 }

 impl Export {
-    /// Execute one single sql query.
-    async fn sql(&self, sql: &str) -> Result<Option<Vec<Vec<Value>>>> {
-        let url = format!(
-            "http://{}/v1/sql?db={}-{}&sql={}",
-            self.addr,
-            self.catalog,
-            self.schema.as_deref().unwrap_or(DEFAULT_SCHEMA_NAME),
-            sql
-        );
+    fn catalog_path(&self) -> PathBuf {
+        PathBuf::from(&self.output_dir).join(&self.catalog)
+    }

-        let mut request = reqwest::Client::new()
-            .get(&url)
-            .header("Content-Type", "application/x-www-form-urlencoded");
-        if let Some(ref auth) = self.auth_header {
-            request = request.header("Authorization", auth);
-        }
+    async fn get_db_names(&self) -> Result<Vec<String>> {
+        let db_names = self.all_db_names().await?;
+        let Some(schema) = &self.schema else {
+            return Ok(db_names);
+        };

-        let response = request.send().await.with_context(|_| HttpQuerySqlSnafu {
-            reason: format!("bad url: {}", url),
-        })?;
-        let response = response
-            .error_for_status()
-            .with_context(|_| HttpQuerySqlSnafu {
-                reason: format!("query failed: {}", sql),
-            })?;
-
-        let text = response.text().await.with_context(|_| HttpQuerySqlSnafu {
-            reason: "cannot get response text".to_string(),
-        })?;
-
-        let body = serde_json::from_str::<GreptimedbV1Response>(&text).context(SerdeJsonSnafu)?;
-        Ok(body.output().first().and_then(|output| match output {
-            GreptimeQueryOutput::Records(records) => Some(records.rows().clone()),
-            GreptimeQueryOutput::AffectedRows(_) => None,
-        }))
+        // Check if the schema exists
+        db_names
+            .into_iter()
+            .find(|db_name| db_name.to_lowercase() == schema.to_lowercase())
+            .map(|name| vec![name])
+            .context(SchemaNotFoundSnafu {
+                catalog: &self.catalog,
+                schema,
+            })
    }

    /// Iterate over all db names.
-    ///
-    /// Newbie: `db_name` is catalog + schema.
-    async fn iter_db_names(&self) -> Result<Vec<(String, String)>> {
-        if let Some(schema) = &self.schema {
-            Ok(vec![(self.catalog.clone(), schema.clone())])
-        } else {
-            let result = self.sql("SHOW DATABASES").await?;
-            let Some(records) = result else {
-                EmptyResultSnafu.fail()?
+    async fn all_db_names(&self) -> Result<Vec<String>> {
+        let records = self
+            .database_client
+            .sql_in_public("SHOW DATABASES")
+            .await?
+            .context(EmptyResultSnafu)?;
+        let mut result = Vec::with_capacity(records.len());
+        for value in records {
+            let Value::String(schema) = &value[0] else {
+                unreachable!()
            };
-            let mut result = Vec::with_capacity(records.len());
-            for value in records {
-                let Value::String(schema) = &value[0] else {
-                    unreachable!()
-                };
-                if schema == common_catalog::consts::INFORMATION_SCHEMA_NAME {
-                    continue;
-                }
-                result.push((self.catalog.clone(), schema.clone()));
+            if schema == common_catalog::consts::INFORMATION_SCHEMA_NAME {
+                continue;
            }
-            Ok(result)
+            if schema == common_catalog::consts::PG_CATALOG_NAME {
+                continue;
+            }
+            result.push(schema.clone());
        }
+        Ok(result)
    }

    /// Return a list of [`TableReference`] to be exported.
@@ -201,7 +170,11 @@ impl Export {
        &self,
        catalog: &str,
        schema: &str,
-    ) -> Result<(Vec<TableReference>, Vec<TableReference>)> {
+    ) -> Result<(
+        Vec<TableReference>,
+        Vec<TableReference>,
+        Vec<TableReference>,
+    )> {
        // Puts all metric table first
        let sql = format!(
            "SELECT table_catalog, table_schema, table_name \
@@ -210,15 +183,16 @@ impl Export {
                and table_catalog = \'{catalog}\' \
                and table_schema = \'{schema}\'"
        );
-        let result = self.sql(&sql).await?;
-        let Some(records) = result else {
-            EmptyResultSnafu.fail()?
-        };
+        let records = self
+            .database_client
+            .sql_in_public(&sql)
+            .await?
+            .context(EmptyResultSnafu)?;
        let mut metric_physical_tables = HashSet::with_capacity(records.len());
        for value in records {
            let mut t = Vec::with_capacity(3);
            for v in &value {
-                let serde_json::Value::String(value) = v else {
+                let Value::String(value) = v else {
                    unreachable!()
                };
                t.push(value);
@@ -226,100 +200,142 @@ impl Export {
            metric_physical_tables.insert((t[0].clone(), t[1].clone(), t[2].clone()));
        }

-        // TODO: SQL injection hurts
        let sql = format!(
-            "SELECT table_catalog, table_schema, table_name \
+            "SELECT table_catalog, table_schema, table_name, table_type \
            FROM information_schema.tables \
-            WHERE table_type = \'BASE TABLE\' \
+            WHERE (table_type = \'BASE TABLE\' OR table_type = \'VIEW\') \
                and table_catalog = \'{catalog}\' \
                and table_schema = \'{schema}\'",
        );
-        let result = self.sql(&sql).await?;
-        let Some(records) = result else {
-            EmptyResultSnafu.fail()?
-        };
+        let records = self
+            .database_client
+            .sql_in_public(&sql)
+            .await?
+            .context(EmptyResultSnafu)?;

-        debug!("Fetched table list: {:?}", records);
+        debug!("Fetched table/view list: {:?}", records);

        if records.is_empty() {
-            return Ok((vec![], vec![]));
+            return Ok((vec![], vec![], vec![]));
        }

        let mut remaining_tables = Vec::with_capacity(records.len());
+        let mut views = Vec::new();
        for value in records {
-            let mut t = Vec::with_capacity(3);
+            let mut t = Vec::with_capacity(4);
            for v in &value {
-                let serde_json::Value::String(value) = v else {
+                let Value::String(value) = v else {
                    unreachable!()
                };
                t.push(value);
            }
            let table = (t[0].clone(), t[1].clone(), t[2].clone());
+            let table_type = t[3].as_str();
            // Ignores the physical table
            if !metric_physical_tables.contains(&table) {
-                remaining_tables.push(table);
+                if table_type == "VIEW" {
+                    views.push(table);
+                } else {
+                    remaining_tables.push(table);
+                }
            }
        }

        Ok((
            metric_physical_tables.into_iter().collect(),
            remaining_tables,
+            views,
        ))
    }

-    async fn show_create_table(&self, catalog: &str, schema: &str, table: &str) -> Result<String> {
-        let sql = format!(
-            r#"SHOW CREATE TABLE "{}"."{}"."{}""#,
-            catalog, schema, table
-        );
-        let result = self.sql(&sql).await?;
-        let Some(records) = result else {
-            EmptyResultSnafu.fail()?
+    async fn show_create(
+        &self,
+        show_type: &str,
+        catalog: &str,
+        schema: &str,
+        table: Option<&str>,
+    ) -> Result<String> {
+        let sql = match table {
+            Some(table) => format!(
+                r#"SHOW CREATE {} "{}"."{}"."{}""#,
+                show_type, catalog, schema, table
+            ),
+            None => format!(r#"SHOW CREATE {} "{}"."{}""#, show_type, catalog, schema),
        };
-        let Value::String(create_table) = &records[0][1] else {
+        let records = self
+            .database_client
+            .sql_in_public(&sql)
+            .await?
+            .context(EmptyResultSnafu)?;
+        let Value::String(create) = &records[0][1] else {
            unreachable!()
        };

-        Ok(format!("{};\n", create_table))
+        Ok(format!("{};\n", create))
+    }
+
+    async fn export_create_database(&self) -> Result<()> {
+        let timer = Instant::now();
+        let db_names = self.get_db_names().await?;
+        let db_count = db_names.len();
+        for schema in db_names {
+            let db_dir = self.catalog_path().join(format!("{schema}/"));
+            tokio::fs::create_dir_all(&db_dir)
+                .await
+                .context(FileIoSnafu)?;
+            let file = db_dir.join("create_database.sql");
+            let mut file = File::create(file).await.context(FileIoSnafu)?;
+            let create_database = self
+                .show_create("DATABASE", &self.catalog, &schema, None)
+                .await?;
+            file.write_all(create_database.as_bytes())
+                .await
+                .context(FileIoSnafu)?;
+        }
+
+        let elapsed = timer.elapsed();
+        info!("Success {db_count} jobs, cost: {elapsed:?}");
+
+        Ok(())
    }

    async fn export_create_table(&self) -> Result<()> {
        let timer = Instant::now();
        let semaphore = Arc::new(Semaphore::new(self.parallelism));
-        let db_names = self.iter_db_names().await?;
+        let db_names = self.get_db_names().await?;
        let db_count = db_names.len();
        let mut tasks = Vec::with_capacity(db_names.len());
-        for (catalog, schema) in db_names {
+        for schema in db_names {
            let semaphore_moved = semaphore.clone();
            tasks.push(async move {
                let _permit = semaphore_moved.acquire().await.unwrap();
-                let (metric_physical_tables, remaining_tables) =
-                    self.get_table_list(&catalog, &schema).await?;
-                let table_count = metric_physical_tables.len() + remaining_tables.len();
-                let output_dir = Path::new(&self.output_dir)
-                    .join(&catalog)
-                    .join(format!("{schema}/"));
-                tokio::fs::create_dir_all(&output_dir)
+                let (metric_physical_tables, remaining_tables, views) =
+                    self.get_table_list(&self.catalog, &schema).await?;
+                let table_count =
+                    metric_physical_tables.len() + remaining_tables.len() + views.len();
+                let db_dir = self.catalog_path().join(format!("{schema}/"));
+                tokio::fs::create_dir_all(&db_dir)
                    .await
                    .context(FileIoSnafu)?;
-                let output_file = Path::new(&output_dir).join("create_tables.sql");
-                let mut file = File::create(output_file).await.context(FileIoSnafu)?;
+                let file = db_dir.join("create_tables.sql");
+                let mut file = File::create(file).await.context(FileIoSnafu)?;
                for (c, s, t) in metric_physical_tables.into_iter().chain(remaining_tables) {
-                    match self.show_create_table(&c, &s, &t).await {
-                        Err(e) => {
-                            error!(e; r#"Failed to export table "{}"."{}"."{}""#, c, s, t)
-                        }
-                        Ok(create_table) => {
-                            file.write_all(create_table.as_bytes())
-                                .await
-                                .context(FileIoSnafu)?;
-                        }
-                    }
+                    let create_table = self.show_create("TABLE", &c, &s, Some(&t)).await?;
+                    file.write_all(create_table.as_bytes())
+                        .await
+                        .context(FileIoSnafu)?;
+                }
+                for (c, s, v) in views {
+                    let create_view = self.show_create("VIEW", &c, &s, Some(&v)).await?;
+                    file.write_all(create_view.as_bytes())
+                        .await
+                        .context(FileIoSnafu)?;
                }

                info!(
-                    "Finished exporting {catalog}.{schema} with {table_count} table schemas to path: {}",
-                    output_dir.to_string_lossy()
+                    "Finished exporting {}.{schema} with {table_count} table schemas to path: {}",
+                    self.catalog,
+                    db_dir.to_string_lossy()
                );

                Ok::<(), Error>(())
@@ -332,14 +348,14 @@ impl Export {
            .filter(|r| match r {
                Ok(_) => true,
                Err(e) => {
-                    error!(e; "export job failed");
+                    error!(e; "export schema job failed");
                    false
                }
            })
            .count();

        let elapsed = timer.elapsed();
-        info!("Success {success}/{db_count} jobs, cost: {:?}", elapsed);
+        info!("Success {success}/{db_count} jobs, cost: {elapsed:?}");

        Ok(())
    }
@@ -347,17 +363,15 @@ impl Export {
    async fn export_database_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_names = self.get_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 mut tasks = Vec::with_capacity(db_count);
+        for schema in db_names {
            let semaphore_moved = semaphore.clone();
            tasks.push(async move {
                let _permit = semaphore_moved.acquire().await.unwrap();
-                let output_dir = Path::new(&self.output_dir)
-                    .join(&catalog)
-                    .join(format!("{schema}/"));
-                tokio::fs::create_dir_all(&output_dir)
+                let db_dir = self.catalog_path().join(format!("{schema}/"));
+                tokio::fs::create_dir_all(&db_dir)
                    .await
                    .context(FileIoSnafu)?;

@@ -379,30 +393,31 @@ impl Export {

                let sql = format!(
                    r#"COPY DATABASE "{}"."{}" TO '{}' {};"#,
-                    catalog,
+                    self.catalog,
                    schema,
-                    output_dir.to_str().unwrap(),
+                    db_dir.to_str().unwrap(),
                    with_options
                );

                info!("Executing sql: {sql}");

-                self.sql(&sql).await?;
+                self.database_client.sql_in_public(&sql).await?;

                info!(
-                    "Finished exporting {catalog}.{schema} data into path: {}",
-                    output_dir.to_string_lossy()
+                    "Finished exporting {}.{schema} data into path: {}",
+                    self.catalog,
+                    db_dir.to_string_lossy()
                );

                // The export copy from sql
-                let copy_from_file = output_dir.join("copy_from.sql");
+                let copy_from_file = db_dir.join("copy_from.sql");
                let mut writer =
                    BufWriter::new(File::create(copy_from_file).await.context(FileIoSnafu)?);
                let copy_database_from_sql = format!(
                    r#"COPY DATABASE "{}"."{}" FROM '{}' WITH (FORMAT='parquet');"#,
-                    catalog,
+                    self.catalog,
                    schema,
-                    output_dir.to_str().unwrap()
+                    db_dir.to_str().unwrap()
                );
                writer
                    .write(copy_database_from_sql.as_bytes())
@@ -410,7 +425,7 @@ impl Export {
                    .context(FileIoSnafu)?;
                writer.flush().await.context(FileIoSnafu)?;

-                info!("Finished exporting {catalog}.{schema} copy_from.sql");
+                info!("Finished exporting {}.{schema} copy_from.sql", self.catalog);

                Ok::<(), Error>(())
            })
@@ -429,20 +444,23 @@ impl Export {
            .count();
        let elapsed = timer.elapsed();

-        info!("Success {success}/{db_count} jobs, costs: {:?}", elapsed);
+        info!("Success {success}/{db_count} jobs, costs: {elapsed:?}");

        Ok(())
    }
 }

-#[allow(deprecated)]
 #[async_trait]
 impl Tool for Export {
    async fn do_work(&self) -> Result<()> {
        match self.target {
-            ExportTarget::Schema => self.export_create_table().await,
+            ExportTarget::Schema => {
+                self.export_create_database().await?;
+                self.export_create_table().await
+            }
            ExportTarget::Data => self.export_database_data().await,
            ExportTarget::All => {
+                self.export_create_database().await?;
                self.export_create_table().await?;
                self.export_database_data().await
            }
@@ -450,20 +468,6 @@ impl Tool for Export {
    }
 }

-/// Split at `-`.
-fn split_database(database: &str) -> Result<(String, Option<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 {
-        Ok((catalog.to_string(), Some(schema.to_string())))
-    }
-}
-
 #[cfg(test)]
 mod tests {
    use clap::Parser;
@@ -471,26 +475,10 @@ 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();
--- a/src/cmd/src/cli/import.rs
+++ b/src/cmd/src/cli/import.rs
@@ -0,0 +1,218 @@
+// 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::path::PathBuf;
+use std::sync::Arc;
+
+use async_trait::async_trait;
+use clap::{Parser, ValueEnum};
+use common_catalog::consts::DEFAULT_SCHEMA_NAME;
+use common_telemetry::{error, info, warn};
+use snafu::{OptionExt, ResultExt};
+use tokio::sync::Semaphore;
+use tokio::time::Instant;
+use tracing_appender::non_blocking::WorkerGuard;
+
+use crate::cli::database::DatabaseClient;
+use crate::cli::{database, Instance, Tool};
+use crate::error::{Error, FileIoSnafu, Result, SchemaNotFoundSnafu};
+
+#[derive(Debug, Default, Clone, ValueEnum)]
+enum ImportTarget {
+    /// Import all table schemas into the database.
+    Schema,
+    /// Import all table data into the database.
+    Data,
+    /// Export all table schemas and data at once.
+    #[default]
+    All,
+}
+
+#[derive(Debug, Default, Parser)]
+pub struct ImportCommand {
+    /// Server address to connect
+    #[clap(long)]
+    addr: String,
+
+    /// Directory of the data. E.g.: /tmp/greptimedb-backup
+    #[clap(long)]
+    input_dir: String,
+
+    /// The name of the catalog to import.
+    #[clap(long, default_value = "greptime-*")]
+    database: String,
+
+    /// Parallelism of the import.
+    #[clap(long, short = 'j', default_value = "1")]
+    import_jobs: usize,
+
+    /// Max retry times for each job.
+    #[clap(long, default_value = "3")]
+    max_retry: usize,
+
+    /// Things to export
+    #[clap(long, short = 't', value_enum, default_value = "all")]
+    target: ImportTarget,
+
+    /// The basic authentication for connecting to the server
+    #[clap(long)]
+    auth_basic: Option<String>,
+}
+
+impl ImportCommand {
+    pub async fn build(&self, guard: Vec<WorkerGuard>) -> Result<Instance> {
+        let (catalog, schema) = database::split_database(&self.database)?;
+        let database_client =
+            DatabaseClient::new(self.addr.clone(), catalog.clone(), self.auth_basic.clone());
+
+        Ok(Instance::new(
+            Box::new(Import {
+                catalog,
+                schema,
+                database_client,
+                input_dir: self.input_dir.clone(),
+                parallelism: self.import_jobs,
+                target: self.target.clone(),
+            }),
+            guard,
+        ))
+    }
+}
+
+pub struct Import {
+    catalog: String,
+    schema: Option<String>,
+    database_client: DatabaseClient,
+    input_dir: String,
+    parallelism: usize,
+    target: ImportTarget,
+}
+
+impl Import {
+    async fn import_create_table(&self) -> Result<()> {
+        // Use default db to creates other dbs
+        self.do_sql_job("create_database.sql", Some(DEFAULT_SCHEMA_NAME))
+            .await?;
+        self.do_sql_job("create_tables.sql", None).await
+    }
+
+    async fn import_database_data(&self) -> Result<()> {
+        self.do_sql_job("copy_from.sql", None).await
+    }
+
+    async fn do_sql_job(&self, filename: &str, exec_db: Option<&str>) -> Result<()> {
+        let timer = Instant::now();
+        let semaphore = Arc::new(Semaphore::new(self.parallelism));
+        let db_names = self.get_db_names().await?;
+        let db_count = db_names.len();
+        let mut tasks = Vec::with_capacity(db_count);
+        for schema in db_names {
+            let semaphore_moved = semaphore.clone();
+            tasks.push(async move {
+                let _permit = semaphore_moved.acquire().await.unwrap();
+                let database_input_dir = self.catalog_path().join(&schema);
+                let sql_file = database_input_dir.join(filename);
+                let sql = tokio::fs::read_to_string(sql_file)
+                    .await
+                    .context(FileIoSnafu)?;
+                if sql.is_empty() {
+                    info!("Empty `{filename}` {database_input_dir:?}");
+                } else {
+                    let db = exec_db.unwrap_or(&schema);
+                    self.database_client.sql(&sql, db).await?;
+                    info!("Imported `{filename}` for database {schema}");
+                }
+
+                Ok::<(), Error>(())
+            })
+        }
+
+        let success = futures::future::join_all(tasks)
+            .await
+            .into_iter()
+            .filter(|r| match r {
+                Ok(_) => true,
+                Err(e) => {
+                    error!(e; "import {filename} job failed");
+                    false
+                }
+            })
+            .count();
+        let elapsed = timer.elapsed();
+        info!("Success {success}/{db_count} `{filename}` jobs, cost: {elapsed:?}");
+
+        Ok(())
+    }
+
+    fn catalog_path(&self) -> PathBuf {
+        PathBuf::from(&self.input_dir).join(&self.catalog)
+    }
+
+    async fn get_db_names(&self) -> Result<Vec<String>> {
+        let db_names = self.all_db_names().await?;
+        let Some(schema) = &self.schema else {
+            return Ok(db_names);
+        };
+
+        // Check if the schema exists
+        db_names
+            .into_iter()
+            .find(|db_name| db_name.to_lowercase() == schema.to_lowercase())
+            .map(|name| vec![name])
+            .context(SchemaNotFoundSnafu {
+                catalog: &self.catalog,
+                schema,
+            })
+    }
+
+    // Get all database names in the input directory.
+    // The directory structure should be like:
+    // /tmp/greptimedb-backup
+    // ├── greptime-1
+    // │   ├── db1
+    // │   └── db2
+    async fn all_db_names(&self) -> Result<Vec<String>> {
+        let mut db_names = vec![];
+        let path = self.catalog_path();
+        let mut entries = tokio::fs::read_dir(path).await.context(FileIoSnafu)?;
+        while let Some(entry) = entries.next_entry().await.context(FileIoSnafu)? {
+            let path = entry.path();
+            if path.is_dir() {
+                let db_name = match path.file_name() {
+                    Some(name) => name.to_string_lossy().to_string(),
+                    None => {
+                        warn!("Failed to get the file name of {:?}", path);
+                        continue;
+                    }
+                };
+                db_names.push(db_name);
+            }
+        }
+        Ok(db_names)
+    }
+}
+
+#[async_trait]
+impl Tool for Import {
+    async fn do_work(&self) -> Result<()> {
+        match self.target {
+            ImportTarget::Schema => self.import_create_table().await,
+            ImportTarget::Data => self.import_database_data().await,
+            ImportTarget::All => {
+                self.import_create_table().await?;
+                self.import_database_data().await
+            }
+        }
+    }
+}
--- a/src/cmd/src/error.rs
+++ b/src/cmd/src/error.rs
@@ -31,13 +31,6 @@ pub enum Error {
        source: common_meta::error::Error,
    },

-    #[snafu(display("Failed to iter stream"))]
-    IterStream {
-        #[snafu(implicit)]
-        location: Location,
-        source: common_meta::error::Error,
-    },
-
    #[snafu(display("Failed to init DDL manager"))]
    InitDdlManager {
        #[snafu(implicit)]
@@ -237,13 +230,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Failed to start catalog manager"))]
-    StartCatalogManager {
-        #[snafu(implicit)]
-        location: Location,
-        source: catalog::error::Error,
-    },
-
    #[snafu(display("Failed to connect to Etcd at {etcd_addr}"))]
    ConnectEtcd {
        etcd_addr: String,
@@ -253,14 +239,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Failed to connect server at {addr}"))]
-    ConnectServer {
-        addr: String,
-        source: client::error::Error,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Failed to serde json"))]
    SerdeJson {
        #[snafu(source)]
@@ -278,12 +256,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Expect data from output, but got another thing"))]
-    NotDataFromOutput {
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Empty result from output"))]
    EmptyResult {
        #[snafu(implicit)]
@@ -346,13 +318,12 @@ pub enum Error {
        source: meta_client::error::Error,
    },

-    #[snafu(display("Tonic transport error: {error:?} with msg: {msg:?}"))]
-    TonicTransport {
+    #[snafu(display("Cannot find schema {schema} in catalog {catalog}"))]
+    SchemaNotFound {
+        catalog: String,
+        schema: String,
        #[snafu(implicit)]
        location: Location,
-        #[snafu(source)]
-        error: tonic::transport::Error,
-        msg: Option<String>,
    },
 }

@@ -370,18 +341,16 @@ impl ErrorExt for Error {
            Error::BuildMetaServer { source, .. } => source.status_code(),
            Error::UnsupportedSelectorType { source, .. } => source.status_code(),

-            Error::IterStream { source, .. }
-            | Error::InitMetadata { source, .. }
-            | Error::InitDdlManager { source, .. } => source.status_code(),
+            Error::InitMetadata { source, .. } | Error::InitDdlManager { source, .. } => {
+                source.status_code()
+            }

-            Error::ConnectServer { source, .. } => source.status_code(),
            Error::MissingConfig { .. }
            | Error::LoadLayeredConfig { .. }
            | Error::IllegalConfig { .. }
            | Error::InvalidReplCommand { .. }
            | Error::InitTimezone { .. }
            | Error::ConnectEtcd { .. }
-            | Error::NotDataFromOutput { .. }
            | Error::CreateDir { .. }
            | Error::EmptyResult { .. } => StatusCode::InvalidArguments,

@@ -399,7 +368,6 @@ impl ErrorExt for Error {
                source.status_code()
            }
            Error::SubstraitEncodeLogicalPlan { source, .. } => source.status_code(),
-            Error::StartCatalogManager { source, .. } => source.status_code(),

            Error::SerdeJson { .. } | Error::FileIo { .. } | Error::SpawnThread { .. } => {
                StatusCode::Unexpected
@@ -414,7 +382,7 @@ impl ErrorExt for Error {
                source.status_code()
            }
            Error::MetaClientInit { source, .. } => source.status_code(),
-            Error::TonicTransport { .. } => StatusCode::Internal,
+            Error::SchemaNotFound { .. } => StatusCode::DatabaseNotFound,
        }
    }

--- a/src/cmd/src/standalone.rs
+++ b/src/cmd/src/standalone.rs
@@ -141,6 +141,8 @@ pub struct StandaloneOptions {
    pub region_engine: Vec<RegionEngineConfig>,
    pub export_metrics: ExportMetricsOption,
    pub tracing: TracingOptions,
+    pub init_regions_in_background: bool,
+    pub init_regions_parallelism: usize,
 }

 impl Default for StandaloneOptions {
@@ -168,6 +170,8 @@ impl Default for StandaloneOptions {
                RegionEngineConfig::File(FileEngineConfig::default()),
            ],
            tracing: TracingOptions::default(),
+            init_regions_in_background: false,
+            init_regions_parallelism: 16,
        }
    }
 }
@@ -218,6 +222,9 @@ impl StandaloneOptions {
            storage: cloned_opts.storage,
            region_engine: cloned_opts.region_engine,
            grpc: cloned_opts.grpc,
+            init_regions_in_background: cloned_opts.init_regions_in_background,
+            init_regions_parallelism: cloned_opts.init_regions_parallelism,
+            mode: Mode::Standalone,
            ..Default::default()
        }
    }
--- a/src/cmd/tests/load_config_test.rs
+++ b/src/cmd/tests/load_config_test.rs
@@ -81,6 +81,7 @@ fn test_load_datanode_example_config() {
                    sst_meta_cache_size: ReadableSize::mb(128),
                    vector_cache_size: ReadableSize::mb(512),
                    page_cache_size: ReadableSize::mb(512),
+                    selector_result_cache_size: ReadableSize::mb(512),
                    max_background_jobs: 4,
                    experimental_write_cache_ttl: Some(Duration::from_secs(60 * 60 * 8)),
                    ..Default::default()
--- a/src/common/catalog/src/error.rs
+++ b/src/common/catalog/src/error.rs
@@ -1,46 +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::any::Any;
-
-use common_error::ext::ErrorExt;
-use common_error::status_code::StatusCode;
-use common_macro::stack_trace_debug;
-use snafu::{Location, Snafu};
-
-#[derive(Snafu)]
-#[snafu(visibility(pub))]
-#[stack_trace_debug]
-pub enum Error {
-    #[snafu(display("Invalid full table name: {}", table_name))]
-    InvalidFullTableName {
-        table_name: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-}
-
-impl ErrorExt for Error {
-    fn status_code(&self) -> StatusCode {
-        match self {
-            Error::InvalidFullTableName { .. } => StatusCode::Unexpected,
-        }
-    }
-
-    fn as_any(&self) -> &dyn Any {
-        self
-    }
-}
-
-pub type Result<T> = std::result::Result<T, Error>;
--- a/src/common/catalog/src/lib.rs
+++ b/src/common/catalog/src/lib.rs
@@ -15,7 +15,6 @@
 use consts::DEFAULT_CATALOG_NAME;

 pub mod consts;
-pub mod error;

 #[inline]
 pub fn format_schema_name(catalog: &str, schema: &str) -> String {
--- a/src/common/function/src/scalars/geo/h3.rs
+++ b/src/common/function/src/scalars/geo/h3.rs
@@ -28,7 +28,9 @@ use snafu::{ensure, ResultExt};

 use crate::function::{Function, FunctionContext};

-/// Function that return h3 encoding string for a given geospatial coordinate.
+/// Function that returns [h3] encoding string for a given geospatial coordinate.
+///
+/// [h3]: https://h3geo.org/
 #[derive(Clone, Debug, Default)]
 pub struct H3Function;

@@ -115,7 +117,7 @@ impl Function for H3Function {
                            ))
                        })
                        .context(error::ExecuteSnafu)?;
-                    let r = Resolution::try_from(r as u8)
+                    let r = Resolution::try_from(r)
                        .map_err(|e| {
                            BoxedError::new(PlainError::new(
                                format!("H3 error: {}", e),
--- a/src/common/grpc-expr/src/error.rs
+++ b/src/common/grpc-expr/src/error.rs
@@ -64,12 +64,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Invalid column proto: {}", err_msg))]
-    InvalidColumnProto {
-        err_msg: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
    #[snafu(display("Failed to create vector"))]
    CreateVector {
        #[snafu(implicit)]
@@ -137,7 +131,6 @@ impl ErrorExt for Error {
            Error::DuplicatedTimestampColumn { .. }
            | Error::DuplicatedColumnName { .. }
            | Error::MissingTimestampColumn { .. } => StatusCode::InvalidArguments,
-            Error::InvalidColumnProto { .. } => StatusCode::InvalidArguments,
            Error::CreateVector { .. } => StatusCode::InvalidArguments,
            Error::MissingField { .. } => StatusCode::InvalidArguments,
            Error::InvalidColumnDef { source, .. } => source.status_code(),
--- a/src/common/meta/src/cache_invalidator.rs
+++ b/src/common/meta/src/cache_invalidator.rs
@@ -24,7 +24,7 @@ use crate::key::table_info::TableInfoKey;
 use crate::key::table_name::TableNameKey;
 use crate::key::table_route::TableRouteKey;
 use crate::key::view_info::ViewInfoKey;
-use crate::key::MetaKey;
+use crate::key::MetadataKey;

 /// KvBackend cache invalidator
 #[async_trait::async_trait]
--- a/src/common/meta/src/ddl/utils.rs
+++ b/src/common/meta/src/ddl/utils.rs
@@ -15,12 +15,12 @@
 use common_catalog::consts::METRIC_ENGINE;
 use common_error::ext::BoxedError;
 use common_procedure::error::Error as ProcedureError;
-use snafu::{ensure, location, OptionExt};
+use snafu::{ensure, OptionExt, ResultExt};
 use store_api::metric_engine_consts::LOGICAL_TABLE_METADATA_KEY;
 use table::metadata::TableId;

 use crate::ddl::DetectingRegion;
-use crate::error::{Error, Result, TableNotFoundSnafu, UnsupportedSnafu};
+use crate::error::{Error, OperateDatanodeSnafu, Result, TableNotFoundSnafu, UnsupportedSnafu};
 use crate::key::table_name::TableNameKey;
 use crate::key::TableMetadataManagerRef;
 use crate::peer::Peer;
@@ -32,11 +32,9 @@ use crate::ClusterId;
 pub fn add_peer_context_if_needed(datanode: Peer) -> impl FnOnce(Error) -> Error {
    move |err| {
        if !err.is_retry_later() {
-            return Error::OperateDatanode {
-                location: location!(),
-                peer: datanode,
-                source: BoxedError::new(err),
-            };
+            return Err::<(), BoxedError>(BoxedError::new(err))
+                .context(OperateDatanodeSnafu { peer: datanode })
+                .unwrap_err();
        }
        err
    }
--- a/src/common/meta/src/error.rs
+++ b/src/common/meta/src/error.rs
@@ -21,7 +21,7 @@ use common_macro::stack_trace_debug;
 use common_wal::options::WalOptions;
 use serde_json::error::Error as JsonError;
 use snafu::{Location, Snafu};
-use store_api::storage::{RegionId, RegionNumber};
+use store_api::storage::RegionId;
 use table::metadata::TableId;

 use crate::peer::Peer;
@@ -49,20 +49,6 @@ pub enum Error {
        region_id: RegionId,
    },

-    #[snafu(display("Invalid result with a txn response: {}", err_msg))]
-    InvalidTxnResult {
-        err_msg: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
-    #[snafu(display("Invalid engine type: {}", engine_type))]
-    InvalidEngineType {
-        engine_type: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Failed to connect to Etcd"))]
    ConnectEtcd {
        #[snafu(source)]
@@ -95,15 +81,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Sequence out of range: {}, start={}, step={}", name, start, step))]
-    SequenceOutOfRange {
-        name: String,
-        start: u64,
-        step: u64,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Unexpected sequence value: {}", err_msg))]
    UnexpectedSequenceValue {
        err_msg: String,
@@ -327,13 +304,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Catalog already exists, catalog: {}", catalog))]
-    CatalogAlreadyExists {
-        catalog: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Schema already exists, catalog:{}, schema: {}", catalog, schema))]
    SchemaAlreadyExists {
        catalog: String,
@@ -385,15 +355,8 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Failed to rename table, reason: {}", reason))]
-    RenameTable {
-        reason: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
-    #[snafu(display("Invalid table metadata, err: {}", err_msg))]
-    InvalidTableMetadata {
+    #[snafu(display("Invalid metadata, err: {}", err_msg))]
+    InvalidMetadata {
        err_msg: String,
        #[snafu(implicit)]
        location: Location,
@@ -423,27 +386,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display(
-        "Failed to move region {} in table {}, err: {}",
-        region,
-        table_id,
-        err_msg
-    ))]
-    MoveRegion {
-        table_id: TableId,
-        region: RegionNumber,
-        err_msg: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
-    #[snafu(display("Invalid catalog value"))]
-    InvalidCatalogValue {
-        source: common_catalog::error::Error,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("External error"))]
    External {
        #[snafu(implicit)]
@@ -612,13 +554,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Delimiter not found, key: {}", key))]
-    DelimiterNotFound {
-        key: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Invalid prefix: {}, key: {}", prefix, key))]
    MismatchPrefix {
        prefix: String,
@@ -702,15 +637,12 @@ impl ErrorExt for Error {
            | ParseOption { .. }
            | RouteInfoCorrupted { .. }
            | InvalidProtoMsg { .. }
-            | InvalidTableMetadata { .. }
-            | MoveRegion { .. }
+            | InvalidMetadata { .. }
            | Unexpected { .. }
            | TableInfoNotFound { .. }
            | NextSequence { .. }
-            | SequenceOutOfRange { .. }
            | UnexpectedSequenceValue { .. }
            | InvalidHeartbeatResponse { .. }
-            | InvalidTxnResult { .. }
            | EncodeJson { .. }
            | DecodeJson { .. }
            | PayloadNotExist { .. }
@@ -734,22 +666,17 @@ impl ErrorExt for Error {
            | MetadataCorruption { .. }
            | StrFromUtf8 { .. } => StatusCode::Unexpected,

-            SendMessage { .. } | GetKvCache { .. } | CacheNotGet { .. } | RenameTable { .. } => {
-                StatusCode::Internal
-            }
+            SendMessage { .. } | GetKvCache { .. } | CacheNotGet { .. } => StatusCode::Internal,

            SchemaAlreadyExists { .. } => StatusCode::DatabaseAlreadyExists,

            ProcedureNotFound { .. }
            | InvalidViewInfo { .. }
            | PrimaryKeyNotFound { .. }
-            | CatalogAlreadyExists { .. }
            | EmptyKey { .. }
-            | InvalidEngineType { .. }
            | AlterLogicalTablesInvalidArguments { .. }
            | CreateLogicalTablesInvalidArguments { .. }
            | MismatchPrefix { .. }
-            | DelimiterNotFound { .. }
            | TlsConfig { .. } => StatusCode::InvalidArguments,

            FlowNotFound { .. } => StatusCode::FlowNotFound,
@@ -767,7 +694,6 @@ impl ErrorExt for Error {
            OperateDatanode { source, .. } => source.status_code(),
            Table { source, .. } => source.status_code(),
            RetryLater { source, .. } => source.status_code(),
-            InvalidCatalogValue { source, .. } => source.status_code(),
            ConvertAlterTableRequest { source, .. } => source.status_code(),

            ParseProcedureId { .. }
--- a/src/common/meta/src/key.rs
+++ b/src/common/meta/src/key.rs
@@ -211,7 +211,7 @@ lazy_static! {
 }

 /// The key of metadata.
-pub trait MetaKey<'a, T> {
+pub trait MetadataKey<'a, T> {
    fn to_bytes(&self) -> Vec<u8>;

    fn from_bytes(bytes: &'a [u8]) -> Result<T>;
@@ -226,7 +226,7 @@ impl From<Vec<u8>> for BytesAdapter {
    }
 }

-impl<'a> MetaKey<'a, BytesAdapter> for BytesAdapter {
+impl<'a> MetadataKey<'a, BytesAdapter> for BytesAdapter {
    fn to_bytes(&self) -> Vec<u8> {
        self.0.clone()
    }
@@ -236,7 +236,7 @@ impl<'a> MetaKey<'a, BytesAdapter> for BytesAdapter {
    }
 }

-pub(crate) trait TableMetaKeyGetTxnOp {
+pub(crate) trait MetadataKeyGetTxnOp {
    fn build_get_op(
        &self,
    ) -> (
@@ -245,7 +245,7 @@ pub(crate) trait TableMetaKeyGetTxnOp {
    );
 }

-pub trait TableMetaValue {
+pub trait MetadataValue {
    fn try_from_raw_value(raw_value: &[u8]) -> Result<Self>
    where
        Self: Sized;
@@ -330,7 +330,7 @@ impl<T: DeserializeOwned + Serialize> Serialize for DeserializedValueWithBytes<T
    }
 }

-impl<'de, T: DeserializeOwned + Serialize + TableMetaValue> Deserialize<'de>
+impl<'de, T: DeserializeOwned + Serialize + MetadataValue> Deserialize<'de>
    for DeserializedValueWithBytes<T>
 {
    /// - Deserialize behaviors:
@@ -359,7 +359,7 @@ impl<T: Serialize + DeserializeOwned + Clone> Clone for DeserializedValueWithByt
    }
 }

-impl<T: Serialize + DeserializeOwned + TableMetaValue> DeserializedValueWithBytes<T> {
+impl<T: Serialize + DeserializeOwned + MetadataValue> DeserializedValueWithBytes<T> {
    /// Returns a struct containing a deserialized value and an original `bytes`.
    /// It accepts original bytes of inner.
    pub fn from_inner_bytes(bytes: Bytes) -> Result<Self> {
@@ -1156,10 +1156,10 @@ impl TableMetadataManager {
 }

 #[macro_export]
-macro_rules! impl_table_meta_value {
+macro_rules! impl_metadata_value {
    ($($val_ty: ty), *) => {
        $(
-            impl $crate::key::TableMetaValue for $val_ty {
+            impl $crate::key::MetadataValue for $val_ty {
                fn try_from_raw_value(raw_value: &[u8]) -> Result<Self> {
                    serde_json::from_slice(raw_value).context(SerdeJsonSnafu)
                }
@@ -1172,10 +1172,10 @@ macro_rules! impl_table_meta_value {
    }
 }

-macro_rules! impl_meta_key_get_txn_op {
+macro_rules! impl_metadata_key_get_txn_op {
    ($($key: ty), *) => {
        $(
-            impl $crate::key::TableMetaKeyGetTxnOp for $key {
+            impl $crate::key::MetadataKeyGetTxnOp for $key {
                /// Returns a [TxnOp] to retrieve the corresponding value
                /// and a filter to retrieve the value from the [TxnOpGetResponseSet]
                fn build_get_op(
@@ -1197,7 +1197,7 @@ macro_rules! impl_meta_key_get_txn_op {
    }
 }

-impl_meta_key_get_txn_op! {
+impl_metadata_key_get_txn_op! {
    TableNameKey<'_>,
    TableInfoKey,
    ViewInfoKey,
@@ -1206,7 +1206,7 @@ impl_meta_key_get_txn_op! {
 }

 #[macro_export]
-macro_rules! impl_optional_meta_value {
+macro_rules! impl_optional_metadata_value {
    ($($val_ty: ty), *) => {
        $(
            impl $val_ty {
@@ -1222,7 +1222,7 @@ macro_rules! impl_optional_meta_value {
    }
 }

-impl_table_meta_value! {
+impl_metadata_value! {
    TableNameValue,
    TableInfoValue,
    ViewInfoValue,
@@ -1233,7 +1233,7 @@ impl_table_meta_value! {
    TableFlowValue
 }

-impl_optional_meta_value! {
+impl_optional_metadata_value! {
    CatalogNameValue,
    SchemaNameValue
 }
--- a/src/common/meta/src/key/catalog_name.rs
+++ b/src/common/meta/src/key/catalog_name.rs
@@ -20,8 +20,8 @@ use futures::stream::BoxStream;
 use serde::{Deserialize, Serialize};
 use snafu::{OptionExt, ResultExt};

-use crate::error::{self, Error, InvalidTableMetadataSnafu, Result};
-use crate::key::{MetaKey, CATALOG_NAME_KEY_PATTERN, CATALOG_NAME_KEY_PREFIX};
+use crate::error::{self, Error, InvalidMetadataSnafu, Result};
+use crate::key::{MetadataKey, CATALOG_NAME_KEY_PATTERN, CATALOG_NAME_KEY_PREFIX};
 use crate::kv_backend::KvBackendRef;
 use crate::range_stream::{PaginationStream, DEFAULT_PAGE_SIZE};
 use crate::rpc::store::RangeRequest;
@@ -56,14 +56,14 @@ impl<'a> CatalogNameKey<'a> {
    }
 }

-impl<'a> MetaKey<'a, CatalogNameKey<'a>> for CatalogNameKey<'_> {
+impl<'a> MetadataKey<'a, CatalogNameKey<'a>> for CatalogNameKey<'_> {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }

    fn from_bytes(bytes: &'a [u8]) -> Result<CatalogNameKey<'a>> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            InvalidTableMetadataSnafu {
+            InvalidMetadataSnafu {
                err_msg: format!(
                    "CatalogNameKey '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -87,7 +87,7 @@ impl<'a> TryFrom<&'a str> for CatalogNameKey<'a> {
    fn try_from(s: &'a str) -> Result<Self> {
        let captures = CATALOG_NAME_KEY_PATTERN
            .captures(s)
-            .context(InvalidTableMetadataSnafu {
+            .context(InvalidMetadataSnafu {
                err_msg: format!("Illegal CatalogNameKey format: '{s}'"),
            })?;

--- a/src/common/meta/src/key/datanode_table.rs
+++ b/src/common/meta/src/key/datanode_table.rs
@@ -22,10 +22,10 @@ use snafu::OptionExt;
 use store_api::storage::RegionNumber;
 use table::metadata::TableId;

-use super::MetaKey;
-use crate::error::{InvalidTableMetadataSnafu, Result};
+use super::MetadataKey;
+use crate::error::{InvalidMetadataSnafu, Result};
 use crate::key::{
-    RegionDistribution, TableMetaValue, DATANODE_TABLE_KEY_PATTERN, DATANODE_TABLE_KEY_PREFIX,
+    MetadataValue, RegionDistribution, DATANODE_TABLE_KEY_PATTERN, DATANODE_TABLE_KEY_PREFIX,
 };
 use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
@@ -77,14 +77,14 @@ impl DatanodeTableKey {
    }
 }

-impl<'a> MetaKey<'a, DatanodeTableKey> for DatanodeTableKey {
+impl<'a> MetadataKey<'a, DatanodeTableKey> for DatanodeTableKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }

    fn from_bytes(bytes: &[u8]) -> Result<DatanodeTableKey> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            InvalidTableMetadataSnafu {
+            InvalidMetadataSnafu {
                err_msg: format!(
                    "DatanodeTableKey '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -92,12 +92,11 @@ impl<'a> MetaKey<'a, DatanodeTableKey> for DatanodeTableKey {
            }
            .build()
        })?;
-        let captures =
-            DATANODE_TABLE_KEY_PATTERN
-                .captures(key)
-                .context(InvalidTableMetadataSnafu {
-                    err_msg: format!("Invalid DatanodeTableKey '{key}'"),
-                })?;
+        let captures = DATANODE_TABLE_KEY_PATTERN
+            .captures(key)
+            .context(InvalidMetadataSnafu {
+                err_msg: format!("Invalid DatanodeTableKey '{key}'"),
+            })?;
        // Safety: pass the regex check above
        let datanode_id = captures[1].parse::<DatanodeId>().unwrap();
        let table_id = captures[2].parse::<TableId>().unwrap();
--- a/src/common/meta/src/key/flow.rs
+++ b/src/common/meta/src/key/flow.rs
@@ -38,7 +38,7 @@ use crate::key::flow::flow_name::FlowNameManager;
 use crate::key::flow::flownode_flow::FlownodeFlowManager;
 pub use crate::key::flow::table_flow::{TableFlowManager, TableFlowManagerRef};
 use crate::key::txn_helper::TxnOpGetResponseSet;
-use crate::key::{FlowId, MetaKey};
+use crate::key::{FlowId, MetadataKey};
 use crate::kv_backend::txn::Txn;
 use crate::kv_backend::KvBackendRef;
 use crate::rpc::store::BatchDeleteRequest;
@@ -66,7 +66,7 @@ impl<T> FlowScoped<T> {
    }
 }

-impl<'a, T: MetaKey<'a, T>> MetaKey<'a, FlowScoped<T>> for FlowScoped<T> {
+impl<'a, T: MetadataKey<'a, T>> MetadataKey<'a, FlowScoped<T>> for FlowScoped<T> {
    fn to_bytes(&self) -> Vec<u8> {
        let prefix = FlowScoped::<T>::PREFIX.as_bytes();
        let inner = self.inner.to_bytes();
@@ -295,7 +295,7 @@ mod tests {
        inner: Vec<u8>,
    }

-    impl<'a> MetaKey<'a, MockKey> for MockKey {
+    impl<'a> MetadataKey<'a, MockKey> for MockKey {
        fn to_bytes(&self) -> Vec<u8> {
            self.inner.clone()
        }
--- a/src/common/meta/src/key/flow/flow_info.rs
+++ b/src/common/meta/src/key/flow/flow_info.rs
@@ -25,7 +25,7 @@ use table::table_name::TableName;
 use crate::error::{self, Result};
 use crate::key::flow::FlowScoped;
 use crate::key::txn_helper::TxnOpGetResponseSet;
-use crate::key::{DeserializedValueWithBytes, FlowId, FlowPartitionId, MetaKey, TableMetaValue};
+use crate::key::{DeserializedValueWithBytes, FlowId, FlowPartitionId, MetadataKey, MetadataValue};
 use crate::kv_backend::txn::Txn;
 use crate::kv_backend::KvBackendRef;
 use crate::FlownodeId;
@@ -42,7 +42,7 @@ lazy_static! {
 /// The layout: `__flow/info/{flow_id}`.
 pub struct FlowInfoKey(FlowScoped<FlowInfoKeyInner>);

-impl<'a> MetaKey<'a, FlowInfoKey> for FlowInfoKey {
+impl<'a> MetadataKey<'a, FlowInfoKey> for FlowInfoKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.0.to_bytes()
    }
@@ -80,14 +80,14 @@ impl FlowInfoKeyInner {
    }
 }

-impl<'a> MetaKey<'a, FlowInfoKeyInner> for FlowInfoKeyInner {
+impl<'a> MetadataKey<'a, FlowInfoKeyInner> for FlowInfoKeyInner {
    fn to_bytes(&self) -> Vec<u8> {
        format!("{FLOW_INFO_KEY_PREFIX}/{}", self.flow_id).into_bytes()
    }

    fn from_bytes(bytes: &'a [u8]) -> Result<FlowInfoKeyInner> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            error::InvalidTableMetadataSnafu {
+            error::InvalidMetadataSnafu {
                err_msg: format!(
                    "FlowInfoKeyInner '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -98,7 +98,7 @@ impl<'a> MetaKey<'a, FlowInfoKeyInner> for FlowInfoKeyInner {
        let captures =
            FLOW_INFO_KEY_PATTERN
                .captures(key)
-                .context(error::InvalidTableMetadataSnafu {
+                .context(error::InvalidMetadataSnafu {
                    err_msg: format!("Invalid FlowInfoKeyInner '{key}'"),
                })?;
        // Safety: pass the regex check above
--- a/src/common/meta/src/key/flow/flow_name.rs
+++ b/src/common/meta/src/key/flow/flow_name.rs
@@ -24,7 +24,7 @@ use crate::error::{self, Result};
 use crate::key::flow::FlowScoped;
 use crate::key::txn_helper::TxnOpGetResponseSet;
 use crate::key::{
-    BytesAdapter, DeserializedValueWithBytes, FlowId, MetaKey, TableMetaValue, NAME_PATTERN,
+    BytesAdapter, DeserializedValueWithBytes, FlowId, MetadataKey, MetadataValue, NAME_PATTERN,
 };
 use crate::kv_backend::txn::Txn;
 use crate::kv_backend::KvBackendRef;
@@ -76,7 +76,7 @@ impl<'a> FlowNameKey<'a> {
    }
 }

-impl<'a> MetaKey<'a, FlowNameKey<'a>> for FlowNameKey<'a> {
+impl<'a> MetadataKey<'a, FlowNameKey<'a>> for FlowNameKey<'a> {
    fn to_bytes(&self) -> Vec<u8> {
        self.0.to_bytes()
    }
@@ -95,7 +95,7 @@ pub struct FlowNameKeyInner<'a> {
    pub flow_name: &'a str,
 }

-impl<'a> MetaKey<'a, FlowNameKeyInner<'a>> for FlowNameKeyInner<'_> {
+impl<'a> MetadataKey<'a, FlowNameKeyInner<'a>> for FlowNameKeyInner<'_> {
    fn to_bytes(&self) -> Vec<u8> {
        format!(
            "{FLOW_NAME_KEY_PREFIX}/{}/{}",
@@ -106,7 +106,7 @@ impl<'a> MetaKey<'a, FlowNameKeyInner<'a>> for FlowNameKeyInner<'_> {

    fn from_bytes(bytes: &'a [u8]) -> Result<FlowNameKeyInner> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            error::InvalidTableMetadataSnafu {
+            error::InvalidMetadataSnafu {
                err_msg: format!(
                    "FlowNameKeyInner '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -117,7 +117,7 @@ impl<'a> MetaKey<'a, FlowNameKeyInner<'a>> for FlowNameKeyInner<'_> {
        let captures =
            FLOW_NAME_KEY_PATTERN
                .captures(key)
-                .context(error::InvalidTableMetadataSnafu {
+                .context(error::InvalidMetadataSnafu {
                    err_msg: format!("Invalid FlowNameKeyInner '{key}'"),
                })?;
        // Safety: pass the regex check above
--- a/src/common/meta/src/key/flow/flow_route.rs
+++ b/src/common/meta/src/key/flow/flow_route.rs
@@ -22,7 +22,7 @@ use snafu::OptionExt;

 use crate::error::{self, Result};
 use crate::key::flow::FlowScoped;
-use crate::key::{BytesAdapter, FlowId, FlowPartitionId, MetaKey, TableMetaValue};
+use crate::key::{BytesAdapter, FlowId, FlowPartitionId, MetadataKey, MetadataValue};
 use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
 use crate::peer::Peer;
@@ -68,7 +68,7 @@ impl FlowRouteKey {
    }
 }

-impl<'a> MetaKey<'a, FlowRouteKey> for FlowRouteKey {
+impl<'a> MetadataKey<'a, FlowRouteKey> for FlowRouteKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.0.to_bytes()
    }
@@ -101,7 +101,7 @@ impl FlowRouteKeyInner {
    }
 }

-impl<'a> MetaKey<'a, FlowRouteKeyInner> for FlowRouteKeyInner {
+impl<'a> MetadataKey<'a, FlowRouteKeyInner> for FlowRouteKeyInner {
    fn to_bytes(&self) -> Vec<u8> {
        format!(
            "{FLOW_ROUTE_KEY_PREFIX}/{}/{}",
@@ -112,7 +112,7 @@ impl<'a> MetaKey<'a, FlowRouteKeyInner> for FlowRouteKeyInner {

    fn from_bytes(bytes: &'a [u8]) -> Result<FlowRouteKeyInner> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            error::InvalidTableMetadataSnafu {
+            error::InvalidMetadataSnafu {
                err_msg: format!(
                    "FlowInfoKeyInner '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -123,7 +123,7 @@ impl<'a> MetaKey<'a, FlowRouteKeyInner> for FlowRouteKeyInner {
        let captures =
            FLOW_ROUTE_KEY_PATTERN
                .captures(key)
-                .context(error::InvalidTableMetadataSnafu {
+                .context(error::InvalidMetadataSnafu {
                    err_msg: format!("Invalid FlowInfoKeyInner '{key}'"),
                })?;
        // Safety: pass the regex check above
@@ -209,7 +209,7 @@ impl FlowRouteManager {
 #[cfg(test)]
 mod tests {
    use super::FlowRouteKey;
-    use crate::key::MetaKey;
+    use crate::key::MetadataKey;

    #[test]
    fn test_key_serialization() {
--- a/src/common/meta/src/key/flow/flownode_flow.rs
+++ b/src/common/meta/src/key/flow/flownode_flow.rs
@@ -22,7 +22,7 @@ use snafu::OptionExt;

 use crate::error::{self, Result};
 use crate::key::flow::FlowScoped;
-use crate::key::{BytesAdapter, FlowId, FlowPartitionId, MetaKey};
+use crate::key::{BytesAdapter, FlowId, FlowPartitionId, MetadataKey};
 use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
 use crate::range_stream::{PaginationStream, DEFAULT_PAGE_SIZE};
@@ -44,7 +44,7 @@ const FLOWNODE_FLOW_KEY_PREFIX: &str = "flownode";
 /// The layout `__flow/flownode/{flownode_id}/{flow_id}/{partition_id}`
 pub struct FlownodeFlowKey(FlowScoped<FlownodeFlowKeyInner>);

-impl<'a> MetaKey<'a, FlownodeFlowKey> for FlownodeFlowKey {
+impl<'a> MetadataKey<'a, FlownodeFlowKey> for FlownodeFlowKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.0.to_bytes()
    }
@@ -113,7 +113,7 @@ impl FlownodeFlowKeyInner {
    }
 }

-impl<'a> MetaKey<'a, FlownodeFlowKeyInner> for FlownodeFlowKeyInner {
+impl<'a> MetadataKey<'a, FlownodeFlowKeyInner> for FlownodeFlowKeyInner {
    fn to_bytes(&self) -> Vec<u8> {
        format!(
            "{FLOWNODE_FLOW_KEY_PREFIX}/{}/{}/{}",
@@ -124,7 +124,7 @@ impl<'a> MetaKey<'a, FlownodeFlowKeyInner> for FlownodeFlowKeyInner {

    fn from_bytes(bytes: &'a [u8]) -> Result<FlownodeFlowKeyInner> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            error::InvalidTableMetadataSnafu {
+            error::InvalidMetadataSnafu {
                err_msg: format!(
                    "FlownodeFlowKeyInner '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -135,7 +135,7 @@ impl<'a> MetaKey<'a, FlownodeFlowKeyInner> for FlownodeFlowKeyInner {
        let captures =
            FLOWNODE_FLOW_KEY_PATTERN
                .captures(key)
-                .context(error::InvalidTableMetadataSnafu {
+                .context(error::InvalidMetadataSnafu {
                    err_msg: format!("Invalid FlownodeFlowKeyInner '{key}'"),
                })?;
        // Safety: pass the regex check above
@@ -208,7 +208,7 @@ impl FlownodeFlowManager {
 #[cfg(test)]
 mod tests {
    use crate::key::flow::flownode_flow::FlownodeFlowKey;
-    use crate::key::MetaKey;
+    use crate::key::MetadataKey;

    #[test]
    fn test_key_serialization() {
--- a/src/common/meta/src/key/flow/table_flow.rs
+++ b/src/common/meta/src/key/flow/table_flow.rs
@@ -23,7 +23,7 @@ use table::metadata::TableId;

 use crate::error::{self, Result};
 use crate::key::flow::FlowScoped;
-use crate::key::{BytesAdapter, FlowId, FlowPartitionId, MetaKey, TableMetaValue};
+use crate::key::{BytesAdapter, FlowId, FlowPartitionId, MetadataKey, MetadataValue};
 use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
 use crate::peer::Peer;
@@ -56,7 +56,7 @@ struct TableFlowKeyInner {
 #[derive(Debug, PartialEq)]
 pub struct TableFlowKey(FlowScoped<TableFlowKeyInner>);

-impl<'a> MetaKey<'a, TableFlowKey> for TableFlowKey {
+impl<'a> MetadataKey<'a, TableFlowKey> for TableFlowKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.0.to_bytes()
    }
@@ -129,7 +129,7 @@ impl TableFlowKeyInner {
    }
 }

-impl<'a> MetaKey<'a, TableFlowKeyInner> for TableFlowKeyInner {
+impl<'a> MetadataKey<'a, TableFlowKeyInner> for TableFlowKeyInner {
    fn to_bytes(&self) -> Vec<u8> {
        format!(
            "{TABLE_FLOW_KEY_PREFIX}/{}/{}/{}/{}",
@@ -140,7 +140,7 @@ impl<'a> MetaKey<'a, TableFlowKeyInner> for TableFlowKeyInner {

    fn from_bytes(bytes: &'a [u8]) -> Result<TableFlowKeyInner> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            error::InvalidTableMetadataSnafu {
+            error::InvalidMetadataSnafu {
                err_msg: format!(
                    "TableFlowKeyInner '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -151,7 +151,7 @@ impl<'a> MetaKey<'a, TableFlowKeyInner> for TableFlowKeyInner {
        let captures =
            TABLE_FLOW_KEY_PATTERN
                .captures(key)
-                .context(error::InvalidTableMetadataSnafu {
+                .context(error::InvalidMetadataSnafu {
                    err_msg: format!("Invalid TableFlowKeyInner '{key}'"),
                })?;
        // Safety: pass the regex check above
--- a/src/common/meta/src/key/schema_name.rs
+++ b/src/common/meta/src/key/schema_name.rs
@@ -23,8 +23,8 @@ use humantime_serde::re::humantime;
 use serde::{Deserialize, Serialize};
 use snafu::{OptionExt, ResultExt};

-use crate::error::{self, Error, InvalidTableMetadataSnafu, ParseOptionSnafu, Result};
-use crate::key::{MetaKey, SCHEMA_NAME_KEY_PATTERN, SCHEMA_NAME_KEY_PREFIX};
+use crate::error::{self, Error, InvalidMetadataSnafu, ParseOptionSnafu, Result};
+use crate::key::{MetadataKey, SCHEMA_NAME_KEY_PATTERN, SCHEMA_NAME_KEY_PREFIX};
 use crate::kv_backend::KvBackendRef;
 use crate::range_stream::{PaginationStream, DEFAULT_PAGE_SIZE};
 use crate::rpc::store::RangeRequest;
@@ -89,6 +89,19 @@ impl TryFrom<&HashMap<String, String>> for SchemaNameValue {
    }
 }

+impl From<SchemaNameValue> for HashMap<String, String> {
+    fn from(value: SchemaNameValue) -> Self {
+        let mut opts = HashMap::new();
+        if let Some(ttl) = value.ttl {
+            opts.insert(
+                OPT_KEY_TTL.to_string(),
+                format!("{}", humantime::format_duration(ttl)),
+            );
+        }
+        opts
+    }
+}
+
 impl<'a> SchemaNameKey<'a> {
    pub fn new(catalog: &'a str, schema: &'a str) -> Self {
        Self { catalog, schema }
@@ -109,14 +122,14 @@ impl Display for SchemaNameKey<'_> {
    }
 }

-impl<'a> MetaKey<'a, SchemaNameKey<'a>> for SchemaNameKey<'_> {
+impl<'a> MetadataKey<'a, SchemaNameKey<'a>> for SchemaNameKey<'_> {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }

    fn from_bytes(bytes: &'a [u8]) -> Result<SchemaNameKey<'a>> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            InvalidTableMetadataSnafu {
+            InvalidMetadataSnafu {
                err_msg: format!(
                    "SchemaNameKey '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -142,7 +155,7 @@ impl<'a> TryFrom<&'a str> for SchemaNameKey<'a> {
    fn try_from(s: &'a str) -> Result<Self> {
        let captures = SCHEMA_NAME_KEY_PATTERN
            .captures(s)
-            .context(InvalidTableMetadataSnafu {
+            .context(InvalidMetadataSnafu {
                err_msg: format!("Illegal SchemaNameKey format: '{s}'"),
            })?;

--- a/src/common/meta/src/key/table_info.rs
+++ b/src/common/meta/src/key/table_info.rs
@@ -23,9 +23,9 @@ use table::table_name::TableName;
 use table::table_reference::TableReference;

 use super::TABLE_INFO_KEY_PATTERN;
-use crate::error::{InvalidTableMetadataSnafu, Result};
+use crate::error::{InvalidMetadataSnafu, Result};
 use crate::key::txn_helper::TxnOpGetResponseSet;
-use crate::key::{DeserializedValueWithBytes, MetaKey, TableMetaValue, TABLE_INFO_KEY_PREFIX};
+use crate::key::{DeserializedValueWithBytes, MetadataKey, MetadataValue, TABLE_INFO_KEY_PREFIX};
 use crate::kv_backend::txn::Txn;
 use crate::kv_backend::KvBackendRef;
 use crate::rpc::store::BatchGetRequest;
@@ -51,14 +51,14 @@ impl Display for TableInfoKey {
    }
 }

-impl<'a> MetaKey<'a, TableInfoKey> for TableInfoKey {
+impl<'a> MetadataKey<'a, TableInfoKey> for TableInfoKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }

    fn from_bytes(bytes: &[u8]) -> Result<TableInfoKey> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            InvalidTableMetadataSnafu {
+            InvalidMetadataSnafu {
                err_msg: format!(
                    "TableInfoKey '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -68,7 +68,7 @@ impl<'a> MetaKey<'a, TableInfoKey> for TableInfoKey {
        })?;
        let captures = TABLE_INFO_KEY_PATTERN
            .captures(key)
-            .context(InvalidTableMetadataSnafu {
+            .context(InvalidMetadataSnafu {
                err_msg: format!("Invalid TableInfoKey '{key}'"),
            })?;
        // Safety: pass the regex check above
--- a/src/common/meta/src/key/table_name.rs
+++ b/src/common/meta/src/key/table_name.rs
@@ -22,8 +22,8 @@ use snafu::OptionExt;
 use table::metadata::TableId;
 use table::table_name::TableName;

-use super::{MetaKey, TableMetaValue, TABLE_NAME_KEY_PATTERN, TABLE_NAME_KEY_PREFIX};
-use crate::error::{Error, InvalidTableMetadataSnafu, Result};
+use super::{MetadataKey, MetadataValue, TABLE_NAME_KEY_PATTERN, TABLE_NAME_KEY_PREFIX};
+use crate::error::{Error, InvalidMetadataSnafu, Result};
 use crate::kv_backend::memory::MemoryKvBackend;
 use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
@@ -63,14 +63,14 @@ impl Display for TableNameKey<'_> {
    }
 }

-impl<'a> MetaKey<'a, TableNameKey<'a>> for TableNameKey<'_> {
+impl<'a> MetadataKey<'a, TableNameKey<'a>> for TableNameKey<'_> {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }

    fn from_bytes(bytes: &'a [u8]) -> Result<TableNameKey<'a>> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            InvalidTableMetadataSnafu {
+            InvalidMetadataSnafu {
                err_msg: format!(
                    "TableNameKey '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -80,7 +80,7 @@ impl<'a> MetaKey<'a, TableNameKey<'a>> for TableNameKey<'_> {
        })?;
        let captures = TABLE_NAME_KEY_PATTERN
            .captures(key)
-            .context(InvalidTableMetadataSnafu {
+            .context(InvalidMetadataSnafu {
                err_msg: format!("Invalid TableNameKey '{key}'"),
            })?;
        let catalog = captures.get(1).unwrap().as_str();
@@ -128,7 +128,7 @@ impl<'a> TryFrom<&'a str> for TableNameKey<'a> {
    fn try_from(s: &'a str) -> Result<Self> {
        let captures = TABLE_NAME_KEY_PATTERN
            .captures(s)
-            .context(InvalidTableMetadataSnafu {
+            .context(InvalidMetadataSnafu {
                err_msg: format!("Illegal TableNameKey format: '{s}'"),
            })?;
        // Safety: pass the regex check above
--- a/src/common/meta/src/key/table_route.rs
+++ b/src/common/meta/src/key/table_route.rs
@@ -22,12 +22,12 @@ use store_api::storage::{RegionId, RegionNumber};
 use table::metadata::TableId;

 use crate::error::{
-    self, InvalidTableMetadataSnafu, MetadataCorruptionSnafu, Result, SerdeJsonSnafu,
+    self, InvalidMetadataSnafu, MetadataCorruptionSnafu, Result, SerdeJsonSnafu,
    TableRouteNotFoundSnafu, UnexpectedLogicalRouteTableSnafu,
 };
 use crate::key::txn_helper::TxnOpGetResponseSet;
 use crate::key::{
-    DeserializedValueWithBytes, MetaKey, RegionDistribution, TableMetaValue,
+    DeserializedValueWithBytes, MetadataKey, MetadataValue, RegionDistribution,
    TABLE_ROUTE_KEY_PATTERN, TABLE_ROUTE_PREFIX,
 };
 use crate::kv_backend::txn::Txn;
@@ -199,7 +199,7 @@ impl TableRouteValue {
    }
 }

-impl TableMetaValue for TableRouteValue {
+impl MetadataValue for TableRouteValue {
    fn try_from_raw_value(raw_value: &[u8]) -> Result<Self> {
        let r = serde_json::from_slice::<TableRouteValue>(raw_value);
        match r {
@@ -244,14 +244,14 @@ impl LogicalTableRouteValue {
    }
 }

-impl<'a> MetaKey<'a, TableRouteKey> for TableRouteKey {
+impl<'a> MetadataKey<'a, TableRouteKey> for TableRouteKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }

    fn from_bytes(bytes: &[u8]) -> Result<TableRouteKey> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
-            InvalidTableMetadataSnafu {
+            InvalidMetadataSnafu {
                err_msg: format!(
                    "TableRouteKey '{}' is not a valid UTF8 string: {e}",
                    String::from_utf8_lossy(bytes)
@@ -259,12 +259,11 @@ impl<'a> MetaKey<'a, TableRouteKey> for TableRouteKey {
            }
            .build()
        })?;
-        let captures =
-            TABLE_ROUTE_KEY_PATTERN
-                .captures(key)
-                .context(InvalidTableMetadataSnafu {
-                    err_msg: format!("Invalid TableRouteKey '{key}'"),
-                })?;
+        let captures = TABLE_ROUTE_KEY_PATTERN
+            .captures(key)
+            .context(InvalidMetadataSnafu {
+                err_msg: format!("Invalid TableRouteKey '{key}'"),
+            })?;
        // Safety: pass the regex check above
        let table_id = captures[1].parse::<TableId>().unwrap();
        Ok(TableRouteKey { table_id })
--- a/src/common/meta/src/key/txn_helper.rs
+++ b/src/common/meta/src/key/txn_helper.rs
@@ -16,7 +16,7 @@ use serde::de::DeserializeOwned;
 use serde::Serialize;

 use crate::error::Result;
-use crate::key::{DeserializedValueWithBytes, TableMetaValue};
+use crate::key::{DeserializedValueWithBytes, MetadataValue};
 use crate::kv_backend::txn::TxnOpResponse;
 use crate::rpc::KeyValue;

@@ -41,7 +41,7 @@ impl TxnOpGetResponseSet {
    ) -> impl FnMut(&mut TxnOpGetResponseSet) -> Result<Option<DeserializedValueWithBytes<T>>>
    where
        F: FnMut(&mut TxnOpGetResponseSet) -> Option<Vec<u8>>,
-        T: Serialize + DeserializeOwned + TableMetaValue,
+        T: Serialize + DeserializeOwned + MetadataValue,
    {
        move |set| {
            f(set)
--- a/src/common/meta/src/key/view_info.rs
+++ b/src/common/meta/src/key/view_info.rs
@@ -24,7 +24,7 @@ use table::table_name::TableName;
 use super::VIEW_INFO_KEY_PATTERN;
 use crate::error::{InvalidViewInfoSnafu, Result};
 use crate::key::txn_helper::TxnOpGetResponseSet;
-use crate::key::{DeserializedValueWithBytes, MetaKey, TableMetaValue, VIEW_INFO_KEY_PREFIX};
+use crate::key::{DeserializedValueWithBytes, MetadataKey, MetadataValue, VIEW_INFO_KEY_PREFIX};
 use crate::kv_backend::txn::Txn;
 use crate::kv_backend::KvBackendRef;
 use crate::rpc::store::BatchGetRequest;
@@ -53,7 +53,7 @@ impl Display for ViewInfoKey {
    }
 }

-impl<'a> MetaKey<'a, ViewInfoKey> for ViewInfoKey {
+impl<'a> MetadataKey<'a, ViewInfoKey> for ViewInfoKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }
--- a/src/common/procedure/src/error.rs
+++ b/src/common/procedure/src/error.rs
@@ -13,7 +13,6 @@
 // limitations under the License.

 use std::any::Any;
-use std::string::FromUtf8Error;
 use std::sync::Arc;

 use common_error::ext::{BoxedError, ErrorExt};
@@ -141,12 +140,6 @@ pub enum Error {
        procedure_id: ProcedureId,
    },

-    #[snafu(display("Corrupted data, error: "))]
-    CorruptedData {
-        #[snafu(source)]
-        error: FromUtf8Error,
-    },
-
    #[snafu(display("Failed to start the remove_outdated_meta method, error"))]
    StartRemoveOutdatedMetaTask {
        source: common_runtime::error::Error,
@@ -161,14 +154,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Subprocedure {} failed", subprocedure_id))]
-    SubprocedureFailed {
-        subprocedure_id: ProcedureId,
-        source: Arc<Error>,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Failed to parse segment key: {key}"))]
    ParseSegmentKey {
        #[snafu(implicit)]
@@ -218,14 +203,11 @@ impl ErrorExt for Error {
                StatusCode::InvalidArguments
            }
            Error::ProcedurePanic { .. }
-            | Error::CorruptedData { .. }
            | Error::ParseSegmentKey { .. }
            | Error::Unexpected { .. } => StatusCode::Unexpected,
            Error::ProcedureExec { source, .. } => source.status_code(),
            Error::StartRemoveOutdatedMetaTask { source, .. }
            | Error::StopRemoveOutdatedMetaTask { source, .. } => source.status_code(),
-
-            Error::SubprocedureFailed { source, .. } => source.status_code(),
        }
    }

--- a/src/common/procedure/src/local/runner.rs
+++ b/src/common/procedure/src/local/runner.rs
@@ -19,10 +19,11 @@ use std::time::Duration;
 use backon::{BackoffBuilder, ExponentialBuilder};
 use common_telemetry::{debug, error, info};
 use rand::Rng;
+use snafu::ResultExt;
 use tokio::time;

 use super::rwlock::OwnedKeyRwLockGuard;
-use crate::error::{self, ProcedurePanicSnafu, Result};
+use crate::error::{self, ProcedurePanicSnafu, Result, RollbackTimesExceededSnafu};
 use crate::local::{ManagerContext, ProcedureMeta, ProcedureMetaRef};
 use crate::procedure::{Output, StringKey};
 use crate::store::{ProcedureMessage, ProcedureStore};
@@ -222,12 +223,12 @@ impl Runner {
                    if let Some(d) = rollback.next() {
                        self.wait_on_err(d, rollback_times).await;
                    } else {
-                        self.meta.set_state(ProcedureState::failed(Arc::new(
-                            Error::RollbackTimesExceeded {
-                                source: error.clone(),
+                        let err = Err::<(), Arc<Error>>(error)
+                            .context(RollbackTimesExceededSnafu {
                                procedure_id: self.meta.id,
-                            },
-                        )));
+                            })
+                            .unwrap_err();
+                        self.meta.set_state(ProcedureState::failed(Arc::new(err)));
                        return;
                    }
                }
--- a/src/common/query/src/error.rs
+++ b/src/common/query/src/error.rs
@@ -127,12 +127,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Not expected to run ExecutionPlan more than once"))]
-    ExecuteRepeatedly {
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("General DataFusion error"))]
    GeneralDataFusion {
        #[snafu(source)]
@@ -193,12 +187,6 @@ pub enum Error {
        source: BoxedError,
    },

-    #[snafu(display("Failed to join thread"))]
-    ThreadJoin {
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Failed to decode logical plan: {source}"))]
    DecodePlan {
        #[snafu(implicit)]
@@ -289,9 +277,7 @@ impl ErrorExt for Error {

            Error::MissingTableMutationHandler { .. }
            | Error::MissingProcedureServiceHandler { .. }
-            | Error::MissingFlowServiceHandler { .. }
-            | Error::ExecuteRepeatedly { .. }
-            | Error::ThreadJoin { .. } => StatusCode::Unexpected,
+            | Error::MissingFlowServiceHandler { .. } => StatusCode::Unexpected,

            Error::UnsupportedInputDataType { .. }
            | Error::TypeCast { .. }
@@ -327,7 +313,6 @@ pub fn datafusion_status_code<T: ErrorExt + 'static>(
    match e {
        DataFusionError::Internal(_) => StatusCode::Internal,
        DataFusionError::NotImplemented(_) => StatusCode::Unsupported,
-        DataFusionError::ResourcesExhausted(_) => StatusCode::RuntimeResourcesExhausted,
        DataFusionError::Plan(_) => StatusCode::PlanQuery,
        DataFusionError::External(e) => {
            if let Some(ext) = (*e).downcast_ref::<T>() {
--- a/src/common/substrait/src/error.rs
+++ b/src/common/substrait/src/error.rs
@@ -17,7 +17,6 @@ use std::any::Any;
 use common_error::ext::{BoxedError, ErrorExt};
 use common_error::status_code::StatusCode;
 use common_macro::stack_trace_debug;
-use datafusion::error::DataFusionError;
 use prost::{DecodeError, EncodeError};
 use snafu::{Location, Snafu};

@@ -41,14 +40,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Internal error from DataFusion"))]
-    DFInternal {
-        #[snafu(source)]
-        error: DataFusionError,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Internal error"))]
    Internal {
        #[snafu(implicit)]
@@ -56,12 +47,6 @@ pub enum Error {
        source: BoxedError,
    },

-    #[snafu(display("Cannot convert plan doesn't belong to GreptimeDB"))]
-    UnknownPlan {
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Failed to encode DataFusion plan"))]
    EncodeDfPlan {
        #[snafu(source)]
@@ -84,10 +69,8 @@ pub type Result<T> = std::result::Result<T, Error>;
 impl ErrorExt for Error {
    fn status_code(&self) -> StatusCode {
        match self {
-            Error::UnknownPlan { .. } | Error::EncodeRel { .. } | Error::DecodeRel { .. } => {
-                StatusCode::InvalidArguments
-            }
-            Error::DFInternal { .. } | Error::Internal { .. } => StatusCode::Internal,
+            Error::EncodeRel { .. } | Error::DecodeRel { .. } => StatusCode::InvalidArguments,
+            Error::Internal { .. } => StatusCode::Internal,
            Error::EncodeDfPlan { .. } | Error::DecodeDfPlan { .. } => StatusCode::Unexpected,
        }
    }
--- a/src/common/telemetry/Cargo.toml
+++ b/src/common/telemetry/Cargo.toml
@@ -34,4 +34,4 @@ tracing = "0.1"
 tracing-appender = "0.2"
 tracing-log = "0.1"
 tracing-opentelemetry = "0.22.0"
-tracing-subscriber = { version = "0.3", features = ["env-filter"] }
+tracing-subscriber = { version = "0.3", features = ["env-filter", "json", "fmt"] }
--- a/src/common/telemetry/src/lib.rs
+++ b/src/common/telemetry/src/lib.rs
@@ -21,7 +21,7 @@ mod panic_hook;
 pub mod tracing_context;
 mod tracing_sampler;

-pub use logging::{init_default_ut_logging, init_global_logging};
+pub use logging::{init_default_ut_logging, init_global_logging, RELOAD_HANDLE};
 pub use metric::dump_metrics;
 pub use panic_hook::set_panic_hook;
-pub use {common_error, tracing};
+pub use {common_error, tracing, tracing_subscriber};
--- a/src/common/telemetry/src/logging.rs
+++ b/src/common/telemetry/src/logging.rs
@@ -16,7 +16,7 @@
 use std::env;
 use std::sync::{Arc, Mutex, Once};

-use once_cell::sync::Lazy;
+use once_cell::sync::{Lazy, OnceCell};
 use opentelemetry::{global, KeyValue};
 use opentelemetry_otlp::WithExportConfig;
 use opentelemetry_sdk::propagation::TraceContextPropagator;
@@ -26,6 +26,7 @@ use serde::{Deserialize, Serialize};
 use tracing_appender::non_blocking::WorkerGuard;
 use tracing_appender::rolling::{RollingFileAppender, Rotation};
 use tracing_log::LogTracer;
+use tracing_subscriber::filter::Targets;
 use tracing_subscriber::fmt::Layer;
 use tracing_subscriber::layer::SubscriberExt;
 use tracing_subscriber::prelude::*;
@@ -35,15 +36,41 @@ use crate::tracing_sampler::{create_sampler, TracingSampleOptions};

 pub const DEFAULT_OTLP_ENDPOINT: &str = "http://localhost:4317";

+// Handle for reloading log level
+pub static RELOAD_HANDLE: OnceCell<tracing_subscriber::reload::Handle<Targets, Registry>> =
+    OnceCell::new();
+
+/// The logging options that used to initialize the logger.
 #[derive(Clone, Debug, Serialize, Deserialize)]
 #[serde(default)]
 pub struct LoggingOptions {
+    /// The directory to store log files. If not set, logs will be written to stdout.
    pub dir: String,
+
+    /// The log level that can be one of "trace", "debug", "info", "warn", "error". Default is "info".
    pub level: Option<String>,
-    pub enable_otlp_tracing: bool,
-    pub otlp_endpoint: Option<String>,
-    pub tracing_sample_ratio: Option<TracingSampleOptions>,
+
+    /// The log format that can be one of "json" or "text". Default is "text".
+    pub log_format: LogFormat,
+
+    /// Whether to append logs to stdout. Default is true.
    pub append_stdout: bool,
+
+    /// Whether to enable tracing with OTLP. Default is false.
+    pub enable_otlp_tracing: bool,
+
+    /// The endpoint of OTLP. Default is "http://localhost:4317".
+    pub otlp_endpoint: Option<String>,
+
+    /// The tracing sample ratio.
+    pub tracing_sample_ratio: Option<TracingSampleOptions>,
+}
+
+#[derive(Clone, Debug, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum LogFormat {
+    Json,
+    Text,
 }

 impl PartialEq for LoggingOptions {
@@ -64,6 +91,7 @@ impl Default for LoggingOptions {
        Self {
            dir: "/tmp/greptimedb/logs".to_string(),
            level: None,
+            log_format: LogFormat::Text,
            enable_otlp_tracing: false,
            otlp_endpoint: None,
            tracing_sample_ratio: None,
@@ -128,62 +156,103 @@ pub fn init_global_logging(
    let mut guards = vec![];

    START.call_once(|| {
-        let dir = &opts.dir;
-        let level = &opts.level;
-        let enable_otlp_tracing = opts.enable_otlp_tracing;
-
        // Enable log compatible layer to convert log record to tracing span.
        LogTracer::init().expect("log tracer must be valid");

-        // stdout log layer.
+        // Configure the stdout logging layer.
        let stdout_logging_layer = if opts.append_stdout {
-            let (stdout_writer, stdout_guard) = tracing_appender::non_blocking(std::io::stdout());
-            guards.push(stdout_guard);
+            let (writer, guard) = tracing_appender::non_blocking(std::io::stdout());
+            guards.push(guard);

-            Some(
-                Layer::new()
-                    .with_writer(stdout_writer)
-                    .with_ansi(atty::is(atty::Stream::Stdout)),
-            )
+            if opts.log_format == LogFormat::Json {
+                Some(
+                    Layer::new()
+                        .json()
+                        .with_writer(writer)
+                        .with_ansi(atty::is(atty::Stream::Stdout))
+                        .boxed(),
+                )
+            } else {
+                Some(
+                    Layer::new()
+                        .with_writer(writer)
+                        .with_ansi(atty::is(atty::Stream::Stdout))
+                        .boxed(),
+                )
+            }
        } else {
            None
        };

-        // file log layer.
-        let rolling_appender = RollingFileAppender::new(Rotation::HOURLY, dir, app_name);
-        let (rolling_writer, rolling_writer_guard) =
-            tracing_appender::non_blocking(rolling_appender);
-        let file_logging_layer = Layer::new().with_writer(rolling_writer).with_ansi(false);
-        guards.push(rolling_writer_guard);
+        // Configure the file logging layer with rolling policy.
+        let file_logging_layer = if !opts.dir.is_empty() {
+            let rolling_appender =
+                RollingFileAppender::new(Rotation::HOURLY, &opts.dir, "greptimedb");
+            let (writer, guard) = tracing_appender::non_blocking(rolling_appender);
+            guards.push(guard);

-        // error file log layer.
-        let err_rolling_appender =
-            RollingFileAppender::new(Rotation::HOURLY, dir, format!("{}-{}", app_name, "err"));
-        let (err_rolling_writer, err_rolling_writer_guard) =
-            tracing_appender::non_blocking(err_rolling_appender);
-        let err_file_logging_layer = Layer::new()
-            .with_writer(err_rolling_writer)
-            .with_ansi(false);
-        guards.push(err_rolling_writer_guard);
+            if opts.log_format == LogFormat::Json {
+                Some(
+                    Layer::new()
+                        .json()
+                        .with_writer(writer)
+                        .with_ansi(false)
+                        .boxed(),
+                )
+            } else {
+                Some(Layer::new().with_writer(writer).with_ansi(false).boxed())
+            }
+        } else {
+            None
+        };
+
+        // Configure the error file logging layer with rolling policy.
+        let err_file_logging_layer = if !opts.dir.is_empty() {
+            let rolling_appender =
+                RollingFileAppender::new(Rotation::HOURLY, &opts.dir, "greptimedb-err");
+            let (writer, guard) = tracing_appender::non_blocking(rolling_appender);
+            guards.push(guard);
+
+            if opts.log_format == LogFormat::Json {
+                Some(
+                    Layer::new()
+                        .json()
+                        .with_writer(writer)
+                        .with_ansi(false)
+                        .with_filter(filter::LevelFilter::ERROR)
+                        .boxed(),
+                )
+            } else {
+                Some(
+                    Layer::new()
+                        .with_writer(writer)
+                        .with_ansi(false)
+                        .with_filter(filter::LevelFilter::ERROR)
+                        .boxed(),
+                )
+            }
+        } else {
+            None
+        };

        // resolve log level settings from:
        // - options from command line or config files
        // - environment variable: RUST_LOG
        // - default settings
-        let rust_log_env = std::env::var(EnvFilter::DEFAULT_ENV).ok();
-        let targets_string = level
+        let filter = opts
+            .level
            .as_deref()
-            .or(rust_log_env.as_deref())
-            .unwrap_or(DEFAULT_LOG_TARGETS);
-        let filter = targets_string
+            .or(env::var(EnvFilter::DEFAULT_ENV).ok().as_deref())
+            .unwrap_or(DEFAULT_LOG_TARGETS)
            .parse::<filter::Targets>()
            .expect("error parsing log level string");
-        let sampler = opts
-            .tracing_sample_ratio
-            .as_ref()
-            .map(create_sampler)
-            .map(Sampler::ParentBased)
-            .unwrap_or(Sampler::ParentBased(Box::new(Sampler::AlwaysOn)));
+
+        let (dyn_filter, reload_handle) = tracing_subscriber::reload::Layer::new(filter.clone());
+
+        RELOAD_HANDLE
+            .set(reload_handle)
+            .expect("reload handle already set, maybe init_global_logging get called twice?");
+
        // Must enable 'tokio_unstable' cfg to use this feature.
        // For example: `RUSTFLAGS="--cfg tokio_unstable" cargo run -F common-telemetry/console -- standalone start`
        #[cfg(feature = "tokio-console")]
@@ -204,59 +273,70 @@ pub fn init_global_logging(
                    None
                };

-            let stdout_logging_layer = stdout_logging_layer.map(|x| x.with_filter(filter.clone()));
-
-            let file_logging_layer = file_logging_layer.with_filter(filter);
-
            Registry::default()
+                .with(dyn_filter)
                .with(tokio_console_layer)
                .with(stdout_logging_layer)
                .with(file_logging_layer)
-                .with(err_file_logging_layer.with_filter(filter::LevelFilter::ERROR))
+                .with(err_file_logging_layer)
        };

-        // consume the `tracing_opts`, to avoid "unused" warnings
+        // consume the `tracing_opts` to avoid "unused" warnings.
        let _ = tracing_opts;

        #[cfg(not(feature = "tokio-console"))]
        let subscriber = Registry::default()
-            .with(filter)
+            .with(dyn_filter)
            .with(stdout_logging_layer)
            .with(file_logging_layer)
-            .with(err_file_logging_layer.with_filter(filter::LevelFilter::ERROR));
+            .with(err_file_logging_layer);

-        if enable_otlp_tracing {
+        if opts.enable_otlp_tracing {
            global::set_text_map_propagator(TraceContextPropagator::new());
-            // otlp exporter
+
+            let sampler = opts
+                .tracing_sample_ratio
+                .as_ref()
+                .map(create_sampler)
+                .map(Sampler::ParentBased)
+                .unwrap_or(Sampler::ParentBased(Box::new(Sampler::AlwaysOn)));
+
+            let trace_config = opentelemetry_sdk::trace::config()
+                .with_sampler(sampler)
+                .with_resource(opentelemetry_sdk::Resource::new(vec![
+                    KeyValue::new(resource::SERVICE_NAME, app_name.to_string()),
+                    KeyValue::new(
+                        resource::SERVICE_INSTANCE_ID,
+                        node_id.unwrap_or("none".to_string()),
+                    ),
+                    KeyValue::new(resource::SERVICE_VERSION, env!("CARGO_PKG_VERSION")),
+                    KeyValue::new(resource::PROCESS_PID, std::process::id().to_string()),
+                ]));
+
+            let exporter = opentelemetry_otlp::new_exporter().tonic().with_endpoint(
+                opts.otlp_endpoint
+                    .as_ref()
+                    .map(|e| {
+                        if e.starts_with("http") {
+                            e.to_string()
+                        } else {
+                            format!("http://{}", e)
+                        }
+                    })
+                    .unwrap_or(DEFAULT_OTLP_ENDPOINT.to_string()),
+            );
+
            let tracer = opentelemetry_otlp::new_pipeline()
                .tracing()
-                .with_exporter(
-                    opentelemetry_otlp::new_exporter().tonic().with_endpoint(
-                        opts.otlp_endpoint
-                            .as_ref()
-                            .map(|e| format!("http://{}", e))
-                            .unwrap_or(DEFAULT_OTLP_ENDPOINT.to_string()),
-                    ),
-                )
-                .with_trace_config(
-                    opentelemetry_sdk::trace::config()
-                        .with_sampler(sampler)
-                        .with_resource(opentelemetry_sdk::Resource::new(vec![
-                            KeyValue::new(resource::SERVICE_NAME, app_name.to_string()),
-                            KeyValue::new(
-                                resource::SERVICE_INSTANCE_ID,
-                                node_id.unwrap_or("none".to_string()),
-                            ),
-                            KeyValue::new(resource::SERVICE_VERSION, env!("CARGO_PKG_VERSION")),
-                            KeyValue::new(resource::PROCESS_PID, std::process::id().to_string()),
-                        ])),
-                )
+                .with_exporter(exporter)
+                .with_trace_config(trace_config)
                .install_batch(opentelemetry_sdk::runtime::Tokio)
                .expect("otlp tracer install failed");
-            let tracing_layer = Some(tracing_opentelemetry::layer().with_tracer(tracer));
-            let subscriber = subscriber.with(tracing_layer);
-            tracing::subscriber::set_global_default(subscriber)
-                .expect("error setting global tracing subscriber");
+
+            tracing::subscriber::set_global_default(
+                subscriber.with(tracing_opentelemetry::layer().with_tracer(tracer)),
+            )
+            .expect("error setting global tracing subscriber");
        } else {
            tracing::subscriber::set_global_default(subscriber)
                .expect("error setting global tracing subscriber");
--- a/src/common/time/src/error.rs
+++ b/src/common/time/src/error.rs
@@ -46,13 +46,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Failed to parse a string into Interval, raw string: {}", raw))]
-    ParseInterval {
-        raw: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Current timestamp overflow"))]
    TimestampOverflow {
        #[snafu(source)]
@@ -115,7 +108,6 @@ impl ErrorExt for Error {
            Error::InvalidDateStr { .. } | Error::ArithmeticOverflow { .. } => {
                StatusCode::InvalidArguments
            }
-            Error::ParseInterval { .. } => StatusCode::InvalidArguments,
        }
    }

--- a/src/datanode/src/config.rs
+++ b/src/datanode/src/config.rs
@@ -47,7 +47,8 @@ pub enum ObjectStoreConfig {
 }

 impl ObjectStoreConfig {
-    pub fn name(&self) -> &'static str {
+    /// Returns the object storage type name, such as `S3`, `Oss` etc.
+    pub fn provider_name(&self) -> &'static str {
        match self {
            Self::File(_) => "File",
            Self::S3(_) => "S3",
@@ -56,6 +57,24 @@ impl ObjectStoreConfig {
            Self::Gcs(_) => "Gcs",
        }
    }
+
+    /// Returns the object storage configuration name, return the provider name if it's empty.
+    pub fn config_name(&self) -> &str {
+        let name = match self {
+            // file storage doesn't support name
+            Self::File(_) => self.provider_name(),
+            Self::S3(s3) => &s3.name,
+            Self::Oss(oss) => &oss.name,
+            Self::Azblob(az) => &az.name,
+            Self::Gcs(gcs) => &gcs.name,
+        };
+
+        if name.trim().is_empty() {
+            return self.provider_name();
+        }
+
+        name
+    }
 }

 /// Storage engine config
@@ -66,6 +85,7 @@ pub struct StorageConfig {
    pub data_home: String,
    #[serde(flatten)]
    pub store: ObjectStoreConfig,
+    /// Object storage providers
    pub providers: Vec<ObjectStoreConfig>,
 }

@@ -95,6 +115,7 @@ pub struct ObjectStorageCacheConfig {
 #[derive(Debug, Clone, Serialize, Deserialize)]
 #[serde(default)]
 pub struct S3Config {
+    pub name: String,
    pub bucket: String,
    pub root: String,
    #[serde(skip_serializing)]
@@ -109,7 +130,8 @@ pub struct S3Config {

 impl PartialEq for S3Config {
    fn eq(&self, other: &Self) -> bool {
-        self.bucket == other.bucket
+        self.name == other.name
+            && self.bucket == other.bucket
            && self.root == other.root
            && self.access_key_id.expose_secret() == other.access_key_id.expose_secret()
            && self.secret_access_key.expose_secret() == other.secret_access_key.expose_secret()
@@ -122,6 +144,7 @@ impl PartialEq for S3Config {
 #[derive(Debug, Clone, Serialize, Deserialize)]
 #[serde(default)]
 pub struct OssConfig {
+    pub name: String,
    pub bucket: String,
    pub root: String,
    #[serde(skip_serializing)]
@@ -135,7 +158,8 @@ pub struct OssConfig {

 impl PartialEq for OssConfig {
    fn eq(&self, other: &Self) -> bool {
-        self.bucket == other.bucket
+        self.name == other.name
+            && self.bucket == other.bucket
            && self.root == other.root
            && self.access_key_id.expose_secret() == other.access_key_id.expose_secret()
            && self.access_key_secret.expose_secret() == other.access_key_secret.expose_secret()
@@ -147,6 +171,7 @@ impl PartialEq for OssConfig {
 #[derive(Debug, Clone, Serialize, Deserialize)]
 #[serde(default)]
 pub struct AzblobConfig {
+    pub name: String,
    pub container: String,
    pub root: String,
    #[serde(skip_serializing)]
@@ -161,7 +186,8 @@ pub struct AzblobConfig {

 impl PartialEq for AzblobConfig {
    fn eq(&self, other: &Self) -> bool {
-        self.container == other.container
+        self.name == other.name
+            && self.container == other.container
            && self.root == other.root
            && self.account_name.expose_secret() == other.account_name.expose_secret()
            && self.account_key.expose_secret() == other.account_key.expose_secret()
@@ -174,6 +200,7 @@ impl PartialEq for AzblobConfig {
 #[derive(Debug, Clone, Serialize, Deserialize)]
 #[serde(default)]
 pub struct GcsConfig {
+    pub name: String,
    pub root: String,
    pub bucket: String,
    pub scope: String,
@@ -188,7 +215,8 @@ pub struct GcsConfig {

 impl PartialEq for GcsConfig {
    fn eq(&self, other: &Self) -> bool {
-        self.root == other.root
+        self.name == other.name
+            && self.root == other.root
            && self.bucket == other.bucket
            && self.scope == other.scope
            && self.credential_path.expose_secret() == other.credential_path.expose_secret()
@@ -201,6 +229,7 @@ impl PartialEq for GcsConfig {
 impl Default for S3Config {
    fn default() -> Self {
        Self {
+            name: String::default(),
            bucket: String::default(),
            root: String::default(),
            access_key_id: SecretString::from(String::default()),
@@ -215,6 +244,7 @@ impl Default for S3Config {
 impl Default for OssConfig {
    fn default() -> Self {
        Self {
+            name: String::default(),
            bucket: String::default(),
            root: String::default(),
            access_key_id: SecretString::from(String::default()),
@@ -228,6 +258,7 @@ impl Default for OssConfig {
 impl Default for AzblobConfig {
    fn default() -> Self {
        Self {
+            name: String::default(),
            container: String::default(),
            root: String::default(),
            account_name: SecretString::from(String::default()),
@@ -242,6 +273,7 @@ impl Default for AzblobConfig {
 impl Default for GcsConfig {
    fn default() -> Self {
        Self {
+            name: String::default(),
            root: String::default(),
            bucket: String::default(),
            scope: String::default(),
@@ -355,6 +387,23 @@ mod tests {
        let _parsed: DatanodeOptions = toml::from_str(&toml_string).unwrap();
    }

+    #[test]
+    fn test_config_name() {
+        let object_store_config = ObjectStoreConfig::default();
+        assert_eq!("File", object_store_config.config_name());
+
+        let s3_config = ObjectStoreConfig::S3(S3Config::default());
+        assert_eq!("S3", s3_config.config_name());
+        assert_eq!("S3", s3_config.provider_name());
+
+        let s3_config = ObjectStoreConfig::S3(S3Config {
+            name: "test".to_string(),
+            ..Default::default()
+        });
+        assert_eq!("test", s3_config.config_name());
+        assert_eq!("S3", s3_config.provider_name());
+    }
+
    #[test]
    fn test_secstr() {
        let toml_str = r#"
--- a/src/datanode/src/datanode.rs
+++ b/src/datanode/src/datanode.rs
@@ -273,11 +273,11 @@ impl DatanodeBuilder {
    /// Builds [ObjectStoreManager] from [StorageConfig].
    pub async fn build_object_store_manager(cfg: &StorageConfig) -> Result<ObjectStoreManagerRef> {
        let object_store = store::new_object_store(cfg.store.clone(), &cfg.data_home).await?;
-        let default_name = cfg.store.name();
+        let default_name = cfg.store.config_name();
        let mut object_store_manager = ObjectStoreManager::new(default_name, object_store);
        for store in &cfg.providers {
            object_store_manager.add(
-                store.name(),
+                store.config_name(),
                store::new_object_store(store.clone(), &cfg.data_home).await?,
            );
        }
--- a/src/datanode/src/error.rs
+++ b/src/datanode/src/error.rs
@@ -98,13 +98,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display(
-        "Columns and values number mismatch, columns: {}, values: {}",
-        columns,
-        values
-    ))]
-    ColumnValuesNumberMismatch { columns: usize, values: usize },
-
    #[snafu(display("Failed to delete value from table: {}", table_name))]
    Delete {
        table_name: String,
@@ -156,13 +149,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Runtime resource error"))]
-    RuntimeResource {
-        #[snafu(implicit)]
-        location: Location,
-        source: common_runtime::error::Error,
-    },
-
    #[snafu(display("Expect KvBackend but not found"))]
    MissingKvBackend {
        #[snafu(implicit)]
@@ -172,16 +158,6 @@ pub enum Error {
    #[snafu(display("Invalid SQL, error: {}", msg))]
    InvalidSql { msg: String },

-    #[snafu(display("Not support SQL, error: {}", msg))]
-    NotSupportSql { msg: String },
-
-    #[snafu(display("Specified timestamp key or primary key column not found: {}", name))]
-    KeyColumnNotFound {
-        name: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Illegal primary keys definition: {}", msg))]
    IllegalPrimaryKeysDef {
        msg: String,
@@ -210,14 +186,6 @@ pub enum Error {
        source: meta_client::error::Error,
    },

-    #[snafu(display(
-        "Table id provider not found, cannot execute SQL directly on datanode in distributed mode"
-    ))]
-    TableIdProviderNotFound {
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Missing node id in Datanode config"))]
    MissingNodeId {
        #[snafu(implicit)]
@@ -231,9 +199,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Cannot find requested database: {}-{}", catalog, schema))]
-    DatabaseNotFound { catalog: String, schema: String },
-
    #[snafu(display(
        "No valid default value can be built automatically, column: {}",
        column,
@@ -264,12 +229,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Missing WAL dir config"))]
-    MissingWalDirConfig {
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Unexpected, violated: {}", violated))]
    Unexpected {
        violated: String,
@@ -320,13 +279,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Unsupported gRPC request, kind: {}", kind))]
-    UnsupportedGrpcRequest {
-        kind: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Unsupported output type, expected: {}", expected))]
    UnsupportedOutput {
        expected: String,
@@ -395,20 +347,6 @@ pub enum Error {
        #[snafu(implicit)]
        location: Location,
    },
-
-    #[snafu(display("Failed to setup plugin"))]
-    SetupPlugin {
-        #[snafu(implicit)]
-        location: Location,
-        source: BoxedError,
-    },
-
-    #[snafu(display("Failed to start plugin"))]
-    StartPlugin {
-        #[snafu(implicit)]
-        location: Location,
-        source: BoxedError,
-    },
 }

 pub type Result<T> = std::result::Result<T, Error>;
@@ -430,19 +368,14 @@ impl ErrorExt for Error {

            Delete { source, .. } => source.status_code(),

-            ColumnValuesNumberMismatch { .. }
-            | InvalidSql { .. }
-            | NotSupportSql { .. }
-            | KeyColumnNotFound { .. }
+            InvalidSql { .. }
            | IllegalPrimaryKeysDef { .. }
            | MissingTimestampColumn { .. }
            | CatalogNotFound { .. }
            | SchemaNotFound { .. }
            | SchemaExists { .. }
-            | DatabaseNotFound { .. }
            | MissingNodeId { .. }
            | ColumnNoneDefaultValue { .. }
-            | MissingWalDirConfig { .. }
            | Catalog { .. }
            | MissingRequiredField { .. }
            | RegionEngineNotFound { .. }
@@ -456,12 +389,9 @@ impl ErrorExt for Error {

            AsyncTaskExecute { source, .. } => source.status_code(),

-            CreateDir { .. }
-            | RemoveDir { .. }
-            | ShutdownInstance { .. }
-            | DataFusion { .. }
-            | SetupPlugin { .. }
-            | StartPlugin { .. } => StatusCode::Internal,
+            CreateDir { .. } | RemoveDir { .. } | ShutdownInstance { .. } | DataFusion { .. } => {
+                StatusCode::Internal
+            }

            RegionNotFound { .. } => StatusCode::RegionNotFound,
            RegionNotReady { .. } => StatusCode::RegionNotReady,
@@ -472,11 +402,8 @@ impl ErrorExt for Error {
            InitBackend { .. } => StatusCode::StorageUnavailable,

            OpenLogStore { source, .. } => source.status_code(),
-            RuntimeResource { .. } => StatusCode::RuntimeResourcesExhausted,
            MetaClientInit { source, .. } => source.status_code(),
-            UnsupportedOutput { .. }
-            | TableIdProviderNotFound { .. }
-            | UnsupportedGrpcRequest { .. } => StatusCode::Unsupported,
+            UnsupportedOutput { .. } => StatusCode::Unsupported,
            HandleRegionRequest { source, .. }
            | GetRegionMetadata { source, .. }
            | HandleBatchOpenRequest { source, .. } => source.status_code(),
--- a/src/datanode/src/heartbeat.rs
+++ b/src/datanode/src/heartbeat.rs
@@ -37,7 +37,7 @@ use crate::alive_keeper::RegionAliveKeeper;
 use crate::config::DatanodeOptions;
 use crate::error::{self, MetaClientInitSnafu, Result};
 use crate::event_listener::RegionServerEventReceiver;
-use crate::metrics;
+use crate::metrics::{self, HEARTBEAT_RECV_COUNT, HEARTBEAT_SENT_COUNT};
 use crate::region_server::RegionServer;

 pub(crate) mod handler;
@@ -231,10 +231,12 @@ impl HeartbeatTask {
                                        mailbox_message: Some(message),
                                        ..Default::default()
                                    };
+                                    HEARTBEAT_RECV_COUNT.with_label_values(&["success"]).inc();
                                    Some(req)
                                }
                                Err(e) => {
                                    error!(e; "Failed to encode mailbox messages!");
+                                    HEARTBEAT_RECV_COUNT.with_label_values(&["error"]).inc();
                                    None
                                }
                            }
@@ -304,6 +306,8 @@ impl HeartbeatTask {
                                error!(e; "Failed to reconnect to metasrv!");
                            }
                        }
+                    } else {
+                        HEARTBEAT_SENT_COUNT.inc();
                    }
                }
            }
--- a/src/datanode/src/metrics.rs
+++ b/src/datanode/src/metrics.rs
@@ -54,4 +54,17 @@ lazy_static! {
        &[REGION_ROLE]
    )
    .unwrap();
+    /// The number of heartbeats send by datanode.
+    pub static ref HEARTBEAT_SENT_COUNT: IntCounter = register_int_counter!(
+        "greptime_datanode_heartbeat_send_count",
+        "datanode heartbeat sent",
+    )
+    .unwrap();
+    /// The number of heartbeats received by datanode, labeled with result type.
+    pub static ref HEARTBEAT_RECV_COUNT: IntCounterVec = register_int_counter_vec!(
+        "greptime_datanode_heartbeat_recv_count",
+        "datanode heartbeat received",
+        &["result"]
+    )
+    .unwrap();
 }
--- a/src/datanode/src/region_server.rs
+++ b/src/datanode/src/region_server.rs
@@ -860,7 +860,7 @@ impl RegionServerInner {
        // complains "higher-ranked lifetime error". Rust can't prove some future is legit.
        // Possible related issue: https://github.com/rust-lang/rust/issues/102211
        //
-        // The walkaround is to put the async functions in the `common_runtime::spawn_global`. Or like
+        // The workaround is to put the async functions in the `common_runtime::spawn_global`. Or like
        // it here, collect the values first then use later separately.

        let regions = self
--- a/src/datatypes/src/value.rs
+++ b/src/datatypes/src/value.rs
@@ -268,7 +268,7 @@ impl Value {
        }
    }

-    /// Cast Value to f32. Return None if it's not castable;
+    /// Cast Value to f64. Return None if it's not castable;
    pub fn as_f64_lossy(&self) -> Option<f64> {
        match self {
            Value::Float32(v) => Some(v.0 as _),
--- a/src/flow/src/adapter.rs
+++ b/src/flow/src/adapter.rs
@@ -49,13 +49,13 @@ use crate::adapter::table_source::TableSource;
 use crate::adapter::util::column_schemas_to_proto;
 use crate::adapter::worker::{create_worker, Worker, WorkerHandle};
 use crate::compute::ErrCollector;
+use crate::df_optimizer::sql_to_flow_plan;
 use crate::error::{ExternalSnafu, InternalSnafu, TableNotFoundSnafu, UnexpectedSnafu};
 use crate::expr::GlobalId;
 use crate::metrics::{
    METRIC_FLOW_INPUT_BUF_SIZE, METRIC_FLOW_INSERT_ELAPSED, METRIC_FLOW_RUN_INTERVAL_MS,
 };
 use crate::repr::{self, DiffRow, Row, BATCH_SIZE};
-use crate::transform::sql_to_flow_plan;

 mod flownode_impl;
 mod parse_expr;
--- a/src/flow/src/compute/render.rs
+++ b/src/flow/src/compute/render.rs
@@ -28,7 +28,7 @@ use super::state::Scheduler;
 use crate::compute::state::DataflowState;
 use crate::compute::types::{Collection, CollectionBundle, ErrCollector, Toff};
 use crate::error::{Error, InvalidQuerySnafu, NotImplementedSnafu};
-use crate::expr::{self, GlobalId, LocalId};
+use crate::expr::{self, Batch, GlobalId, LocalId};
 use crate::plan::{Plan, TypedPlan};
 use crate::repr::{self, DiffRow};

@@ -87,9 +87,38 @@ impl<'referred, 'df> Context<'referred, 'df> {
 }

 impl<'referred, 'df> Context<'referred, 'df> {
-    /// Interpret and execute plan
+    /// Like `render_plan` but in Batch Mode
+    pub fn render_plan_batch(&mut self, plan: TypedPlan) -> Result<CollectionBundle<Batch>, Error> {
+        match plan.plan {
+            Plan::Constant { rows } => Ok(self.render_constant_batch(rows)),
+            Plan::Get { .. } => NotImplementedSnafu {
+                reason: "Get is still WIP in batchmode",
+            }
+            .fail(),
+            Plan::Let { .. } => NotImplementedSnafu {
+                reason: "Let is still WIP in batchmode",
+            }
+            .fail(),
+            Plan::Mfp { input, mfp } => self.render_mfp_batch(input, mfp),
+            Plan::Reduce {
+                input,
+                key_val_plan,
+                reduce_plan,
+            } => self.render_reduce_batch(input, &key_val_plan, &reduce_plan, &plan.schema.typ),
+            Plan::Join { .. } => NotImplementedSnafu {
+                reason: "Join is still WIP",
+            }
+            .fail(),
+            Plan::Union { .. } => NotImplementedSnafu {
+                reason: "Union is still WIP",
+            }
+            .fail(),
+        }
+    }
+
+    /// Interpret plan to dataflow and prepare them for execution
    ///
-    /// return the output of this plan
+    /// return the output handler of this plan
    pub fn render_plan(&mut self, plan: TypedPlan) -> Result<CollectionBundle, Error> {
        match plan.plan {
            Plan::Constant { rows } => Ok(self.render_constant(rows)),
@@ -112,17 +141,61 @@ impl<'referred, 'df> Context<'referred, 'df> {
        }
    }

+    /// render Constant, take all rows that have a timestamp not greater than the current time
+    /// This function is primarily used for testing
+    /// Always assume input is sorted by timestamp
+    pub fn render_constant_batch(&mut self, rows: Vec<DiffRow>) -> CollectionBundle<Batch> {
+        let (send_port, recv_port) = self.df.make_edge::<_, Toff<Batch>>("constant_batch");
+        let mut per_time: BTreeMap<repr::Timestamp, Vec<DiffRow>> = Default::default();
+        for (key, group) in &rows.into_iter().group_by(|(_row, ts, _diff)| *ts) {
+            per_time.entry(key).or_default().extend(group);
+        }
+
+        let now = self.compute_state.current_time_ref();
+        // TODO(discord9): better way to schedule future run
+        let scheduler = self.compute_state.get_scheduler();
+        let scheduler_inner = scheduler.clone();
+        let err_collector = self.err_collector.clone();
+
+        let subgraph_id =
+            self.df
+                .add_subgraph_source("ConstantBatch", send_port, move |_ctx, send_port| {
+                    // find the first timestamp that is greater than now
+                    // use filter_map
+
+                    let mut after = per_time.split_off(&(*now.borrow() + 1));
+                    // swap
+                    std::mem::swap(&mut per_time, &mut after);
+                    let not_great_than_now = after;
+
+                    not_great_than_now.into_iter().for_each(|(_ts, rows)| {
+                        err_collector.run(|| {
+                            let rows = rows.into_iter().map(|(row, _ts, _diff)| row).collect();
+                            let batch = Batch::try_from_rows(rows)?;
+                            send_port.give(vec![batch]);
+                            Ok(())
+                        });
+                    });
+                    // schedule the next run
+                    if let Some(next_run_time) = per_time.keys().next().copied() {
+                        scheduler_inner.schedule_at(next_run_time);
+                    }
+                });
+        scheduler.set_cur_subgraph(subgraph_id);
+
+        CollectionBundle::from_collection(Collection::from_port(recv_port))
+    }
+
    /// render Constant, take all rows that have a timestamp not greater than the current time
    ///
    /// Always assume input is sorted by timestamp
    pub fn render_constant(&mut self, rows: Vec<DiffRow>) -> CollectionBundle {
        let (send_port, recv_port) = self.df.make_edge::<_, Toff>("constant");
-        let mut per_time: BTreeMap<repr::Timestamp, Vec<DiffRow>> = rows
-            .into_iter()
-            .group_by(|(_row, ts, _diff)| *ts)
-            .into_iter()
-            .map(|(k, v)| (k, v.into_iter().collect_vec()))
-            .collect();
+        let mut per_time: BTreeMap<repr::Timestamp, Vec<DiffRow>> = Default::default();
+        for (key, group) in &rows.into_iter().group_by(|(_row, ts, _diff)| *ts) {
+            per_time.entry(key).or_default().extend(group);
+        }
+
        let now = self.compute_state.current_time_ref();
        // TODO(discord9): better way to schedule future run
        let scheduler = self.compute_state.get_scheduler();
--- a/src/flow/src/compute/render/map.rs
+++ b/src/flow/src/compute/render/map.rs
@@ -23,12 +23,59 @@ use crate::compute::render::Context;
 use crate::compute::state::Scheduler;
 use crate::compute::types::{Arranged, Collection, CollectionBundle, ErrCollector, Toff};
 use crate::error::{Error, PlanSnafu};
-use crate::expr::{EvalError, MapFilterProject, MfpPlan, ScalarExpr};
+use crate::expr::{Batch, EvalError, MapFilterProject, MfpPlan, ScalarExpr};
 use crate::plan::TypedPlan;
 use crate::repr::{self, DiffRow, KeyValDiffRow, Row};
 use crate::utils::ArrangeHandler;

 impl<'referred, 'df> Context<'referred, 'df> {
+    /// Like `render_mfp` but in batch mode
+    pub fn render_mfp_batch(
+        &mut self,
+        input: Box<TypedPlan>,
+        mfp: MapFilterProject,
+    ) -> Result<CollectionBundle<Batch>, Error> {
+        let input = self.render_plan_batch(*input)?;
+
+        let (out_send_port, out_recv_port) = self.df.make_edge::<_, Toff<Batch>>("mfp_batch");
+
+        // This closure capture following variables:
+        let mfp_plan = MfpPlan::create_from(mfp)?;
+
+        let err_collector = self.err_collector.clone();
+
+        // TODO(discord9): better way to schedule future run
+        let scheduler = self.compute_state.get_scheduler();
+
+        let subgraph = self.df.add_subgraph_in_out(
+            "mfp_batch",
+            input.collection.into_inner(),
+            out_send_port,
+            move |_ctx, recv, send| {
+                // mfp only need to passively receive updates from recvs
+                let src_data = recv.take_inner().into_iter().flat_map(|v| v.into_iter());
+
+                let output_batches = src_data
+                    .filter_map(|mut input_batch| {
+                        err_collector.run(|| {
+                            let res_batch = mfp_plan.mfp.eval_batch_into(&mut input_batch)?;
+                            Ok(res_batch)
+                        })
+                    })
+                    .collect_vec();
+
+                send.give(output_batches);
+            },
+        );
+
+        // register current subgraph in scheduler for future scheduling
+        scheduler.set_cur_subgraph(subgraph);
+
+        let bundle =
+            CollectionBundle::from_collection(Collection::<Batch>::from_port(out_recv_port));
+        Ok(bundle)
+    }
+
    /// render MapFilterProject, will only emit the `rows` once. Assume all incoming row's sys time being `now`` and ignore the row's stated sys time
    /// TODO(discord9): schedule mfp operator to run when temporal filter need
    ///
--- a/src/flow/src/compute/render/reduce.rs
+++ b/src/flow/src/compute/render/reduce.rs
@@ -14,23 +14,247 @@

 use std::collections::BTreeMap;
 use std::ops::Range;
+use std::sync::Arc;

 use datatypes::data_type::ConcreteDataType;
+use datatypes::prelude::DataType;
 use datatypes::value::{ListValue, Value};
+use datatypes::vectors::NullVector;
 use hydroflow::scheduled::graph_ext::GraphExt;
 use itertools::Itertools;
 use snafu::{ensure, OptionExt, ResultExt};

 use crate::compute::render::{Context, SubgraphArg};
 use crate::compute::types::{Arranged, Collection, CollectionBundle, ErrCollector, Toff};
-use crate::error::{Error, PlanSnafu};
+use crate::error::{Error, NotImplementedSnafu, PlanSnafu};
 use crate::expr::error::{DataAlreadyExpiredSnafu, DataTypeSnafu, InternalSnafu};
-use crate::expr::{EvalError, ScalarExpr};
+use crate::expr::{Batch, EvalError, ScalarExpr};
 use crate::plan::{AccumulablePlan, AggrWithIndex, KeyValPlan, ReducePlan, TypedPlan};
 use crate::repr::{self, DiffRow, KeyValDiffRow, RelationType, Row};
 use crate::utils::{ArrangeHandler, ArrangeReader, ArrangeWriter, KeyExpiryManager};

 impl<'referred, 'df> Context<'referred, 'df> {
+    const REDUCE_BATCH: &'static str = "reduce_batch";
+    /// Like `render_reduce`, but for batch mode, and only barebone implementation
+    /// no support for distinct aggregation for now
+    // There is a false positive in using `Vec<ScalarExpr>` as key due to `Value` have `bytes` variant
+    #[allow(clippy::mutable_key_type)]
+    pub fn render_reduce_batch(
+        &mut self,
+        input: Box<TypedPlan>,
+        key_val_plan: &KeyValPlan,
+        reduce_plan: &ReducePlan,
+        output_type: &RelationType,
+    ) -> Result<CollectionBundle<Batch>, Error> {
+        let accum_plan = if let ReducePlan::Accumulable(accum_plan) = reduce_plan {
+            if !accum_plan.distinct_aggrs.is_empty() {
+                NotImplementedSnafu {
+                    reason: "Distinct aggregation is not supported in batch mode",
+                }
+                .fail()?
+            }
+            accum_plan.clone()
+        } else {
+            NotImplementedSnafu {
+                reason: "Only accumulable reduce plan is supported in batch mode",
+            }
+            .fail()?
+        };
+
+        let input = self.render_plan_batch(*input)?;
+
+        // first assembly key&val to separate key and val columns(since this is batch mode)
+        // Then stream kvs through a reduce operator
+
+        // the output is concat from key and val
+        let output_key_arity = key_val_plan.key_plan.output_arity();
+
+        // TODO(discord9): config global expire time from self
+        let arrange_handler = self.compute_state.new_arrange(None);
+
+        if let (Some(time_index), Some(expire_after)) =
+            (output_type.time_index, self.compute_state.expire_after())
+        {
+            let expire_man =
+                KeyExpiryManager::new(Some(expire_after), Some(ScalarExpr::Column(time_index)));
+            arrange_handler.write().set_expire_state(expire_man);
+        }
+
+        // reduce need full arrangement to be able to query all keys
+        let arrange_handler_inner = arrange_handler.clone_full_arrange().context(PlanSnafu {
+            reason: "No write is expected at this point",
+        })?;
+        let key_val_plan = key_val_plan.clone();
+
+        let now = self.compute_state.current_time_ref();
+
+        let err_collector = self.err_collector.clone();
+
+        // TODO(discord9): better way to schedule future run
+        let scheduler = self.compute_state.get_scheduler();
+
+        let (out_send_port, out_recv_port) =
+            self.df.make_edge::<_, Toff<Batch>>(Self::REDUCE_BATCH);
+
+        let subgraph =
+            self.df.add_subgraph_in_out(
+                Self::REDUCE_BATCH,
+                input.collection.into_inner(),
+                out_send_port,
+                move |_ctx, recv, send| {
+                    let now = *(now.borrow());
+                    let arrange = arrange_handler_inner.clone();
+                    // mfp only need to passively receive updates from recvs
+                    let src_data = recv
+                        .take_inner()
+                        .into_iter()
+                        .flat_map(|v| v.into_iter())
+                        .collect_vec();
+
+                    let mut key_to_many_vals = BTreeMap::<Row, Batch>::new();
+                    for batch in src_data {
+                        err_collector.run(|| {
+                            let (key_batch, val_batch) =
+                                batch_split_by_key_val(&batch, &key_val_plan, &err_collector);
+                            ensure!(
+                                key_batch.row_count() == val_batch.row_count(),
+                                InternalSnafu {
+                                    reason: format!(
+                                        "Key and val batch should have the same row count, found {} and {}", 
+                                        key_batch.row_count(),
+                                        val_batch.row_count()
+                                    )
+                                }
+                            );
+
+                            for row_idx in 0..key_batch.row_count() {
+                                let key_row = key_batch.get_row(row_idx).unwrap();
+                                let val_row = val_batch.slice(row_idx, 1)?;
+                                let val_batch =
+                                    key_to_many_vals.entry(Row::new(key_row)).or_default();
+                                val_batch.append_batch(val_row)?;
+                            }
+
+                            Ok(())
+                        });
+                    }
+
+                    // write lock the arrange for the rest of the function body
+                    // to prevent wired race condition
+                    let mut arrange = arrange.write();
+                    let mut all_arrange_updates = Vec::with_capacity(key_to_many_vals.len());
+                    let mut all_output_rows = Vec::with_capacity(key_to_many_vals.len());
+
+                    for (key, val_batch) in key_to_many_vals {
+                        err_collector.run(|| -> Result<(), _> {
+                            let (accums, _, _) = arrange.get(now, &key).unwrap_or_default();
+                            let accum_list = from_accum_values_to_live_accums(
+                                accums.unpack(),
+                                accum_plan.simple_aggrs.len(),
+                            )?;
+
+                            let mut accum_output = AccumOutput::new();
+                            for AggrWithIndex {
+                                expr,
+                                input_idx,
+                                output_idx,
+                            } in accum_plan.simple_aggrs.iter()
+                            {
+                                let cur_old_accum = accum_list.get(*output_idx).cloned().unwrap_or_default();
+                                // if batch is empty, input null instead
+                                let cur_input = val_batch.batch().get(*input_idx).cloned().unwrap_or_else(||Arc::new(NullVector::new(val_batch.row_count())));
+
+                                let (output, new_accum) =
+                                    expr.func.eval_batch(cur_old_accum, cur_input, None)?;
+
+                                accum_output.insert_accum(*output_idx, new_accum);
+                                accum_output.insert_output(*output_idx, output);
+                            }
+
+                            let (new_accums, res_val_row) = accum_output.into_accum_output()?;
+
+                            let arrange_update = ((key.clone(), Row::new(new_accums)), now, 1);
+                            all_arrange_updates.push(arrange_update);
+
+                            let mut key_val = key;
+                            key_val.extend(res_val_row);
+                            all_output_rows.push((key_val, now, 1));
+
+                            Ok(())
+                        });
+                    }
+
+                    err_collector.run(|| {
+                        arrange.apply_updates(now, all_arrange_updates)?;
+                        arrange.compact_to(now)
+                    });
+
+                    // this output part is not supposed to be resource intensive
+                    // (because for every batch there wouldn't usually be as many output row?), 
+                    // so we can do some costly operation here
+                    let output_types = all_output_rows.first().map(|(row, _, _)| {
+                        row.iter()
+                            .map(|v| v.data_type())
+                            .collect::<Vec<ConcreteDataType>>()
+                    });
+
+                    if let Some(output_types) = output_types {
+                        err_collector.run(|| {
+                            let column_cnt = output_types.len();
+                            let row_cnt = all_output_rows.len();
+
+                            let mut output_builder = output_types
+                                .into_iter()
+                                .map(|t| t.create_mutable_vector(row_cnt))
+                                .collect_vec();
+
+                            for (row, _, _) in all_output_rows {
+                                for (i, v) in row.into_iter().enumerate() {
+                                    output_builder
+                                    .get_mut(i)
+                                    .context(InternalSnafu{
+                                        reason: format!(
+                                            "Output builder should have the same length as the row, expected at most {} but got {}", 
+                                            column_cnt-1,
+                                            i
+                                        )
+                                    })?
+                                    .try_push_value_ref(v.as_value_ref())
+                                    .context(DataTypeSnafu {
+                                        msg: "Failed to push value",
+                                    })?;
+                                }
+                            }
+
+                            let output_columns = output_builder
+                                .into_iter()
+                                .map(|mut b| b.to_vector())
+                                .collect_vec();
+
+                            let output_batch = Batch::try_new(output_columns, row_cnt)?;
+                            send.give(vec![output_batch]);
+
+                            Ok(())
+                        });
+                    }
+                },
+            );
+
+        scheduler.set_cur_subgraph(subgraph);
+
+        // by default the key of output arrange
+        let arranged = BTreeMap::from([(
+            (0..output_key_arity).map(ScalarExpr::Column).collect_vec(),
+            Arranged::new(arrange_handler),
+        )]);
+
+        let bundle = CollectionBundle {
+            collection: Collection::from_port(out_recv_port),
+            arranged,
+        };
+        Ok(bundle)
+    }
+
    const REDUCE: &'static str = "reduce";
    /// render `Plan::Reduce` into executable dataflow
    // There is a false positive in using `Vec<ScalarExpr>` as key due to `Value` have `bytes` variant
@@ -151,6 +375,18 @@ impl<'referred, 'df> Context<'referred, 'df> {
    }
 }

+fn from_accum_values_to_live_accums(
+    accums: Vec<Value>,
+    len: usize,
+) -> Result<Vec<Vec<Value>>, EvalError> {
+    let accum_ranges = from_val_to_slice_idx(accums.first().cloned(), len)?;
+    let mut accum_list = vec![];
+    for range in accum_ranges.iter() {
+        accum_list.push(accums.get(range.clone()).unwrap_or_default().to_vec());
+    }
+    Ok(accum_list)
+}
+
 /// All arrange(aka state) used in reduce operator
 pub struct ReduceArrange {
    /// The output arrange of reduce operator
@@ -160,33 +396,40 @@ pub struct ReduceArrange {
    distinct_input: Option<Vec<ArrangeHandler>>,
 }

-/// split a row into key and val by evaluate the key and val plan
-fn split_row_to_key_val(
-    row: Row,
-    sys_time: repr::Timestamp,
-    diff: repr::Diff,
+fn batch_split_by_key_val(
+    batch: &Batch,
    key_val_plan: &KeyValPlan,
-    row_buf: &mut Row,
-) -> Result<Option<KeyValDiffRow>, EvalError> {
-    if let Some(key) = key_val_plan
-        .key_plan
-        .evaluate_into(&mut row.inner.clone(), row_buf)?
-    {
-        // val_plan is not supported to carry any filter predicate,
-        let val = key_val_plan
-            .val_plan
-            .evaluate_into(&mut row.inner.clone(), row_buf)?
-            .context(InternalSnafu {
-                reason: "val_plan should not contain any filter predicate",
-            })?;
-        Ok(Some(((key, val), sys_time, diff)))
-    } else {
-        Ok(None)
+    err_collector: &ErrCollector,
+) -> (Batch, Batch) {
+    let row_count = batch.row_count();
+    let mut key_batch = Batch::empty();
+    let mut val_batch = Batch::empty();
+
+    err_collector.run(|| {
+        if key_val_plan.key_plan.output_arity() != 0 {
+            key_batch = key_val_plan.key_plan.eval_batch_into(&mut batch.clone())?;
+        }
+
+        if key_val_plan.val_plan.output_arity() != 0 {
+            val_batch = key_val_plan.val_plan.eval_batch_into(&mut batch.clone())?;
+        }
+        Ok(())
+    });
+
+    // deal with empty key or val
+    if key_batch.row_count() == 0 && key_batch.column_count() == 0 {
+        key_batch.set_row_count(row_count);
    }
+
+    if val_batch.row_count() == 0 && val_batch.column_count() == 0 {
+        val_batch.set_row_count(row_count);
+    }
+
+    (key_batch, val_batch)
 }

 /// split a row into key and val by evaluate the key and val plan
-fn batch_split_rows_to_key_val(
+fn split_rows_to_key_val(
    rows: impl IntoIterator<Item = DiffRow>,
    key_val_plan: KeyValPlan,
    err_collector: ErrCollector,
@@ -235,7 +478,7 @@ fn reduce_subgraph(
        send,
    }: SubgraphArg,
 ) {
-    let key_val = batch_split_rows_to_key_val(data, key_val_plan.clone(), err_collector.clone());
+    let key_val = split_rows_to_key_val(data, key_val_plan.clone(), err_collector.clone());
    // from here for distinct reduce and accum reduce, things are drastically different
    // for distinct reduce the arrange store the output,
    // but for accum reduce the arrange store the accum state, and output is
@@ -1127,6 +1370,105 @@ mod test {
        run_and_check(&mut state, &mut df, 6..7, expected, output);
    }

+    /// Batch Mode Reduce Evaluation
+    /// SELECT SUM(col) FROM table
+    ///
+    /// table schema:
+    /// | name | type  |
+    /// |------|-------|
+    /// | col  | Int64 |
+    #[test]
+    fn test_basic_batch_reduce_accum() {
+        let mut df = Hydroflow::new();
+        let mut state = DataflowState::default();
+        let now = state.current_time_ref();
+        let mut ctx = harness_test_ctx(&mut df, &mut state);
+
+        let rows = vec![
+            (Row::new(vec![1i64.into()]), 1, 1),
+            (Row::new(vec![2i64.into()]), 2, 1),
+            (Row::new(vec![3i64.into()]), 3, 1),
+            (Row::new(vec![1i64.into()]), 4, 1),
+            (Row::new(vec![2i64.into()]), 5, 1),
+            (Row::new(vec![3i64.into()]), 6, 1),
+        ];
+        let input_plan = Plan::Constant { rows: rows.clone() };
+
+        let typ = RelationType::new(vec![ColumnType::new_nullable(
+            ConcreteDataType::int64_datatype(),
+        )]);
+        let key_val_plan = KeyValPlan {
+            key_plan: MapFilterProject::new(1).project([]).unwrap().into_safe(),
+            val_plan: MapFilterProject::new(1).project([0]).unwrap().into_safe(),
+        };
+
+        let simple_aggrs = vec![AggrWithIndex::new(
+            AggregateExpr {
+                func: AggregateFunc::SumInt64,
+                expr: ScalarExpr::Column(0),
+                distinct: false,
+            },
+            0,
+            0,
+        )];
+        let accum_plan = AccumulablePlan {
+            full_aggrs: vec![AggregateExpr {
+                func: AggregateFunc::SumInt64,
+                expr: ScalarExpr::Column(0),
+                distinct: false,
+            }],
+            simple_aggrs,
+            distinct_aggrs: vec![],
+        };
+
+        let reduce_plan = ReducePlan::Accumulable(accum_plan);
+        let bundle = ctx
+            .render_reduce_batch(
+                Box::new(input_plan.with_types(typ.into_unnamed())),
+                &key_val_plan,
+                &reduce_plan,
+                &RelationType::empty(),
+            )
+            .unwrap();
+
+        {
+            let now_inner = now.clone();
+            let expected = BTreeMap::<i64, Vec<i64>>::from([
+                (1, vec![1i64]),
+                (2, vec![3i64]),
+                (3, vec![6i64]),
+                (4, vec![7i64]),
+                (5, vec![9i64]),
+                (6, vec![12i64]),
+            ]);
+            let collection = bundle.collection;
+            ctx.df
+                .add_subgraph_sink("test_sink", collection.into_inner(), move |_ctx, recv| {
+                    let now = *now_inner.borrow();
+                    let data = recv.take_inner();
+                    let res = data.into_iter().flat_map(|v| v.into_iter()).collect_vec();
+
+                    if let Some(expected) = expected.get(&now) {
+                        let batch = expected.iter().map(|v| Value::from(*v)).collect_vec();
+                        let batch = Batch::try_from_rows(vec![batch.into()]).unwrap();
+                        assert_eq!(res.first(), Some(&batch));
+                    }
+                });
+            drop(ctx);
+
+            for now in 1..7 {
+                state.set_current_ts(now);
+                state.run_available_with_schedule(&mut df);
+                if !state.get_err_collector().is_empty() {
+                    panic!(
+                        "Errors occur: {:?}",
+                        state.get_err_collector().get_all_blocking()
+                    )
+                }
+            }
+        }
+    }
+
    /// SELECT SUM(col) FROM table
    ///
    /// table schema:
--- a/src/flow/src/compute/render/src_sink.rs
+++ b/src/flow/src/compute/render/src_sink.rs
@@ -27,11 +27,67 @@ use crate::compute::render::Context;
 use crate::compute::types::{Arranged, Collection, CollectionBundle, Toff};
 use crate::error::{Error, PlanSnafu};
 use crate::expr::error::InternalSnafu;
-use crate::expr::EvalError;
+use crate::expr::{Batch, EvalError};
 use crate::repr::{DiffRow, Row, BROADCAST_CAP};

 #[allow(clippy::mutable_key_type)]
 impl<'referred, 'df> Context<'referred, 'df> {
+    /// simply send the batch to downstream, without fancy features like buffering
+    pub fn render_source_batch(
+        &mut self,
+        mut src_recv: broadcast::Receiver<Batch>,
+    ) -> Result<CollectionBundle<Batch>, Error> {
+        debug!("Rendering Source Batch");
+        let (send_port, recv_port) = self.df.make_edge::<_, Toff<Batch>>("source_batch");
+
+        let schd = self.compute_state.get_scheduler();
+        let inner_schd = schd.clone();
+        let now = self.compute_state.current_time_ref();
+        let err_collector = self.err_collector.clone();
+
+        let sub = self
+            .df
+            .add_subgraph_source("source_batch", send_port, move |_ctx, send| {
+                loop {
+                    match src_recv.try_recv() {
+                        Ok(batch) => {
+                            send.give(vec![batch]);
+                        }
+                        Err(TryRecvError::Empty) => {
+                            break;
+                        }
+                        Err(TryRecvError::Lagged(lag_offset)) => {
+                            // use `err_collector` instead of `error!` to locate which operator caused the error
+                            err_collector.run(|| -> Result<(), EvalError> {
+                                InternalSnafu {
+                                    reason: format!("Flow missing {} rows behind", lag_offset),
+                                }
+                                .fail()
+                            });
+                            break;
+                        }
+                        Err(TryRecvError::Closed) => {
+                            err_collector.run(|| -> Result<(), EvalError> {
+                                InternalSnafu {
+                                    reason: "Source Batch Channel is closed".to_string(),
+                                }
+                                .fail()
+                            });
+                            break;
+                        }
+                    }
+                }
+
+                let now = *now.borrow();
+                // always schedule source to run at now so we can
+                // repeatedly run source if needed
+                inner_schd.schedule_at(now);
+            });
+        schd.set_cur_subgraph(sub);
+        let bundle = CollectionBundle::from_collection(Collection::<Batch>::from_port(recv_port));
+        Ok(bundle)
+    }
+
    /// Render a source which comes from brocast channel into the dataflow
    /// will immediately send updates not greater than `now` and buffer the rest in arrangement
    pub fn render_source(
@@ -114,6 +170,32 @@ impl<'referred, 'df> Context<'referred, 'df> {
        })
    }

+    pub fn render_unbounded_sink_batch(
+        &mut self,
+        bundle: CollectionBundle<Batch>,
+        sender: mpsc::UnboundedSender<Batch>,
+    ) {
+        let CollectionBundle {
+            collection,
+            arranged: _,
+        } = bundle;
+
+        let _sink = self.df.add_subgraph_sink(
+            "UnboundedSinkBatch",
+            collection.into_inner(),
+            move |_ctx, recv| {
+                let data = recv.take_inner();
+                for batch in data.into_iter().flat_map(|i| i.into_iter()) {
+                    // if the sender is closed unexpectedly, stop sending
+                    if sender.is_closed() || sender.send(batch).is_err() {
+                        common_telemetry::error!("UnboundedSinkBatch is closed");
+                        break;
+                    }
+                }
+            },
+        );
+    }
+
    pub fn render_unbounded_sink(
        &mut self,
        bundle: CollectionBundle,
--- a/src/flow/src/compute/types.rs
+++ b/src/flow/src/compute/types.rs
@@ -105,11 +105,13 @@ impl Arranged {
 /// This type maintains the invariant that it does contain at least one(or both) valid
 /// source of data, either a collection or at least one arrangement. This is for convenience
 /// of reading the data from the collection.
-pub struct CollectionBundle {
+///
+// TODO(discord9): make T default to Batch and obsolete the Row Mode
+pub struct CollectionBundle<T: 'static = DiffRow> {
    /// This is useful for passively reading the new updates from the collection
    ///
    /// Invariant: the timestamp of the updates should always not greater than now, since future updates should be stored in the arrangement
-    pub collection: Collection<DiffRow>,
+    pub collection: Collection<T>,
    /// the key [`ScalarExpr`] indicate how the keys(also a [`Row`]) used in Arranged is extract from collection's [`Row`]
    /// So it is the "index" of the arrangement
    ///
@@ -121,13 +123,16 @@ pub struct CollectionBundle {
    pub arranged: BTreeMap<Vec<ScalarExpr>, Arranged>,
 }

-impl CollectionBundle {
-    pub fn from_collection(collection: Collection<DiffRow>) -> Self {
+impl<T: 'static> CollectionBundle<T> {
+    pub fn from_collection(collection: Collection<T>) -> Self {
        Self {
            collection,
            arranged: BTreeMap::default(),
        }
    }
+}
+
+impl<T: 'static + Clone> CollectionBundle<T> {
    pub fn clone(&self, df: &mut Hydroflow) -> Self {
        Self {
            collection: self.collection.clone(df),
--- a/src/flow/src/df_optimizer.rs
+++ b/src/flow/src/df_optimizer.rs
@@ -0,0 +1,604 @@
+// 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.
+
+//! Datafusion optimizer for flow plan
+
+#![warn(unused)]
+
+use std::collections::{HashMap, HashSet};
+use std::sync::Arc;
+
+use common_error::ext::BoxedError;
+use common_telemetry::debug;
+use datafusion::config::ConfigOptions;
+use datafusion::error::DataFusionError;
+use datafusion::optimizer::analyzer::type_coercion::TypeCoercion;
+use datafusion::optimizer::common_subexpr_eliminate::CommonSubexprEliminate;
+use datafusion::optimizer::optimize_projections::OptimizeProjections;
+use datafusion::optimizer::simplify_expressions::SimplifyExpressions;
+use datafusion::optimizer::unwrap_cast_in_comparison::UnwrapCastInComparison;
+use datafusion::optimizer::utils::NamePreserver;
+use datafusion::optimizer::{Analyzer, AnalyzerRule, Optimizer, OptimizerContext};
+use datafusion_common::tree_node::{
+    Transformed, TreeNode, TreeNodeRecursion, TreeNodeRewriter, TreeNodeVisitor,
+};
+use datafusion_common::{Column, DFSchema, ScalarValue};
+use datafusion_expr::aggregate_function::AggregateFunction;
+use datafusion_expr::expr::AggregateFunctionDefinition;
+use datafusion_expr::utils::merge_schema;
+use datafusion_expr::{
+    BinaryExpr, Expr, Operator, Projection, ScalarUDFImpl, Signature, TypeSignature, Volatility,
+};
+use query::parser::QueryLanguageParser;
+use query::plan::LogicalPlan;
+use query::query_engine::DefaultSerializer;
+use query::QueryEngine;
+use snafu::ResultExt;
+/// note here we are using the `substrait_proto_df` crate from the `substrait` module and
+/// rename it to `substrait_proto`
+use substrait::DFLogicalSubstraitConvertor;
+
+use crate::adapter::FlownodeContext;
+use crate::error::{DatafusionSnafu, Error, ExternalSnafu, UnexpectedSnafu};
+use crate::expr::{TUMBLE_END, TUMBLE_START};
+use crate::plan::TypedPlan;
+
+// TODO(discord9): use `Analyzer` to manage rules if more `AnalyzerRule` is needed
+pub async fn apply_df_optimizer(
+    plan: datafusion_expr::LogicalPlan,
+) -> Result<datafusion_expr::LogicalPlan, Error> {
+    let cfg = ConfigOptions::new();
+    let analyzer = Analyzer::with_rules(vec![
+        Arc::new(AvgExpandRule::new()),
+        Arc::new(TumbleExpandRule::new()),
+        Arc::new(CheckGroupByRule::new()),
+        Arc::new(TypeCoercion::new()),
+    ]);
+    let plan = analyzer
+        .execute_and_check(plan, &cfg, |p, r| {
+            debug!("After apply rule {}, get plan: \n{:?}", r.name(), p);
+        })
+        .context(DatafusionSnafu {
+            context: "Fail to apply analyzer",
+        })?;
+
+    let ctx = OptimizerContext::new();
+    let optimizer = Optimizer::with_rules(vec![
+        Arc::new(OptimizeProjections::new()),
+        Arc::new(CommonSubexprEliminate::new()),
+        Arc::new(SimplifyExpressions::new()),
+        Arc::new(UnwrapCastInComparison::new()),
+    ]);
+    let plan = optimizer
+        .optimize(plan, &ctx, |_, _| {})
+        .context(DatafusionSnafu {
+            context: "Fail to apply optimizer",
+        })?;
+
+    Ok(plan)
+}
+
+/// To reuse existing code for parse sql, the sql is first parsed into a datafusion logical plan,
+/// then to a substrait plan, and finally to a flow plan.
+pub async fn sql_to_flow_plan(
+    ctx: &mut FlownodeContext,
+    engine: &Arc<dyn QueryEngine>,
+    sql: &str,
+) -> Result<TypedPlan, Error> {
+    let query_ctx = ctx.query_context.clone().ok_or_else(|| {
+        UnexpectedSnafu {
+            reason: "Query context is missing",
+        }
+        .build()
+    })?;
+    let stmt = QueryLanguageParser::parse_sql(sql, &query_ctx)
+        .map_err(BoxedError::new)
+        .context(ExternalSnafu)?;
+    let plan = engine
+        .planner()
+        .plan(stmt, query_ctx)
+        .await
+        .map_err(BoxedError::new)
+        .context(ExternalSnafu)?;
+    let LogicalPlan::DfPlan(plan) = plan;
+
+    let opted_plan = apply_df_optimizer(plan).await?;
+
+    // TODO(discord9): add df optimization
+    let sub_plan = DFLogicalSubstraitConvertor {}
+        .to_sub_plan(&opted_plan, DefaultSerializer)
+        .map_err(BoxedError::new)
+        .context(ExternalSnafu)?;
+
+    let flow_plan = TypedPlan::from_substrait_plan(ctx, &sub_plan).await?;
+
+    Ok(flow_plan)
+}
+
+struct AvgExpandRule {}
+
+impl AvgExpandRule {
+    pub fn new() -> Self {
+        Self {}
+    }
+}
+
+impl AnalyzerRule for AvgExpandRule {
+    fn analyze(
+        &self,
+        plan: datafusion_expr::LogicalPlan,
+        _config: &ConfigOptions,
+    ) -> datafusion_common::Result<datafusion_expr::LogicalPlan> {
+        let transformed = plan
+            .transform_up_with_subqueries(expand_avg_analyzer)?
+            .data
+            .transform_down_with_subqueries(put_aggr_to_proj_analyzer)?
+            .data;
+        Ok(transformed)
+    }
+
+    fn name(&self) -> &str {
+        "avg_expand"
+    }
+}
+
+/// lift aggr's composite aggr_expr to outer proj, and leave aggr only with simple direct aggr expr
+/// i.e.
+/// ```ignore
+/// proj: avg(x)
+/// -- aggr: [sum(x)/count(x) as avg(x)]
+/// ```
+/// becomes:
+/// ```ignore
+/// proj: sum(x)/count(x) as avg(x)
+/// -- aggr: [sum(x), count(x)]
+/// ```
+fn put_aggr_to_proj_analyzer(
+    plan: datafusion_expr::LogicalPlan,
+) -> Result<Transformed<datafusion_expr::LogicalPlan>, DataFusionError> {
+    if let datafusion_expr::LogicalPlan::Projection(proj) = &plan {
+        if let datafusion_expr::LogicalPlan::Aggregate(aggr) = proj.input.as_ref() {
+            let mut replace_old_proj_exprs = HashMap::new();
+            let mut expanded_aggr_exprs = vec![];
+            for aggr_expr in &aggr.aggr_expr {
+                let mut is_composite = false;
+                if let Expr::AggregateFunction(_) = &aggr_expr {
+                    expanded_aggr_exprs.push(aggr_expr.clone());
+                } else {
+                    let old_name = aggr_expr.name_for_alias()?;
+                    let new_proj_expr = aggr_expr
+                        .clone()
+                        .transform(|ch| {
+                            if let Expr::AggregateFunction(_) = &ch {
+                                is_composite = true;
+                                expanded_aggr_exprs.push(ch.clone());
+                                Ok(Transformed::yes(Expr::Column(Column::from_qualified_name(
+                                    ch.name_for_alias()?,
+                                ))))
+                            } else {
+                                Ok(Transformed::no(ch))
+                            }
+                        })?
+                        .data;
+                    replace_old_proj_exprs.insert(old_name, new_proj_expr);
+                }
+            }
+
+            if expanded_aggr_exprs.len() > aggr.aggr_expr.len() {
+                let mut aggr = aggr.clone();
+                aggr.aggr_expr = expanded_aggr_exprs;
+                let mut aggr_plan = datafusion_expr::LogicalPlan::Aggregate(aggr);
+                // important to recompute schema after changing aggr_expr
+                aggr_plan = aggr_plan.recompute_schema()?;
+
+                // reconstruct proj with new proj_exprs
+                let mut new_proj_exprs = proj.expr.clone();
+                for proj_expr in new_proj_exprs.iter_mut() {
+                    if let Some(new_proj_expr) =
+                        replace_old_proj_exprs.get(&proj_expr.name_for_alias()?)
+                    {
+                        *proj_expr = new_proj_expr.clone();
+                    }
+                    *proj_expr = proj_expr
+                        .clone()
+                        .transform(|expr| {
+                            if let Some(new_expr) =
+                                replace_old_proj_exprs.get(&expr.name_for_alias()?)
+                            {
+                                Ok(Transformed::yes(new_expr.clone()))
+                            } else {
+                                Ok(Transformed::no(expr))
+                            }
+                        })?
+                        .data;
+                }
+                let proj = datafusion_expr::LogicalPlan::Projection(Projection::try_new(
+                    new_proj_exprs,
+                    Arc::new(aggr_plan),
+                )?);
+                return Ok(Transformed::yes(proj));
+            }
+        }
+    }
+    Ok(Transformed::no(plan))
+}
+
+/// expand `avg(<expr>)` function into `cast(sum((<expr>) AS f64)/count((<expr>)`
+fn expand_avg_analyzer(
+    plan: datafusion_expr::LogicalPlan,
+) -> Result<Transformed<datafusion_expr::LogicalPlan>, DataFusionError> {
+    let mut schema = merge_schema(plan.inputs());
+
+    if let datafusion_expr::LogicalPlan::TableScan(ts) = &plan {
+        let source_schema =
+            DFSchema::try_from_qualified_schema(ts.table_name.clone(), &ts.source.schema())?;
+        schema.merge(&source_schema);
+    }
+
+    let mut expr_rewrite = ExpandAvgRewriter::new(&schema);
+
+    let name_preserver = NamePreserver::new(&plan);
+    // apply coercion rewrite all expressions in the plan individually
+    plan.map_expressions(|expr| {
+        let original_name = name_preserver.save(&expr)?;
+        expr.rewrite(&mut expr_rewrite)?
+            .map_data(|expr| original_name.restore(expr))
+    })?
+    .map_data(|plan| plan.recompute_schema())
+}
+
+/// rewrite `avg(<expr>)` function into `CASE WHEN count(<expr>) !=0 THEN  cast(sum((<expr>) AS avg_return_type)/count((<expr>) ELSE 0`
+///
+/// TODO(discord9): support avg return type decimal128
+///
+/// see impl details at https://github.com/apache/datafusion/blob/4ad4f90d86c57226a4e0fb1f79dfaaf0d404c273/datafusion/expr/src/type_coercion/aggregates.rs#L457-L462
+pub(crate) struct ExpandAvgRewriter<'a> {
+    /// schema of the plan
+    #[allow(unused)]
+    pub(crate) schema: &'a DFSchema,
+}
+
+impl<'a> ExpandAvgRewriter<'a> {
+    fn new(schema: &'a DFSchema) -> Self {
+        Self { schema }
+    }
+}
+
+impl<'a> TreeNodeRewriter for ExpandAvgRewriter<'a> {
+    type Node = Expr;
+
+    fn f_up(&mut self, expr: Expr) -> Result<Transformed<Expr>, DataFusionError> {
+        if let Expr::AggregateFunction(aggr_func) = &expr {
+            if let AggregateFunctionDefinition::BuiltIn(AggregateFunction::Avg) =
+                &aggr_func.func_def
+            {
+                let sum_expr = {
+                    let mut tmp = aggr_func.clone();
+                    tmp.func_def = AggregateFunctionDefinition::BuiltIn(AggregateFunction::Sum);
+                    Expr::AggregateFunction(tmp)
+                };
+                let sum_cast = {
+                    let mut tmp = sum_expr.clone();
+                    tmp = Expr::Cast(datafusion_expr::Cast {
+                        expr: Box::new(tmp),
+                        data_type: arrow_schema::DataType::Float64,
+                    });
+                    tmp
+                };
+
+                let count_expr = {
+                    let mut tmp = aggr_func.clone();
+                    tmp.func_def = AggregateFunctionDefinition::BuiltIn(AggregateFunction::Count);
+
+                    Expr::AggregateFunction(tmp)
+                };
+                let count_expr_ref =
+                    Expr::Column(Column::from_qualified_name(count_expr.name_for_alias()?));
+
+                let div =
+                    BinaryExpr::new(Box::new(sum_cast), Operator::Divide, Box::new(count_expr));
+                let div_expr = Box::new(Expr::BinaryExpr(div));
+
+                let zero = Box::new(Expr::Literal(ScalarValue::Int64(Some(0))));
+                let not_zero =
+                    BinaryExpr::new(Box::new(count_expr_ref), Operator::NotEq, zero.clone());
+                let not_zero = Box::new(Expr::BinaryExpr(not_zero));
+                let null = Box::new(Expr::Literal(ScalarValue::Null));
+
+                let case_when =
+                    datafusion_expr::Case::new(None, vec![(not_zero, div_expr)], Some(null));
+                let case_when_expr = Expr::Case(case_when);
+
+                return Ok(Transformed::yes(case_when_expr));
+            }
+        }
+
+        Ok(Transformed::no(expr))
+    }
+}
+
+/// expand tumble in aggr expr to tumble_start and tumble_end with column name like `window_start`
+struct TumbleExpandRule {}
+
+impl TumbleExpandRule {
+    pub fn new() -> Self {
+        Self {}
+    }
+}
+
+impl AnalyzerRule for TumbleExpandRule {
+    fn analyze(
+        &self,
+        plan: datafusion_expr::LogicalPlan,
+        _config: &ConfigOptions,
+    ) -> datafusion_common::Result<datafusion_expr::LogicalPlan> {
+        let transformed = plan
+            .transform_up_with_subqueries(expand_tumble_analyzer)?
+            .data;
+        Ok(transformed)
+    }
+
+    fn name(&self) -> &str {
+        "tumble_expand"
+    }
+}
+
+/// expand `tumble` in aggr expr to `tumble_start` and `tumble_end`, also expand related alias and column ref
+///
+/// will add `tumble_start` and `tumble_end` to outer projection if not exist before
+fn expand_tumble_analyzer(
+    plan: datafusion_expr::LogicalPlan,
+) -> Result<Transformed<datafusion_expr::LogicalPlan>, DataFusionError> {
+    if let datafusion_expr::LogicalPlan::Projection(proj) = &plan {
+        if let datafusion_expr::LogicalPlan::Aggregate(aggr) = proj.input.as_ref() {
+            let mut new_group_expr = vec![];
+            let mut alias_to_expand = HashMap::new();
+            let mut encountered_tumble = false;
+            for expr in aggr.group_expr.iter() {
+                match expr {
+                    datafusion_expr::Expr::ScalarFunction(func) if func.name() == "tumble" => {
+                        encountered_tumble = true;
+
+                        let tumble_start = TumbleExpand::new(TUMBLE_START);
+                        let tumble_start = datafusion_expr::expr::ScalarFunction::new_udf(
+                            Arc::new(tumble_start.into()),
+                            func.args.clone(),
+                        );
+                        let tumble_start = datafusion_expr::Expr::ScalarFunction(tumble_start);
+                        let start_col_name = tumble_start.name_for_alias()?;
+                        new_group_expr.push(tumble_start);
+
+                        let tumble_end = TumbleExpand::new(TUMBLE_END);
+                        let tumble_end = datafusion_expr::expr::ScalarFunction::new_udf(
+                            Arc::new(tumble_end.into()),
+                            func.args.clone(),
+                        );
+                        let tumble_end = datafusion_expr::Expr::ScalarFunction(tumble_end);
+                        let end_col_name = tumble_end.name_for_alias()?;
+                        new_group_expr.push(tumble_end);
+
+                        alias_to_expand
+                            .insert(expr.name_for_alias()?, (start_col_name, end_col_name));
+                    }
+                    _ => new_group_expr.push(expr.clone()),
+                }
+            }
+            if !encountered_tumble {
+                return Ok(Transformed::no(plan));
+            }
+            let mut new_aggr = aggr.clone();
+            new_aggr.group_expr = new_group_expr;
+            let new_aggr = datafusion_expr::LogicalPlan::Aggregate(new_aggr).recompute_schema()?;
+            // replace alias in projection if needed, and add new column ref if necessary
+            let mut new_proj_expr = vec![];
+            let mut have_expanded = false;
+
+            for proj_expr in proj.expr.iter() {
+                if let Some((start_col_name, end_col_name)) =
+                    alias_to_expand.get(&proj_expr.name_for_alias()?)
+                {
+                    let start_col = Column::from_qualified_name(start_col_name);
+                    let end_col = Column::from_qualified_name(end_col_name);
+                    new_proj_expr.push(datafusion_expr::Expr::Column(start_col));
+                    new_proj_expr.push(datafusion_expr::Expr::Column(end_col));
+                    have_expanded = true;
+                } else {
+                    new_proj_expr.push(proj_expr.clone());
+                }
+            }
+
+            // append to end of projection if not exist
+            if !have_expanded {
+                for (start_col_name, end_col_name) in alias_to_expand.values() {
+                    let start_col = Column::from_qualified_name(start_col_name);
+                    let end_col = Column::from_qualified_name(end_col_name);
+                    new_proj_expr
+                        .push(datafusion_expr::Expr::Column(start_col).alias("window_start"));
+                    new_proj_expr.push(datafusion_expr::Expr::Column(end_col).alias("window_end"));
+                }
+            }
+
+            let new_proj = datafusion_expr::LogicalPlan::Projection(Projection::try_new(
+                new_proj_expr,
+                Arc::new(new_aggr),
+            )?);
+            return Ok(Transformed::yes(new_proj));
+        }
+    }
+
+    Ok(Transformed::no(plan))
+}
+
+/// This is a placeholder for tumble_start and tumble_end function, so that datafusion can
+/// recognize them as scalar function
+#[derive(Debug)]
+pub struct TumbleExpand {
+    signature: Signature,
+    name: String,
+}
+
+impl TumbleExpand {
+    pub fn new(name: &str) -> Self {
+        Self {
+            signature: Signature::new(TypeSignature::UserDefined, Volatility::Immutable),
+            name: name.to_string(),
+        }
+    }
+}
+
+impl ScalarUDFImpl for TumbleExpand {
+    fn as_any(&self) -> &dyn std::any::Any {
+        self
+    }
+
+    fn name(&self) -> &str {
+        &self.name
+    }
+
+    /// elide the signature for now
+    fn signature(&self) -> &Signature {
+        &self.signature
+    }
+
+    fn coerce_types(
+        &self,
+        arg_types: &[arrow_schema::DataType],
+    ) -> datafusion_common::Result<Vec<arrow_schema::DataType>> {
+        match (arg_types.first(), arg_types.get(1), arg_types.get(2)) {
+            (Some(ts), Some(window), opt) => {
+                use arrow_schema::DataType::*;
+                if !matches!(ts, Date32 | Date64 | Timestamp(_, _)) {
+                    return Err(DataFusionError::Plan(
+                        format!("Expect timestamp column as first arg for tumble_start, found {:?}", ts)
+                    ));
+                }
+                if !matches!(window, Utf8 | Interval(_)) {
+                    return Err(DataFusionError::Plan(
+                        format!("Expect second arg for window size's type being interval for tumble_start, found {:?}", window),
+                    ));
+                }
+
+                if let Some(start_time) = opt{
+                    if !matches!(start_time,  Utf8 | Date32 | Date64 | Timestamp(_, _)){
+                        return Err(DataFusionError::Plan(
+                            format!("Expect start_time to either be date, timestampe or string, found {:?}", start_time)
+                        ));
+                    }
+                }
+
+                Ok(arg_types.to_vec())
+            }
+            _ => Err(DataFusionError::Plan(
+                "Expect tumble function have at least two arg(timestamp column and window size) and a third optional arg for starting time".to_string(),
+            )),
+        }
+    }
+
+    fn return_type(
+        &self,
+        arg_types: &[arrow_schema::DataType],
+    ) -> Result<arrow_schema::DataType, DataFusionError> {
+        arg_types.first().cloned().ok_or_else(|| {
+            DataFusionError::Plan(
+                "Expect tumble function have at least two arg(timestamp column and window size)"
+                    .to_string(),
+            )
+        })
+    }
+
+    fn invoke(
+        &self,
+        _args: &[datafusion_expr::ColumnarValue],
+    ) -> Result<datafusion_expr::ColumnarValue, DataFusionError> {
+        Err(DataFusionError::Plan(
+            "This function should not be executed by datafusion".to_string(),
+        ))
+    }
+}
+
+/// This rule check all group by exprs, and make sure they are also in select clause in a aggr query
+struct CheckGroupByRule {}
+
+impl CheckGroupByRule {
+    pub fn new() -> Self {
+        Self {}
+    }
+}
+
+impl AnalyzerRule for CheckGroupByRule {
+    fn analyze(
+        &self,
+        plan: datafusion_expr::LogicalPlan,
+        _config: &ConfigOptions,
+    ) -> datafusion_common::Result<datafusion_expr::LogicalPlan> {
+        let transformed = plan
+            .transform_up_with_subqueries(check_group_by_analyzer)?
+            .data;
+        Ok(transformed)
+    }
+
+    fn name(&self) -> &str {
+        "check_groupby"
+    }
+}
+
+/// make sure everything in group by's expr is in select
+fn check_group_by_analyzer(
+    plan: datafusion_expr::LogicalPlan,
+) -> Result<Transformed<datafusion_expr::LogicalPlan>, DataFusionError> {
+    if let datafusion_expr::LogicalPlan::Projection(proj) = &plan {
+        if let datafusion_expr::LogicalPlan::Aggregate(aggr) = proj.input.as_ref() {
+            let mut found_column_used = FindColumn::new();
+            proj.expr
+                .iter()
+                .map(|i| i.visit(&mut found_column_used))
+                .count();
+            for expr in aggr.group_expr.iter() {
+                if !found_column_used
+                    .names_for_alias
+                    .contains(&expr.name_for_alias()?)
+                {
+                    return Err(DataFusionError::Plan(format!("Expect {} expr in group by also exist in select list, but select list only contain {:?}",expr.name_for_alias()?, found_column_used.names_for_alias)));
+                }
+            }
+        }
+    }
+
+    Ok(Transformed::no(plan))
+}
+
+/// Find all column names in a plan
+#[derive(Debug, Default)]
+struct FindColumn {
+    names_for_alias: HashSet<String>,
+}
+
+impl FindColumn {
+    fn new() -> Self {
+        Default::default()
+    }
+}
+
+impl TreeNodeVisitor<'_> for FindColumn {
+    type Node = datafusion_expr::Expr;
+    fn f_down(
+        &mut self,
+        node: &datafusion_expr::Expr,
+    ) -> Result<TreeNodeRecursion, DataFusionError> {
+        if let datafusion_expr::Expr::Column(_) = node {
+            self.names_for_alias.insert(node.name_for_alias()?);
+        }
+        Ok(TreeNodeRecursion::Continue)
+    }
+}
--- a/src/flow/src/error.rs
+++ b/src/flow/src/error.rs
@@ -21,7 +21,6 @@ use common_error::ext::BoxedError;
 use common_macro::stack_trace_debug;
 use common_telemetry::common_error::ext::ErrorExt;
 use common_telemetry::common_error::status_code::StatusCode;
-use datatypes::value::Value;
 use snafu::{Location, Snafu};

 use crate::adapter::FlowId;
@@ -69,13 +68,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Table already exist: {name}"))]
-    TableAlreadyExist {
-        name: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Flow not found, id={id}"))]
    FlowNotFound {
        id: FlowId,
@@ -106,13 +98,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Invalid query: prost can't decode substrait plan: {inner}"))]
-    InvalidQueryProst {
-        inner: api::DecodeError,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Invalid query: {reason}"))]
    InvalidQuery {
        reason: String,
@@ -120,13 +105,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("No protobuf type for value: {value}"))]
-    NoProtoType {
-        value: Value,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Not implement in flow: {reason}"))]
    NotImplemented {
        reason: String,
@@ -214,21 +192,18 @@ pub type Result<T> = std::result::Result<T, Error>;
 impl ErrorExt for Error {
    fn status_code(&self) -> StatusCode {
        match self {
-            Self::Eval { .. } | &Self::JoinTask { .. } | &Self::Datafusion { .. } => {
+            Self::Eval { .. } | Self::JoinTask { .. } | Self::Datafusion { .. } => {
                StatusCode::Internal
            }
-            &Self::TableAlreadyExist { .. } | Self::FlowAlreadyExist { .. } => {
-                StatusCode::TableAlreadyExists
-            }
+            Self::FlowAlreadyExist { .. } => StatusCode::TableAlreadyExists,
            Self::TableNotFound { .. }
            | Self::TableNotFoundMeta { .. }
            | Self::FlowNotFound { .. }
            | Self::ListFlows { .. } => StatusCode::TableNotFound,
-            Self::InvalidQueryProst { .. }
-            | &Self::InvalidQuery { .. }
-            | &Self::Plan { .. }
-            | &Self::Datatypes { .. } => StatusCode::PlanQuery,
-            Self::NoProtoType { .. } | Self::Unexpected { .. } => StatusCode::Unexpected,
+            Self::InvalidQuery { .. } | Self::Plan { .. } | Self::Datatypes { .. } => {
+                StatusCode::PlanQuery
+            }
+            Self::Unexpected { .. } => StatusCode::Unexpected,
            Self::NotImplemented { .. } | Self::UnsupportedTemporalFilter { .. } => {
                StatusCode::Unsupported
            }
--- a/src/flow/src/expr.rs
+++ b/src/flow/src/expr.rs
@@ -24,6 +24,7 @@ mod scalar;
 mod signature;

 use datatypes::prelude::DataType;
+use datatypes::value::Value;
 use datatypes::vectors::VectorRef;
 pub(crate) use df_func::{DfScalarFunction, RawDfScalarFn};
 pub(crate) use error::{EvalError, InvalidArgumentSnafu};
@@ -37,42 +38,168 @@ use snafu::{ensure, ResultExt};

 use crate::expr::error::DataTypeSnafu;

+pub const TUMBLE_START: &str = "tumble_start";
+pub const TUMBLE_END: &str = "tumble_end";
+
 /// A batch of vectors with the same length but without schema, only useful in dataflow
+///
+/// somewhere cheap to clone since it just contains a list of VectorRef(which is a `Arc`).
+#[derive(Debug, Clone)]
 pub struct Batch {
    batch: Vec<VectorRef>,
    row_count: usize,
+    /// describe if corresponding rows in batch is insert or delete, None means all rows are insert
+    diffs: Option<VectorRef>,
+}
+
+impl PartialEq for Batch {
+    fn eq(&self, other: &Self) -> bool {
+        let mut batch_eq = true;
+        if self.batch.len() != other.batch.len() {
+            return false;
+        }
+        for (left, right) in self.batch.iter().zip(other.batch.iter()) {
+            batch_eq = batch_eq
+                && <dyn arrow::array::Array>::eq(&left.to_arrow_array(), &right.to_arrow_array());
+        }
+
+        let diff_eq = match (&self.diffs, &other.diffs) {
+            (Some(left), Some(right)) => {
+                <dyn arrow::array::Array>::eq(&left.to_arrow_array(), &right.to_arrow_array())
+            }
+            (None, None) => true,
+            _ => false,
+        };
+        batch_eq && diff_eq && self.row_count == other.row_count
+    }
+}
+
+impl Eq for Batch {}
+
+impl Default for Batch {
+    fn default() -> Self {
+        Self::empty()
+    }
 }

 impl Batch {
-    pub fn new(batch: Vec<VectorRef>, row_count: usize) -> Self {
-        Self { batch, row_count }
+    pub fn try_from_rows(rows: Vec<crate::repr::Row>) -> Result<Self, EvalError> {
+        if rows.is_empty() {
+            return Ok(Self::empty());
+        }
+        let len = rows.len();
+        let mut builder = rows
+            .first()
+            .unwrap()
+            .iter()
+            .map(|v| v.data_type().create_mutable_vector(len))
+            .collect_vec();
+        for row in rows {
+            ensure!(
+                row.len() == builder.len(),
+                InvalidArgumentSnafu {
+                    reason: format!(
+                        "row length not match, expect {}, found {}",
+                        builder.len(),
+                        row.len()
+                    )
+                }
+            );
+            for (idx, value) in row.iter().enumerate() {
+                builder[idx]
+                    .try_push_value_ref(value.as_value_ref())
+                    .context(DataTypeSnafu {
+                        msg: "Failed to convert rows to columns",
+                    })?;
+            }
+        }
+
+        let columns = builder.into_iter().map(|mut b| b.to_vector()).collect_vec();
+        let batch = Self::try_new(columns, len)?;
+        Ok(batch)
+    }
+
+    pub fn empty() -> Self {
+        Self {
+            batch: vec![],
+            row_count: 0,
+            diffs: None,
+        }
+    }
+    pub fn try_new(batch: Vec<VectorRef>, row_count: usize) -> Result<Self, EvalError> {
+        ensure!(
+            batch.iter().map(|v| v.len()).all_equal()
+                && batch.first().map(|v| v.len() == row_count).unwrap_or(true),
+            InvalidArgumentSnafu {
+                reason: "All columns should have same length".to_string()
+            }
+        );
+        Ok(Self {
+            batch,
+            row_count,
+            diffs: None,
+        })
+    }
+
+    pub fn new_unchecked(batch: Vec<VectorRef>, row_count: usize) -> Self {
+        Self {
+            batch,
+            row_count,
+            diffs: None,
+        }
    }

    pub fn batch(&self) -> &[VectorRef] {
        &self.batch
    }

+    pub fn batch_mut(&mut self) -> &mut Vec<VectorRef> {
+        &mut self.batch
+    }
+
    pub fn row_count(&self) -> usize {
        self.row_count
    }

+    pub fn set_row_count(&mut self, row_count: usize) {
+        self.row_count = row_count;
+    }
+
+    pub fn column_count(&self) -> usize {
+        self.batch.len()
+    }
+
+    pub fn get_row(&self, idx: usize) -> Result<Vec<Value>, EvalError> {
+        ensure!(
+            idx < self.row_count,
+            InvalidArgumentSnafu {
+                reason: format!(
+                    "Expect row index to be less than {}, found {}",
+                    self.row_count, idx
+                )
+            }
+        );
+        Ok(self.batch.iter().map(|v| v.get(idx)).collect_vec())
+    }
+
    /// Slices the `Batch`, returning a new `Batch`.
-    ///
-    /// # Panics
-    /// This function panics if `offset + length > self.row_count()`.
-    pub fn slice(&self, offset: usize, length: usize) -> Batch {
+    pub fn slice(&self, offset: usize, length: usize) -> Result<Batch, EvalError> {
        let batch = self
            .batch()
            .iter()
            .map(|v| v.slice(offset, length))
            .collect_vec();
-        Batch::new(batch, length)
+        Batch::try_new(batch, length)
    }

    /// append another batch to self
+    ///
+    /// NOTE: This is expensive since it will create new vectors for each column
    pub fn append_batch(&mut self, other: Batch) -> Result<(), EvalError> {
        ensure!(
-            self.batch.len() == other.batch.len(),
+            self.batch.len() == other.batch.len()
+                || self.batch.is_empty()
+                || other.batch.is_empty(),
            InvalidArgumentSnafu {
                reason: format!(
                    "Expect two batch to have same numbers of column, found {} and {} columns",
@@ -82,21 +209,31 @@ impl Batch {
            }
        );

-        let batch_builders = self
-            .batch
+        if self.batch.is_empty() {
+            self.batch = other.batch;
+            self.row_count = other.row_count;
+            return Ok(());
+        } else if other.batch.is_empty() {
+            return Ok(());
+        }
+
+        let dts = if self.batch.is_empty() {
+            other.batch.iter().map(|v| v.data_type()).collect_vec()
+        } else {
+            self.batch.iter().map(|v| v.data_type()).collect_vec()
+        };
+
+        let batch_builders = dts
            .iter()
-            .map(|v| {
-                v.data_type()
-                    .create_mutable_vector(self.row_count() + other.row_count())
-            })
+            .map(|dt| dt.create_mutable_vector(self.row_count() + other.row_count()))
            .collect_vec();

        let mut result = vec![];
-        let zelf_row_count = self.row_count();
+        let self_row_count = self.row_count();
        let other_row_count = other.row_count();
        for (idx, mut builder) in batch_builders.into_iter().enumerate() {
            builder
-                .extend_slice_of(self.batch()[idx].as_ref(), 0, zelf_row_count)
+                .extend_slice_of(self.batch()[idx].as_ref(), 0, self_row_count)
                .context(DataTypeSnafu {
                    msg: "Failed to extend vector",
                })?;
@@ -108,7 +245,7 @@ impl Batch {
            result.push(builder.to_vector());
        }
        self.batch = result;
-        self.row_count = zelf_row_count + other_row_count;
+        self.row_count = self_row_count + other_row_count;
        Ok(())
    }
 }
--- a/src/flow/src/expr/func.rs
+++ b/src/flow/src/expr/func.rs
@@ -35,13 +35,13 @@ use snafu::{ensure, OptionExt, ResultExt};
 use strum::{EnumIter, IntoEnumIterator};
 use substrait::df_logical_plan::consumer::name_to_op;

-use crate::error::{Error, ExternalSnafu, InvalidQuerySnafu, PlanSnafu};
+use crate::error::{Error, ExternalSnafu, InvalidQuerySnafu, PlanSnafu, UnexpectedSnafu};
 use crate::expr::error::{
    ArrowSnafu, CastValueSnafu, DataTypeSnafu, DivisionByZeroSnafu, EvalError, OverflowSnafu,
    TryFromValueSnafu, TypeMismatchSnafu,
 };
 use crate::expr::signature::{GenericFn, Signature};
-use crate::expr::{Batch, InvalidArgumentSnafu, ScalarExpr, TypedExpr};
+use crate::expr::{Batch, InvalidArgumentSnafu, ScalarExpr, TypedExpr, TUMBLE_END, TUMBLE_START};
 use crate::repr::{self, value_to_internal_ts};

 /// UnmaterializableFunc is a function that can't be eval independently,
@@ -87,42 +87,10 @@ impl UnmaterializableFunc {
    }

    /// Create a UnmaterializableFunc from a string of the function name
-    pub fn from_str_args(name: &str, args: Vec<TypedExpr>) -> Result<Self, Error> {
+    pub fn from_str_args(name: &str, _args: Vec<TypedExpr>) -> Result<Self, Error> {
        match name.to_lowercase().as_str() {
            "now" => Ok(Self::Now),
            "current_schema" => Ok(Self::CurrentSchema),
-            "tumble" => {
-                let ts = args.first().context(InvalidQuerySnafu {
-                    reason: "Tumble window function requires a timestamp argument",
-                })?;
-                let window_size = args
-                    .get(1)
-                    .and_then(|expr| expr.expr.as_literal())
-                    .context(InvalidQuerySnafu {
-                        reason: "Tumble window function requires a window size argument"
-                    })?.as_string() // TODO(discord9): since df to substrait convertor does not support interval type yet, we need to take a string and cast it to interval instead
-                    .map(|s|cast(Value::from(s), &ConcreteDataType::interval_month_day_nano_datatype())).transpose().map_err(BoxedError::new).context(
-                        ExternalSnafu
-                    )?.and_then(|v|v.as_interval())
-                    .with_context(||InvalidQuerySnafu {
-                        reason: format!("Tumble window function requires window size argument to be a string describe a interval, found {:?}", args.get(1))
-                    })?;
-                let start_time = match args.get(2) {
-                    Some(start_time) => start_time.expr.as_literal(),
-                    None => None,
-                }
-                .map(|s| cast(s.clone(), &ConcreteDataType::datetime_datatype())).transpose().map_err(BoxedError::new).context(ExternalSnafu)?.map(|v|v.as_datetime().with_context(
-                    ||InvalidQuerySnafu {
-                        reason: format!("Tumble window function requires start time argument to be a datetime describe in string, found {:?}", args.get(2))
-                    }
-                )).transpose()?;
-
-                Ok(Self::TumbleWindow {
-                    ts: Box::new(ts.clone()),
-                    window_size,
-                    start_time,
-                })
-            }
            _ => InvalidQuerySnafu {
                reason: format!("Unknown unmaterializable function: {}", name),
            }
@@ -347,6 +315,96 @@ impl UnaryFunc {
        }
    }

+    pub fn from_tumble_func(name: &str, args: &[TypedExpr]) -> Result<(Self, TypedExpr), Error> {
+        match name.to_lowercase().as_str() {
+            TUMBLE_START | TUMBLE_END => {
+                let ts = args.first().context(InvalidQuerySnafu {
+                    reason: "Tumble window function requires a timestamp argument",
+                })?;
+                let window_size = {
+                    let window_size_untyped = args
+                        .get(1)
+                        .and_then(|expr| expr.expr.as_literal())
+                        .context(InvalidQuerySnafu {
+                        reason: "Tumble window function requires a window size argument",
+                    })?;
+                    if let Some(window_size) = window_size_untyped.as_string() {
+                        // cast as interval
+                        cast(
+                            Value::from(window_size),
+                            &ConcreteDataType::interval_month_day_nano_datatype(),
+                        )
+                        .map_err(BoxedError::new)
+                        .context(ExternalSnafu)?
+                        .as_interval()
+                        .context(UnexpectedSnafu {
+                            reason: "Expect window size arg to be interval after successful cast"
+                                .to_string(),
+                        })?
+                    } else if let Some(interval) = window_size_untyped.as_interval() {
+                        interval
+                    } else {
+                        InvalidQuerySnafu {
+                                reason: format!(
+                                    "Tumble window function requires window size argument to be either a interval or a string describe a interval, found {:?}",
+                                    window_size_untyped
+                                )
+                            }.fail()?
+                    }
+                };
+
+                // start time argument is optional
+                let start_time = match args.get(2) {
+                    Some(start_time) => {
+                        if let Some(value) = start_time.expr.as_literal() {
+                            // cast as DateTime
+                            let ret = cast(value, &ConcreteDataType::datetime_datatype())
+                                .map_err(BoxedError::new)
+                                .context(ExternalSnafu)?
+                                .as_datetime()
+                                .context(UnexpectedSnafu {
+                                    reason:
+                                        "Expect start time arg to be datetime after successful cast"
+                                            .to_string(),
+                                })?;
+                            Some(ret)
+                        } else {
+                            UnexpectedSnafu {
+                                reason: "Expect start time arg to be literal",
+                            }
+                            .fail()?
+                        }
+                    }
+                    None => None,
+                };
+
+                if name == TUMBLE_START {
+                    Ok((
+                        Self::TumbleWindowFloor {
+                            window_size,
+                            start_time,
+                        },
+                        ts.clone(),
+                    ))
+                } else if name == TUMBLE_END {
+                    Ok((
+                        Self::TumbleWindowCeiling {
+                            window_size,
+                            start_time,
+                        },
+                        ts.clone(),
+                    ))
+                } else {
+                    unreachable!()
+                }
+            }
+            _ => crate::error::InternalSnafu {
+                reason: format!("Unknown tumble kind function: {}", name),
+            }
+            .fail()?,
+        }
+    }
+
    /// Evaluate the function with given values and expression
    ///
    /// # Arguments
@@ -712,8 +770,8 @@ impl BinaryFunc {
                    t1 == t2,
                    InvalidQuerySnafu {
                        reason: format!(
-                            "Binary function {:?} requires both arguments to have the same type",
-                            generic
+                            "Binary function {:?} requires both arguments to have the same type, left={:?}, right={:?}",
+                            generic, t1, t2
                        ),
                    }
                );
--- a/src/flow/src/expr/linear.rs
+++ b/src/flow/src/expr/linear.rs
@@ -16,13 +16,18 @@

 use std::collections::{BTreeMap, BTreeSet};

+use arrow::array::BooleanArray;
+use arrow::compute::FilterBuilder;
 use common_telemetry::debug;
+use datatypes::prelude::ConcreteDataType;
 use datatypes::value::Value;
-use snafu::ensure;
+use datatypes::vectors::{BooleanVector, Helper};
+use itertools::Itertools;
+use snafu::{ensure, OptionExt, ResultExt};

 use crate::error::{Error, InvalidQuerySnafu};
-use crate::expr::error::{EvalError, InternalSnafu};
-use crate::expr::{InvalidArgumentSnafu, ScalarExpr};
+use crate::expr::error::{ArrowSnafu, DataTypeSnafu, EvalError, InternalSnafu, TypeMismatchSnafu};
+use crate::expr::{Batch, InvalidArgumentSnafu, ScalarExpr};
 use crate::repr::{self, value_to_internal_ts, Diff, Row};

 /// A compound operator that can be applied row-by-row.
@@ -473,6 +478,85 @@ impl SafeMfpPlan {
        self.mfp.permute(map, new_arity)
    }

+    /// similar to [`MapFilterProject::evaluate_into`], just in batch, and rows that don't pass the predicates are not included in the output.
+    ///
+    /// so it's not guaranteed that the output will have the same number of rows as the input.
+    pub fn eval_batch_into(&self, batch: &mut Batch) -> Result<Batch, EvalError> {
+        ensure!(
+            batch.column_count() == self.mfp.input_arity,
+            InvalidArgumentSnafu {
+                reason: format!(
+                    "batch column length {} is not equal to input_arity {}",
+                    batch.column_count(),
+                    self.mfp.input_arity
+                ),
+            }
+        );
+
+        let passed_predicates = self.eval_batch_inner(batch)?;
+        let filter = FilterBuilder::new(passed_predicates.as_boolean_array());
+        let pred = filter.build();
+        let mut result = vec![];
+        for col in batch.batch() {
+            let filtered = pred
+                .filter(col.to_arrow_array().as_ref())
+                .context(ArrowSnafu {
+                    context: format!("failed to filter column for mfp operator {:?}", self),
+                })?;
+            result.push(Helper::try_into_vector(filtered).context(DataTypeSnafu {
+                msg: "Failed to convert arrow array to vector",
+            })?);
+        }
+        let projected = self
+            .mfp
+            .projection
+            .iter()
+            .map(|c| result[*c].clone())
+            .collect_vec();
+        let row_count = pred.count();
+
+        Batch::try_new(projected, row_count)
+    }
+
+    /// similar to [`MapFilterProject::evaluate_into`], just in batch.
+    pub fn eval_batch_inner(&self, batch: &mut Batch) -> Result<BooleanVector, EvalError> {
+        // mark the columns that have been evaluated and appended to the `batch`
+        let mut expression = 0;
+        // preds default to true and will be updated as we evaluate each predicate
+        let mut all_preds = BooleanVector::from(vec![Some(true); batch.row_count()]);
+
+        // to compute predicate, need to first compute all expressions used in predicates
+        for (support, predicate) in self.mfp.predicates.iter() {
+            while self.mfp.input_arity + expression < *support {
+                let expr_eval = self.mfp.expressions[expression].eval_batch(batch)?;
+                batch.batch_mut().push(expr_eval);
+                expression += 1;
+            }
+            let pred_vec = predicate.eval_batch(batch)?;
+            let pred_arr = pred_vec.to_arrow_array();
+            let pred_arr = pred_arr.as_any().downcast_ref::<BooleanArray>().context({
+                TypeMismatchSnafu {
+                    expected: ConcreteDataType::boolean_datatype(),
+                    actual: pred_vec.data_type(),
+                }
+            })?;
+            let all_arr = all_preds.as_boolean_array();
+            let res_arr = arrow::compute::and(all_arr, pred_arr).context(ArrowSnafu {
+                context: format!("failed to compute predicate for mfp operator {:?}", self),
+            })?;
+            all_preds = BooleanVector::from(res_arr);
+        }
+
+        // while evaluated expressions are less than total expressions, keep evaluating
+        while expression < self.mfp.expressions.len() {
+            let expr_eval = self.mfp.expressions[expression].eval_batch(batch)?;
+            batch.batch_mut().push(expr_eval);
+            expression += 1;
+        }
+
+        Ok(all_preds)
+    }
+
    /// Evaluates the linear operator on a supplied list of datums.
    ///
    /// The arguments are the initial datums associated with the row,
@@ -735,10 +819,15 @@ impl MfpPlan {

 #[cfg(test)]
 mod test {
+    use std::sync::Arc;
+
    use datatypes::data_type::ConcreteDataType;
+    use datatypes::vectors::{Int32Vector, Int64Vector};
+    use pretty_assertions::assert_eq;

    use super::*;
    use crate::expr::{BinaryFunc, UnaryFunc, UnmaterializableFunc};
+
    #[test]
    fn test_mfp_with_time() {
        use crate::expr::func::BinaryFunc;
@@ -844,6 +933,21 @@ mod test {
            .unwrap()
            .unwrap();
        assert_eq!(ret, Row::pack(vec![Value::from(false), Value::from(true)]));
+
+        // batch mode
+        let mut batch = Batch::try_from_rows(vec![Row::from(vec![
+            Value::from(4),
+            Value::from(2),
+            Value::from(3),
+        ])])
+        .unwrap();
+        let ret = safe_mfp.eval_batch_into(&mut batch).unwrap();
+
+        assert_eq!(
+            ret,
+            Batch::try_from_rows(vec![Row::from(vec![Value::from(false), Value::from(true)])])
+                .unwrap()
+        );
    }

    #[test]
@@ -865,7 +969,7 @@ mod test {
                BinaryFunc::Gt,
            )])
            .unwrap();
-        let mut input1 = vec![
+        let input1 = vec![
            Value::from(4),
            Value::from(2),
            Value::from(3),
@@ -873,19 +977,34 @@ mod test {
        ];
        let safe_mfp = SafeMfpPlan { mfp };
        let ret = safe_mfp
-            .evaluate_into(&mut input1, &mut Row::empty())
+            .evaluate_into(&mut input1.clone(), &mut Row::empty())
            .unwrap();
        assert_eq!(ret, None);
-        let mut input2 = vec![
+
+        let mut input1_batch = Batch::try_from_rows(vec![Row::new(input1)]).unwrap();
+        let ret_batch = safe_mfp.eval_batch_into(&mut input1_batch).unwrap();
+        assert_eq!(
+            ret_batch,
+            Batch::try_new(vec![Arc::new(Int32Vector::from_vec(vec![]))], 0).unwrap()
+        );
+
+        let input2 = vec![
            Value::from(5),
            Value::from(2),
            Value::from(4),
            Value::from("abc"),
        ];
        let ret = safe_mfp
-            .evaluate_into(&mut input2, &mut Row::empty())
+            .evaluate_into(&mut input2.clone(), &mut Row::empty())
            .unwrap();
        assert_eq!(ret, Some(Row::pack(vec![Value::from(11)])));
+
+        let mut input2_batch = Batch::try_from_rows(vec![Row::new(input2)]).unwrap();
+        let ret_batch = safe_mfp.eval_batch_into(&mut input2_batch).unwrap();
+        assert_eq!(
+            ret_batch,
+            Batch::try_new(vec![Arc::new(Int32Vector::from_vec(vec![11]))], 1).unwrap()
+        );
    }

    #[test]
@@ -923,27 +1042,50 @@ mod test {
            .unwrap()
            .project([0, 1, 2])
            .unwrap();
-        let mut input1 = vec![
+        let input1 = vec![
            Value::from(4i64),
            Value::from(2),
            Value::from(3),
            Value::from(53),
        ];
        let safe_mfp = SafeMfpPlan { mfp };
-        let ret = safe_mfp.evaluate_into(&mut input1, &mut Row::empty());
+        let ret = safe_mfp.evaluate_into(&mut input1.clone(), &mut Row::empty());
        assert!(matches!(ret, Err(EvalError::InvalidArgument { .. })));

+        let mut input1_batch = Batch::try_from_rows(vec![Row::new(input1)]).unwrap();
+        let ret_batch = safe_mfp.eval_batch_into(&mut input1_batch);
+        assert!(matches!(ret_batch, Err(EvalError::InvalidArgument { .. })));
+
        let input2 = vec![Value::from(4i64), Value::from(2), Value::from(3)];
        let ret = safe_mfp
            .evaluate_into(&mut input2.clone(), &mut Row::empty())
            .unwrap();
-        assert_eq!(ret, Some(Row::new(input2)));
+        assert_eq!(ret, Some(Row::new(input2.clone())));

-        let mut input3 = vec![Value::from(4i64), Value::from(5), Value::from(2)];
+        let input2_batch = Batch::try_from_rows(vec![Row::new(input2)]).unwrap();
+        let ret_batch = safe_mfp.eval_batch_into(&mut input2_batch.clone()).unwrap();
+        assert_eq!(ret_batch, input2_batch);
+
+        let input3 = vec![Value::from(4i64), Value::from(5), Value::from(2)];
        let ret = safe_mfp
-            .evaluate_into(&mut input3, &mut Row::empty())
+            .evaluate_into(&mut input3.clone(), &mut Row::empty())
            .unwrap();
        assert_eq!(ret, None);
+
+        let input3_batch = Batch::try_from_rows(vec![Row::new(input3)]).unwrap();
+        let ret_batch = safe_mfp.eval_batch_into(&mut input3_batch.clone()).unwrap();
+        assert_eq!(
+            ret_batch,
+            Batch::try_new(
+                vec![
+                    Arc::new(Int64Vector::from_vec(Default::default())),
+                    Arc::new(Int32Vector::from_vec(Default::default())),
+                    Arc::new(Int32Vector::from_vec(Default::default()))
+                ],
+                0
+            )
+            .unwrap()
+        );
    }

    #[test]
@@ -961,10 +1103,18 @@ mod test {
            .unwrap()
            .project(vec![3])
            .unwrap();
-        let mut input1 = vec![Value::from(2), Value::from(3), Value::from(4)];
+        let input1 = vec![Value::from(2), Value::from(3), Value::from(4)];
        let safe_mfp = SafeMfpPlan { mfp };
-        let ret = safe_mfp.evaluate_into(&mut input1, &mut Row::empty());
+        let ret = safe_mfp.evaluate_into(&mut input1.clone(), &mut Row::empty());
        assert_eq!(ret.unwrap(), Some(Row::new(vec![Value::from(false)])));
+
+        let mut input1_batch = Batch::try_from_rows(vec![Row::new(input1)]).unwrap();
+        let ret_batch = safe_mfp.eval_batch_into(&mut input1_batch).unwrap();
+
+        assert_eq!(
+            ret_batch,
+            Batch::try_new(vec![Arc::new(BooleanVector::from(vec![false]))], 1).unwrap()
+        );
    }
    #[test]
    fn test_mfp_chore() {
--- a/src/flow/src/expr/relation/func.rs
+++ b/src/flow/src/expr/relation/func.rs
@@ -18,15 +18,17 @@ use std::sync::OnceLock;

 use datatypes::prelude::ConcreteDataType;
 use datatypes::value::Value;
+use datatypes::vectors::VectorRef;
 use serde::{Deserialize, Serialize};
 use smallvec::smallvec;
-use snafu::{IntoError, OptionExt};
+use snafu::{ensure, IntoError, OptionExt};
 use strum::{EnumIter, IntoEnumIterator};

 use crate::error::{DatafusionSnafu, Error, InvalidQuerySnafu};
 use crate::expr::error::EvalError;
 use crate::expr::relation::accum::{Accum, Accumulator};
 use crate::expr::signature::{GenericFn, Signature};
+use crate::expr::InvalidArgumentSnafu;
 use crate::repr::Diff;

 /// Aggregate functions that can be applied to a group of rows.
@@ -131,6 +133,98 @@ impl AggregateFunc {
        let res = accum.eval(self)?;
        Ok((res, accum.into_state()))
    }
+
+    /// return output value and new accumulator state
+    pub fn eval_batch<A>(
+        &self,
+        accum: A,
+        vector: VectorRef,
+        diff: Option<VectorRef>,
+    ) -> Result<(Value, Vec<Value>), EvalError>
+    where
+        A: IntoIterator<Item = Value>,
+    {
+        let mut accum = accum.into_iter().peekable();
+
+        let mut accum = if accum.peek().is_none() {
+            Accum::new_accum(self)?
+        } else {
+            Accum::try_from_iter(self, &mut accum)?
+        };
+
+        let vector_diff = VectorDiff::try_new(vector, diff)?;
+
+        accum.update_batch(self, vector_diff)?;
+
+        let res = accum.eval(self)?;
+        Ok((res, accum.into_state()))
+    }
+}
+
+struct VectorDiff {
+    vector: VectorRef,
+    diff: Option<VectorRef>,
+}
+
+impl VectorDiff {
+    fn len(&self) -> usize {
+        self.vector.len()
+    }
+
+    fn try_new(vector: VectorRef, diff: Option<VectorRef>) -> Result<Self, EvalError> {
+        ensure!(
+            diff.as_ref()
+                .map_or(true, |diff| diff.len() == vector.len()),
+            InvalidArgumentSnafu {
+                reason: "Length of vector and diff should be the same"
+            }
+        );
+        Ok(Self { vector, diff })
+    }
+}
+
+impl IntoIterator for VectorDiff {
+    type Item = (Value, Diff);
+    type IntoIter = VectorDiffIter;
+
+    fn into_iter(self) -> Self::IntoIter {
+        VectorDiffIter {
+            vector: self.vector,
+            diff: self.diff,
+            idx: 0,
+        }
+    }
+}
+
+struct VectorDiffIter {
+    vector: VectorRef,
+    diff: Option<VectorRef>,
+    idx: usize,
+}
+
+impl std::iter::Iterator for VectorDiffIter {
+    type Item = (Value, Diff);
+
+    fn next(&mut self) -> Option<Self::Item> {
+        if self.idx >= self.vector.len() {
+            return None;
+        }
+        let value = self.vector.get(self.idx);
+        // +1 means insert, -1 means delete, and default to +1 insert when diff is not provided
+        let diff = if let Some(diff) = self.diff.as_ref() {
+            if let Ok(diff_at) = diff.get(self.idx).try_into() {
+                diff_at
+            } else {
+                common_telemetry::warn!("Invalid diff value at index {}", self.idx);
+                return None;
+            }
+        } else {
+            1
+        };
+
+        self.idx += 1;
+        Some((value, diff))
+    }
 }

 /// Generate signature for each aggregate function
--- a/src/flow/src/expr/scalar.rs
+++ b/src/flow/src/expr/scalar.rs
@@ -30,7 +30,7 @@ use crate::expr::error::{
 };
 use crate::expr::func::{BinaryFunc, UnaryFunc, UnmaterializableFunc, VariadicFunc};
 use crate::expr::{Batch, DfScalarFunction};
-use crate::repr::{ColumnType, RelationType};
+use crate::repr::ColumnType;
 /// A scalar expression with a known type.
 #[derive(Ord, PartialOrd, Clone, Debug, Eq, PartialEq, Hash)]
 pub struct TypedExpr {
@@ -46,77 +46,6 @@ impl TypedExpr {
    }
 }

-impl TypedExpr {
-    /// expand multi-value expression to multiple expressions with new indices
-    ///
-    /// Currently it just mean expand `TumbleWindow` to `TumbleWindowFloor` and `TumbleWindowCeiling`
-    ///
-    /// TODO(discord9): test if nested reduce combine with df scalar function would cause problem
-    pub fn expand_multi_value(
-        input_typ: &RelationType,
-        exprs: &[TypedExpr],
-    ) -> Result<Vec<TypedExpr>, Error> {
-        // old indices in mfp, expanded expr
-        let mut ret = vec![];
-        let input_arity = input_typ.column_types.len();
-        for (old_idx, expr) in exprs.iter().enumerate() {
-            if let ScalarExpr::CallUnmaterializable(UnmaterializableFunc::TumbleWindow {
-                ts,
-                window_size,
-                start_time,
-            }) = &expr.expr
-            {
-                let floor = UnaryFunc::TumbleWindowFloor {
-                    window_size: *window_size,
-                    start_time: *start_time,
-                };
-                let ceil = UnaryFunc::TumbleWindowCeiling {
-                    window_size: *window_size,
-                    start_time: *start_time,
-                };
-                let floor = ScalarExpr::CallUnary {
-                    func: floor,
-                    expr: Box::new(ts.expr.clone()),
-                }
-                .with_type(ts.typ.clone());
-                ret.push((None, floor));
-
-                let ceil = ScalarExpr::CallUnary {
-                    func: ceil,
-                    expr: Box::new(ts.expr.clone()),
-                }
-                .with_type(ts.typ.clone());
-                ret.push((None, ceil));
-            } else {
-                ret.push((Some(input_arity + old_idx), expr.clone()))
-            }
-        }
-
-        // get shuffled index(old_idx -> new_idx)
-        // note index is offset by input_arity because mfp is designed to be first include input columns then intermediate columns
-        let shuffle = ret
-            .iter()
-            .map(|(old_idx, _)| *old_idx) // [Option<opt_idx>]
-            .enumerate()
-            .map(|(new, old)| (old, new + input_arity))
-            .flat_map(|(old, new)| old.map(|o| (o, new)))
-            .chain((0..input_arity).map(|i| (i, i))) // also remember to chain the input columns as not changed
-            .collect::<BTreeMap<_, _>>();
-
-        // shuffle expr's index
-        let exprs = ret
-            .into_iter()
-            .map(|(_, mut expr)| {
-                // invariant: it is expect that no expr will try to refer the column being expanded
-                expr.expr.permute_map(&shuffle)?;
-                Ok(expr)
-            })
-            .collect::<Result<Vec<_>, _>>()?;
-
-        Ok(exprs)
-    }
-}
-
 /// A scalar expression, which can be evaluated to a value.
 #[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
 pub enum ScalarExpr {
@@ -210,6 +139,13 @@ impl ScalarExpr {
 }

 impl ScalarExpr {
+    pub fn cast(self, typ: ConcreteDataType) -> Self {
+        ScalarExpr::CallUnary {
+            func: UnaryFunc::Cast(typ),
+            expr: Box::new(self),
+        }
+    }
+
    /// apply optimization to the expression, like flatten variadic function
    pub fn optimize(&mut self) {
        self.flatten_varidic_fn();
@@ -341,7 +277,7 @@ impl ScalarExpr {
                // put a slice to corresponding batch
                let slice_offset = prev_cond_idx;
                let slice_length = idx - prev_cond_idx;
-                let to_be_append = batch.slice(slice_offset, slice_length);
+                let to_be_append = batch.slice(slice_offset, slice_length)?;

                let to_put_back = match prev_cond {
                    Some(true) => (
@@ -364,7 +300,7 @@ impl ScalarExpr {
        if let Some(slice_offset) = prev_start_idx {
            let prev_cond = prev_cond.unwrap();
            let slice_length = bool_conds.len() - slice_offset;
-            let to_be_append = batch.slice(slice_offset, slice_length);
+            let to_be_append = batch.slice(slice_offset, slice_length)?;
            let to_put_back = match prev_cond {
                Some(true) => (
                    Some(true),
@@ -876,7 +812,7 @@ mod test {
            let raw_len = raw.len();
            let vectors = vec![Int32Vector::from(raw).slice(0, raw_len)];

-            let batch = Batch::new(vectors, raw_len);
+            let batch = Batch::try_new(vectors, raw_len).unwrap();
            let expected = Int32Vector::from(vec![
                None,
                Some(42),
@@ -895,7 +831,7 @@ mod test {
            let raw_len = raw.len();
            let vectors = vec![Int32Vector::from(raw).slice(0, raw_len)];

-            let batch = Batch::new(vectors, raw_len);
+            let batch = Batch::try_new(vectors, raw_len).unwrap();
            let expected = Int32Vector::from(vec![Some(42)]).slice(0, raw_len);
            assert_eq!(expr.eval_batch(&batch).unwrap(), expected);

@@ -903,7 +839,7 @@ mod test {
            let raw_len = raw.len();
            let vectors = vec![Int32Vector::from(raw).slice(0, raw_len)];

-            let batch = Batch::new(vectors, raw_len);
+            let batch = Batch::try_new(vectors, raw_len).unwrap();
            let expected = NullVector::new(raw_len).slice(0, raw_len);
            assert_eq!(expr.eval_batch(&batch).unwrap(), expected);
        }
--- a/src/flow/src/expr/signature.rs
+++ b/src/flow/src/expr/signature.rs
@@ -19,6 +19,8 @@ use serde::{Deserialize, Serialize};
 use smallvec::SmallVec;

 /// Function signature
+///
+/// TODO(discord9): use `common_query::signature::Signature` crate
 #[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Deserialize, Serialize, Hash)]
 pub struct Signature {
    /// the input types, usually not great than two input arg
--- a/src/flow/src/lib.rs
+++ b/src/flow/src/lib.rs
@@ -20,9 +20,14 @@
 #![allow(dead_code)]
 #![warn(clippy::missing_docs_in_private_items)]
 #![warn(clippy::too_many_lines)]
+
+// TODO(discord9): enable this lint to handle out of bound access
+// #![cfg_attr(not(test), warn(clippy::indexing_slicing))]
+
 // allow unused for now because it should be use later
 mod adapter;
 mod compute;
+mod df_optimizer;
 pub mod error;
 mod expr;
 pub mod heartbeat;
--- a/src/flow/src/plan.rs
+++ b/src/flow/src/plan.rs
@@ -115,6 +115,8 @@ impl TypedPlan {

 /// TODO(discord9): support `TableFunc`（by define FlatMap that map 1 to n)
 /// Plan describe how to transform data in dataflow
+///
+/// This can be considered as a physical plan in dataflow, which describe how to transform data in a streaming manner.
 #[derive(Debug, Clone, Eq, PartialEq, Ord, PartialOrd)]
 pub enum Plan {
    /// A constant collection of rows.
--- a/src/flow/src/repr.rs
+++ b/src/flow/src/repr.rs
@@ -177,6 +177,12 @@ impl Row {
    }
 }

+impl From<Vec<Value>> for Row {
+    fn from(row: Vec<Value>) -> Self {
+        Row::new(row)
+    }
+}
+
 impl From<ProtoRow> for Row {
    fn from(row: ProtoRow) -> Self {
        Row::pack(
--- a/src/flow/src/repr/relation.rs
+++ b/src/flow/src/repr/relation.rs
@@ -374,10 +374,8 @@ impl RelationDesc {
            .collect();
        let arrow_schema = arrow_schema::Schema::new(fields);

-        DFSchema::try_from(arrow_schema.clone()).context({
-            DatafusionSnafu {
-                context: format!("Error when converting to DFSchema: {:?}", arrow_schema),
-            }
+        DFSchema::try_from(arrow_schema.clone()).with_context(|_e| DatafusionSnafu {
+            context: format!("Error when converting to DFSchema: {:?}", arrow_schema),
        })
    }

--- a/src/flow/src/transform.rs
+++ b/src/flow/src/transform.rs
@@ -17,24 +17,19 @@ use std::collections::{BTreeMap, HashMap};
 use std::sync::Arc;

 use common_error::ext::BoxedError;
-use datafusion::optimizer::simplify_expressions::SimplifyExpressions;
-use datafusion::optimizer::{OptimizerContext, OptimizerRule};
 use datatypes::data_type::ConcreteDataType as CDT;
-use query::parser::QueryLanguageParser;
-use query::plan::LogicalPlan;
-use query::query_engine::DefaultSerializer;
 use query::QueryEngine;
 use serde::{Deserialize, Serialize};
 use snafu::ResultExt;
 /// note here we are using the `substrait_proto_df` crate from the `substrait` module and
 /// rename it to `substrait_proto`
-use substrait::{substrait_proto_df as substrait_proto, DFLogicalSubstraitConvertor};
+use substrait::substrait_proto_df as substrait_proto;
 use substrait_proto::proto::extensions::simple_extension_declaration::MappingType;
 use substrait_proto::proto::extensions::SimpleExtensionDeclaration;

 use crate::adapter::FlownodeContext;
-use crate::error::{DatafusionSnafu, Error, ExternalSnafu, NotImplementedSnafu, UnexpectedSnafu};
-use crate::plan::TypedPlan;
+use crate::error::{Error, NotImplementedSnafu, UnexpectedSnafu};
+use crate::expr::{TUMBLE_END, TUMBLE_START};
 /// a simple macro to generate a not implemented error
 macro_rules! not_impl_err {
    ($($arg:tt)*)  => {
@@ -102,68 +97,39 @@ impl FunctionExtensions {
    }
 }

-/// To reuse existing code for parse sql, the sql is first parsed into a datafusion logical plan,
-/// then to a substrait plan, and finally to a flow plan.
-pub async fn sql_to_flow_plan(
-    ctx: &mut FlownodeContext,
-    engine: &Arc<dyn QueryEngine>,
-    sql: &str,
-) -> Result<TypedPlan, Error> {
-    let query_ctx = ctx.query_context.clone().ok_or_else(|| {
-        UnexpectedSnafu {
-            reason: "Query context is missing",
-        }
-        .build()
-    })?;
-    let stmt = QueryLanguageParser::parse_sql(sql, &query_ctx)
-        .map_err(BoxedError::new)
-        .context(ExternalSnafu)?;
-    let plan = engine
-        .planner()
-        .plan(stmt, query_ctx)
-        .await
-        .map_err(BoxedError::new)
-        .context(ExternalSnafu)?;
-    let LogicalPlan::DfPlan(plan) = plan;
-    let plan = SimplifyExpressions::new()
-        .rewrite(plan, &OptimizerContext::default())
-        .context(DatafusionSnafu {
-            context: "Fail to apply `SimplifyExpressions` optimization",
-        })?
-        .data;
-    let sub_plan = DFLogicalSubstraitConvertor {}
-        .to_sub_plan(&plan, DefaultSerializer)
-        .map_err(BoxedError::new)
-        .context(ExternalSnafu)?;
-
-    let flow_plan = TypedPlan::from_substrait_plan(ctx, &sub_plan).await?;
-
-    Ok(flow_plan)
-}
-
 /// register flow-specific functions to the query engine
 pub fn register_function_to_query_engine(engine: &Arc<dyn QueryEngine>) {
-    engine.register_function(Arc::new(TumbleFunction {}));
+    engine.register_function(Arc::new(TumbleFunction::new("tumble")));
+    engine.register_function(Arc::new(TumbleFunction::new(TUMBLE_START)));
+    engine.register_function(Arc::new(TumbleFunction::new(TUMBLE_END)));
 }

 #[derive(Debug)]
-pub struct TumbleFunction {}
+pub struct TumbleFunction {
+    name: String,
+}

-const TUMBLE_NAME: &str = "tumble";
+impl TumbleFunction {
+    fn new(name: &str) -> Self {
+        Self {
+            name: name.to_string(),
+        }
+    }
+}

 impl std::fmt::Display for TumbleFunction {
    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
-        write!(f, "{}", TUMBLE_NAME.to_ascii_uppercase())
+        write!(f, "{}", self.name.to_ascii_uppercase())
    }
 }

 impl common_function::function::Function for TumbleFunction {
    fn name(&self) -> &str {
-        TUMBLE_NAME
+        &self.name
    }

    fn return_type(&self, _input_types: &[CDT]) -> common_query::error::Result<CDT> {
-        Ok(CDT::datetime_datatype())
+        Ok(CDT::timestamp_millisecond_datatype())
    }

    fn signature(&self) -> common_query::prelude::Signature {
@@ -198,6 +164,7 @@ mod test {
    use prost::Message;
    use query::parser::QueryLanguageParser;
    use query::plan::LogicalPlan;
+    use query::query_engine::DefaultSerializer;
    use query::QueryEngine;
    use session::context::QueryContext;
    use substrait::{DFLogicalSubstraitConvertor, SubstraitPlan};
@@ -207,6 +174,7 @@ mod test {

    use super::*;
    use crate::adapter::node_context::IdToNameMap;
+    use crate::df_optimizer::apply_df_optimizer;
    use crate::expr::GlobalId;
    use crate::repr::{ColumnType, RelationType};

@@ -292,7 +260,7 @@ mod test {
        let factory = query::QueryEngineFactory::new(catalog_list, None, None, None, None, false);

        let engine = factory.query_engine();
-        engine.register_function(Arc::new(TumbleFunction {}));
+        register_function_to_query_engine(&engine);

        assert_eq!("datafusion", engine.name());
        engine
@@ -307,6 +275,7 @@ mod test {
            .await
            .unwrap();
        let LogicalPlan::DfPlan(plan) = plan;
+        let plan = apply_df_optimizer(plan).await.unwrap();

        // encode then decode so to rely on the impl of conversion from logical plan to substrait plan
        let bytes = DFLogicalSubstraitConvertor {}
@@ -315,4 +284,22 @@ mod test {

        proto::Plan::decode(bytes).unwrap()
    }
+
+    /// TODO(discord9): add more illegal sql tests
+    #[tokio::test]
+    async fn test_missing_key_check() {
+        let engine = create_test_query_engine();
+        let sql = "SELECT avg(number) FROM numbers_with_ts GROUP BY tumble(ts, '1 hour'), number";
+
+        let stmt = QueryLanguageParser::parse_sql(sql, &QueryContext::arc()).unwrap();
+        let plan = engine
+            .planner()
+            .plan(stmt, QueryContext::arc())
+            .await
+            .unwrap();
+        let LogicalPlan::DfPlan(plan) = plan;
+        let plan = apply_df_optimizer(plan).await;
+
+        assert!(plan.is_err());
+    }
 }
--- a/src/flow/src/transform/aggr.rs
+++ b/src/flow/src/transform/aggr.rs
--- a/src/flow/src/transform/expr.rs
+++ b/src/flow/src/transform/expr.rs
@@ -20,7 +20,7 @@ use common_error::ext::BoxedError;
 use common_telemetry::debug;
 use datafusion_physical_expr::PhysicalExpr;
 use datatypes::data_type::ConcreteDataType as CDT;
-use snafu::{OptionExt, ResultExt};
+use snafu::{ensure, OptionExt, ResultExt};
 use substrait_proto::proto::expression::field_reference::ReferenceType::DirectReference;
 use substrait_proto::proto::expression::reference_segment::ReferenceType::StructField;
 use substrait_proto::proto::expression::{IfThen, RexType, ScalarFunction};
@@ -33,7 +33,7 @@ use crate::error::{
 };
 use crate::expr::{
    BinaryFunc, DfScalarFunction, RawDfScalarFn, ScalarExpr, TypedExpr, UnaryFunc,
-    UnmaterializableFunc, VariadicFunc,
+    UnmaterializableFunc, VariadicFunc, TUMBLE_END, TUMBLE_START,
 };
 use crate::repr::{ColumnType, RelationDesc, RelationType};
 use crate::transform::literal::{
@@ -167,6 +167,16 @@ fn rewrite_scalar_function(
    arg_typed_exprs: &[TypedExpr],
 ) -> Result<ScalarFunction, Error> {
    let mut f_rewrite = f.clone();
+    ensure!(
+        f_rewrite.arguments.len() == arg_typed_exprs.len(),
+        crate::error::InternalSnafu {
+            reason: format!(
+                "Expect `f_rewrite` and `arg_typed_expr` to be same length, found {} and {}",
+                f_rewrite.arguments.len(),
+                arg_typed_exprs.len()
+            )
+        }
+    );
    for (idx, raw_expr) in f_rewrite.arguments.iter_mut().enumerate() {
        // only replace it with col(idx) if it is not literal
        // will try best to determine if it is literal, i.e. for function like `cast(<literal>)` will try
@@ -351,7 +361,13 @@ impl TypedExpr {
                Ok(TypedExpr::new(ret_expr, ret_type))
            }
            _var => {
-                if VariadicFunc::is_valid_func_name(fn_name) {
+                if fn_name == TUMBLE_START || fn_name == TUMBLE_END {
+                    let (func, arg) = UnaryFunc::from_tumble_func(fn_name, &arg_typed_exprs)?;
+
+                    let ret_type = ColumnType::new_nullable(func.signature().output.clone());
+
+                    Ok(TypedExpr::new(arg.expr.call_unary(func), ret_type))
+                } else if VariadicFunc::is_valid_func_name(fn_name) {
                    let func = VariadicFunc::from_str_and_types(fn_name, &arg_types)?;
                    let ret_type = ColumnType::new_nullable(func.signature().output.clone());
                    let mut expr = ScalarExpr::CallVariadic {
@@ -521,7 +537,6 @@ impl TypedExpr {

 #[cfg(test)]
 mod test {
-    use common_time::{DateTime, Interval};
    use datatypes::prelude::ConcreteDataType;
    use datatypes::value::Value;
    use pretty_assertions::assert_eq;
@@ -562,7 +577,7 @@ mod test {
        };
        let expected = TypedPlan {
            schema: RelationType::new(vec![ColumnType::new(CDT::uint32_datatype(), false)])
-                .into_named(vec![Some("number".to_string())]),
+                .into_named(vec![Some("numbers.number".to_string())]),
            plan: Plan::Mfp {
                input: Box::new(
                    Plan::Get {
@@ -576,13 +591,7 @@ mod test {
                        .into_named(vec![Some("number".to_string())]),
                    ),
                ),
-                mfp: MapFilterProject::new(1)
-                    .map(vec![ScalarExpr::Column(0)])
-                    .unwrap()
-                    .filter(vec![filter])
-                    .unwrap()
-                    .project(vec![1])
-                    .unwrap(),
+                mfp: MapFilterProject::new(1).filter(vec![filter]).unwrap(),
            },
        };
        assert_eq!(flow_plan.unwrap(), expected);
@@ -600,7 +609,7 @@ mod test {

        let expected = TypedPlan {
            schema: RelationType::new(vec![ColumnType::new(CDT::boolean_datatype(), true)])
-                .into_unnamed(),
+                .into_named(vec![Some("Int64(1) + Int64(1) * Int64(2) - Int64(1) / Int64(1) + Int64(1) % Int64(2) = Int64(3)".to_string())]),
            plan: Plan::Constant {
                rows: vec![(
                    repr::Row::new(vec![Value::from(true)]),
@@ -624,8 +633,8 @@ mod test {
        let flow_plan = TypedPlan::from_substrait_plan(&mut ctx, &plan).await;

        let expected = TypedPlan {
-            schema: RelationType::new(vec![ColumnType::new(CDT::uint32_datatype(), true)])
-                .into_unnamed(),
+            schema: RelationType::new(vec![ColumnType::new(CDT::int64_datatype(), true)])
+                .into_named(vec![Some("numbers.number + Int64(1)".to_string())]),
            plan: Plan::Mfp {
                input: Box::new(
                    Plan::Get {
@@ -640,10 +649,12 @@ mod test {
                    ),
                ),
                mfp: MapFilterProject::new(1)
-                    .map(vec![ScalarExpr::Column(0).call_binary(
-                        ScalarExpr::Literal(Value::from(1u32), CDT::uint32_datatype()),
-                        BinaryFunc::AddUInt32,
-                    )])
+                    .map(vec![ScalarExpr::Column(0)
+                        .call_unary(UnaryFunc::Cast(CDT::int64_datatype()))
+                        .call_binary(
+                            ScalarExpr::Literal(Value::from(1i64), CDT::int64_datatype()),
+                            BinaryFunc::AddInt64,
+                        )])
                    .unwrap()
                    .project(vec![1])
                    .unwrap(),
@@ -663,7 +674,9 @@ mod test {

        let expected = TypedPlan {
            schema: RelationType::new(vec![ColumnType::new(CDT::int16_datatype(), true)])
-                .into_unnamed(),
+                .into_named(vec![Some(
+                    "arrow_cast(Int64(1),Utf8(\"Int16\"))".to_string(),
+                )]),
            plan: Plan::Constant {
                // cast of literal is constant folded
                rows: vec![(repr::Row::new(vec![Value::from(1i16)]), i64::MIN, 1)],
@@ -683,7 +696,7 @@ mod test {

        let expected = TypedPlan {
            schema: RelationType::new(vec![ColumnType::new(CDT::uint32_datatype(), true)])
-                .into_unnamed(),
+                .into_named(vec![Some("numbers.number + numbers.number".to_string())]),
            plan: Plan::Mfp {
                input: Box::new(
                    Plan::Get {
@@ -780,65 +793,5 @@ mod test {
                },
            }
        );
-
-        let f = substrait_proto::proto::expression::ScalarFunction {
-            function_reference: 0,
-            arguments: vec![proto_col(0), lit("1 second"), lit("2021-07-01 00:00:00")],
-            options: vec![],
-            output_type: None,
-            ..Default::default()
-        };
-        let input_schema = RelationType::new(vec![
-            ColumnType::new(CDT::timestamp_nanosecond_datatype(), false),
-            ColumnType::new(CDT::string_datatype(), false),
-        ])
-        .into_unnamed();
-        let extensions = FunctionExtensions::from_iter(vec![(0, "tumble".to_string())]);
-        let res = TypedExpr::from_substrait_scalar_func(&f, &input_schema, &extensions)
-            .await
-            .unwrap();
-
-        assert_eq!(
-            res,
-            ScalarExpr::CallUnmaterializable(UnmaterializableFunc::TumbleWindow {
-                ts: Box::new(
-                    ScalarExpr::Column(0)
-                        .with_type(ColumnType::new(CDT::timestamp_nanosecond_datatype(), false))
-                ),
-                window_size: Interval::from_month_day_nano(0, 0, 1_000_000_000),
-                start_time: Some(DateTime::new(1625097600000))
-            })
-            .with_type(ColumnType::new(CDT::timestamp_millisecond_datatype(), true)),
-        );
-
-        let f = substrait_proto::proto::expression::ScalarFunction {
-            function_reference: 0,
-            arguments: vec![proto_col(0), lit("1 second")],
-            options: vec![],
-            output_type: None,
-            ..Default::default()
-        };
-        let input_schema = RelationType::new(vec![
-            ColumnType::new(CDT::timestamp_nanosecond_datatype(), false),
-            ColumnType::new(CDT::string_datatype(), false),
-        ])
-        .into_unnamed();
-        let extensions = FunctionExtensions::from_iter(vec![(0, "tumble".to_string())]);
-        let res = TypedExpr::from_substrait_scalar_func(&f, &input_schema, &extensions)
-            .await
-            .unwrap();
-
-        assert_eq!(
-            res,
-            ScalarExpr::CallUnmaterializable(UnmaterializableFunc::TumbleWindow {
-                ts: Box::new(
-                    ScalarExpr::Column(0)
-                        .with_type(ColumnType::new(CDT::timestamp_nanosecond_datatype(), false))
-                ),
-                window_size: Interval::from_month_day_nano(0, 0, 1_000_000_000),
-                start_time: None
-            })
-            .with_type(ColumnType::new(CDT::timestamp_millisecond_datatype(), true)),
-        )
    }
 }
--- a/src/flow/src/transform/literal.rs
+++ b/src/flow/src/transform/literal.rs
@@ -340,6 +340,8 @@ pub fn from_substrait_type(null_type: &substrait_proto::proto::Type) -> Result<C

 #[cfg(test)]
 mod test {
+    use pretty_assertions::assert_eq;
+
    use super::*;
    use crate::plan::{Plan, TypedPlan};
    use crate::repr::{self, ColumnType, RelationType};
@@ -356,7 +358,7 @@ mod test {

        let expected = TypedPlan {
            schema: RelationType::new(vec![ColumnType::new(CDT::int64_datatype(), true)])
-                .into_unnamed(),
+                .into_named(vec![Some("Int64(1)".to_string())]),
            plan: Plan::Constant {
                rows: vec![(
                    repr::Row::new(vec![Value::Int64(1)]),
--- a/src/flow/src/transform/plan.rs
+++ b/src/flow/src/transform/plan.rs
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use std::collections::{BTreeMap, BTreeSet, HashSet};
+use std::collections::HashSet;

 use itertools::Itertools;
 use snafu::OptionExt;
@@ -23,9 +23,9 @@ use substrait_proto::proto::rel::RelType;
 use substrait_proto::proto::{plan_rel, Plan as SubPlan, ProjectRel, Rel};

 use crate::error::{Error, InvalidQuerySnafu, NotImplementedSnafu, PlanSnafu, UnexpectedSnafu};
-use crate::expr::{MapFilterProject, ScalarExpr, TypedExpr, UnaryFunc};
-use crate::plan::{KeyValPlan, Plan, TypedPlan};
-use crate::repr::{self, RelationDesc, RelationType};
+use crate::expr::{MapFilterProject, TypedExpr};
+use crate::plan::{Plan, TypedPlan};
+use crate::repr::{self, RelationType};
 use crate::transform::{substrait_proto, FlownodeContext, FunctionExtensions};

 impl TypedPlan {
@@ -49,7 +49,14 @@ impl TypedPlan {
                        let input = root.input.as_ref().with_context(|| InvalidQuerySnafu {
                            reason: "Root relation without input",
                        })?;
-                        Ok(TypedPlan::from_substrait_rel(ctx, input, &function_extension).await?)
+
+                        let mut ret = TypedPlan::from_substrait_rel(ctx, input, &function_extension).await?;
+
+                        if !root.names.is_empty() {
+                            ret.schema = ret.schema.clone().try_with_names(root.names.clone())?;
+                        }
+
+                        Ok(ret)
                    }
                },
                None => plan_err!("Cannot parse plan relation: None")
@@ -115,17 +122,6 @@ impl TypedPlan {
                plan,
            })
        } else {
-            match input.plan.clone() {
-                Plan::Reduce { key_val_plan, .. } => {
-                    rewrite_projection_after_reduce(key_val_plan, &input.schema, &mut exprs)?;
-                }
-                Plan::Mfp { input, mfp: _ } => {
-                    if let Plan::Reduce { key_val_plan, .. } = input.plan {
-                        rewrite_projection_after_reduce(key_val_plan, &input.schema, &mut exprs)?;
-                    }
-                }
-                _ => (),
-            }
            input.projection(exprs)
        }
    }
@@ -233,120 +229,13 @@ impl TypedPlan {
    }
 }

-/// if reduce_plan contains the special function like tumble floor/ceiling, add them to the proj_exprs
-/// so the effect is the window_start, window_end column are auto added to output rows
-///
-/// This is to fix a problem that we have certain functions that return two values, but since substrait doesn't know that, it will assume it return one value
-/// this function fix that and rewrite `proj_exprs` to correct form
-fn rewrite_projection_after_reduce(
-    key_val_plan: KeyValPlan,
-    reduce_output_type: &RelationDesc,
-    proj_exprs: &mut Vec<TypedExpr>,
-) -> Result<(), Error> {
-    // TODO(discord9): get keys correctly
-    let key_exprs = key_val_plan
-        .key_plan
-        .projection
-        .clone()
-        .into_iter()
-        .map(|i| {
-            if i < key_val_plan.key_plan.input_arity {
-                ScalarExpr::Column(i)
-            } else {
-                key_val_plan.key_plan.expressions[i - key_val_plan.key_plan.input_arity].clone()
-            }
-        })
-        .collect_vec();
-    let mut shift_offset = 0;
-    let mut shuffle: BTreeMap<usize, usize> = BTreeMap::new();
-    let special_keys = key_exprs
-        .clone()
-        .into_iter()
-        .enumerate()
-        .filter(|(idx, p)| {
-            shuffle.insert(*idx, *idx + shift_offset);
-            if matches!(
-                p,
-                ScalarExpr::CallUnary {
-                    func: UnaryFunc::TumbleWindowFloor { .. },
-                    ..
-                } | ScalarExpr::CallUnary {
-                    func: UnaryFunc::TumbleWindowCeiling { .. },
-                    ..
-                }
-            ) {
-                if matches!(
-                    p,
-                    ScalarExpr::CallUnary {
-                        func: UnaryFunc::TumbleWindowFloor { .. },
-                        ..
-                    }
-                ) {
-                    shift_offset += 1;
-                }
-                true
-            } else {
-                false
-            }
-        })
-        .collect_vec();
-    let spec_key_arity = special_keys.len();
-    if spec_key_arity == 0 {
-        return Ok(());
-    }
-
-    // shuffle proj_exprs
-    // because substrait use offset while assume `tumble` only return one value
-    for proj_expr in proj_exprs.iter_mut() {
-        proj_expr.expr.permute_map(&shuffle)?;
-    } // add key to the end
-    for (key_idx, _key_expr) in special_keys {
-        // here we assume the output type of reduce operator(`reduce_output_type`) is just first keys columns, then append value columns
-        // so we can use `key_idx` to index `reduce_output_type` and get the keys we need to append to `proj_exprs`
-        proj_exprs.push(
-            ScalarExpr::Column(key_idx)
-                .with_type(reduce_output_type.typ().column_types[key_idx].clone()),
-        );
-    }
-
-    // check if normal expr in group exprs are all in proj_exprs
-    let all_cols_ref_in_proj: BTreeSet<usize> = proj_exprs
-        .iter()
-        .filter_map(|e| {
-            if let ScalarExpr::Column(i) = &e.expr {
-                Some(*i)
-            } else {
-                None
-            }
-        })
-        .collect();
-    for (key_idx, key_expr) in key_exprs.iter().enumerate() {
-        if let ScalarExpr::Column(_) = key_expr {
-            if !all_cols_ref_in_proj.contains(&key_idx) {
-                let err_msg = format!(
-                    "Expect normal column in group by also appear in projection, but column {}(name is {}) is missing", 
-                    key_idx,
-                    reduce_output_type
-                        .get_name(key_idx)
-                        .clone()
-                        .map(|s|format!("'{}'",s))
-                        .unwrap_or("unknown".to_string())
-            );
-                return InvalidQuerySnafu { reason: err_msg }.fail();
-            }
-        }
-    }
-
-    Ok(())
-}
-
 #[cfg(test)]
 mod test {
    use datatypes::prelude::ConcreteDataType;
    use pretty_assertions::assert_eq;

    use super::*;
-    use crate::expr::{GlobalId, ScalarExpr};
+    use crate::expr::GlobalId;
    use crate::plan::{Plan, TypedPlan};
    use crate::repr::{ColumnType, RelationType};
    use crate::transform::test::{create_test_ctx, create_test_query_engine, sql_to_substrait};
@@ -363,7 +252,7 @@ mod test {

        let expected = TypedPlan {
            schema: RelationType::new(vec![ColumnType::new(CDT::uint32_datatype(), false)])
-                .into_named(vec![Some("number".to_string())]),
+                .into_named(vec![Some("numbers.number".to_string())]),
            plan: Plan::Mfp {
                input: Box::new(
                    Plan::Get {
@@ -377,11 +266,7 @@ mod test {
                        .into_named(vec![Some("number".to_string())]),
                    ),
                ),
-                mfp: MapFilterProject::new(1)
-                    .map(vec![ScalarExpr::Column(0)])
-                    .unwrap()
-                    .project(vec![1])
-                    .unwrap(),
+                mfp: MapFilterProject::new(1),
            },
        };

--- a/src/flow/src/utils.rs
+++ b/src/flow/src/utils.rs
@@ -513,7 +513,7 @@ pub type ArrangeReader<'a> = tokio::sync::RwLockReadGuard<'a, Arrangement>;
 pub type ArrangeWriter<'a> = tokio::sync::RwLockWriteGuard<'a, Arrangement>;

 /// A handler to the inner Arrangement, can be cloned and shared, useful for query it's inner state
-#[derive(Debug)]
+#[derive(Debug, Clone)]
 pub struct ArrangeHandler {
    inner: Arc<RwLock<Arrangement>>,
 }
--- a/src/frontend/src/error.rs
+++ b/src/frontend/src/error.rs
@@ -20,7 +20,6 @@ use common_error::ext::{BoxedError, ErrorExt};
 use common_error::status_code::StatusCode;
 use common_macro::stack_trace_debug;
 use snafu::{Location, Snafu};
-use store_api::storage::RegionNumber;

 #[derive(Snafu)]
 #[snafu(visibility(pub))]
@@ -61,13 +60,6 @@ pub enum Error {
        source: common_meta::error::Error,
    },

-    #[snafu(display("Runtime resource error"))]
-    RuntimeResource {
-        #[snafu(implicit)]
-        location: Location,
-        source: common_runtime::error::Error,
-    },
-
    #[snafu(display("Failed to start server"))]
    StartServer {
        #[snafu(implicit)]
@@ -96,13 +88,6 @@ pub enum Error {
        source: sql::error::Error,
    },

-    #[snafu(display("Failed to convert vector to GRPC column, reason: {}", reason))]
-    VectorToGrpcColumn {
-        reason: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Invalid SQL, error: {}", err_msg))]
    InvalidSql {
        err_msg: String,
@@ -117,13 +102,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Failed to find Datanode by region: {:?}", region))]
-    FindDatanode {
-        region: RegionNumber,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Invalid InsertRequest, reason: {}", reason))]
    InvalidInsertRequest {
        reason: String,
@@ -220,20 +198,6 @@ pub enum Error {
        source: servers::error::Error,
    },

-    #[snafu(display("Failed to find leaders when altering table, table: {}", table))]
-    LeaderNotFound {
-        table: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
-    #[snafu(display("Failed to found context value: {}", key))]
-    ContextValueNotFound {
-        key: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Not supported: {}", feat))]
    NotSupported { feat: String },

@@ -282,14 +246,6 @@ pub enum Error {
        source: script::error::Error,
    },

-    #[snafu(display("Failed to copy table: {}", table_name))]
-    CopyTable {
-        table_name: String,
-        #[snafu(implicit)]
-        location: Location,
-        source: table::error::Error,
-    },
-
    #[snafu(display("Failed to insert value into table: {}", table_name))]
    Insert {
        table_name: String,
@@ -312,13 +268,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Empty data: {}", msg))]
-    EmptyData {
-        msg: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display(
        "No valid default value can be built automatically, column: {}",
        column,
@@ -364,20 +313,6 @@ pub enum Error {
        #[snafu(implicit)]
        location: Location,
    },
-
-    #[snafu(display("Failed to setup plugin"))]
-    SetupPlugin {
-        #[snafu(implicit)]
-        location: Location,
-        source: BoxedError,
-    },
-
-    #[snafu(display("Failed to start plugin"))]
-    StartPlugin {
-        #[snafu(implicit)]
-        location: Location,
-        source: BoxedError,
-    },
 }

 pub type Result<T> = std::result::Result<T, Error>;
@@ -395,7 +330,6 @@ impl ErrorExt for Error {
            | Error::ColumnNotFound { .. }
            | Error::UnsupportedFormat { .. }
            | Error::IllegalAuthConfig { .. }
-            | Error::EmptyData { .. }
            | Error::ColumnNoneDefaultValue { .. }
            | Error::IncompleteGrpcRequest { .. }
            | Error::InvalidTlsConfig { .. } => StatusCode::InvalidArguments,
@@ -408,7 +342,6 @@ impl ErrorExt for Error {

            Error::HandleHeartbeatResponse { source, .. } => source.status_code(),

-            Error::RuntimeResource { source, .. } => source.status_code(),
            Error::PromStoreRemoteQueryPlan { source, .. }
            | Error::ExecutePromql { source, .. } => source.status_code(),

@@ -420,25 +353,16 @@ impl ErrorExt for Error {

            Error::InvalidateTableCache { source, .. } => source.status_code(),

-            Error::Table { source, .. }
-            | Error::CopyTable { source, .. }
-            | Error::Insert { source, .. } => source.status_code(),
+            Error::Table { source, .. } | Error::Insert { source, .. } => source.status_code(),

            Error::OpenRaftEngineBackend { .. } => StatusCode::StorageUnavailable,

            Error::RequestQuery { source, .. } => source.status_code(),

-            Error::FindDatanode { .. } => StatusCode::RegionNotReady,
-
-            Error::VectorToGrpcColumn { .. }
-            | Error::CacheRequired { .. }
-            | Error::SetupPlugin { .. }
-            | Error::StartPlugin { .. } => StatusCode::Internal,
+            Error::CacheRequired { .. } => StatusCode::Internal,

            Error::InvalidRegionRequest { .. } => StatusCode::IllegalState,

-            Error::ContextValueNotFound { .. } => StatusCode::Unexpected,
-
            Error::TableNotFound { .. } => StatusCode::TableNotFound,

            Error::Catalog { source, .. } => source.status_code(),
@@ -450,7 +374,6 @@ impl ErrorExt for Error {
            | Error::ReadTable { source, .. }
            | Error::ExecLogicalPlan { source, .. } => source.status_code(),

-            Error::LeaderNotFound { .. } => StatusCode::StorageUnavailable,
            Error::InvokeRegionServer { source, .. } => source.status_code(),

            Error::External { source, .. } => source.status_code(),
--- a/src/frontend/src/heartbeat.rs
+++ b/src/frontend/src/heartbeat.rs
@@ -32,6 +32,7 @@ use tokio::time::{Duration, Instant};
 use crate::error;
 use crate::error::Result;
 use crate::frontend::FrontendOptions;
+use crate::metrics::{HEARTBEAT_RECV_COUNT, HEARTBEAT_SENT_COUNT};

 pub mod handler;

@@ -94,10 +95,16 @@ impl HeartbeatTask {
                        let ctx = HeartbeatResponseHandlerContext::new(mailbox.clone(), resp);
                        if let Err(e) = capture_self.handle_response(ctx).await {
                            error!(e; "Error while handling heartbeat response");
+                            HEARTBEAT_RECV_COUNT
+                                .with_label_values(&["processing_error"])
+                                .inc();
+                        } else {
+                            HEARTBEAT_RECV_COUNT.with_label_values(&["success"]).inc();
                        }
                    }
                    Ok(None) => break,
                    Err(e) => {
+                        HEARTBEAT_RECV_COUNT.with_label_values(&["error"]).inc();
                        error!(e; "Occur error while reading heartbeat response");
                        capture_self
                            .start_with_retry(Duration::from_millis(retry_interval))
@@ -180,6 +187,7 @@ impl HeartbeatTask {
                        error!(e; "Failed to send heartbeat to metasrv");
                        break;
                    } else {
+                        HEARTBEAT_SENT_COUNT.inc();
                        debug!("Send a heartbeat request to metasrv, content: {:?}", req);
                    }
                }
--- a/src/frontend/src/heartbeat/handler/tests.rs
+++ b/src/frontend/src/heartbeat/handler/tests.rs
@@ -24,7 +24,7 @@ use common_meta::heartbeat::mailbox::{HeartbeatMailbox, MessageMeta};
 use common_meta::instruction::{CacheIdent, Instruction};
 use common_meta::key::schema_name::{SchemaName, SchemaNameKey};
 use common_meta::key::table_info::TableInfoKey;
-use common_meta::key::MetaKey;
+use common_meta::key::MetadataKey;
 use partition::manager::TableRouteCacheInvalidator;
 use table::metadata::TableId;
 use tokio::sync::mpsc;
--- a/src/frontend/src/instance.rs
+++ b/src/frontend/src/instance.rs
@@ -79,9 +79,9 @@ pub use standalone::StandaloneDatanodeManager;

 use self::prom_store::ExportMetricHandler;
 use crate::error::{
-    self, Error, ExecLogicalPlanSnafu, ExecutePromqlSnafu, ExternalSnafu, ParseSqlSnafu,
-    PermissionSnafu, PlanStatementSnafu, Result, SqlExecInterceptedSnafu, StartServerSnafu,
-    TableOperationSnafu,
+    self, Error, ExecLogicalPlanSnafu, ExecutePromqlSnafu, ExternalSnafu, InvalidSqlSnafu,
+    ParseSqlSnafu, PermissionSnafu, PlanStatementSnafu, Result, SqlExecInterceptedSnafu,
+    StartServerSnafu, TableOperationSnafu,
 };
 use crate::frontend::FrontendOptions;
 use crate::heartbeat::HeartbeatTask;
@@ -452,6 +452,9 @@ pub fn check_permission(
        | Statement::DropDatabase(_)
        | Statement::DropFlow(_)
        | Statement::Use(_) => {}
+        Statement::ShowCreateDatabase(stmt) => {
+            validate_database(&stmt.database_name, query_ctx)?;
+        }
        Statement::ShowCreateTable(stmt) => {
            validate_param(&stmt.table_name, query_ctx)?;
        }
@@ -527,8 +530,8 @@ pub fn check_permission(
        },
        Statement::Copy(sql::statements::copy::Copy::CopyDatabase(copy_database)) => {
            match copy_database {
-                CopyDatabase::To(stmt) => validate_param(&stmt.database_name, query_ctx)?,
-                CopyDatabase::From(stmt) => validate_param(&stmt.database_name, query_ctx)?,
+                CopyDatabase::To(stmt) => validate_database(&stmt.database_name, query_ctx)?,
+                CopyDatabase::From(stmt) => validate_database(&stmt.database_name, query_ctx)?,
            }
        }
        Statement::TruncateTable(stmt) => {
@@ -548,6 +551,26 @@ fn validate_param(name: &ObjectName, query_ctx: &QueryContextRef) -> Result<()>
        .context(SqlExecInterceptedSnafu)
 }

+fn validate_database(name: &ObjectName, query_ctx: &QueryContextRef) -> Result<()> {
+    let (catalog, schema) = match &name.0[..] {
+        [schema] => (
+            query_ctx.current_catalog().to_string(),
+            schema.value.clone(),
+        ),
+        [catalog, schema] => (catalog.value.clone(), schema.value.clone()),
+        _ => InvalidSqlSnafu {
+            err_msg: format!(
+                "expect database name to be <catalog>.<schema> or <schema>, actual: {name}",
+            ),
+        }
+        .fail()?,
+    };
+
+    validate_catalog_and_schema(&catalog, &schema, query_ctx)
+        .map_err(BoxedError::new)
+        .context(SqlExecInterceptedSnafu)
+}
+
 #[cfg(test)]
 mod tests {
    use std::collections::HashMap;
--- a/src/frontend/src/metrics.rs
+++ b/src/frontend/src/metrics.rs
@@ -51,4 +51,17 @@ lazy_static! {
        "frontend otlp traces rows"
    )
    .unwrap();
+    /// The number of heartbeats send by frontend node.
+    pub static ref HEARTBEAT_SENT_COUNT: IntCounter = register_int_counter!(
+        "greptime_frontend_heartbeat_send_count",
+        "frontend heartbeat sent",
+    )
+    .unwrap();
+    /// The number of heartbeats received by frontend node, labeled with result type.
+    pub static ref HEARTBEAT_RECV_COUNT: IntCounterVec = register_int_counter_vec!(
+        "greptime_frontend_heartbeat_recv_count",
+        "frontend heartbeat received",
+        &["result"]
+    )
+    .unwrap();
 }
--- a/src/index/src/inverted_index/create/sort/intermediate_rw/codec_v1.rs
+++ b/src/index/src/inverted_index/create/sort/intermediate_rw/codec_v1.rs
@@ -17,9 +17,9 @@ use std::io;
 use asynchronous_codec::{BytesMut, Decoder, Encoder};
 use bytes::{Buf, BufMut};
 use common_base::BitVec;
-use snafu::location;
+use snafu::ResultExt;

-use crate::inverted_index::error::{Error, Result};
+use crate::inverted_index::error::{CommonIoSnafu, Error, Result};
 use crate::inverted_index::Bytes;

 const U64_LENGTH: usize = std::mem::size_of::<u64>();
@@ -105,10 +105,9 @@ impl Decoder for IntermediateItemDecoderV1 {
 /// Required for [`Encoder`] and [`Decoder`] implementations.
 impl From<io::Error> for Error {
    fn from(error: io::Error) -> Self {
-        Error::CommonIoError {
-            error,
-            location: location!(),
-        }
+        Err::<(), io::Error>(error)
+            .context(CommonIoSnafu)
+            .unwrap_err()
    }
 }

--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Yingwen	506dc20765	fix: last non null iter not init (#4687 )	2024-09-06 04:13:23 +00:00
Lei, HUANG	114772ba87	chore: bump version v0.9.3 (#4684 )	2024-09-06 02:31:41 +00:00
liyang	89a3da8a3a	chore(dockerfile): remove mysql and postgresql clients in greptimedb image (#4685 )	2024-09-05 16:00:53 +00:00
jeremyhi	8814695b58	feat: invalidate cache via invalidator on region migration (#4682 ) feat: invalidate table via invalidator on region migration	2024-09-05 06:15:38 +00:00
Lanqing Yang	86cef648cd	feat: add more spans to mito engine (#4643 ) feat: add more span on mito engine	2024-09-05 06:13:22 +00:00
Ning Sun	e476e36647	feat: add geohash and h3 as built-in functions (#4656 ) * feat: add built-in functions h3 and geohash * tests: add sqlness tests for geo functions * doc: correct h3 comment * fix: lint error * fix: toml format * refactor: address review comments * test: add more sqlness cases * Apply suggestions from code review Co-authored-by: Ruihang Xia <waynestxia@gmail.com> --------- Co-authored-by: Ruihang Xia <waynestxia@gmail.com>	2024-09-05 04:42:29 +00:00
shuiyisong	4781b327f3	fix: ref to auth err (#4681 ) * fix: ref to auth err * fix: typo	2024-09-05 04:05:39 +00:00
LFC	3e4a69017d	build: add mysql and postgresql clients to greptimedb image (#4677 )	2024-09-04 11:38:47 +00:00
LFC	d43e31c7ed	feat: schedule compaction when adding sst files by editing region (#4648 ) * feat: schedule compaction when adding sst files by editing region * add minimum time interval for two successive compactions * resolve PR comments	2024-09-04 10:10:07 +00:00
discord9	19e2a9d44b	feat: change log level dynamically (#4653 ) * feat: add dyn_log handle * feat: use reload handle * chore: per review	2024-09-04 07:54:50 +00:00
zyy17	8453df1392	refactor: make `init_global_logging()` clean and add `log_format` (#4657 ) refactor: refine the code logic of init_global_logging and add json output format	2024-09-04 03:04:51 +00:00
Ruihang Xia	8ca35a4a1a	fix: use number of partitions as parallilism in region scanner (#4669 ) * fix: use number of partitions as parallilism in region scanner Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * add sqlness Signed-off-by: Ruihang Xia <waynestxia@gmail.com> Co-authored-by: Lei HUANG <mrsatangel@gmail.com> * order by ts Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * debug pring time range Signed-off-by: Ruihang Xia <waynestxia@gmail.com> --------- Signed-off-by: Ruihang Xia <waynestxia@gmail.com> Co-authored-by: Lei HUANG <mrsatangel@gmail.com>	2024-09-03 13:42:38 +00:00
Ruihang Xia	93f202694c	refactor: remove unused error variants (#4666 ) * add python script Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * remove unused errors Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * fix all negative cases Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * setup CI Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * add license header Signed-off-by: Ruihang Xia <waynestxia@gmail.com> --------- Signed-off-by: Ruihang Xia <waynestxia@gmail.com>	2024-09-03 13:19:38 +00:00
Weny Xu	b52e3c694a	chore(ci): set etcd resources limits (#4665 )	2024-09-03 07:25:23 +00:00
dennis zhuang	a612b67470	feat: supports name in object storage config (#4630 ) * feat: supports name in object storage config * fix: integration test * fix: integration test * fix: update sample config * fix: config api test	2024-09-03 07:02:55 +00:00
jeremyhi	9b03940e03	chore: refactor metadata key value trait (#4664 )	2024-09-03 07:00:24 +00:00
jeremyhi	8d6cd8ae16	feat: export import database (#4654 ) * feat: export database create sql * feat: import create database * Update src/cmd/src/cli/export.rs Co-authored-by: Weny Xu <wenymedia@gmail.com> * Update src/cmd/src/cli/import.rs Co-authored-by: Weny Xu <wenymedia@gmail.com> * Update src/cmd/src/error.rs Co-authored-by: Ruihang Xia <waynestxia@gmail.com> * chore: make show create fail fast --------- Co-authored-by: Weny Xu <wenymedia@gmail.com> Co-authored-by: Ruihang Xia <waynestxia@gmail.com>	2024-09-03 04:45:25 +00:00
dennis zhuang	8f4ec536de	feat: grpc writing supports TTL hint (#4651 )	2024-09-03 02:15:01 +00:00
zyy17	f0e2d6e663	fix: use 'target' for 'actions-rust-lang/setup-rust-toolchain' to fix cross build failed (#4661 )	2024-09-02 06:11:12 +00:00
Weny Xu	306bd25c64	fix: expose missing options for initializing regions (#4660 ) * fix: expose `init_regions_in_background` and `init_regions_parallelism` opts * fix: ci	2024-09-02 03:11:18 +00:00
zyy17	ddafcc678c	ci: disable macos integration test and some minor refactoring (#4658 )	2024-09-02 03:06:17 +00:00
Weny Xu	2564b5daee	fix: correct otlp endpoint formatting (#4646 )	2024-09-02 02:59:50 +00:00
Lei, HUANG	37dcf34bb9	fix(mito): avoid caching empty batches in row group (#4652 ) * fix: avoid caching empty batches in row group * fix: clippy * Update tests/cases/standalone/common/select/last_value.sql * fix: sqlness	2024-09-02 02:43:00 +00:00
Yingwen	8eda36bfe3	feat: remove files from the write cache in purger (#4655 ) * feat: remove files from the write cache in purger * chore: fix typo	2024-08-31 04:19:52 +00:00
Ruihang Xia	68b59e0e5e	feat: remove the requirement that partition column must be PK (#4647 ) Signed-off-by: Ruihang Xia <waynestxia@gmail.com>	2024-08-31 03:16:01 +00:00
Ruihang Xia	a37aeb2814	feat: initialize partition range from ScanInput (#4635 ) * feat: initialize partition range from ScanInput Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * use num_rows instead Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * add todo Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * setup unordered scan Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * Update src/mito2/src/read/scan_region.rs Co-authored-by: jeremyhi <jiachun_feng@proton.me> * leave unordered scan unchanged Signed-off-by: Ruihang Xia <waynestxia@gmail.com> --------- Signed-off-by: Ruihang Xia <waynestxia@gmail.com> Co-authored-by: jeremyhi <jiachun_feng@proton.me>	2024-08-30 07:30:37 +00:00
jeremyhi	f641c562c2	feat: show create database (#4642 ) * feat: show create database * feat: add sqlness test * chore: reorder mod and use * feat: show create schema * Update src/frontend/src/instance.rs	2024-08-30 03:58:11 +00:00
Lanqing Yang	9286e963e7	chore: adding heartbeat sent/recv counts in greptimedb nodes (#4624 ) obs: adding heartbeat sent/recv counts in greptimedb nodes	2024-08-30 03:57:16 +00:00
LFC	8ea4f67e4b	refactor: reduce a object store "stat" call (#4645 )	2024-08-30 03:31:19 +00:00
jeremyhi	5e4bac2633	feat: import cli tool (#4639 ) * feat: import create tables * feat: import databasse * fix: export view schema	2024-08-29 09:32:21 +00:00
LFC	d45b04180c	feat: pre-download the ingested sst (#4636 ) * refactor: pre-read the ingested sst file in object store to fill the local cache to accelerate first query * feat: pre-download the ingested SST from remote to accelerate following reads * resolve PR comments * resolve PR comments	2024-08-29 08:36:41 +00:00
discord9	8c8499ce53	perf(flow): Map&Reduce Operator use batch to reduce alloc (#4567 ) * feat: partial impl mfp * feat: eval batch inner * chore: fmt * feat: mfp eval_batch * WIP * feat: Collection generic over row&Batch * feat: render source batch * chore: chore * feat: render mfp batch * feat: render reduce batch(WIP) * feat(WIP): render reduce * feat: reduce batch * feat: render sink batch * feat: render constant batch * chore: error handling& mfp batch test * test: mfp batch * chore: rm import * test: render reduce batch * chore: add TODO * chore: per bot review * refactor: per review * chore: cmt * chore: rename * docs: update no panic	2024-08-29 07:28:13 +00:00
Weny Xu	79f40a762b	fix: set `selector_result_cache_size` in unit test again (#4641 )	2024-08-29 07:14:40 +00:00
jeremyhi	b062d8515d	feat: copy database ignores view and temporary tables (#4640 ) feat: copy database ingores view and temporary tables	2024-08-29 06:17:51 +00:00
discord9	9f9c1dab60	feat(flow): use DataFusion's optimizer (#4489 ) * feat: use datafusion optimization refactor: mv `sql_to_flow_plan` elsewhere feat(WIP): use df optimization WIP analyzer rule feat(WIP): avg expander fix: transform avg expander fix: avg expand feat: names from substrait fix: avg rewrite test: update `test_avg`&`test_avg_group_by` test: fix `test_sum` test: fix some tests chore: remove unused flow plan transform feat: tumble expander test: update tests * chore: clippy * fix: tumble lose `group expr` * test: sqlness test update * test: rm unused cast * test: simplify sqlness * refactor: per review * chore: after rebase * fix: remove a outdated test * test: add comment * fix: report error when not literal * chore: update sqlness test after rebase * refactor: per review	2024-08-29 02:52:00 +00:00