assign partition_ranges

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

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

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

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

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

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

@@ -8,7 +8,7 @@ inputs:
     default: 2
     description: "Number of Datanode replicas"
   meta-replicas:
-    default: 1
+    default: 2
     description: "Number of Metasrv replicas"
   image-registry: 
     default: "docker.io"

.github/workflows/dev-build.yml

@@ -238,6 +238,13 @@ jobs:
           version: ${{ needs.allocate-runners.outputs.version }}
           push-latest-tag: false # Don't push the latest tag to registry.
           dev-mode: true # Only build the standard images.
+          
+      - name: Echo Docker image tag to step summary
+        run: |
+          echo "## Docker Image Tag" >> $GITHUB_STEP_SUMMARY
+          echo "Image Tag: \`${{ needs.allocate-runners.outputs.version }}\`" >> $GITHUB_STEP_SUMMARY
+          echo "Full Image Name: \`docker.io/${{ vars.IMAGE_NAMESPACE }}/${{ vars.DEV_BUILD_IMAGE_NAME }}:${{ needs.allocate-runners.outputs.version }}\`" >> $GITHUB_STEP_SUMMARY
+          echo "Pull Command: \`docker pull docker.io/${{ vars.IMAGE_NAMESPACE }}/${{ vars.DEV_BUILD_IMAGE_NAME }}:${{ needs.allocate-runners.outputs.version }}\`" >> $GITHUB_STEP_SUMMARY
 
       - name: Set build result
         id: set-build-result

.github/workflows/develop.yml

@@ -111,7 +111,7 @@ jobs:
       - name: Build greptime binaries
         shell: bash
         # `cargo gc` will invoke `cargo build` with specified args
-        run: cargo gc -- --bin greptime --bin sqlness-runner --features pg_kvbackend
+        run: cargo gc -- --bin greptime --bin sqlness-runner --features "pg_kvbackend,mysql_kvbackend"
       - name: Pack greptime binaries
         shell: bash
         run: |
@@ -270,7 +270,7 @@ jobs:
       - name: Build greptime bianry
         shell: bash
         # `cargo gc` will invoke `cargo build` with specified args
-        run: cargo gc --profile ci -- --bin greptime --features pg_kvbackend
+        run: cargo gc --profile ci -- --bin greptime --features "pg_kvbackend,mysql_kvbackend"
       - name: Pack greptime binary
         shell: bash
         run: |
@@ -576,9 +576,12 @@ jobs:
           - name: "Remote WAL"
             opts: "-w kafka -k 127.0.0.1:9092"
             kafka: true
-          - name: "Pg Kvbackend"
+          - name: "PostgreSQL KvBackend"
             opts: "--setup-pg"
             kafka: false
+          - name: "MySQL Kvbackend"
+            opts: "--setup-mysql"
+            kafka: false
     timeout-minutes: 60
     steps:
       - uses: actions/checkout@v4
@@ -687,7 +690,7 @@ jobs:
         working-directory: tests-integration/fixtures
         run: docker compose up -d --wait
       - name: Run nextest cases
-        run: cargo nextest run --workspace -F dashboard -F pg_kvbackend
+        run: cargo nextest run --workspace -F dashboard -F pg_kvbackend -F mysql_kvbackend
         env:
           CARGO_BUILD_RUSTFLAGS: "-C link-arg=-fuse-ld=mold"
           RUST_BACKTRACE: 1
@@ -704,6 +707,7 @@ jobs:
           GT_MINIO_ENDPOINT_URL: http://127.0.0.1:9000
           GT_ETCD_ENDPOINTS: http://127.0.0.1:2379
           GT_POSTGRES_ENDPOINTS: postgres://greptimedb:admin@127.0.0.1:5432/postgres
+          GT_MYSQL_ENDPOINTS: mysql://greptimedb:admin@127.0.0.1:3306/mysql
           GT_KAFKA_ENDPOINTS: 127.0.0.1:9092
           GT_KAFKA_SASL_ENDPOINTS: 127.0.0.1:9093
           UNITTEST_LOG_DIR: "__unittest_logs"
@@ -739,7 +743,7 @@ jobs:
         working-directory: tests-integration/fixtures
         run: docker compose up -d --wait
       - name: Run nextest cases
-        run: cargo llvm-cov nextest --workspace --lcov --output-path lcov.info -F dashboard -F pg_kvbackend
+        run: cargo llvm-cov nextest --workspace --lcov --output-path lcov.info -F dashboard -F pg_kvbackend -F mysql_kvbackend
         env:
           CARGO_BUILD_RUSTFLAGS: "-C link-arg=-fuse-ld=mold"
           RUST_BACKTRACE: 1
@@ -755,6 +759,7 @@ jobs:
           GT_MINIO_ENDPOINT_URL: http://127.0.0.1:9000
           GT_ETCD_ENDPOINTS: http://127.0.0.1:2379
           GT_POSTGRES_ENDPOINTS: postgres://greptimedb:admin@127.0.0.1:5432/postgres
+          GT_MYSQL_ENDPOINTS: mysql://greptimedb:admin@127.0.0.1:3306/mysql
           GT_KAFKA_ENDPOINTS: 127.0.0.1:9092
           GT_KAFKA_SASL_ENDPOINTS: 127.0.0.1:9093
           UNITTEST_LOG_DIR: "__unittest_logs"

.github/workflows/nightly-ci.yml

@@ -107,7 +107,6 @@ jobs:
           CARGO_BUILD_RUSTFLAGS: "-C linker=lld-link"
           RUST_BACKTRACE: 1
           CARGO_INCREMENTAL: 0
-          RUSTUP_WINDOWS_PATH_ADD_BIN: 1 # Workaround for https://github.com/nextest-rs/nextest/issues/1493
           GT_S3_BUCKET: ${{ vars.AWS_CI_TEST_BUCKET }}
           GT_S3_ACCESS_KEY_ID: ${{ secrets.AWS_CI_TEST_ACCESS_KEY_ID }}
           GT_S3_ACCESS_KEY: ${{ secrets.AWS_CI_TEST_SECRET_ACCESS_KEY }}

.github/workflows/release.yml

@@ -91,7 +91,7 @@ env:
   # The scheduled version is '${{ env.NEXT_RELEASE_VERSION }}-nightly-YYYYMMDD', like v0.2.0-nigthly-20230313;
   NIGHTLY_RELEASE_PREFIX: nightly
   # Note: The NEXT_RELEASE_VERSION should be modified manually by every formal release.
-  NEXT_RELEASE_VERSION: v0.13.0
+  NEXT_RELEASE_VERSION: v0.14.0
 
 jobs:
   allocate-runners:

.gitignore

@@ -54,3 +54,6 @@ tests-fuzz/corpus/
 # Nix
 .direnv
 .envrc
+
+## default data home
+greptimedb_data

Cargo.toml

@@ -29,6 +29,7 @@ members = [
     "src/common/query",
     "src/common/recordbatch",
     "src/common/runtime",
+    "src/common/session",
     "src/common/substrait",
     "src/common/telemetry",
     "src/common/test-util",
@@ -67,7 +68,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "0.13.0"
+version = "0.14.0"
 edition = "2021"
 license = "Apache-2.0"
 
@@ -88,7 +89,7 @@ rust.unexpected_cfgs = { level = "warn", check-cfg = ['cfg(tokio_unstable)'] }
 #
 # See for more detaiils: https://github.com/rust-lang/cargo/issues/11329
 ahash = { version = "0.8", features = ["compile-time-rng"] }
-aquamarine = "0.3"
+aquamarine = "0.6"
 arrow = { version = "53.0.0", features = ["prettyprint"] }
 arrow-array = { version = "53.0.0", default-features = false, features = ["chrono-tz"] }
 arrow-flight = "53.0"
@@ -99,9 +100,9 @@ async-trait = "0.1"
 # Remember to update axum-extra, axum-macros when updating axum
 axum = "0.8"
 axum-extra = "0.10"
-axum-macros = "0.4"
+axum-macros = "0.5"
 backon = "1"
-base64 = "0.21"
+base64 = "0.22"
 bigdecimal = "0.4.2"
 bitflags = "2.4.1"
 bytemuck = "1.12"
@@ -111,7 +112,7 @@ chrono-tz = "0.10.1"
 clap = { version = "4.4", features = ["derive"] }
 config = "0.13.0"
 crossbeam-utils = "0.8"
-dashmap = "5.4"
+dashmap = "6.1"
 datafusion = { git = "https://github.com/apache/datafusion.git", rev = "2464703c84c400a09cc59277018813f0e797bb4e" }
 datafusion-common = { git = "https://github.com/apache/datafusion.git", rev = "2464703c84c400a09cc59277018813f0e797bb4e" }
 datafusion-expr = { git = "https://github.com/apache/datafusion.git", rev = "2464703c84c400a09cc59277018813f0e797bb4e" }
@@ -121,31 +122,31 @@ datafusion-physical-expr = { git = "https://github.com/apache/datafusion.git", r
 datafusion-physical-plan = { git = "https://github.com/apache/datafusion.git", rev = "2464703c84c400a09cc59277018813f0e797bb4e" }
 datafusion-sql = { git = "https://github.com/apache/datafusion.git", rev = "2464703c84c400a09cc59277018813f0e797bb4e" }
 datafusion-substrait = { git = "https://github.com/apache/datafusion.git", rev = "2464703c84c400a09cc59277018813f0e797bb4e" }
-deadpool = "0.10"
-deadpool-postgres = "0.12"
-derive_builder = "0.12"
+deadpool = "0.12"
+deadpool-postgres = "0.14"
+derive_builder = "0.20"
 dotenv = "0.15"
 etcd-client = "0.14"
 fst = "0.4.7"
 futures = "0.3"
 futures-util = "0.3"
-greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "d92c9ac4e90ef4abdcf5c2eaf5a164e18ba09486" }
+greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "dd4a1996982534636734674db66e44464b0c0d83" }
 hex = "0.4"
 http = "1"
 humantime = "2.1"
 humantime-serde = "1.1"
 hyper = "1.1"
 hyper-util = "0.1"
-itertools = "0.10"
+itertools = "0.14"
 jsonb = { git = "https://github.com/databendlabs/jsonb.git", rev = "8c8d2fc294a39f3ff08909d60f718639cfba3875", default-features = false }
 lazy_static = "1.4"
 local-ip-address = "0.6"
 loki-proto = { git = "https://github.com/GreptimeTeam/loki-proto.git", rev = "1434ecf23a2654025d86188fb5205e7a74b225d3" }
 meter-core = { git = "https://github.com/GreptimeTeam/greptime-meter.git", rev = "5618e779cf2bb4755b499c630fba4c35e91898cb" }
-mockall = "0.11.4"
+mockall = "0.13"
 moka = "0.12"
 nalgebra = "0.33"
-notify = "6.1"
+notify = "8.0"
 num_cpus = "1.16"
 once_cell = "1.18"
 opentelemetry-proto = { version = "0.27", features = [
@@ -163,8 +164,8 @@ prometheus = { version = "0.13.3", features = ["process"] }
 promql-parser = { version = "0.5", features = ["ser"] }
 prost = "0.13"
 raft-engine = { version = "0.4.1", default-features = false }
-rand = "0.8"
-ratelimit = "0.9"
+rand = "0.9"
+ratelimit = "0.10"
 regex = "1.8"
 regex-automata = "0.4"
 reqwest = { version = "0.12", default-features = false, features = [
@@ -176,29 +177,37 @@ reqwest = { version = "0.12", default-features = false, features = [
 rskafka = { git = "https://github.com/influxdata/rskafka.git", rev = "75535b5ad9bae4a5dbb582c82e44dfd81ec10105", features = [
     "transport-tls",
 ] }
-rstest = "0.21"
+rstest = "0.25"
 rstest_reuse = "0.7"
 rust_decimal = "1.33"
 rustc-hash = "2.0"
-rustls = { version = "0.23.20", default-features = false } # override by patch, see [patch.crates-io]
+# It is worth noting that we should try to avoid using aws-lc-rs until it can be compiled on various platforms.
+rustls = { version = "0.23.25", default-features = false }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = { version = "1.0", features = ["float_roundtrip"] }
 serde_with = "3"
-shadow-rs = "0.38"
+shadow-rs = "1.1"
+simd-json = "0.15"
 similar-asserts = "1.6.0"
 smallvec = { version = "1", features = ["serde"] }
 snafu = "0.8"
-sysinfo = "0.30"
+sqlx = { version = "0.8", features = [
+    "runtime-tokio-rustls",
+    "mysql",
+    "postgres",
+    "chrono",
+] }
+sysinfo = "0.33"
 # on branch v0.52.x
 sqlparser = { git = "https://github.com/GreptimeTeam/sqlparser-rs.git", rev = "71dd86058d2af97b9925093d40c4e03360403170", features = [
     "visitor",
     "serde",
 ] } # on branch v0.44.x
-strum = { version = "0.25", features = ["derive"] }
+strum = { version = "0.27", features = ["derive"] }
 tempfile = "3"
 tokio = { version = "1.40", features = ["full"] }
 tokio-postgres = "0.7"
-tokio-rustls = { version = "0.26.0", default-features = false } # override by patch, see [patch.crates-io]
+tokio-rustls = { version = "0.26.2", default-features = false }
 tokio-stream = "0.1"
 tokio-util = { version = "0.7", features = ["io-util", "compat"] }
 toml = "0.8.8"
@@ -241,6 +250,7 @@ common-procedure-test = { path = "src/common/procedure-test" }
 common-query = { path = "src/common/query" }
 common-recordbatch = { path = "src/common/recordbatch" }
 common-runtime = { path = "src/common/runtime" }
+common-session = { path = "src/common/session" }
 common-telemetry = { path = "src/common/telemetry" }
 common-test-util = { path = "src/common/test-util" }
 common-time = { path = "src/common/time" }
@@ -273,15 +283,6 @@ store-api = { path = "src/store-api" }
 substrait = { path = "src/common/substrait" }
 table = { path = "src/table" }
 
-[patch.crates-io]
-# change all rustls dependencies to use our fork to default to `ring` to make it "just work"
-hyper-rustls = { git = "https://github.com/GreptimeTeam/hyper-rustls", rev = "a951e03" } # version = "0.27.5" with ring patch
-rustls = { git = "https://github.com/GreptimeTeam/rustls", rev = "34fd0c6" }             # version = "0.23.20" with ring patch
-tokio-rustls = { git = "https://github.com/GreptimeTeam/tokio-rustls", rev = "4604ca6" } # version = "0.26.0" with ring patch
-# This is commented, since we are not using aws-lc-sys, if we need to use it, we need to uncomment this line or use a release after this commit, or it wouldn't compile with gcc < 8.1
-# see https://github.com/aws/aws-lc-rs/pull/526
-# aws-lc-sys = { git ="https://github.com/aws/aws-lc-rs", rev = "556558441e3494af4b156ae95ebc07ebc2fd38aa" }
-
 [workspace.dependencies.meter-macros]
 git = "https://github.com/GreptimeTeam/greptime-meter.git"
 rev = "5618e779cf2bb4755b499c630fba4c35e91898cb"

README.md

@@ -6,7 +6,7 @@
   </picture>
 </p>
 
-<h2 align="center">Unified & Cost-Effective Time Series Database for Metrics, Logs, and Events</h2>
+<h2 align="center">Unified & Cost-Effective Observerability Database for Metrics, Logs, and Events</h2>
 
 <div align="center">
 <h3 align="center">
@@ -62,15 +62,19 @@
 
 ## Introduction
 
-**GreptimeDB** is an open-source unified & cost-effective time-series database for **Metrics**, **Logs**, and **Events** (also **Traces** in plan). You can gain real-time insights from Edge to Cloud at Any Scale.
+**GreptimeDB** is an open-source unified & cost-effective observerability database for **Metrics**, **Logs**, and **Events** (also **Traces** in plan). You can gain real-time insights from Edge to Cloud at Any Scale.
+
+## News
+
+**[GreptimeDB tops JSONBench's billion-record cold run test!](https://greptime.com/blogs/2025-03-18-jsonbench-greptimedb-performance)**
 
 ## Why GreptimeDB
 
-Our core developers have been building time-series data platforms for years. Based on our best practices, GreptimeDB was born to give you:
+Our core developers have been building observerability data platforms for years. Based on our best practices, GreptimeDB was born to give you:
 
 * **Unified Processing of Metrics, Logs, and Events**
 
-  GreptimeDB unifies time series data processing by treating all data - whether metrics, logs, or events - as timestamped events with context. Users can analyze this data using either [SQL](https://docs.greptime.com/user-guide/query-data/sql) or [PromQL](https://docs.greptime.com/user-guide/query-data/promql) and leverage stream processing ([Flow](https://docs.greptime.com/user-guide/flow-computation/overview)) to enable continuous aggregation. [Read more](https://docs.greptime.com/user-guide/concepts/data-model).
+  GreptimeDB unifies observerability data processing by treating all data - whether metrics, logs, or events - as timestamped events with context. Users can analyze this data using either [SQL](https://docs.greptime.com/user-guide/query-data/sql) or [PromQL](https://docs.greptime.com/user-guide/query-data/promql) and leverage stream processing ([Flow](https://docs.greptime.com/user-guide/flow-computation/overview)) to enable continuous aggregation. [Read more](https://docs.greptime.com/user-guide/concepts/data-model).
 
 * **Cloud-native Distributed Database**
 
@@ -112,7 +116,7 @@ Start a GreptimeDB container with:
 
 ```shell
 docker run -p 127.0.0.1:4000-4003:4000-4003 \
-  -v "$(pwd)/greptimedb:/tmp/greptimedb" \
+  -v "$(pwd)/greptimedb:./greptimedb_data" \
   --name greptime --rm \
   greptime/greptimedb:latest standalone start \
   --http-addr 0.0.0.0:4000 \

config/config.md

@@ -12,7 +12,6 @@
 
 | Key | Type | Default | Descriptions |
 | --- | -----| ------- | ----------- |
-| `mode` | String | `standalone` | The running mode of the datanode. It can be `standalone` or `distributed`. |
 | `default_timezone` | String | Unset | 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. |
@@ -24,7 +23,7 @@
 | `runtime.compact_rt_size` | Integer | `4` | The number of threads to execute the runtime for global write operations. |
 | `http` | -- | -- | The HTTP server options. |
 | `http.addr` | String | `127.0.0.1:4000` | The address to bind the HTTP server. |
-| `http.timeout` | String | `30s` | HTTP request timeout. Set to 0 to disable timeout. |
+| `http.timeout` | String | `0s` | HTTP request timeout. Set to 0 to disable timeout. |
 | `http.body_limit` | String | `64MB` | HTTP request body limit.<br/>The following units are supported: `B`, `KB`, `KiB`, `MB`, `MiB`, `GB`, `GiB`, `TB`, `TiB`, `PB`, `PiB`.<br/>Set to 0 to disable limit. |
 | `http.enable_cors` | Bool | `true` | HTTP CORS support, it's turned on by default<br/>This allows browser to access http APIs without CORS restrictions |
 | `http.cors_allowed_origins` | Array | Unset | Customize allowed origins for HTTP CORS. |
@@ -86,10 +85,6 @@
 | `wal.create_topic_timeout` | String | `30s` | Above which a topic creation operation will be cancelled.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.max_batch_bytes` | String | `1MB` | The max size of a single producer batch.<br/>Warning: Kafka has a default limit of 1MB per message in a topic.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.consumer_wait_timeout` | String | `100ms` | The consumer wait timeout.<br/>**It's only used when the provider is `kafka`**. |
-| `wal.backoff_init` | String | `500ms` | The initial backoff delay.<br/>**It's only used when the provider is `kafka`**. |
-| `wal.backoff_max` | String | `10s` | The maximum backoff delay.<br/>**It's only used when the provider is `kafka`**. |
-| `wal.backoff_base` | Integer | `2` | The exponential backoff rate, i.e. next backoff = base * current backoff.<br/>**It's only used when the provider is `kafka`**. |
-| `wal.backoff_deadline` | String | `5mins` | The deadline of retries.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.overwrite_entry_start_id` | Bool | `false` | Ignore missing entries during read WAL.<br/>**It's only used when the provider is `kafka`**.<br/><br/>This option ensures that when Kafka messages are deleted, the system<br/>can still successfully replay memtable data without throwing an<br/>out-of-range error.<br/>However, enabling this option might lead to unexpected data loss,<br/>as the system will skip over missing entries instead of treating<br/>them as critical errors. |
 | `metadata_store` | -- | -- | Metadata storage options. |
 | `metadata_store.file_size` | String | `64MB` | The size of the metadata store log file. |
@@ -98,10 +93,11 @@
 | `procedure` | -- | -- | Procedure storage options. |
 | `procedure.max_retry_times` | Integer | `3` | Procedure max retry time. |
 | `procedure.retry_delay` | String | `500ms` | Initial retry delay of procedures, increases exponentially |
+| `procedure.max_running_procedures` | Integer | `128` | Max running procedures.<br/>The maximum number of procedures that can be running at the same time.<br/>If the number of running procedures exceeds this limit, the procedure will be rejected. |
 | `flow` | -- | -- | flow engine options. |
 | `flow.num_workers` | Integer | `0` | The number of flow worker in flownode.<br/>Not setting(or set to 0) this value will use the number of CPU cores divided by 2. |
 | `storage` | -- | -- | The data storage options. |
-| `storage.data_home` | String | `/tmp/greptimedb/` | The working home directory. |
+| `storage.data_home` | String | `./greptimedb_data/` | The working home directory. |
 | `storage.type` | String | `File` | The storage type used to store the data.<br/>- `File`: the data is stored in the local file system.<br/>- `S3`: the data is stored in the S3 object storage.<br/>- `Gcs`: the data is stored in the Google Cloud Storage.<br/>- `Azblob`: the data is stored in the Azure Blob Storage.<br/>- `Oss`: the data is stored in the Aliyun OSS. |
 | `storage.cache_path` | String | Unset | Read cache configuration for object storage such as 'S3' etc, it's configured by default when using object storage. It is recommended to configure it when using object storage for better performance.<br/>A local file directory, defaults to `{data_home}`. An empty string means disabling. |
 | `storage.cache_capacity` | String | Unset | The local file cache capacity in bytes. If your disk space is sufficient, it is recommended to set it larger. |
@@ -181,7 +177,7 @@
 | `region_engine.metric` | -- | -- | Metric engine options. |
 | `region_engine.metric.experimental_sparse_primary_key_encoding` | Bool | `false` | Whether to enable the experimental sparse primary key encoding. |
 | `logging` | -- | -- | The logging options. |
-| `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.dir` | String | `./greptimedb_data/logs` | The directory to store the log files. If set to empty, logs will not be written to files. |
 | `logging.level` | String | Unset | 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. |
@@ -222,7 +218,7 @@
 | `heartbeat.retry_interval` | String | `3s` | Interval for retrying to send heartbeat messages to the metasrv. |
 | `http` | -- | -- | The HTTP server options. |
 | `http.addr` | String | `127.0.0.1:4000` | The address to bind the HTTP server. |
-| `http.timeout` | String | `30s` | HTTP request timeout. Set to 0 to disable timeout. |
+| `http.timeout` | String | `0s` | HTTP request timeout. Set to 0 to disable timeout. |
 | `http.body_limit` | String | `64MB` | HTTP request body limit.<br/>The following units are supported: `B`, `KB`, `KiB`, `MB`, `MiB`, `GB`, `GiB`, `TB`, `TiB`, `PB`, `PiB`.<br/>Set to 0 to disable limit. |
 | `http.enable_cors` | Bool | `true` | HTTP CORS support, it's turned on by default<br/>This allows browser to access http APIs without CORS restrictions |
 | `http.cors_allowed_origins` | Array | Unset | Customize allowed origins for HTTP CORS. |
@@ -279,7 +275,7 @@
 | `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. If set to empty, logs will not be written to files. |
+| `logging.dir` | String | `./greptimedb_data/logs` | The directory to store the log files. If set to empty, logs will not be written to files. |
 | `logging.level` | String | Unset | 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. |
@@ -308,7 +304,7 @@
 
 | Key | Type | Default | Descriptions |
 | --- | -----| ------- | ----------- |
-| `data_home` | String | `/tmp/metasrv/` | The working home directory. |
+| `data_home` | String | `./greptimedb_data/metasrv/` | The working home directory. |
 | `bind_addr` | String | `127.0.0.1:3002` | The bind address of metasrv. |
 | `server_addr` | String | `127.0.0.1:3002` | The communication server address for the frontend and datanode to connect to metasrv.<br/>If left empty or unset, the server will automatically use the IP address of the first network interface<br/>on the host, with the same port number as the one specified in `bind_addr`. |
 | `store_addrs` | Array | -- | Store server address default to etcd store.<br/>For postgres store, the format is:<br/>"password=password dbname=postgres user=postgres host=localhost port=5432"<br/>For etcd store, the format is:<br/>"127.0.0.1:2379" |
@@ -328,6 +324,7 @@
 | `procedure.max_retry_times` | Integer | `12` | Procedure max retry time. |
 | `procedure.retry_delay` | String | `500ms` | Initial retry delay of procedures, increases exponentially |
 | `procedure.max_metadata_value_size` | String | `1500KiB` | Auto split large value<br/>GreptimeDB procedure uses etcd as the default metadata storage backend.<br/>The etcd the maximum size of any request is 1.5 MiB<br/>1500KiB = 1536KiB (1.5MiB) - 36KiB (reserved size of key)<br/>Comments out the `max_metadata_value_size`, for don't split large value (no limit). |
+| `procedure.max_running_procedures` | Integer | `128` | Max running procedures.<br/>The maximum number of procedures that can be running at the same time.<br/>If the number of running procedures exceeds this limit, the procedure will be rejected. |
 | `failure_detector` | -- | -- | -- |
 | `failure_detector.threshold` | Float | `8.0` | The threshold value used by the failure detector to determine failure conditions. |
 | `failure_detector.min_std_deviation` | String | `100ms` | The minimum standard deviation of the heartbeat intervals, used to calculate acceptable variations. |
@@ -347,12 +344,8 @@
 | `wal.topic_name_prefix` | String | `greptimedb_wal_topic` | A Kafka topic is constructed by concatenating `topic_name_prefix` and `topic_id`.<br/>Only accepts strings that match the following regular expression pattern:<br/>[a-zA-Z_:-][a-zA-Z0-9_:\-\.@#]*<br/>i.g., greptimedb_wal_topic_0, greptimedb_wal_topic_1. |
 | `wal.replication_factor` | Integer | `1` | Expected number of replicas of each partition. |
 | `wal.create_topic_timeout` | String | `30s` | Above which a topic creation operation will be cancelled. |
-| `wal.backoff_init` | String | `500ms` | The initial backoff for kafka clients. |
-| `wal.backoff_max` | String | `10s` | The maximum backoff for kafka clients. |
-| `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. If set to empty, logs will not be written to files. |
+| `logging.dir` | String | `./greptimedb_data/logs` | The directory to store the log files. If set to empty, logs will not be written to files. |
 | `logging.level` | String | Unset | 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. |
@@ -381,7 +374,6 @@
 
 | Key | Type | Default | Descriptions |
 | --- | -----| ------- | ----------- |
-| `mode` | String | `standalone` | The running mode of the datanode. It can be `standalone` or `distributed`. |
 | `node_id` | Integer | Unset | The datanode identifier and should be unique in the cluster. |
 | `require_lease_before_startup` | Bool | `false` | Start services after regions have obtained leases.<br/>It will block the datanode start if it can't receive leases in the heartbeat from metasrv. |
 | `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. |
@@ -390,7 +382,7 @@
 | `enable_telemetry` | Bool | `true` | Enable telemetry to collect anonymous usage data. Enabled by default. |
 | `http` | -- | -- | The HTTP server options. |
 | `http.addr` | String | `127.0.0.1:4000` | The address to bind the HTTP server. |
-| `http.timeout` | String | `30s` | HTTP request timeout. Set to 0 to disable timeout. |
+| `http.timeout` | String | `0s` | HTTP request timeout. Set to 0 to disable timeout. |
 | `http.body_limit` | String | `64MB` | HTTP request body limit.<br/>The following units are supported: `B`, `KB`, `KiB`, `MB`, `MiB`, `GB`, `GiB`, `TB`, `TiB`, `PB`, `PiB`.<br/>Set to 0 to disable limit. |
 | `grpc` | -- | -- | The gRPC server options. |
 | `grpc.bind_addr` | String | `127.0.0.1:3001` | The address to bind the gRPC server. |
@@ -434,15 +426,11 @@
 | `wal.broker_endpoints` | Array | -- | The Kafka broker endpoints.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.max_batch_bytes` | String | `1MB` | The max size of a single producer batch.<br/>Warning: Kafka has a default limit of 1MB per message in a topic.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.consumer_wait_timeout` | String | `100ms` | The consumer wait timeout.<br/>**It's only used when the provider is `kafka`**. |
-| `wal.backoff_init` | String | `500ms` | The initial backoff delay.<br/>**It's only used when the provider is `kafka`**. |
-| `wal.backoff_max` | String | `10s` | The maximum backoff delay.<br/>**It's only used when the provider is `kafka`**. |
-| `wal.backoff_base` | Integer | `2` | The exponential backoff rate, i.e. next backoff = base * current backoff.<br/>**It's only used when the provider is `kafka`**. |
-| `wal.backoff_deadline` | String | `5mins` | The deadline of retries.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.create_index` | Bool | `true` | Whether to enable WAL index creation.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.dump_index_interval` | String | `60s` | The interval for dumping WAL indexes.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.overwrite_entry_start_id` | Bool | `false` | Ignore missing entries during read WAL.<br/>**It's only used when the provider is `kafka`**.<br/><br/>This option ensures that when Kafka messages are deleted, the system<br/>can still successfully replay memtable data without throwing an<br/>out-of-range error.<br/>However, enabling this option might lead to unexpected data loss,<br/>as the system will skip over missing entries instead of treating<br/>them as critical errors. |
 | `storage` | -- | -- | The data storage options. |
-| `storage.data_home` | String | `/tmp/greptimedb/` | The working home directory. |
+| `storage.data_home` | String | `./greptimedb_data/` | The working home directory. |
 | `storage.type` | String | `File` | The storage type used to store the data.<br/>- `File`: the data is stored in the local file system.<br/>- `S3`: the data is stored in the S3 object storage.<br/>- `Gcs`: the data is stored in the Google Cloud Storage.<br/>- `Azblob`: the data is stored in the Azure Blob Storage.<br/>- `Oss`: the data is stored in the Aliyun OSS. |
 | `storage.cache_path` | String | Unset | Read cache configuration for object storage such as 'S3' etc, it's configured by default when using object storage. It is recommended to configure it when using object storage for better performance.<br/>A local file directory, defaults to `{data_home}`. An empty string means disabling. |
 | `storage.cache_capacity` | String | Unset | The local file cache capacity in bytes. If your disk space is sufficient, it is recommended to set it larger. |
@@ -522,7 +510,7 @@
 | `region_engine.metric` | -- | -- | Metric engine options. |
 | `region_engine.metric.experimental_sparse_primary_key_encoding` | Bool | `false` | Whether to enable the experimental sparse primary key encoding. |
 | `logging` | -- | -- | The logging options. |
-| `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.dir` | String | `./greptimedb_data/logs` | The directory to store the log files. If set to empty, logs will not be written to files. |
 | `logging.level` | String | Unset | 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. |
@@ -551,7 +539,6 @@
 
 | Key | Type | Default | Descriptions |
 | --- | -----| ------- | ----------- |
-| `mode` | String | `distributed` | The running mode of the flownode. It can be `standalone` or `distributed`. |
 | `node_id` | Integer | Unset | The flownode identifier and should be unique in the cluster. |
 | `flow` | -- | -- | flow engine options. |
 | `flow.num_workers` | Integer | `0` | The number of flow worker in flownode.<br/>Not setting(or set to 0) this value will use the number of CPU cores divided by 2. |
@@ -563,7 +550,7 @@
 | `grpc.max_send_message_size` | String | `512MB` | The maximum send message size for gRPC server. |
 | `http` | -- | -- | The HTTP server options. |
 | `http.addr` | String | `127.0.0.1:4000` | The address to bind the HTTP server. |
-| `http.timeout` | String | `30s` | HTTP request timeout. Set to 0 to disable timeout. |
+| `http.timeout` | String | `0s` | HTTP request timeout. Set to 0 to disable timeout. |
 | `http.body_limit` | String | `64MB` | HTTP request body limit.<br/>The following units are supported: `B`, `KB`, `KiB`, `MB`, `MiB`, `GB`, `GiB`, `TB`, `TiB`, `PB`, `PiB`.<br/>Set to 0 to disable limit. |
 | `meta_client` | -- | -- | The metasrv client options. |
 | `meta_client.metasrv_addrs` | Array | -- | The addresses of the metasrv. |
@@ -579,7 +566,7 @@
 | `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. If set to empty, logs will not be written to files. |
+| `logging.dir` | String | `./greptimedb_data/logs` | The directory to store the log files. If set to empty, logs will not be written to files. |
 | `logging.level` | String | Unset | 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. |

config/datanode.example.toml

@@ -1,6 +1,3 @@
-## The running mode of the datanode. It can be `standalone` or `distributed`.
-mode = "standalone"
-
 ## The datanode identifier and should be unique in the cluster.
 ## @toml2docs:none-default
 node_id = 42
@@ -27,7 +24,7 @@ max_concurrent_queries = 0
 ## The address to bind the HTTP server.
 addr = "127.0.0.1:4000"
 ## HTTP request timeout. Set to 0 to disable timeout.
-timeout = "30s"
+timeout = "0s"
 ## HTTP request body limit.
 ## The following units are supported: `B`, `KB`, `KiB`, `MB`, `MiB`, `GB`, `GiB`, `TB`, `TiB`, `PB`, `PiB`.
 ## Set to 0 to disable limit.
@@ -119,7 +116,7 @@ provider = "raft_engine"
 ## The directory to store the WAL files.
 ## **It's only used when the provider is `raft_engine`**.
 ## @toml2docs:none-default
-dir = "/tmp/greptimedb/wal"
+dir = "./greptimedb_data/wal"
 
 ## The size of the WAL segment file.
 ## **It's only used when the provider is `raft_engine`**.
@@ -169,22 +166,6 @@ max_batch_bytes = "1MB"
 ## **It's only used when the provider is `kafka`**.
 consumer_wait_timeout = "100ms"
 
-## The initial backoff delay.
-## **It's only used when the provider is `kafka`**.
-backoff_init = "500ms"
-
-## The maximum backoff delay.
-## **It's only used when the provider is `kafka`**.
-backoff_max = "10s"
-
-## The exponential backoff rate, i.e. next backoff = base * current backoff.
-## **It's only used when the provider is `kafka`**.
-backoff_base = 2
-
-## The deadline of retries.
-## **It's only used when the provider is `kafka`**.
-backoff_deadline = "5mins"
-
 ## Whether to enable WAL index creation.
 ## **It's only used when the provider is `kafka`**.
 create_index = true
@@ -231,6 +212,7 @@ overwrite_entry_start_id = false
 # secret_access_key = "123456"
 # endpoint = "https://s3.amazonaws.com"
 # region = "us-west-2"
+# enable_virtual_host_style = false
 
 # Example of using Oss as the storage.
 # [storage]
@@ -264,7 +246,7 @@ overwrite_entry_start_id = false
 ## The data storage options.
 [storage]
 ## The working home directory.
-data_home = "/tmp/greptimedb/"
+data_home = "./greptimedb_data/"
 
 ## The storage type used to store the data.
 ## - `File`: the data is stored in the local file system.
@@ -617,7 +599,7 @@ experimental_sparse_primary_key_encoding = false
 ## The logging options.
 [logging]
 ## The directory to store the log files. If set to empty, logs will not be written to files.
-dir = "/tmp/greptimedb/logs"
+dir = "./greptimedb_data/logs"
 
 ## The log level. Can be `info`/`debug`/`warn`/`error`.
 ## @toml2docs:none-default

config/flownode.example.toml

@@ -1,6 +1,3 @@
-## The running mode of the flownode. It can be `standalone` or `distributed`.
-mode = "distributed"
-
 ## The flownode identifier and should be unique in the cluster.
 ## @toml2docs:none-default
 node_id = 14
@@ -30,7 +27,7 @@ max_send_message_size = "512MB"
 ## The address to bind the HTTP server.
 addr = "127.0.0.1:4000"
 ## HTTP request timeout. Set to 0 to disable timeout.
-timeout = "30s"
+timeout = "0s"
 ## HTTP request body limit.
 ## The following units are supported: `B`, `KB`, `KiB`, `MB`, `MiB`, `GB`, `GiB`, `TB`, `TiB`, `PB`, `PiB`.
 ## Set to 0 to disable limit.
@@ -76,7 +73,7 @@ retry_interval = "3s"
 ## The logging options.
 [logging]
 ## The directory to store the log files. If set to empty, logs will not be written to files.
-dir = "/tmp/greptimedb/logs"
+dir = "./greptimedb_data/logs"
 
 ## The log level. Can be `info`/`debug`/`warn`/`error`.
 ## @toml2docs:none-default
@@ -121,4 +118,3 @@ sample_ratio = 1.0
 ## The tokio console address.
 ## @toml2docs:none-default
 #+ tokio_console_addr = "127.0.0.1"
-

config/frontend.example.toml

@@ -26,7 +26,7 @@ retry_interval = "3s"
 ## The address to bind the HTTP server.
 addr = "127.0.0.1:4000"
 ## HTTP request timeout. Set to 0 to disable timeout.
-timeout = "30s"
+timeout = "0s"
 ## HTTP request body limit.
 ## The following units are supported: `B`, `KB`, `KiB`, `MB`, `MiB`, `GB`, `GiB`, `TB`, `TiB`, `PB`, `PiB`.
 ## Set to 0 to disable limit.
@@ -189,7 +189,7 @@ tcp_nodelay = true
 ## The logging options.
 [logging]
 ## The directory to store the log files. If set to empty, logs will not be written to files.
-dir = "/tmp/greptimedb/logs"
+dir = "./greptimedb_data/logs"
 
 ## The log level. Can be `info`/`debug`/`warn`/`error`.
 ## @toml2docs:none-default

config/metasrv.example.toml

@@ -1,5 +1,5 @@
 ## The working home directory.
-data_home = "/tmp/metasrv/"
+data_home = "./greptimedb_data/metasrv/"
 
 ## The bind address of metasrv.
 bind_addr = "127.0.0.1:3002"
@@ -79,6 +79,11 @@ retry_delay = "500ms"
 ## Comments out the `max_metadata_value_size`, for don't split large value (no limit).
 max_metadata_value_size = "1500KiB"
 
+## Max running procedures.
+## The maximum number of procedures that can be running at the same time.
+## If the number of running procedures exceeds this limit, the procedure will be rejected.
+max_running_procedures = 128
+
 # Failure detectors options.
 [failure_detector]
 
@@ -144,17 +149,6 @@ replication_factor = 1
 
 ## Above which a topic creation operation will be cancelled.
 create_topic_timeout = "30s"
-## The initial backoff for kafka clients.
-backoff_init = "500ms"
-
-## The maximum backoff for kafka clients.
-backoff_max = "10s"
-
-## Exponential backoff rate, i.e. next backoff = base * current backoff.
-backoff_base = 2
-
-## Stop reconnecting if the total wait time reaches the deadline. If this config is missing, the reconnecting won't terminate.
-backoff_deadline = "5mins"
 
 # The Kafka SASL configuration.
 # **It's only used when the provider is `kafka`**.
@@ -177,7 +171,7 @@ backoff_deadline = "5mins"
 ## The logging options.
 [logging]
 ## The directory to store the log files. If set to empty, logs will not be written to files.
-dir = "/tmp/greptimedb/logs"
+dir = "./greptimedb_data/logs"
 
 ## The log level. Can be `info`/`debug`/`warn`/`error`.
 ## @toml2docs:none-default

config/standalone.example.toml

@@ -1,6 +1,3 @@
-## The running mode of the datanode. It can be `standalone` or `distributed`.
-mode = "standalone"
-
 ## The default timezone of the server.
 ## @toml2docs:none-default
 default_timezone = "UTC"
@@ -34,7 +31,7 @@ max_concurrent_queries = 0
 ## The address to bind the HTTP server.
 addr = "127.0.0.1:4000"
 ## HTTP request timeout. Set to 0 to disable timeout.
-timeout = "30s"
+timeout = "0s"
 ## HTTP request body limit.
 ## The following units are supported: `B`, `KB`, `KiB`, `MB`, `MiB`, `GB`, `GiB`, `TB`, `TiB`, `PB`, `PiB`.
 ## Set to 0 to disable limit.
@@ -164,7 +161,7 @@ provider = "raft_engine"
 ## The directory to store the WAL files.
 ## **It's only used when the provider is `raft_engine`**.
 ## @toml2docs:none-default
-dir = "/tmp/greptimedb/wal"
+dir = "./greptimedb_data/wal"
 
 ## The size of the WAL segment file.
 ## **It's only used when the provider is `raft_engine`**.
@@ -242,22 +239,6 @@ max_batch_bytes = "1MB"
 ## **It's only used when the provider is `kafka`**.
 consumer_wait_timeout = "100ms"
 
-## The initial backoff delay.
-## **It's only used when the provider is `kafka`**.
-backoff_init = "500ms"
-
-## The maximum backoff delay.
-## **It's only used when the provider is `kafka`**.
-backoff_max = "10s"
-
-## The exponential backoff rate, i.e. next backoff = base * current backoff.
-## **It's only used when the provider is `kafka`**.
-backoff_base = 2
-
-## The deadline of retries.
-## **It's only used when the provider is `kafka`**.
-backoff_deadline = "5mins"
-
 ## Ignore missing entries during read WAL.
 ## **It's only used when the provider is `kafka`**.
 ##
@@ -302,6 +283,10 @@ purge_interval = "1m"
 max_retry_times = 3
 ## Initial retry delay of procedures, increases exponentially
 retry_delay = "500ms"
+## Max running procedures.
+## The maximum number of procedures that can be running at the same time.
+## If the number of running procedures exceeds this limit, the procedure will be rejected.
+max_running_procedures = 128
 
 ## flow engine options.
 [flow]
@@ -318,6 +303,7 @@ retry_delay = "500ms"
 # secret_access_key = "123456"
 # endpoint = "https://s3.amazonaws.com"
 # region = "us-west-2"
+# enable_virtual_host_style = false
 
 # Example of using Oss as the storage.
 # [storage]
@@ -351,7 +337,7 @@ retry_delay = "500ms"
 ## The data storage options.
 [storage]
 ## The working home directory.
-data_home = "/tmp/greptimedb/"
+data_home = "./greptimedb_data/"
 
 ## The storage type used to store the data.
 ## - `File`: the data is stored in the local file system.
@@ -704,7 +690,7 @@ experimental_sparse_primary_key_encoding = false
 ## The logging options.
 [logging]
 ## The directory to store the log files. If set to empty, logs will not be written to files.
-dir = "/tmp/greptimedb/logs"
+dir = "./greptimedb_data/logs"
 
 ## The log level. Can be `info`/`debug`/`warn`/`error`.
 ## @toml2docs:none-default

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

@@ -25,7 +25,7 @@ services:
       - --initial-cluster-state=new
       - *etcd_initial_cluster_token
     volumes:
-      - /tmp/greptimedb-cluster-docker-compose/etcd0:/var/lib/etcd
+      - ./greptimedb-cluster-docker-compose/etcd0:/var/lib/etcd
     healthcheck:
       test: [ "CMD", "etcdctl", "--endpoints=http://etcd0:2379", "endpoint", "health" ]
       interval: 5s
@@ -68,12 +68,13 @@ services:
       - datanode
       - start
       - --node-id=0
+      - --data-home=/greptimedb_data
       - --rpc-bind-addr=0.0.0.0:3001
       - --rpc-server-addr=datanode0:3001
       - --metasrv-addrs=metasrv:3002
       - --http-addr=0.0.0.0:5000
     volumes:
-      - /tmp/greptimedb-cluster-docker-compose/datanode0:/tmp/greptimedb
+      - ./greptimedb-cluster-docker-compose/datanode0:/greptimedb_data
     healthcheck:
       test: [ "CMD", "curl", "-fv", "http://datanode0:5000/health" ]
       interval: 5s

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

@@ -3,7 +3,7 @@
 This document introduces how to write fuzz tests in GreptimeDB.
 
 ## What is a fuzz test
-Fuzz test is tool that leverage deterministic random generation to assist in finding bugs. The goal of fuzz tests is to identify inputs generated by the fuzzer that cause system panics, crashes, or unexpected behaviors to occur. And we are using the [cargo-fuzz](https://github.com/rust-fuzz/cargo-fuzz) to run our fuzz test targets. 
+Fuzz test is tool that leverage deterministic random generation to assist in finding bugs. The goal of fuzz tests is to identify inputs generated by the fuzzer that cause system panics, crashes, or unexpected behaviors to occur. And we are using the [cargo-fuzz](https://github.com/rust-fuzz/cargo-fuzz) to run our fuzz test targets.
 
 ## Why we need them
 - Find bugs by leveraging random generation
@@ -13,7 +13,7 @@ Fuzz test is tool that leverage deterministic random generation to assist in fin
 All fuzz test-related resources are located in the `/tests-fuzz` directory.
 There are two types of resources: (1) fundamental components and (2) test targets.
 
-### Fundamental components 
+### Fundamental components
 They are located in the `/tests-fuzz/src` directory. The fundamental components define how to generate SQLs (including dialects for different protocols) and validate execution results (e.g., column attribute validation), etc.
 
 ### Test targets
@@ -21,25 +21,25 @@ They are located in the `/tests-fuzz/targets` directory, with each file represen
 
 Figure 1 illustrates the fundamental components of the fuzz test provide the ability to generate random SQLs. It utilizes a Random Number Generator (Rng) to generate the Intermediate Representation (IR), then employs a DialectTranslator to produce specified dialects for different protocols. Finally, the fuzz tests send the generated SQL via the specified protocol and verify that the execution results meet expectations.
 ```
-                            Rng                                 
-                             |                                  
-                             |                                  
-                             v                                  
-                       ExprGenerator                            
-                             |                                  
-                             |                                  
-                             v                                  
-               Intermediate representation (IR)                 
-                             |                                  
-                             |                                  
-      +----------------------+----------------------+           
-      |                      |                      |           
-      v                      v                      v           
+                            Rng
+                             |
+                             |
+                             v
+                       ExprGenerator
+                             |
+                             |
+                             v
+               Intermediate representation (IR)
+                             |
+                             |
+      +----------------------+----------------------+
+      |                      |                      |
+      v                      v                      v
 MySQLTranslator    PostgreSQLTranslator   OtherDialectTranslator
-      |                      |                      |           
-      |                      |                      |           
-      v                      v                      v           
-SQL(MySQL Dialect)         .....                  .....         
+      |                      |                      |
+      |                      |                      |
+      v                      v                      v
+SQL(MySQL Dialect)         .....                  .....
       |
       |
       v
@@ -133,4 +133,4 @@ fuzz_target!(|input: FuzzInput| {
     cargo fuzz run <fuzz-target> --fuzz-dir tests-fuzz
 ```
 
-For more details, please refer to this [document](/tests-fuzz/README.md).
+For more details, please refer to this [document](/tests-fuzz/README.md).

docs/rfcs/2025-02-06-remote-wal-purge.md

@@ -0,0 +1,77 @@
+---
+Feature Name: Remote WAL Purge
+Tracking Issue: https://github.com/GreptimeTeam/greptimedb/issues/5474
+Date: 2025-02-06
+Author: "Yuhan Wang <profsyb@gmail.com>"
+---
+
+# Summary
+
+This RFC proposes a method for purging remote WAL in the database.
+
+# Motivation
+
+Currently only local wal entries are purged when flushing, while remote wal does nothing.
+
+# Details
+
+```mermaid
+sequenceDiagram
+    Region0->>Kafka: Last entry id of the topic in use
+    Region0->>WALPruner: Heartbeat with last entry id
+    WALPruner->>+WALPruner: Time Loop
+    WALPruner->>+ProcedureManager: Submit purge procedure
+    ProcedureManager->>Region0: Flush request
+    ProcedureManager->>Kafka: Prune WAL entries
+    Region0->>Region0: Flush
+```
+
+## Steps
+
+### Before purge
+
+Before purging remote WAL, metasrv needs to know:
+
+1. `last_entry_id` of each region.
+2. `kafka_topic_last_entry_id` which is the last entry id of the topic in use. Can be lazily updated and needed when region has empty memtable.
+3. Kafka topics that each region uses.
+
+The states are maintained through:
+1. Heartbeat: Datanode sends `last_entry_id` to metasrv in heartbeat. As for regions with empty memtable, `last_entry_id` should equals to `kafka_topic_last_entry_id`.
+2. Metasrv maintains a topic-region map to know which region uses which topic.
+
+`kafka_topic_last_entry_id` will be maintained by the region itself. Region will update the value after `k` heartbeats if the memtable is empty.
+
+### Purge procedure
+
+We can better handle locks utilizing current procedure. It's quite similar to the region migration procedure.
+
+After a period of time, metasrv will submit a purge procedure to ProcedureManager. The purge will apply to all topics.
+
+The procedure is divided into following stages:
+
+1. Preparation:
+   - Retrieve `last_entry_id` of each region kvbackend.
+   - Choose regions that have a relatively small `last_entry_id` as candidate regions, which means we need to send a flush request to these regions.
+2. Communication:
+   - Send flush requests to candidate regions.
+3. Purge:
+   - Choose proper entry id to delete for each topic. The entry should be the smallest `last_entry_id - 1` among all regions. 
+   - Delete legacy entries in Kafka.
+   - Store the `last_purged_entry_id` in kvbackend. It should be locked to prevent other regions from replaying the purged entries.
+
+### After purge
+
+After purge, there may be some regions that have `last_entry_id` smaller than the entry we just deleted. It's legal since we only delete the entries that are not needed anymore.
+
+When restarting a region, it should query the `last_purged_entry_id` from metasrv and replay from `min(last_entry_id, last_purged_entry_id)`.
+
+### Error handling
+
+No persisted states are needed since all states are maintained in kvbackend.
+
+Retry when failed to retrieving metadata from kvbackend.
+
+# Alternatives
+
+Purge time can depend on the size of the WAL entries instead of a fixed period of time, which may be more efficient.

grafana/greptimedb-cluster.json

@@ -4782,7 +4782,7 @@
         "type": "prometheus",
         "uid": "${DS_PROMETHEUS}"
       },
-      "description": "Current counts for stalled write requests by instance\n\nWrite stalls when memtable is full and pending for flush\n\n",
+      "description": "Ingestion size by row counts.",
       "fieldConfig": {
         "defaults": {
           "color": {
@@ -4844,7 +4844,7 @@
         "x": 12,
         "y": 138
       },
-      "id": 221,
+      "id": 277,
       "options": {
         "legend": {
           "calcs": [],
@@ -4864,14 +4864,14 @@
             "uid": "${DS_PROMETHEUS}"
           },
           "editorMode": "code",
-          "expr": "sum by(pod) (greptime_mito_write_stall_total{pod=~\"$datanode\"})",
+          "expr": "rate(greptime_mito_write_rows_total{pod=~\"$datanode\"}[$__rate_interval])",
           "instant": false,
           "legendFormat": "{{pod}}",
           "range": true,
           "refId": "A"
         }
       ],
-      "title": "Write Stall per Instance",
+      "title": "Write Rows per Instance",
       "type": "timeseries"
     },
     {
@@ -4976,7 +4976,7 @@
         "type": "prometheus",
         "uid": "${DS_PROMETHEUS}"
       },
-      "description": "Cache size by instance.\n",
+      "description": "Current counts for stalled write requests by instance\n\nWrite stalls when memtable is full and pending for flush\n\n",
       "fieldConfig": {
         "defaults": {
           "color": {
@@ -5028,7 +5028,7 @@
               }
             ]
           },
-          "unit": "decbytes"
+          "unit": "none"
         },
         "overrides": []
       },
@@ -5038,7 +5038,7 @@
         "x": 12,
         "y": 146
       },
-      "id": 229,
+      "id": 221,
       "options": {
         "legend": {
           "calcs": [],
@@ -5058,14 +5058,14 @@
             "uid": "${DS_PROMETHEUS}"
           },
           "editorMode": "code",
-          "expr": "greptime_mito_cache_bytes{pod=~\"$datanode\"}",
+          "expr": "sum by(pod) (greptime_mito_write_stall_total{pod=~\"$datanode\"})",
           "instant": false,
-          "legendFormat": "{{pod}}-{{type}}",
+          "legendFormat": "{{pod}}",
           "range": true,
           "refId": "A"
         }
       ],
-      "title": "Cached Bytes per Instance",
+      "title": "Write Stall per Instance",
       "type": "timeseries"
     },
     {
@@ -5172,7 +5172,7 @@
         "type": "prometheus",
         "uid": "${DS_PROMETHEUS}"
       },
-      "description": "P99 latency of each type of reads by instance",
+      "description": "Cache size by instance.\n",
       "fieldConfig": {
         "defaults": {
           "color": {
@@ -5224,7 +5224,7 @@
               }
             ]
           },
-          "unit": "s"
+          "unit": "decbytes"
         },
         "overrides": []
       },
@@ -5234,17 +5234,13 @@
         "x": 12,
         "y": 154
       },
-      "id": 228,
+      "id": 229,
       "options": {
         "legend": {
-          "calcs": [
-            "lastNotNull"
-          ],
+          "calcs": [],
           "displayMode": "table",
           "placement": "bottom",
-          "showLegend": true,
-          "sortBy": "Last *",
-          "sortDesc": true
+          "showLegend": true
         },
         "tooltip": {
           "mode": "single",
@@ -5258,14 +5254,14 @@
             "uid": "${DS_PROMETHEUS}"
           },
           "editorMode": "code",
-          "expr": "histogram_quantile(0.99, sum by(pod, le, stage) (rate(greptime_mito_read_stage_elapsed_bucket{pod=~\"$datanode\"}[$__rate_interval])))",
+          "expr": "greptime_mito_cache_bytes{pod=~\"$datanode\"}",
           "instant": false,
-          "legendFormat": "{{pod}}-{{stage}}-p99",
+          "legendFormat": "{{pod}}-{{type}}",
           "range": true,
           "refId": "A"
         }
       ],
-      "title": "Read Stage P99 per Instance",
+      "title": "Cached Bytes per Instance",
       "type": "timeseries"
     },
     {
@@ -5317,7 +5313,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": null
               },
               {
                 "color": "red",
@@ -5370,7 +5367,7 @@
         "type": "prometheus",
         "uid": "${DS_PROMETHEUS}"
       },
-      "description": "Latency of compaction task, at p99",
+      "description": "P99 latency of each type of reads by instance",
       "fieldConfig": {
         "defaults": {
           "color": {
@@ -5414,7 +5411,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": null
               },
               {
                 "color": "red",
@@ -5432,7 +5430,7 @@
         "x": 12,
         "y": 162
       },
-      "id": 230,
+      "id": 228,
       "options": {
         "legend": {
           "calcs": [
@@ -5440,7 +5438,9 @@
           ],
           "displayMode": "table",
           "placement": "bottom",
-          "showLegend": true
+          "showLegend": true,
+          "sortBy": "Last *",
+          "sortDesc": true
         },
         "tooltip": {
           "mode": "single",
@@ -5454,14 +5454,14 @@
             "uid": "${DS_PROMETHEUS}"
           },
           "editorMode": "code",
-          "expr": "histogram_quantile(0.99, sum by(pod, le) (rate(greptime_mito_compaction_total_elapsed_bucket{pod=~\"$datanode\"}[$__rate_interval])))",
+          "expr": "histogram_quantile(0.99, sum by(pod, le, stage) (rate(greptime_mito_read_stage_elapsed_bucket{pod=~\"$datanode\"}[$__rate_interval])))",
           "instant": false,
-          "legendFormat": "[{{pod}}]-compaction-p99",
+          "legendFormat": "{{pod}}-{{stage}}-p99",
           "range": true,
           "refId": "A"
         }
       ],
-      "title": "Compaction P99 per Instance",
+      "title": "Read Stage P99 per Instance",
       "type": "timeseries"
     },
     {
@@ -5570,7 +5570,7 @@
         "type": "prometheus",
         "uid": "${DS_PROMETHEUS}"
       },
-      "description": "Compaction latency by stage",
+      "description": "Latency of compaction task, at p99",
       "fieldConfig": {
         "defaults": {
           "color": {
@@ -5632,7 +5632,7 @@
         "x": 12,
         "y": 170
       },
-      "id": 232,
+      "id": 230,
       "options": {
         "legend": {
           "calcs": [
@@ -5654,9 +5654,9 @@
             "uid": "${DS_PROMETHEUS}"
           },
           "editorMode": "code",
-          "expr": "histogram_quantile(0.99, sum by(pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{pod=~\"$datanode\"}[$__rate_interval])))",
+          "expr": "histogram_quantile(0.99, sum by(pod, le) (rate(greptime_mito_compaction_total_elapsed_bucket{pod=~\"$datanode\"}[$__rate_interval])))",
           "instant": false,
-          "legendFormat": "{{pod}}-{{stage}}-p99",
+          "legendFormat": "[{{pod}}]-compaction-p99",
           "range": true,
           "refId": "A"
         }
@@ -5794,7 +5794,7 @@
         "type": "prometheus",
         "uid": "${DS_PROMETHEUS}"
       },
-      "description": "Write-ahead log operations latency at p99",
+      "description": "Compaction latency by stage",
       "fieldConfig": {
         "defaults": {
           "color": {
@@ -5856,13 +5856,13 @@
         "x": 12,
         "y": 178
       },
-      "id": 269,
+      "id": 232,
       "options": {
         "legend": {
           "calcs": [
             "lastNotNull"
           ],
-          "displayMode": "list",
+          "displayMode": "table",
           "placement": "bottom",
           "showLegend": true
         },
@@ -5878,14 +5878,14 @@
             "uid": "${DS_PROMETHEUS}"
           },
           "editorMode": "code",
-          "expr": "histogram_quantile(0.99, sum by(le,logstore,optype,pod) (rate(greptime_logstore_op_elapsed_bucket[$__rate_interval])))",
+          "expr": "histogram_quantile(0.99, sum by(pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{pod=~\"$datanode\"}[$__rate_interval])))",
           "instant": false,
-          "legendFormat": "{{pod}}-{{logstore}}-{{optype}}-p99",
+          "legendFormat": "{{pod}}-{{stage}}-p99",
           "range": true,
           "refId": "A"
         }
       ],
-      "title": "Log Store op duration seconds",
+      "title": "Compaction P99 per Instance",
       "type": "timeseries"
     },
     {
@@ -5993,7 +5993,7 @@
         "type": "prometheus",
         "uid": "${DS_PROMETHEUS}"
       },
-      "description": "Ongoing compaction task count",
+      "description": "Write-ahead log operations latency at p99",
       "fieldConfig": {
         "defaults": {
           "color": {
@@ -6045,7 +6045,7 @@
               }
             ]
           },
-          "unit": "none"
+          "unit": "s"
         },
         "overrides": []
       },
@@ -6055,13 +6055,13 @@
         "x": 12,
         "y": 186
       },
-      "id": 271,
+      "id": 269,
       "options": {
         "legend": {
           "calcs": [
             "lastNotNull"
           ],
-          "displayMode": "table",
+          "displayMode": "list",
           "placement": "bottom",
           "showLegend": true
         },
@@ -6078,14 +6078,14 @@
             "uid": "${DS_PROMETHEUS}"
           },
           "editorMode": "code",
-          "expr": "greptime_mito_inflight_compaction_count",
+          "expr": "histogram_quantile(0.99, sum by(le,logstore,optype,pod) (rate(greptime_logstore_op_elapsed_bucket[$__rate_interval])))",
           "instant": false,
-          "legendFormat": "{{pod}}",
+          "legendFormat": "{{pod}}-{{logstore}}-{{optype}}-p99",
           "range": true,
           "refId": "A"
         }
       ],
-      "title": "Inflight Compaction",
+      "title": "Log Store op duration seconds",
       "type": "timeseries"
     },
     {
@@ -6188,6 +6188,105 @@
       "title": "Inflight Flush",
       "type": "timeseries"
     },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "${DS_PROMETHEUS}"
+      },
+      "description": "Ongoing compaction task count",
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "axisBorderShow": false,
+            "axisCenteredZero": false,
+            "axisColorMode": "text",
+            "axisLabel": "",
+            "axisPlacement": "auto",
+            "barAlignment": 0,
+            "drawStyle": "points",
+            "fillOpacity": 0,
+            "gradientMode": "none",
+            "hideFrom": {
+              "legend": false,
+              "tooltip": false,
+              "viz": false
+            },
+            "insertNulls": false,
+            "lineInterpolation": "linear",
+            "lineWidth": 1,
+            "pointSize": 5,
+            "scaleDistribution": {
+              "type": "linear"
+            },
+            "showPoints": "auto",
+            "spanNulls": false,
+            "stacking": {
+              "group": "A",
+              "mode": "none"
+            },
+            "thresholdsStyle": {
+              "mode": "off"
+            }
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green"
+              },
+              {
+                "color": "red",
+                "value": 80
+              }
+            ]
+          },
+          "unit": "none"
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 8,
+        "w": 12,
+        "x": 12,
+        "y": 194
+      },
+      "id": 271,
+      "options": {
+        "legend": {
+          "calcs": [
+            "lastNotNull"
+          ],
+          "displayMode": "table",
+          "placement": "bottom",
+          "showLegend": true
+        },
+        "tooltip": {
+          "mode": "single",
+          "sort": "none"
+        }
+      },
+      "pluginVersion": "11.1.3",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "${DS_PROMETHEUS}"
+          },
+          "editorMode": "code",
+          "expr": "greptime_mito_inflight_compaction_count",
+          "instant": false,
+          "legendFormat": "{{pod}}",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Inflight Compaction",
+      "type": "timeseries"
+    },
     {
       "collapsed": false,
       "gridPos": {

src/api/src/helper.rs

@@ -19,9 +19,7 @@ use common_decimal::decimal128::{DECIMAL128_DEFAULT_SCALE, DECIMAL128_MAX_PRECIS
 use common_decimal::Decimal128;
 use common_time::time::Time;
 use common_time::timestamp::TimeUnit;
-use common_time::{
-    Date, DateTime, IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth, Timestamp,
-};
+use common_time::{Date, IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth, Timestamp};
 use datatypes::prelude::{ConcreteDataType, ValueRef};
 use datatypes::scalars::ScalarVector;
 use datatypes::types::{
@@ -29,8 +27,8 @@ use datatypes::types::{
 };
 use datatypes::value::{OrderedF32, OrderedF64, Value};
 use datatypes::vectors::{
-    BinaryVector, BooleanVector, DateTimeVector, DateVector, Decimal128Vector, Float32Vector,
-    Float64Vector, Int32Vector, Int64Vector, IntervalDayTimeVector, IntervalMonthDayNanoVector,
+    BinaryVector, BooleanVector, DateVector, Decimal128Vector, Float32Vector, Float64Vector,
+    Int32Vector, Int64Vector, IntervalDayTimeVector, IntervalMonthDayNanoVector,
     IntervalYearMonthVector, PrimitiveVector, StringVector, TimeMicrosecondVector,
     TimeMillisecondVector, TimeNanosecondVector, TimeSecondVector, TimestampMicrosecondVector,
     TimestampMillisecondVector, TimestampNanosecondVector, TimestampSecondVector, UInt32Vector,
@@ -118,7 +116,7 @@ impl From<ColumnDataTypeWrapper> for ConcreteDataType {
             ColumnDataType::Json => ConcreteDataType::json_datatype(),
             ColumnDataType::String => ConcreteDataType::string_datatype(),
             ColumnDataType::Date => ConcreteDataType::date_datatype(),
-            ColumnDataType::Datetime => ConcreteDataType::datetime_datatype(),
+            ColumnDataType::Datetime => ConcreteDataType::timestamp_microsecond_datatype(),
             ColumnDataType::TimestampSecond => ConcreteDataType::timestamp_second_datatype(),
             ColumnDataType::TimestampMillisecond => {
                 ConcreteDataType::timestamp_millisecond_datatype()
@@ -271,7 +269,6 @@ impl TryFrom<ConcreteDataType> for ColumnDataTypeWrapper {
             ConcreteDataType::Binary(_) => ColumnDataType::Binary,
             ConcreteDataType::String(_) => ColumnDataType::String,
             ConcreteDataType::Date(_) => ColumnDataType::Date,
-            ConcreteDataType::DateTime(_) => ColumnDataType::Datetime,
             ConcreteDataType::Timestamp(t) => match t {
                 TimestampType::Second(_) => ColumnDataType::TimestampSecond,
                 TimestampType::Millisecond(_) => ColumnDataType::TimestampMillisecond,
@@ -476,7 +473,6 @@ pub fn push_vals(column: &mut Column, origin_count: usize, vector: VectorRef) {
         Value::String(val) => values.string_values.push(val.as_utf8().to_string()),
         Value::Binary(val) => values.binary_values.push(val.to_vec()),
         Value::Date(val) => values.date_values.push(val.val()),
-        Value::DateTime(val) => values.datetime_values.push(val.val()),
         Value::Timestamp(val) => match val.unit() {
             TimeUnit::Second => values.timestamp_second_values.push(val.value()),
             TimeUnit::Millisecond => values.timestamp_millisecond_values.push(val.value()),
@@ -577,12 +573,11 @@ pub fn pb_value_to_value_ref<'a>(
         ValueData::BinaryValue(bytes) => ValueRef::Binary(bytes.as_slice()),
         ValueData::StringValue(string) => ValueRef::String(string.as_str()),
         ValueData::DateValue(d) => ValueRef::Date(Date::from(*d)),
-        ValueData::DatetimeValue(d) => ValueRef::DateTime(DateTime::new(*d)),
         ValueData::TimestampSecondValue(t) => ValueRef::Timestamp(Timestamp::new_second(*t)),
         ValueData::TimestampMillisecondValue(t) => {
             ValueRef::Timestamp(Timestamp::new_millisecond(*t))
         }
-        ValueData::TimestampMicrosecondValue(t) => {
+        ValueData::DatetimeValue(t) | ValueData::TimestampMicrosecondValue(t) => {
             ValueRef::Timestamp(Timestamp::new_microsecond(*t))
         }
         ValueData::TimestampNanosecondValue(t) => {
@@ -651,7 +646,6 @@ pub fn pb_values_to_vector_ref(data_type: &ConcreteDataType, values: Values) ->
         ConcreteDataType::Binary(_) => Arc::new(BinaryVector::from(values.binary_values)),
         ConcreteDataType::String(_) => Arc::new(StringVector::from_vec(values.string_values)),
         ConcreteDataType::Date(_) => Arc::new(DateVector::from_vec(values.date_values)),
-        ConcreteDataType::DateTime(_) => Arc::new(DateTimeVector::from_vec(values.datetime_values)),
         ConcreteDataType::Timestamp(unit) => match unit {
             TimestampType::Second(_) => Arc::new(TimestampSecondVector::from_vec(
                 values.timestamp_second_values,
@@ -787,11 +781,6 @@ pub fn pb_values_to_values(data_type: &ConcreteDataType, values: Values) -> Vec<
             .into_iter()
             .map(|val| val.into())
             .collect(),
-        ConcreteDataType::DateTime(_) => values
-            .datetime_values
-            .into_iter()
-            .map(|v| Value::DateTime(v.into()))
-            .collect(),
         ConcreteDataType::Date(_) => values
             .date_values
             .into_iter()
@@ -947,9 +936,6 @@ pub fn to_proto_value(value: Value) -> Option<v1::Value> {
         Value::Date(v) => v1::Value {
             value_data: Some(ValueData::DateValue(v.val())),
         },
-        Value::DateTime(v) => v1::Value {
-            value_data: Some(ValueData::DatetimeValue(v.val())),
-        },
         Value::Timestamp(v) => match v.unit() {
             TimeUnit::Second => v1::Value {
                 value_data: Some(ValueData::TimestampSecondValue(v.value())),
@@ -1066,7 +1052,6 @@ pub fn value_to_grpc_value(value: Value) -> GrpcValue {
             Value::String(v) => Some(ValueData::StringValue(v.as_utf8().to_string())),
             Value::Binary(v) => Some(ValueData::BinaryValue(v.to_vec())),
             Value::Date(v) => Some(ValueData::DateValue(v.val())),
-            Value::DateTime(v) => Some(ValueData::DatetimeValue(v.val())),
             Value::Timestamp(v) => Some(match v.unit() {
                 TimeUnit::Second => ValueData::TimestampSecondValue(v.value()),
                 TimeUnit::Millisecond => ValueData::TimestampMillisecondValue(v.value()),
@@ -1248,7 +1233,7 @@ mod tests {
             ColumnDataTypeWrapper::date_datatype().into()
         );
         assert_eq!(
-            ConcreteDataType::datetime_datatype(),
+            ConcreteDataType::timestamp_microsecond_datatype(),
             ColumnDataTypeWrapper::datetime_datatype().into()
         );
         assert_eq!(
@@ -1339,10 +1324,6 @@ mod tests {
             ColumnDataTypeWrapper::date_datatype(),
             ConcreteDataType::date_datatype().try_into().unwrap()
         );
-        assert_eq!(
-            ColumnDataTypeWrapper::datetime_datatype(),
-            ConcreteDataType::datetime_datatype().try_into().unwrap()
-        );
         assert_eq!(
             ColumnDataTypeWrapper::timestamp_millisecond_datatype(),
             ConcreteDataType::timestamp_millisecond_datatype()
@@ -1830,17 +1811,6 @@ mod tests {
         ]
     );
 
-    test_convert_values!(
-        datetime,
-        vec![1.into(), 2.into(), 3.into()],
-        datetime,
-        vec![
-            Value::DateTime(1.into()),
-            Value::DateTime(2.into()),
-            Value::DateTime(3.into())
-        ]
-    );
-
     #[test]
     fn test_vectors_to_rows_for_different_types() {
         let boolean_vec = BooleanVector::from_vec(vec![true, false, true]);

src/api/src/v1/column_def.rs

@@ -15,10 +15,13 @@
 use std::collections::HashMap;
 
 use datatypes::schema::{
-    ColumnDefaultConstraint, ColumnSchema, FulltextAnalyzer, FulltextOptions, SkippingIndexOptions,
-    SkippingIndexType, COMMENT_KEY, FULLTEXT_KEY, INVERTED_INDEX_KEY, SKIPPING_INDEX_KEY,
+    ColumnDefaultConstraint, ColumnSchema, FulltextAnalyzer, FulltextBackend, FulltextOptions,
+    SkippingIndexOptions, SkippingIndexType, COMMENT_KEY, FULLTEXT_KEY, INVERTED_INDEX_KEY,
+    SKIPPING_INDEX_KEY,
+};
+use greptime_proto::v1::{
+    Analyzer, FulltextBackend as PbFulltextBackend, SkippingIndexType as PbSkippingIndexType,
 };
-use greptime_proto::v1::{Analyzer, SkippingIndexType as PbSkippingIndexType};
 use snafu::ResultExt;
 
 use crate::error::{self, Result};
@@ -132,14 +135,31 @@ pub fn options_from_skipping(skipping: &SkippingIndexOptions) -> Result<Option<C
     Ok((!options.options.is_empty()).then_some(options))
 }
 
+/// Tries to construct a `ColumnOptions` for inverted index.
+pub fn options_from_inverted() -> ColumnOptions {
+    let mut options = ColumnOptions::default();
+    options
+        .options
+        .insert(INVERTED_INDEX_GRPC_KEY.to_string(), "true".to_string());
+    options
+}
+
 /// Tries to construct a `FulltextAnalyzer` from the given analyzer.
-pub fn as_fulltext_option(analyzer: Analyzer) -> FulltextAnalyzer {
+pub fn as_fulltext_option_analyzer(analyzer: Analyzer) -> FulltextAnalyzer {
     match analyzer {
         Analyzer::English => FulltextAnalyzer::English,
         Analyzer::Chinese => FulltextAnalyzer::Chinese,
     }
 }
 
+/// Tries to construct a `FulltextBackend` from the given backend.
+pub fn as_fulltext_option_backend(backend: PbFulltextBackend) -> FulltextBackend {
+    match backend {
+        PbFulltextBackend::Bloom => FulltextBackend::Bloom,
+        PbFulltextBackend::Tantivy => FulltextBackend::Tantivy,
+    }
+}
+
 /// Tries to construct a `SkippingIndexType` from the given skipping index type.
 pub fn as_skipping_index_type(skipping_index_type: PbSkippingIndexType) -> SkippingIndexType {
     match skipping_index_type {
@@ -151,7 +171,7 @@ pub fn as_skipping_index_type(skipping_index_type: PbSkippingIndexType) -> Skipp
 mod tests {
 
     use datatypes::data_type::ConcreteDataType;
-    use datatypes::schema::FulltextAnalyzer;
+    use datatypes::schema::{FulltextAnalyzer, FulltextBackend};
 
     use super::*;
     use crate::v1::ColumnDataType;
@@ -210,13 +230,14 @@ mod tests {
                 enable: true,
                 analyzer: FulltextAnalyzer::English,
                 case_sensitive: false,
+                backend: FulltextBackend::Bloom,
             })
             .unwrap();
         schema.set_inverted_index(true);
         let options = options_from_column_schema(&schema).unwrap();
         assert_eq!(
             options.options.get(FULLTEXT_GRPC_KEY).unwrap(),
-            "{\"enable\":true,\"analyzer\":\"English\",\"case-sensitive\":false}"
+            "{\"enable\":true,\"analyzer\":\"English\",\"case-sensitive\":false,\"backend\":\"bloom\"}"
         );
         assert_eq!(
             options.options.get(INVERTED_INDEX_GRPC_KEY).unwrap(),
@@ -230,11 +251,12 @@ mod tests {
             enable: true,
             analyzer: FulltextAnalyzer::English,
             case_sensitive: false,
+            backend: FulltextBackend::Bloom,
         };
         let options = options_from_fulltext(&fulltext).unwrap().unwrap();
         assert_eq!(
             options.options.get(FULLTEXT_GRPC_KEY).unwrap(),
-            "{\"enable\":true,\"analyzer\":\"English\",\"case-sensitive\":false}"
+            "{\"enable\":true,\"analyzer\":\"English\",\"case-sensitive\":false,\"backend\":\"bloom\"}"
         );
     }
 

src/catalog/src/system_schema.rs

@@ -77,7 +77,7 @@ trait SystemSchemaProviderInner {
     fn system_table(&self, name: &str) -> Option<SystemTableRef>;
 
     fn table_info(catalog_name: String, table: &SystemTableRef) -> TableInfoRef {
-        let table_meta = TableMetaBuilder::default()
+        let table_meta = TableMetaBuilder::empty()
             .schema(table.schema())
             .primary_key_indices(vec![])
             .next_column_id(0)

src/catalog/src/system_schema/information_schema.rs

@@ -19,7 +19,7 @@ mod information_memory_table;
 pub mod key_column_usage;
 mod partitions;
 mod procedure_info;
-mod region_peers;
+pub mod region_peers;
 mod region_statistics;
 mod runtime_metrics;
 pub mod schemata;

src/catalog/src/system_schema/information_schema/columns.rs

@@ -56,6 +56,8 @@ pub const TABLE_CATALOG: &str = "table_catalog";
 pub const TABLE_SCHEMA: &str = "table_schema";
 pub const TABLE_NAME: &str = "table_name";
 pub const COLUMN_NAME: &str = "column_name";
+pub const REGION_ID: &str = "region_id";
+pub const PEER_ID: &str = "peer_id";
 const ORDINAL_POSITION: &str = "ordinal_position";
 const CHARACTER_MAXIMUM_LENGTH: &str = "character_maximum_length";
 const CHARACTER_OCTET_LENGTH: &str = "character_octet_length";
@@ -365,10 +367,6 @@ impl InformationSchemaColumnsBuilder {
             self.numeric_scales.push(None);
 
             match &column_schema.data_type {
-                ConcreteDataType::DateTime(datetime_type) => {
-                    self.datetime_precisions
-                        .push(Some(datetime_type.precision() as i64));
-                }
                 ConcreteDataType::Timestamp(ts_type) => {
                     self.datetime_precisions
                         .push(Some(ts_type.precision() as i64));

src/catalog/src/system_schema/information_schema/flows.rs

@@ -28,16 +28,19 @@ use datafusion::physical_plan::streaming::PartitionStream as DfPartitionStream;
 use datatypes::prelude::ConcreteDataType as CDT;
 use datatypes::scalars::ScalarVectorBuilder;
 use datatypes::schema::{ColumnSchema, Schema, SchemaRef};
+use datatypes::timestamp::TimestampMillisecond;
 use datatypes::value::Value;
 use datatypes::vectors::{
-    Int64VectorBuilder, StringVectorBuilder, UInt32VectorBuilder, UInt64VectorBuilder, VectorRef,
+    Int64VectorBuilder, StringVectorBuilder, TimestampMillisecondVectorBuilder,
+    UInt32VectorBuilder, UInt64VectorBuilder, VectorRef,
 };
 use futures::TryStreamExt;
 use snafu::{OptionExt, ResultExt};
 use store_api::storage::{ScanRequest, TableId};
 
 use crate::error::{
-    CreateRecordBatchSnafu, FlowInfoNotFoundSnafu, InternalSnafu, JsonSnafu, ListFlowsSnafu, Result,
+    CreateRecordBatchSnafu, FlowInfoNotFoundSnafu, InternalSnafu, JsonSnafu, ListFlowsSnafu,
+    Result, UpgradeWeakCatalogManagerRefSnafu,
 };
 use crate::information_schema::{Predicates, FLOWS};
 use crate::system_schema::information_schema::InformationTable;
@@ -59,6 +62,10 @@ pub const SOURCE_TABLE_IDS: &str = "source_table_ids";
 pub const SINK_TABLE_NAME: &str = "sink_table_name";
 pub const FLOWNODE_IDS: &str = "flownode_ids";
 pub const OPTIONS: &str = "options";
+pub const CREATED_TIME: &str = "created_time";
+pub const UPDATED_TIME: &str = "updated_time";
+pub const LAST_EXECUTION_TIME: &str = "last_execution_time";
+pub const SOURCE_TABLE_NAMES: &str = "source_table_names";
 
 /// The `information_schema.flows` to provides information about flows in databases.
 #[derive(Debug)]
@@ -99,6 +106,14 @@ impl InformationSchemaFlows {
                 (SINK_TABLE_NAME, CDT::string_datatype(), false),
                 (FLOWNODE_IDS, CDT::string_datatype(), true),
                 (OPTIONS, CDT::string_datatype(), true),
+                (CREATED_TIME, CDT::timestamp_millisecond_datatype(), false),
+                (UPDATED_TIME, CDT::timestamp_millisecond_datatype(), false),
+                (
+                    LAST_EXECUTION_TIME,
+                    CDT::timestamp_millisecond_datatype(),
+                    true,
+                ),
+                (SOURCE_TABLE_NAMES, CDT::string_datatype(), true),
             ]
             .into_iter()
             .map(|(name, ty, nullable)| ColumnSchema::new(name, ty, nullable))
@@ -170,6 +185,10 @@ struct InformationSchemaFlowsBuilder {
     sink_table_names: StringVectorBuilder,
     flownode_id_groups: StringVectorBuilder,
     option_groups: StringVectorBuilder,
+    created_time: TimestampMillisecondVectorBuilder,
+    updated_time: TimestampMillisecondVectorBuilder,
+    last_execution_time: TimestampMillisecondVectorBuilder,
+    source_table_names: StringVectorBuilder,
 }
 
 impl InformationSchemaFlowsBuilder {
@@ -196,6 +215,10 @@ impl InformationSchemaFlowsBuilder {
             sink_table_names: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             flownode_id_groups: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             option_groups: StringVectorBuilder::with_capacity(INIT_CAPACITY),
+            created_time: TimestampMillisecondVectorBuilder::with_capacity(INIT_CAPACITY),
+            updated_time: TimestampMillisecondVectorBuilder::with_capacity(INIT_CAPACITY),
+            last_execution_time: TimestampMillisecondVectorBuilder::with_capacity(INIT_CAPACITY),
+            source_table_names: StringVectorBuilder::with_capacity(INIT_CAPACITY),
         }
     }
 
@@ -235,13 +258,14 @@ impl InformationSchemaFlowsBuilder {
                     catalog_name: catalog_name.to_string(),
                     flow_name: flow_name.to_string(),
                 })?;
-            self.add_flow(&predicates, flow_id.flow_id(), flow_info, &flow_stat)?;
+            self.add_flow(&predicates, flow_id.flow_id(), flow_info, &flow_stat)
+                .await?;
         }
 
         self.finish()
     }
 
-    fn add_flow(
+    async fn add_flow(
         &mut self,
         predicates: &Predicates,
         flow_id: FlowId,
@@ -290,6 +314,36 @@ impl InformationSchemaFlowsBuilder {
                     input: format!("{:?}", flow_info.options()),
                 },
             )?));
+        self.created_time
+            .push(Some(flow_info.created_time().timestamp_millis().into()));
+        self.updated_time
+            .push(Some(flow_info.updated_time().timestamp_millis().into()));
+        self.last_execution_time
+            .push(flow_stat.as_ref().and_then(|state| {
+                state
+                    .last_exec_time_map
+                    .get(&flow_id)
+                    .map(|v| TimestampMillisecond::new(*v))
+            }));
+
+        let mut source_table_names = vec![];
+        let catalog_name = self.catalog_name.clone();
+        let catalog_manager = self
+            .catalog_manager
+            .upgrade()
+            .context(UpgradeWeakCatalogManagerRefSnafu)?;
+        for schema_name in catalog_manager.schema_names(&catalog_name, None).await? {
+            source_table_names.extend(
+                catalog_manager
+                    .tables_by_ids(&catalog_name, &schema_name, flow_info.source_table_ids())
+                    .await?
+                    .into_iter()
+                    .map(|table| table.table_info().full_table_name()),
+            );
+        }
+
+        let source_table_names = source_table_names.join(",");
+        self.source_table_names.push(Some(&source_table_names));
 
         Ok(())
     }
@@ -307,6 +361,10 @@ impl InformationSchemaFlowsBuilder {
             Arc::new(self.sink_table_names.finish()),
             Arc::new(self.flownode_id_groups.finish()),
             Arc::new(self.option_groups.finish()),
+            Arc::new(self.created_time.finish()),
+            Arc::new(self.updated_time.finish()),
+            Arc::new(self.last_execution_time.finish()),
+            Arc::new(self.source_table_names.finish()),
         ];
         RecordBatch::new(self.schema.clone(), columns).context(CreateRecordBatchSnafu)
     }

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

@@ -20,7 +20,7 @@ use datatypes::vectors::{Int64Vector, StringVector, VectorRef};
 
 use super::table_names::*;
 use crate::system_schema::utils::tables::{
-    bigint_column, datetime_column, string_column, string_columns,
+    bigint_column, string_column, string_columns, timestamp_micro_column,
 };
 
 const NO_VALUE: &str = "NO";
@@ -163,17 +163,17 @@ pub(super) fn get_schema_columns(table_name: &str) -> (SchemaRef, Vec<VectorRef>
                 string_column("EVENT_BODY"),
                 string_column("EVENT_DEFINITION"),
                 string_column("EVENT_TYPE"),
-                datetime_column("EXECUTE_AT"),
+                timestamp_micro_column("EXECUTE_AT"),
                 bigint_column("INTERVAL_VALUE"),
                 string_column("INTERVAL_FIELD"),
                 string_column("SQL_MODE"),
-                datetime_column("STARTS"),
-                datetime_column("ENDS"),
+                timestamp_micro_column("STARTS"),
+                timestamp_micro_column("ENDS"),
                 string_column("STATUS"),
                 string_column("ON_COMPLETION"),
-                datetime_column("CREATED"),
-                datetime_column("LAST_ALTERED"),
-                datetime_column("LAST_EXECUTED"),
+                timestamp_micro_column("CREATED"),
+                timestamp_micro_column("LAST_ALTERED"),
+                timestamp_micro_column("LAST_EXECUTED"),
                 string_column("EVENT_COMMENT"),
                 bigint_column("ORIGINATOR"),
                 string_column("CHARACTER_SET_CLIENT"),
@@ -204,10 +204,10 @@ pub(super) fn get_schema_columns(table_name: &str) -> (SchemaRef, Vec<VectorRef>
                 bigint_column("INITIAL_SIZE"),
                 bigint_column("MAXIMUM_SIZE"),
                 bigint_column("AUTOEXTEND_SIZE"),
-                datetime_column("CREATION_TIME"),
-                datetime_column("LAST_UPDATE_TIME"),
-                datetime_column("LAST_ACCESS_TIME"),
-                datetime_column("RECOVER_TIME"),
+                timestamp_micro_column("CREATION_TIME"),
+                timestamp_micro_column("LAST_UPDATE_TIME"),
+                timestamp_micro_column("LAST_ACCESS_TIME"),
+                timestamp_micro_column("RECOVER_TIME"),
                 bigint_column("TRANSACTION_COUNTER"),
                 string_column("VERSION"),
                 string_column("ROW_FORMAT"),
@@ -217,9 +217,9 @@ pub(super) fn get_schema_columns(table_name: &str) -> (SchemaRef, Vec<VectorRef>
                 bigint_column("MAX_DATA_LENGTH"),
                 bigint_column("INDEX_LENGTH"),
                 bigint_column("DATA_FREE"),
-                datetime_column("CREATE_TIME"),
-                datetime_column("UPDATE_TIME"),
-                datetime_column("CHECK_TIME"),
+                timestamp_micro_column("CREATE_TIME"),
+                timestamp_micro_column("UPDATE_TIME"),
+                timestamp_micro_column("CHECK_TIME"),
                 string_column("CHECKSUM"),
                 string_column("STATUS"),
                 string_column("EXTRA"),
@@ -330,8 +330,8 @@ pub(super) fn get_schema_columns(table_name: &str) -> (SchemaRef, Vec<VectorRef>
                 string_column("SQL_DATA_ACCESS"),
                 string_column("SQL_PATH"),
                 string_column("SECURITY_TYPE"),
-                datetime_column("CREATED"),
-                datetime_column("LAST_ALTERED"),
+                timestamp_micro_column("CREATED"),
+                timestamp_micro_column("LAST_ALTERED"),
                 string_column("SQL_MODE"),
                 string_column("ROUTINE_COMMENT"),
                 string_column("DEFINER"),
@@ -383,7 +383,7 @@ pub(super) fn get_schema_columns(table_name: &str) -> (SchemaRef, Vec<VectorRef>
                 string_column("ACTION_REFERENCE_NEW_TABLE"),
                 string_column("ACTION_REFERENCE_OLD_ROW"),
                 string_column("ACTION_REFERENCE_NEW_ROW"),
-                datetime_column("CREATED"),
+                timestamp_micro_column("CREATED"),
                 string_column("SQL_MODE"),
                 string_column("DEFINER"),
                 string_column("CHARACTER_SET_CLIENT"),

src/catalog/src/system_schema/information_schema/partitions.rs

@@ -20,17 +20,18 @@ use common_catalog::consts::INFORMATION_SCHEMA_PARTITIONS_TABLE_ID;
 use common_error::ext::BoxedError;
 use common_recordbatch::adapter::RecordBatchStreamAdapter;
 use common_recordbatch::{RecordBatch, SendableRecordBatchStream};
-use common_time::datetime::DateTime;
 use datafusion::execution::TaskContext;
 use datafusion::physical_plan::stream::RecordBatchStreamAdapter as DfRecordBatchStreamAdapter;
 use datafusion::physical_plan::streaming::PartitionStream as DfPartitionStream;
 use datafusion::physical_plan::SendableRecordBatchStream as DfSendableRecordBatchStream;
 use datatypes::prelude::{ConcreteDataType, ScalarVectorBuilder, VectorRef};
 use datatypes::schema::{ColumnSchema, Schema, SchemaRef};
+use datatypes::timestamp::TimestampMicrosecond;
 use datatypes::value::Value;
 use datatypes::vectors::{
-    ConstantVector, DateTimeVector, DateTimeVectorBuilder, Int64Vector, Int64VectorBuilder,
-    MutableVector, StringVector, StringVectorBuilder, UInt64VectorBuilder,
+    ConstantVector, Int64Vector, Int64VectorBuilder, MutableVector, StringVector,
+    StringVectorBuilder, TimestampMicrosecondVector, TimestampMicrosecondVectorBuilder,
+    UInt64VectorBuilder,
 };
 use futures::{StreamExt, TryStreamExt};
 use partition::manager::PartitionInfo;
@@ -127,9 +128,21 @@ impl InformationSchemaPartitions {
             ColumnSchema::new("max_data_length", ConcreteDataType::int64_datatype(), true),
             ColumnSchema::new("index_length", ConcreteDataType::int64_datatype(), true),
             ColumnSchema::new("data_free", ConcreteDataType::int64_datatype(), true),
-            ColumnSchema::new("create_time", ConcreteDataType::datetime_datatype(), true),
-            ColumnSchema::new("update_time", ConcreteDataType::datetime_datatype(), true),
-            ColumnSchema::new("check_time", ConcreteDataType::datetime_datatype(), true),
+            ColumnSchema::new(
+                "create_time",
+                ConcreteDataType::timestamp_microsecond_datatype(),
+                true,
+            ),
+            ColumnSchema::new(
+                "update_time",
+                ConcreteDataType::timestamp_microsecond_datatype(),
+                true,
+            ),
+            ColumnSchema::new(
+                "check_time",
+                ConcreteDataType::timestamp_microsecond_datatype(),
+                true,
+            ),
             ColumnSchema::new("checksum", ConcreteDataType::int64_datatype(), true),
             ColumnSchema::new(
                 "partition_comment",
@@ -200,7 +213,7 @@ struct InformationSchemaPartitionsBuilder {
     partition_names: StringVectorBuilder,
     partition_ordinal_positions: Int64VectorBuilder,
     partition_expressions: StringVectorBuilder,
-    create_times: DateTimeVectorBuilder,
+    create_times: TimestampMicrosecondVectorBuilder,
     partition_ids: UInt64VectorBuilder,
 }
 
@@ -220,7 +233,7 @@ impl InformationSchemaPartitionsBuilder {
             partition_names: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             partition_ordinal_positions: Int64VectorBuilder::with_capacity(INIT_CAPACITY),
             partition_expressions: StringVectorBuilder::with_capacity(INIT_CAPACITY),
-            create_times: DateTimeVectorBuilder::with_capacity(INIT_CAPACITY),
+            create_times: TimestampMicrosecondVectorBuilder::with_capacity(INIT_CAPACITY),
             partition_ids: UInt64VectorBuilder::with_capacity(INIT_CAPACITY),
         }
     }
@@ -324,7 +337,7 @@ impl InformationSchemaPartitionsBuilder {
             };
 
             self.partition_expressions.push(expressions.as_deref());
-            self.create_times.push(Some(DateTime::from(
+            self.create_times.push(Some(TimestampMicrosecond::from(
                 table_info.meta.created_on.timestamp_millis(),
             )));
             self.partition_ids.push(Some(partition.id.as_u64()));
@@ -342,8 +355,8 @@ impl InformationSchemaPartitionsBuilder {
             Arc::new(Int64Vector::from(vec![None])),
             rows_num,
         ));
-        let null_datetime_vector = Arc::new(ConstantVector::new(
-            Arc::new(DateTimeVector::from(vec![None])),
+        let null_timestampmicrosecond_vector = Arc::new(ConstantVector::new(
+            Arc::new(TimestampMicrosecondVector::from(vec![None])),
             rows_num,
         ));
         let partition_methods = Arc::new(ConstantVector::new(
@@ -373,8 +386,8 @@ impl InformationSchemaPartitionsBuilder {
             null_i64_vector.clone(),
             Arc::new(self.create_times.finish()),
             // TODO(dennis): supports update_time
-            null_datetime_vector.clone(),
-            null_datetime_vector,
+            null_timestampmicrosecond_vector.clone(),
+            null_timestampmicrosecond_vector,
             null_i64_vector,
             null_string_vector.clone(),
             null_string_vector.clone(),

src/catalog/src/system_schema/information_schema/region_peers.rs

@@ -21,6 +21,7 @@ use common_error::ext::BoxedError;
 use common_meta::rpc::router::RegionRoute;
 use common_recordbatch::adapter::RecordBatchStreamAdapter;
 use common_recordbatch::{RecordBatch, SendableRecordBatchStream};
+use datafusion::common::HashMap;
 use datafusion::execution::TaskContext;
 use datafusion::physical_plan::stream::RecordBatchStreamAdapter as DfRecordBatchStreamAdapter;
 use datafusion::physical_plan::streaming::PartitionStream as DfPartitionStream;
@@ -43,16 +44,22 @@ use crate::kvbackend::KvBackendCatalogManager;
 use crate::system_schema::information_schema::{InformationTable, Predicates};
 use crate::CatalogManager;
 
-const REGION_ID: &str = "region_id";
-const PEER_ID: &str = "peer_id";
+pub const TABLE_CATALOG: &str = "table_catalog";
+pub const TABLE_SCHEMA: &str = "table_schema";
+pub const TABLE_NAME: &str = "table_name";
+pub const REGION_ID: &str = "region_id";
+pub const PEER_ID: &str = "peer_id";
 const PEER_ADDR: &str = "peer_addr";
-const IS_LEADER: &str = "is_leader";
+pub const IS_LEADER: &str = "is_leader";
 const STATUS: &str = "status";
 const DOWN_SECONDS: &str = "down_seconds";
 const INIT_CAPACITY: usize = 42;
 
 /// The `REGION_PEERS` table provides information about the region distribution and routes. Including fields:
 ///
+/// - `table_catalog`: the table catalog name
+/// - `table_schema`: the table schema name
+/// - `table_name`: the table name
 /// - `region_id`: the region id
 /// - `peer_id`: the region storage datanode peer id
 /// - `peer_addr`: the region storage datanode gRPC peer address
@@ -77,6 +84,9 @@ impl InformationSchemaRegionPeers {
 
     pub(crate) fn schema() -> SchemaRef {
         Arc::new(Schema::new(vec![
+            ColumnSchema::new(TABLE_CATALOG, ConcreteDataType::string_datatype(), false),
+            ColumnSchema::new(TABLE_SCHEMA, ConcreteDataType::string_datatype(), false),
+            ColumnSchema::new(TABLE_NAME, ConcreteDataType::string_datatype(), false),
             ColumnSchema::new(REGION_ID, ConcreteDataType::uint64_datatype(), false),
             ColumnSchema::new(PEER_ID, ConcreteDataType::uint64_datatype(), true),
             ColumnSchema::new(PEER_ADDR, ConcreteDataType::string_datatype(), true),
@@ -134,6 +144,9 @@ struct InformationSchemaRegionPeersBuilder {
     catalog_name: String,
     catalog_manager: Weak<dyn CatalogManager>,
 
+    table_catalogs: StringVectorBuilder,
+    table_schemas: StringVectorBuilder,
+    table_names: StringVectorBuilder,
     region_ids: UInt64VectorBuilder,
     peer_ids: UInt64VectorBuilder,
     peer_addrs: StringVectorBuilder,
@@ -152,6 +165,9 @@ impl InformationSchemaRegionPeersBuilder {
             schema,
             catalog_name,
             catalog_manager,
+            table_catalogs: StringVectorBuilder::with_capacity(INIT_CAPACITY),
+            table_schemas: StringVectorBuilder::with_capacity(INIT_CAPACITY),
+            table_names: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             region_ids: UInt64VectorBuilder::with_capacity(INIT_CAPACITY),
             peer_ids: UInt64VectorBuilder::with_capacity(INIT_CAPACITY),
             peer_addrs: StringVectorBuilder::with_capacity(INIT_CAPACITY),
@@ -177,24 +193,28 @@ impl InformationSchemaRegionPeersBuilder {
         let predicates = Predicates::from_scan_request(&request);
 
         for schema_name in catalog_manager.schema_names(&catalog_name, None).await? {
-            let table_id_stream = catalog_manager
+            let table_stream = catalog_manager
                 .tables(&catalog_name, &schema_name, None)
                 .try_filter_map(|t| async move {
                     let table_info = t.table_info();
                     if table_info.table_type == TableType::Temporary {
                         Ok(None)
                     } else {
-                        Ok(Some(table_info.ident.table_id))
+                        Ok(Some((
+                            table_info.ident.table_id,
+                            table_info.name.to_string(),
+                        )))
                     }
                 });
 
             const BATCH_SIZE: usize = 128;
 
-            // Split table ids into chunks
-            let mut table_id_chunks = pin!(table_id_stream.ready_chunks(BATCH_SIZE));
+            // Split tables into chunks
+            let mut table_chunks = pin!(table_stream.ready_chunks(BATCH_SIZE));
 
-            while let Some(table_ids) = table_id_chunks.next().await {
-                let table_ids = table_ids.into_iter().collect::<Result<Vec<_>>>()?;
+            while let Some(tables) = table_chunks.next().await {
+                let tables = tables.into_iter().collect::<Result<HashMap<_, _>>>()?;
+                let table_ids = tables.keys().cloned().collect::<Vec<_>>();
 
                 let table_routes = if let Some(partition_manager) = &partition_manager {
                     partition_manager
@@ -206,7 +226,16 @@ impl InformationSchemaRegionPeersBuilder {
                 };
 
                 for (table_id, routes) in table_routes {
-                    self.add_region_peers(&predicates, table_id, &routes);
+                    // Safety: table_id is guaranteed to be in the map
+                    let table_name = tables.get(&table_id).unwrap();
+                    self.add_region_peers(
+                        &catalog_name,
+                        &schema_name,
+                        table_name,
+                        &predicates,
+                        table_id,
+                        &routes,
+                    );
                 }
             }
         }
@@ -216,6 +245,9 @@ impl InformationSchemaRegionPeersBuilder {
 
     fn add_region_peers(
         &mut self,
+        table_catalog: &str,
+        table_schema: &str,
+        table_name: &str,
         predicates: &Predicates,
         table_id: TableId,
         routes: &[RegionRoute],
@@ -231,13 +263,20 @@ impl InformationSchemaRegionPeersBuilder {
                 Some("ALIVE".to_string())
             };
 
-            let row = [(REGION_ID, &Value::from(region_id))];
+            let row = [
+                (TABLE_CATALOG, &Value::from(table_catalog)),
+                (TABLE_SCHEMA, &Value::from(table_schema)),
+                (TABLE_NAME, &Value::from(table_name)),
+                (REGION_ID, &Value::from(region_id)),
+            ];
 
             if !predicates.eval(&row) {
                 return;
             }
 
-            // TODO(dennis): adds followers.
+            self.table_catalogs.push(Some(table_catalog));
+            self.table_schemas.push(Some(table_schema));
+            self.table_names.push(Some(table_name));
             self.region_ids.push(Some(region_id));
             self.peer_ids.push(peer_id);
             self.peer_addrs.push(peer_addr.as_deref());
@@ -245,11 +284,26 @@ impl InformationSchemaRegionPeersBuilder {
             self.statuses.push(state.as_deref());
             self.down_seconds
                 .push(route.leader_down_millis().map(|m| m / 1000));
+
+            for follower in &route.follower_peers {
+                self.table_catalogs.push(Some(table_catalog));
+                self.table_schemas.push(Some(table_schema));
+                self.table_names.push(Some(table_name));
+                self.region_ids.push(Some(region_id));
+                self.peer_ids.push(Some(follower.id));
+                self.peer_addrs.push(Some(follower.addr.as_str()));
+                self.is_leaders.push(Some("No"));
+                self.statuses.push(None);
+                self.down_seconds.push(None);
+            }
         }
     }
 
     fn finish(&mut self) -> Result<RecordBatch> {
         let columns: Vec<VectorRef> = vec![
+            Arc::new(self.table_catalogs.finish()),
+            Arc::new(self.table_schemas.finish()),
+            Arc::new(self.table_names.finish()),
             Arc::new(self.region_ids.finish()),
             Arc::new(self.peer_ids.finish()),
             Arc::new(self.peer_addrs.finish()),

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

@@ -30,7 +30,8 @@ use datatypes::prelude::{ConcreteDataType, ScalarVectorBuilder, VectorRef};
 use datatypes::schema::{ColumnSchema, Schema, SchemaRef};
 use datatypes::value::Value;
 use datatypes::vectors::{
-    DateTimeVectorBuilder, StringVectorBuilder, UInt32VectorBuilder, UInt64VectorBuilder,
+    StringVectorBuilder, TimestampMicrosecondVectorBuilder, UInt32VectorBuilder,
+    UInt64VectorBuilder,
 };
 use futures::TryStreamExt;
 use snafu::{OptionExt, ResultExt};
@@ -105,9 +106,21 @@ impl InformationSchemaTables {
             ColumnSchema::new(TABLE_ROWS, ConcreteDataType::uint64_datatype(), true),
             ColumnSchema::new(DATA_FREE, ConcreteDataType::uint64_datatype(), true),
             ColumnSchema::new(AUTO_INCREMENT, ConcreteDataType::uint64_datatype(), true),
-            ColumnSchema::new(CREATE_TIME, ConcreteDataType::datetime_datatype(), true),
-            ColumnSchema::new(UPDATE_TIME, ConcreteDataType::datetime_datatype(), true),
-            ColumnSchema::new(CHECK_TIME, ConcreteDataType::datetime_datatype(), true),
+            ColumnSchema::new(
+                CREATE_TIME,
+                ConcreteDataType::timestamp_microsecond_datatype(),
+                true,
+            ),
+            ColumnSchema::new(
+                UPDATE_TIME,
+                ConcreteDataType::timestamp_microsecond_datatype(),
+                true,
+            ),
+            ColumnSchema::new(
+                CHECK_TIME,
+                ConcreteDataType::timestamp_microsecond_datatype(),
+                true,
+            ),
             ColumnSchema::new(TABLE_COLLATION, ConcreteDataType::string_datatype(), true),
             ColumnSchema::new(CHECKSUM, ConcreteDataType::uint64_datatype(), true),
             ColumnSchema::new(CREATE_OPTIONS, ConcreteDataType::string_datatype(), true),
@@ -182,9 +195,9 @@ struct InformationSchemaTablesBuilder {
     max_index_length: UInt64VectorBuilder,
     data_free: UInt64VectorBuilder,
     auto_increment: UInt64VectorBuilder,
-    create_time: DateTimeVectorBuilder,
-    update_time: DateTimeVectorBuilder,
-    check_time: DateTimeVectorBuilder,
+    create_time: TimestampMicrosecondVectorBuilder,
+    update_time: TimestampMicrosecondVectorBuilder,
+    check_time: TimestampMicrosecondVectorBuilder,
     table_collation: StringVectorBuilder,
     checksum: UInt64VectorBuilder,
     create_options: StringVectorBuilder,
@@ -219,9 +232,9 @@ impl InformationSchemaTablesBuilder {
             max_index_length: UInt64VectorBuilder::with_capacity(INIT_CAPACITY),
             data_free: UInt64VectorBuilder::with_capacity(INIT_CAPACITY),
             auto_increment: UInt64VectorBuilder::with_capacity(INIT_CAPACITY),
-            create_time: DateTimeVectorBuilder::with_capacity(INIT_CAPACITY),
-            update_time: DateTimeVectorBuilder::with_capacity(INIT_CAPACITY),
-            check_time: DateTimeVectorBuilder::with_capacity(INIT_CAPACITY),
+            create_time: TimestampMicrosecondVectorBuilder::with_capacity(INIT_CAPACITY),
+            update_time: TimestampMicrosecondVectorBuilder::with_capacity(INIT_CAPACITY),
+            check_time: TimestampMicrosecondVectorBuilder::with_capacity(INIT_CAPACITY),
             table_collation: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             checksum: UInt64VectorBuilder::with_capacity(INIT_CAPACITY),
             create_options: StringVectorBuilder::with_capacity(INIT_CAPACITY),

src/catalog/src/system_schema/utils/tables.rs

@@ -51,10 +51,10 @@ pub fn bigint_column(name: &str) -> ColumnSchema {
     )
 }
 
-pub fn datetime_column(name: &str) -> ColumnSchema {
+pub fn timestamp_micro_column(name: &str) -> ColumnSchema {
     ColumnSchema::new(
         str::to_lowercase(name),
-        ConcreteDataType::datetime_datatype(),
+        ConcreteDataType::timestamp_microsecond_datatype(),
         false,
     )
 }

src/cli/Cargo.toml

@@ -6,6 +6,7 @@ license.workspace = true
 
 [features]
 pg_kvbackend = ["common-meta/pg_kvbackend"]
+mysql_kvbackend = ["common-meta/mysql_kvbackend"]
 
 [lints]
 workspace = true
@@ -43,6 +44,10 @@ futures.workspace = true
 humantime.workspace = true
 meta-client.workspace = true
 nu-ansi-term = "0.46"
+opendal = { version = "0.51.1", features = [
+    "services-fs",
+    "services-s3",
+] }
 query.workspace = true
 rand.workspace = true
 reqwest.workspace = true

src/cli/src/bench.rs

@@ -23,6 +23,8 @@ use common_error::ext::BoxedError;
 use common_meta::key::{TableMetadataManager, TableMetadataManagerRef};
 use common_meta::kv_backend::etcd::EtcdStore;
 use common_meta::kv_backend::memory::MemoryKvBackend;
+#[cfg(feature = "mysql_kvbackend")]
+use common_meta::kv_backend::rds::MySqlStore;
 #[cfg(feature = "pg_kvbackend")]
 use common_meta::kv_backend::rds::PgStore;
 use common_meta::peer::Peer;
@@ -63,6 +65,9 @@ pub struct BenchTableMetadataCommand {
     #[cfg(feature = "pg_kvbackend")]
     #[clap(long)]
     postgres_addr: Option<String>,
+    #[cfg(feature = "mysql_kvbackend")]
+    #[clap(long)]
+    mysql_addr: Option<String>,
     #[clap(long)]
     count: u32,
 }
@@ -86,6 +91,16 @@ impl BenchTableMetadataCommand {
             kv_backend
         };
 
+        #[cfg(feature = "mysql_kvbackend")]
+        let kv_backend = if let Some(mysql_addr) = &self.mysql_addr {
+            info!("Using mysql as kv backend");
+            MySqlStore::with_url(mysql_addr, "greptime_metakv", 128)
+                .await
+                .unwrap()
+        } else {
+            kv_backend
+        };
+
         let table_metadata_manager = Arc::new(TableMetadataManager::new(kv_backend));
 
         let tool = BenchTableMetadata {
@@ -162,7 +177,7 @@ fn create_table_info(table_id: TableId, table_name: TableName) -> RawTableInfo {
 
 fn create_region_routes(regions: Vec<RegionNumber>) -> Vec<RegionRoute> {
     let mut region_routes = Vec::with_capacity(100);
-    let mut rng = rand::thread_rng();
+    let mut rng = rand::rng();
 
     for region_id in regions.into_iter().map(u64::from) {
         region_routes.push(RegionRoute {
@@ -173,7 +188,7 @@ fn create_region_routes(regions: Vec<RegionNumber>) -> Vec<RegionRoute> {
                 attrs: BTreeMap::new(),
             },
             leader_peer: Some(Peer {
-                id: rng.gen_range(0..10),
+                id: rng.random_range(0..10),
                 addr: String::new(),
             }),
             follower_peers: vec![],

src/cli/src/error.rs

@@ -276,6 +276,24 @@ pub enum Error {
         #[snafu(implicit)]
         location: Location,
     },
+
+    #[snafu(display("OpenDAL operator failed"))]
+    OpenDal {
+        #[snafu(implicit)]
+        location: Location,
+        #[snafu(source)]
+        error: opendal::Error,
+    },
+    #[snafu(display("S3 config need be set"))]
+    S3ConfigNotSet {
+        #[snafu(implicit)]
+        location: Location,
+    },
+    #[snafu(display("Output directory not set"))]
+    OutputDirNotSet {
+        #[snafu(implicit)]
+        location: Location,
+    },
 }
 
 pub type Result<T> = std::result::Result<T, Error>;
@@ -319,6 +337,9 @@ impl ErrorExt for Error {
             | Error::BuildClient { .. } => StatusCode::Unexpected,
 
             Error::Other { source, .. } => source.status_code(),
+            Error::OpenDal { .. } => StatusCode::Internal,
+            Error::S3ConfigNotSet { .. } => StatusCode::InvalidArguments,
+            Error::OutputDirNotSet { .. } => StatusCode::InvalidArguments,
 
             Error::BuildRuntime { source, .. } => source.status_code(),
 

src/cli/src/export.rs

@@ -21,15 +21,18 @@ use async_trait::async_trait;
 use clap::{Parser, ValueEnum};
 use common_error::ext::BoxedError;
 use common_telemetry::{debug, error, info};
+use opendal::layers::LoggingLayer;
+use opendal::{services, Operator};
 use serde_json::Value;
 use snafu::{OptionExt, ResultExt};
-use tokio::fs::File;
-use tokio::io::{AsyncWriteExt, BufWriter};
 use tokio::sync::Semaphore;
 use tokio::time::Instant;
 
 use crate::database::{parse_proxy_opts, DatabaseClient};
-use crate::error::{EmptyResultSnafu, Error, FileIoSnafu, Result, SchemaNotFoundSnafu};
+use crate::error::{
+    EmptyResultSnafu, Error, OpenDalSnafu, OutputDirNotSetSnafu, Result, S3ConfigNotSetSnafu,
+    SchemaNotFoundSnafu,
+};
 use crate::{database, Tool};
 
 type TableReference = (String, String, String);
@@ -52,8 +55,9 @@ pub struct ExportCommand {
     addr: String,
 
     /// Directory to put the exported data. E.g.: /tmp/greptimedb-export
+    /// for local export.
     #[clap(long)]
-    output_dir: String,
+    output_dir: Option<String>,
 
     /// The name of the catalog to export.
     #[clap(long, default_value = "greptime-*")]
@@ -101,10 +105,51 @@ pub struct ExportCommand {
     /// Disable proxy server, if set, will not use any proxy.
     #[clap(long)]
     no_proxy: bool,
+
+    /// if export data to s3
+    #[clap(long)]
+    s3: bool,
+
+    /// The s3 bucket name
+    /// if s3 is set, this is required
+    #[clap(long)]
+    s3_bucket: Option<String>,
+
+    /// The s3 endpoint
+    /// if s3 is set, this is required
+    #[clap(long)]
+    s3_endpoint: Option<String>,
+
+    /// The s3 access key
+    /// if s3 is set, this is required
+    #[clap(long)]
+    s3_access_key: Option<String>,
+
+    /// The s3 secret key
+    /// if s3 is set, this is required
+    #[clap(long)]
+    s3_secret_key: Option<String>,
+
+    /// The s3 region
+    /// if s3 is set, this is required
+    #[clap(long)]
+    s3_region: Option<String>,
 }
 
 impl ExportCommand {
     pub async fn build(&self) -> std::result::Result<Box<dyn Tool>, BoxedError> {
+        if self.s3
+            && (self.s3_bucket.is_none()
+                || self.s3_endpoint.is_none()
+                || self.s3_access_key.is_none()
+                || self.s3_secret_key.is_none()
+                || self.s3_region.is_none())
+        {
+            return Err(BoxedError::new(S3ConfigNotSetSnafu {}.build()));
+        }
+        if !self.s3 && self.output_dir.is_none() {
+            return Err(BoxedError::new(OutputDirNotSetSnafu {}.build()));
+        }
         let (catalog, schema) =
             database::split_database(&self.database).map_err(BoxedError::new)?;
         let proxy = parse_proxy_opts(self.proxy.clone(), self.no_proxy)?;
@@ -126,24 +171,43 @@ impl ExportCommand {
             target: self.target.clone(),
             start_time: self.start_time.clone(),
             end_time: self.end_time.clone(),
+            s3: self.s3,
+            s3_bucket: self.s3_bucket.clone(),
+            s3_endpoint: self.s3_endpoint.clone(),
+            s3_access_key: self.s3_access_key.clone(),
+            s3_secret_key: self.s3_secret_key.clone(),
+            s3_region: self.s3_region.clone(),
         }))
     }
 }
 
+#[derive(Clone)]
 pub struct Export {
     catalog: String,
     schema: Option<String>,
     database_client: DatabaseClient,
-    output_dir: String,
+    output_dir: Option<String>,
     parallelism: usize,
     target: ExportTarget,
     start_time: Option<String>,
     end_time: Option<String>,
+    s3: bool,
+    s3_bucket: Option<String>,
+    s3_endpoint: Option<String>,
+    s3_access_key: Option<String>,
+    s3_secret_key: Option<String>,
+    s3_region: Option<String>,
 }
 
 impl Export {
     fn catalog_path(&self) -> PathBuf {
-        PathBuf::from(&self.output_dir).join(&self.catalog)
+        if self.s3 {
+            PathBuf::from(&self.catalog)
+        } else if let Some(dir) = &self.output_dir {
+            PathBuf::from(dir).join(&self.catalog)
+        } else {
+            unreachable!("catalog_path: output_dir must be set when not using s3")
+        }
     }
 
     async fn get_db_names(&self) -> Result<Vec<String>> {
@@ -300,19 +364,23 @@ impl Export {
         let timer = Instant::now();
         let db_names = self.get_db_names().await?;
         let db_count = db_names.len();
+        let operator = self.build_operator().await?;
+
         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 file_path = self.get_file_path(&schema, "create_database.sql");
+            self.write_to_storage(&operator, &file_path, create_database.into_bytes())
+                .await?;
+
+            info!(
+                "Exported {}.{} database creation SQL to {}",
+                self.catalog,
+                schema,
+                self.format_output_path(&file_path)
+            );
         }
 
         let elapsed = timer.elapsed();
@@ -326,149 +394,267 @@ impl Export {
         let semaphore = Arc::new(Semaphore::new(self.parallelism));
         let db_names = self.get_db_names().await?;
         let db_count = db_names.len();
+        let operator = Arc::new(self.build_operator().await?);
         let mut tasks = Vec::with_capacity(db_names.len());
+
         for schema in db_names {
             let semaphore_moved = semaphore.clone();
+            let export_self = self.clone();
+            let operator = operator.clone();
             tasks.push(async move {
                 let _permit = semaphore_moved.acquire().await.unwrap();
-                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 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) {
-                    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)?;
+                let (metric_physical_tables, remaining_tables, views) = export_self
+                    .get_table_list(&export_self.catalog, &schema)
+                    .await?;
+
+                // Create directory if needed for file system storage
+                if !export_self.s3 {
+                    let db_dir = format!("{}/{}/", export_self.catalog, schema);
+                    operator.create_dir(&db_dir).await.context(OpenDalSnafu)?;
                 }
 
+                let file_path = export_self.get_file_path(&schema, "create_tables.sql");
+                let mut content = Vec::new();
+
+                // Add table creation SQL
+                for (c, s, t) in metric_physical_tables.iter().chain(&remaining_tables) {
+                    let create_table = export_self.show_create("TABLE", c, s, Some(t)).await?;
+                    content.extend_from_slice(create_table.as_bytes());
+                }
+
+                // Add view creation SQL
+                for (c, s, v) in &views {
+                    let create_view = export_self.show_create("VIEW", c, s, Some(v)).await?;
+                    content.extend_from_slice(create_view.as_bytes());
+                }
+
+                // Write to storage
+                export_self
+                    .write_to_storage(&operator, &file_path, content)
+                    .await?;
+
                 info!(
-                    "Finished exporting {}.{schema} with {table_count} table schemas to path: {}",
-                    self.catalog,
-                    db_dir.to_string_lossy()
+                    "Finished exporting {}.{schema} with {} table schemas to path: {}",
+                    export_self.catalog,
+                    metric_physical_tables.len() + remaining_tables.len() + views.len(),
+                    export_self.format_output_path(&file_path)
                 );
 
                 Ok::<(), Error>(())
             });
         }
 
-        let success = futures::future::join_all(tasks)
-            .await
-            .into_iter()
-            .filter(|r| match r {
-                Ok(_) => true,
-                Err(e) => {
-                    error!(e; "export schema job failed");
-                    false
-                }
-            })
-            .count();
-
+        let success = self.execute_tasks(tasks).await;
         let elapsed = timer.elapsed();
         info!("Success {success}/{db_count} jobs, cost: {elapsed:?}");
 
         Ok(())
     }
 
+    async fn build_operator(&self) -> Result<Operator> {
+        if self.s3 {
+            self.build_s3_operator().await
+        } else {
+            self.build_fs_operator().await
+        }
+    }
+
+    async fn build_s3_operator(&self) -> Result<Operator> {
+        let mut builder = services::S3::default().root("").bucket(
+            self.s3_bucket
+                .as_ref()
+                .expect("s3_bucket must be provided when s3 is enabled"),
+        );
+
+        if let Some(endpoint) = self.s3_endpoint.as_ref() {
+            builder = builder.endpoint(endpoint);
+        }
+
+        if let Some(region) = self.s3_region.as_ref() {
+            builder = builder.region(region);
+        }
+
+        if let Some(key_id) = self.s3_access_key.as_ref() {
+            builder = builder.access_key_id(key_id);
+        }
+
+        if let Some(secret_key) = self.s3_secret_key.as_ref() {
+            builder = builder.secret_access_key(secret_key);
+        }
+
+        let op = Operator::new(builder)
+            .context(OpenDalSnafu)?
+            .layer(LoggingLayer::default())
+            .finish();
+        Ok(op)
+    }
+
+    async fn build_fs_operator(&self) -> Result<Operator> {
+        let root = self
+            .output_dir
+            .as_ref()
+            .context(OutputDirNotSetSnafu)?
+            .clone();
+        let op = Operator::new(services::Fs::default().root(&root))
+            .context(OpenDalSnafu)?
+            .layer(LoggingLayer::default())
+            .finish();
+        Ok(op)
+    }
+
     async fn export_database_data(&self) -> 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);
+        let operator = Arc::new(self.build_operator().await?);
+        let with_options = build_with_options(&self.start_time, &self.end_time);
+
         for schema in db_names {
             let semaphore_moved = semaphore.clone();
+            let export_self = self.clone();
+            let with_options_clone = with_options.clone();
+            let operator = operator.clone();
+
             tasks.push(async move {
                 let _permit = semaphore_moved.acquire().await.unwrap();
-                let db_dir = self.catalog_path().join(format!("{schema}/"));
-                tokio::fs::create_dir_all(&db_dir)
-                    .await
-                    .context(FileIoSnafu)?;
 
-                let with_options = match (&self.start_time, &self.end_time) {
-                    (Some(start_time), Some(end_time)) => {
-                        format!(
-                            "WITH (FORMAT='parquet', start_time='{}', end_time='{}')",
-                            start_time, end_time
-                        )
-                    }
-                    (Some(start_time), None) => {
-                        format!("WITH (FORMAT='parquet', start_time='{}')", start_time)
-                    }
-                    (None, Some(end_time)) => {
-                        format!("WITH (FORMAT='parquet', end_time='{}')", end_time)
-                    }
-                    (None, None) => "WITH (FORMAT='parquet')".to_string(),
-                };
+                // Create directory if not using S3
+                if !export_self.s3 {
+                    let db_dir = format!("{}/{}/", export_self.catalog, schema);
+                    operator.create_dir(&db_dir).await.context(OpenDalSnafu)?;
+                }
 
+                let (path, connection_part) = export_self.get_storage_params(&schema);
+
+                // Execute COPY DATABASE TO command
                 let sql = format!(
-                    r#"COPY DATABASE "{}"."{}" TO '{}' {};"#,
-                    self.catalog,
-                    schema,
-                    db_dir.to_str().unwrap(),
-                    with_options
+                    r#"COPY DATABASE "{}"."{}" TO '{}' WITH ({}){};"#,
+                    export_self.catalog, schema, path, with_options_clone, connection_part
+                );
+                info!("Executing sql: {sql}");
+                export_self.database_client.sql_in_public(&sql).await?;
+                info!(
+                    "Finished exporting {}.{} data to {}",
+                    export_self.catalog, schema, path
                 );
 
-                info!("Executing sql: {sql}");
+                // Create copy_from.sql file
+                let copy_database_from_sql = format!(
+                    r#"COPY DATABASE "{}"."{}" FROM '{}' WITH ({}){};"#,
+                    export_self.catalog, schema, path, with_options_clone, connection_part
+                );
 
-                self.database_client.sql_in_public(&sql).await?;
+                let copy_from_path = export_self.get_file_path(&schema, "copy_from.sql");
+                export_self
+                    .write_to_storage(
+                        &operator,
+                        &copy_from_path,
+                        copy_database_from_sql.into_bytes(),
+                    )
+                    .await?;
 
                 info!(
-                    "Finished exporting {}.{schema} data into path: {}",
-                    self.catalog,
-                    db_dir.to_string_lossy()
-                );
-
-                // The export 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');"#,
-                    self.catalog,
+                    "Finished exporting {}.{} copy_from.sql to {}",
+                    export_self.catalog,
                     schema,
-                    db_dir.to_str().unwrap()
+                    export_self.format_output_path(&copy_from_path)
                 );
-                writer
-                    .write(copy_database_from_sql.as_bytes())
-                    .await
-                    .context(FileIoSnafu)?;
-                writer.flush().await.context(FileIoSnafu)?;
-
-                info!("Finished exporting {}.{schema} copy_from.sql", self.catalog);
 
                 Ok::<(), Error>(())
-            })
+            });
         }
 
-        let success = futures::future::join_all(tasks)
+        let success = self.execute_tasks(tasks).await;
+        let elapsed = timer.elapsed();
+        info!("Success {success}/{db_count} jobs, costs: {elapsed:?}");
+
+        Ok(())
+    }
+
+    fn get_file_path(&self, schema: &str, file_name: &str) -> String {
+        format!("{}/{}/{}", self.catalog, schema, file_name)
+    }
+
+    fn format_output_path(&self, file_path: &str) -> String {
+        if self.s3 {
+            format!(
+                "s3://{}/{}",
+                self.s3_bucket.as_ref().unwrap_or(&String::new()),
+                file_path
+            )
+        } else {
+            format!(
+                "{}/{}",
+                self.output_dir.as_ref().unwrap_or(&String::new()),
+                file_path
+            )
+        }
+    }
+
+    async fn write_to_storage(
+        &self,
+        op: &Operator,
+        file_path: &str,
+        content: Vec<u8>,
+    ) -> Result<()> {
+        op.write(file_path, content).await.context(OpenDalSnafu)
+    }
+
+    fn get_storage_params(&self, schema: &str) -> (String, String) {
+        if self.s3 {
+            let s3_path = format!(
+                "s3://{}/{}/{}/",
+                // Safety: s3_bucket is required when s3 is enabled
+                self.s3_bucket.as_ref().unwrap(),
+                self.catalog,
+                schema
+            );
+
+            // endpoint is optional
+            let endpoint_option = if let Some(endpoint) = self.s3_endpoint.as_ref() {
+                format!(", ENDPOINT='{}'", endpoint)
+            } else {
+                String::new()
+            };
+
+            // Safety: All s3 options are required
+            let connection_options = format!(
+                "ACCESS_KEY_ID='{}', SECRET_ACCESS_KEY='{}', REGION='{}'{}",
+                self.s3_access_key.as_ref().unwrap(),
+                self.s3_secret_key.as_ref().unwrap(),
+                self.s3_region.as_ref().unwrap(),
+                endpoint_option
+            );
+
+            (s3_path, format!(" CONNECTION ({})", connection_options))
+        } else {
+            (
+                self.catalog_path()
+                    .join(format!("{schema}/"))
+                    .to_string_lossy()
+                    .to_string(),
+                String::new(),
+            )
+        }
+    }
+
+    async fn execute_tasks(
+        &self,
+        tasks: Vec<impl std::future::Future<Output = Result<()>>>,
+    ) -> usize {
+        futures::future::join_all(tasks)
             .await
             .into_iter()
             .filter(|r| match r {
                 Ok(_) => true,
                 Err(e) => {
-                    error!(e; "export database job failed");
+                    error!(e; "export job failed");
                     false
                 }
             })
-            .count();
-        let elapsed = timer.elapsed();
-
-        info!("Success {success}/{db_count} jobs, costs: {elapsed:?}");
-
-        Ok(())
+            .count()
     }
 }
 
@@ -493,3 +679,15 @@ impl Tool for Export {
         }
     }
 }
+
+/// Builds the WITH options string for SQL commands, assuming consistent syntax across S3 and local exports.
+fn build_with_options(start_time: &Option<String>, end_time: &Option<String>) -> String {
+    let mut options = vec!["format = 'parquet'".to_string()];
+    if let Some(start) = start_time {
+        options.push(format!("start_time = '{}'", start));
+    }
+    if let Some(end) = end_time {
+        options.push(format!("end_time = '{}'", end));
+    }
+    options.join(", ")
+}

src/client/src/load_balance.rs

@@ -13,7 +13,7 @@
 // limitations under the License.
 
 use enum_dispatch::enum_dispatch;
-use rand::seq::SliceRandom;
+use rand::seq::IndexedRandom;
 
 #[enum_dispatch]
 pub trait LoadBalance {
@@ -37,7 +37,7 @@ pub struct Random;
 
 impl LoadBalance for Random {
     fn get_peer<'a>(&self, peers: &'a [String]) -> Option<&'a String> {
-        peers.choose(&mut rand::thread_rng())
+        peers.choose(&mut rand::rng())
     }
 }
 

src/cmd/src/datanode.rs

@@ -30,7 +30,7 @@ use datanode::datanode::{Datanode, DatanodeBuilder};
 use datanode::service::DatanodeServiceBuilder;
 use meta_client::{MetaClientOptions, MetaClientType};
 use servers::Mode;
-use snafu::{OptionExt, ResultExt};
+use snafu::{ensure, OptionExt, ResultExt};
 use tracing_appender::non_blocking::WorkerGuard;
 
 use crate::error::{
@@ -223,15 +223,14 @@ impl StartCommand {
                 .get_or_insert_with(MetaClientOptions::default)
                 .metasrv_addrs
                 .clone_from(metasrv_addrs);
-            opts.mode = Mode::Distributed;
         }
 
-        if let (Mode::Distributed, None) = (&opts.mode, &opts.node_id) {
-            return MissingConfigSnafu {
-                msg: "Missing node id option",
+        ensure!(
+            opts.node_id.is_some(),
+            MissingConfigSnafu {
+                msg: "Missing node id option"
             }
-            .fail();
-        }
+        );
 
         if let Some(data_home) = &self.data_home {
             opts.storage.data_home.clone_from(data_home);
@@ -295,10 +294,13 @@ impl StartCommand {
             msg: "'meta_client_options'",
         })?;
 
-        let meta_client =
-            meta_client::create_meta_client(MetaClientType::Datanode { member_id }, meta_config)
-                .await
-                .context(MetaClientInitSnafu)?;
+        let meta_client = meta_client::create_meta_client(
+            MetaClientType::Datanode { member_id },
+            meta_config,
+            None,
+        )
+        .await
+        .context(MetaClientInitSnafu)?;
 
         let meta_backend = Arc::new(MetaKvBackend {
             client: meta_client.clone(),
@@ -311,7 +313,7 @@ impl StartCommand {
                 .build(),
         );
 
-        let mut datanode = DatanodeBuilder::new(opts.clone(), plugins)
+        let mut datanode = DatanodeBuilder::new(opts.clone(), plugins, Mode::Distributed)
             .with_meta_client(meta_client)
             .with_kv_backend(meta_backend)
             .with_cache_registry(layered_cache_registry)
@@ -333,6 +335,7 @@ impl StartCommand {
 
 #[cfg(test)]
 mod tests {
+    use std::assert_matches::assert_matches;
     use std::io::Write;
     use std::time::Duration;
 
@@ -340,7 +343,6 @@ mod tests {
     use common_test_util::temp_dir::create_named_temp_file;
     use datanode::config::{FileConfig, GcsConfig, ObjectStoreConfig, S3Config};
     use servers::heartbeat_options::HeartbeatOptions;
-    use servers::Mode;
 
     use super::*;
     use crate::options::GlobalOptions;
@@ -406,7 +408,7 @@ mod tests {
             sync_write = false
 
             [storage]
-            data_home = "/tmp/greptimedb/"
+            data_home = "./greptimedb_data/"
             type = "File"
 
             [[storage.providers]]
@@ -420,7 +422,7 @@ mod tests {
 
             [logging]
             level = "debug"
-            dir = "/tmp/greptimedb/test/logs"
+            dir = "./greptimedb_data/test/logs"
         "#;
         write!(file, "{}", toml_str).unwrap();
 
@@ -467,7 +469,7 @@ mod tests {
         assert_eq!(10000, ddl_timeout.as_millis());
         assert_eq!(3000, timeout.as_millis());
         assert!(tcp_nodelay);
-        assert_eq!("/tmp/greptimedb/", options.storage.data_home);
+        assert_eq!("./greptimedb_data/", options.storage.data_home);
         assert!(matches!(
             &options.storage.store,
             ObjectStoreConfig::File(FileConfig { .. })
@@ -483,27 +485,14 @@ mod tests {
         ));
 
         assert_eq!("debug", options.logging.level.unwrap());
-        assert_eq!("/tmp/greptimedb/test/logs".to_string(), options.logging.dir);
+        assert_eq!(
+            "./greptimedb_data/test/logs".to_string(),
+            options.logging.dir
+        );
     }
 
     #[test]
     fn test_try_from_cmd() {
-        let opt = StartCommand::default()
-            .load_options(&GlobalOptions::default())
-            .unwrap()
-            .component;
-        assert_eq!(Mode::Standalone, opt.mode);
-
-        let opt = (StartCommand {
-            node_id: Some(42),
-            metasrv_addrs: Some(vec!["127.0.0.1:3002".to_string()]),
-            ..Default::default()
-        })
-        .load_options(&GlobalOptions::default())
-        .unwrap()
-        .component;
-        assert_eq!(Mode::Distributed, opt.mode);
-
         assert!((StartCommand {
             metasrv_addrs: Some(vec!["127.0.0.1:3002".to_string()]),
             ..Default::default()
@@ -522,11 +511,23 @@ mod tests {
 
     #[test]
     fn test_load_log_options_from_cli() {
-        let cmd = StartCommand::default();
+        let mut cmd = StartCommand::default();
+
+        let result = cmd.load_options(&GlobalOptions {
+            log_dir: Some("./greptimedb_data/test/logs".to_string()),
+            log_level: Some("debug".to_string()),
+
+            #[cfg(feature = "tokio-console")]
+            tokio_console_addr: None,
+        });
+        // Missing node_id.
+        assert_matches!(result, Err(crate::error::Error::MissingConfig { .. }));
+
+        cmd.node_id = Some(42);
 
         let options = cmd
             .load_options(&GlobalOptions {
-                log_dir: Some("/tmp/greptimedb/test/logs".to_string()),
+                log_dir: Some("./greptimedb_data/test/logs".to_string()),
                 log_level: Some("debug".to_string()),
 
                 #[cfg(feature = "tokio-console")]
@@ -536,7 +537,7 @@ mod tests {
             .component;
 
         let logging_opt = options.logging;
-        assert_eq!("/tmp/greptimedb/test/logs", logging_opt.dir);
+        assert_eq!("./greptimedb_data/test/logs", logging_opt.dir);
         assert_eq!("debug", logging_opt.level.as_ref().unwrap());
     }
 
@@ -565,11 +566,11 @@ mod tests {
 
             [storage]
             type = "File"
-            data_home = "/tmp/greptimedb/"
+            data_home = "./greptimedb_data/"
 
             [logging]
             level = "debug"
-            dir = "/tmp/greptimedb/test/logs"
+            dir = "./greptimedb_data/test/logs"
         "#;
         write!(file, "{}", toml_str).unwrap();
 

src/cmd/src/error.rs

@@ -100,6 +100,13 @@ pub enum Error {
         source: flow::Error,
     },
 
+    #[snafu(display("Servers error"))]
+    Servers {
+        #[snafu(implicit)]
+        location: Location,
+        source: servers::error::Error,
+    },
+
     #[snafu(display("Failed to start frontend"))]
     StartFrontend {
         #[snafu(implicit)]
@@ -365,6 +372,7 @@ impl ErrorExt for Error {
             Error::ShutdownFrontend { source, .. } => source.status_code(),
             Error::StartMetaServer { source, .. } => source.status_code(),
             Error::ShutdownMetaServer { source, .. } => source.status_code(),
+            Error::Servers { source, .. } => source.status_code(),
             Error::BuildMetaServer { source, .. } => source.status_code(),
             Error::UnsupportedSelectorType { source, .. } => source.status_code(),
             Error::BuildCli { source, .. } => source.status_code(),

src/cmd/src/flownode.rs

@@ -32,10 +32,9 @@ use common_meta::key::TableMetadataManager;
 use common_telemetry::info;
 use common_telemetry::logging::TracingOptions;
 use common_version::{short_version, version};
-use flow::{FlownodeBuilder, FlownodeInstance, FrontendClient, FrontendInvoker};
+use flow::{FlownodeBuilder, FlownodeInstance, FrontendInvoker};
 use meta_client::{MetaClientOptions, MetaClientType};
-use servers::Mode;
-use snafu::{OptionExt, ResultExt};
+use snafu::{ensure, OptionExt, ResultExt};
 use tracing_appender::non_blocking::WorkerGuard;
 
 use crate::error::{
@@ -203,7 +202,6 @@ impl StartCommand {
                 .get_or_insert_with(MetaClientOptions::default)
                 .metasrv_addrs
                 .clone_from(metasrv_addrs);
-            opts.mode = Mode::Distributed;
         }
 
         if let Some(http_addr) = &self.http_addr {
@@ -214,12 +212,12 @@ impl StartCommand {
             opts.http.timeout = Duration::from_secs(http_timeout);
         }
 
-        if let (Mode::Distributed, None) = (&opts.mode, &opts.node_id) {
-            return MissingConfigSnafu {
-                msg: "Missing node id option",
+        ensure!(
+            opts.node_id.is_some(),
+            MissingConfigSnafu {
+                msg: "Missing node id option"
             }
-            .fail();
-        }
+        );
 
         Ok(())
     }
@@ -249,10 +247,13 @@ impl StartCommand {
             msg: "'meta_client_options'",
         })?;
 
-        let meta_client =
-            meta_client::create_meta_client(MetaClientType::Flownode { member_id }, meta_config)
-                .await
-                .context(MetaClientInitSnafu)?;
+        let meta_client = meta_client::create_meta_client(
+            MetaClientType::Flownode { member_id },
+            meta_config,
+            None,
+        )
+        .await
+        .context(MetaClientInitSnafu)?;
 
         let cache_max_capacity = meta_config.metadata_cache_max_capacity;
         let cache_ttl = meta_config.metadata_cache_ttl;
@@ -311,8 +312,6 @@ impl StartCommand {
             Arc::new(executor),
         );
 
-        let frontend_client = FrontendClient::from_meta_client(meta_client.clone());
-
         let flow_metadata_manager = Arc::new(FlowMetadataManager::new(cached_meta_backend.clone()));
         let flownode_builder = FlownodeBuilder::new(
             opts,
@@ -320,7 +319,6 @@ impl StartCommand {
             table_metadata_manager,
             catalog_manager.clone(),
             flow_metadata_manager,
-            Arc::new(frontend_client),
         )
         .with_heartbeat_task(heartbeat_task);
 

src/cmd/src/frontend.rs

@@ -32,28 +32,25 @@ use common_telemetry::info;
 use common_telemetry::logging::TracingOptions;
 use common_time::timezone::set_default_timezone;
 use common_version::{short_version, version};
+use frontend::frontend::Frontend;
 use frontend::heartbeat::HeartbeatTask;
 use frontend::instance::builder::FrontendBuilder;
-use frontend::instance::{FrontendInstance, Instance as FeInstance};
 use frontend::server::Services;
 use meta_client::{MetaClientOptions, MetaClientType};
 use query::stats::StatementStatistics;
+use servers::export_metrics::ExportMetricsTask;
 use servers::tls::{TlsMode, TlsOption};
 use snafu::{OptionExt, ResultExt};
 use tracing_appender::non_blocking::WorkerGuard;
 
-use crate::error::{
-    self, InitTimezoneSnafu, LoadLayeredConfigSnafu, MetaClientInitSnafu, MissingConfigSnafu,
-    Result, StartFrontendSnafu,
-};
+use crate::error::{self, Result};
 use crate::options::{GlobalOptions, GreptimeOptions};
 use crate::{log_versions, App};
 
 type FrontendOptions = GreptimeOptions<frontend::frontend::FrontendOptions>;
 
 pub struct Instance {
-    frontend: FeInstance,
-
+    frontend: Frontend,
     // Keep the logging guard to prevent the worker from being dropped.
     _guard: Vec<WorkerGuard>,
 }
@@ -61,20 +58,17 @@ pub struct Instance {
 pub const APP_NAME: &str = "greptime-frontend";
 
 impl Instance {
-    pub fn new(frontend: FeInstance, guard: Vec<WorkerGuard>) -> Self {
-        Self {
-            frontend,
-            _guard: guard,
-        }
+    pub fn new(frontend: Frontend, _guard: Vec<WorkerGuard>) -> Self {
+        Self { frontend, _guard }
     }
 
-    pub fn mut_inner(&mut self) -> &mut FeInstance {
-        &mut self.frontend
-    }
-
-    pub fn inner(&self) -> &FeInstance {
+    pub fn inner(&self) -> &Frontend {
         &self.frontend
     }
+
+    pub fn mut_inner(&mut self) -> &mut Frontend {
+        &mut self.frontend
+    }
 }
 
 #[async_trait]
@@ -84,11 +78,15 @@ impl App for Instance {
     }
 
     async fn start(&mut self) -> Result<()> {
-        plugins::start_frontend_plugins(self.frontend.plugins().clone())
+        let plugins = self.frontend.instance.plugins().clone();
+        plugins::start_frontend_plugins(plugins)
             .await
-            .context(StartFrontendSnafu)?;
+            .context(error::StartFrontendSnafu)?;
 
-        self.frontend.start().await.context(StartFrontendSnafu)
+        self.frontend
+            .start()
+            .await
+            .context(error::StartFrontendSnafu)
     }
 
     async fn stop(&self) -> Result<()> {
@@ -178,7 +176,7 @@ impl StartCommand {
             self.config_file.as_deref(),
             self.env_prefix.as_ref(),
         )
-        .context(LoadLayeredConfigSnafu)?;
+        .context(error::LoadLayeredConfigSnafu)?;
 
         self.merge_with_cli_options(global_options, &mut opts)?;
 
@@ -283,22 +281,28 @@ impl StartCommand {
         let mut plugins = Plugins::new();
         plugins::setup_frontend_plugins(&mut plugins, &plugin_opts, &opts)
             .await
-            .context(StartFrontendSnafu)?;
+            .context(error::StartFrontendSnafu)?;
 
-        set_default_timezone(opts.default_timezone.as_deref()).context(InitTimezoneSnafu)?;
+        set_default_timezone(opts.default_timezone.as_deref()).context(error::InitTimezoneSnafu)?;
 
-        let meta_client_options = opts.meta_client.as_ref().context(MissingConfigSnafu {
-            msg: "'meta_client'",
-        })?;
+        let meta_client_options = opts
+            .meta_client
+            .as_ref()
+            .context(error::MissingConfigSnafu {
+                msg: "'meta_client'",
+            })?;
 
         let cache_max_capacity = meta_client_options.metadata_cache_max_capacity;
         let cache_ttl = meta_client_options.metadata_cache_ttl;
         let cache_tti = meta_client_options.metadata_cache_tti;
 
-        let meta_client =
-            meta_client::create_meta_client(MetaClientType::Frontend, meta_client_options)
-                .await
-                .context(MetaClientInitSnafu)?;
+        let meta_client = meta_client::create_meta_client(
+            MetaClientType::Frontend,
+            meta_client_options,
+            Some(&plugins),
+        )
+        .await
+        .context(error::MetaClientInitSnafu)?;
 
         // TODO(discord9): add helper function to ease the creation of cache registry&such
         let cached_meta_backend =
@@ -345,6 +349,7 @@ impl StartCommand {
             opts.heartbeat.clone(),
             Arc::new(executor),
         );
+        let heartbeat_task = Some(heartbeat_task);
 
         // frontend to datanode need not timeout.
         // Some queries are expected to take long time.
@@ -356,7 +361,7 @@ impl StartCommand {
         };
         let client = NodeClients::new(channel_config);
 
-        let mut instance = FrontendBuilder::new(
+        let instance = FrontendBuilder::new(
             opts.clone(),
             cached_meta_backend.clone(),
             layered_cache_registry.clone(),
@@ -367,20 +372,27 @@ impl StartCommand {
         )
         .with_plugin(plugins.clone())
         .with_local_cache_invalidator(layered_cache_registry)
-        .with_heartbeat_task(heartbeat_task)
         .try_build()
         .await
-        .context(StartFrontendSnafu)?;
+        .context(error::StartFrontendSnafu)?;
+        let instance = Arc::new(instance);
 
-        let servers = Services::new(opts, Arc::new(instance.clone()), plugins)
+        let export_metrics_task = ExportMetricsTask::try_new(&opts.export_metrics, Some(&plugins))
+            .context(error::ServersSnafu)?;
+
+        let servers = Services::new(opts, instance.clone(), plugins)
             .build()
             .await
-            .context(StartFrontendSnafu)?;
-        instance
-            .build_servers(servers)
-            .context(StartFrontendSnafu)?;
+            .context(error::StartFrontendSnafu)?;
 
-        Ok(Instance::new(instance, guard))
+        let frontend = Frontend {
+            instance,
+            servers,
+            heartbeat_task,
+            export_metrics_task,
+        };
+
+        Ok(Instance::new(frontend, guard))
     }
 }
 
@@ -440,7 +452,7 @@ mod tests {
 
             [http]
             addr = "127.0.0.1:4000"
-            timeout = "30s"
+            timeout = "0s"
             body_limit = "2GB"
 
             [opentsdb]
@@ -448,7 +460,7 @@ mod tests {
 
             [logging]
             level = "debug"
-            dir = "/tmp/greptimedb/test/logs"
+            dir = "./greptimedb_data/test/logs"
         "#;
         write!(file, "{}", toml_str).unwrap();
 
@@ -461,12 +473,15 @@ mod tests {
         let fe_opts = command.load_options(&Default::default()).unwrap().component;
 
         assert_eq!("127.0.0.1:4000".to_string(), fe_opts.http.addr);
-        assert_eq!(Duration::from_secs(30), fe_opts.http.timeout);
+        assert_eq!(Duration::from_secs(0), fe_opts.http.timeout);
 
         assert_eq!(ReadableSize::gb(2), fe_opts.http.body_limit);
 
         assert_eq!("debug", fe_opts.logging.level.as_ref().unwrap());
-        assert_eq!("/tmp/greptimedb/test/logs".to_string(), fe_opts.logging.dir);
+        assert_eq!(
+            "./greptimedb_data/test/logs".to_string(),
+            fe_opts.logging.dir
+        );
         assert!(!fe_opts.opentsdb.enable);
     }
 
@@ -505,7 +520,7 @@ mod tests {
 
         let options = cmd
             .load_options(&GlobalOptions {
-                log_dir: Some("/tmp/greptimedb/test/logs".to_string()),
+                log_dir: Some("./greptimedb_data/test/logs".to_string()),
                 log_level: Some("debug".to_string()),
 
                 #[cfg(feature = "tokio-console")]
@@ -515,7 +530,7 @@ mod tests {
             .component;
 
         let logging_opt = options.logging;
-        assert_eq!("/tmp/greptimedb/test/logs", logging_opt.dir);
+        assert_eq!("./greptimedb_data/test/logs", logging_opt.dir);
         assert_eq!("debug", logging_opt.level.as_ref().unwrap());
     }
 

src/cmd/src/metasrv.rs

@@ -337,7 +337,7 @@ mod tests {
 
             [logging]
             level = "debug"
-            dir = "/tmp/greptimedb/test/logs"
+            dir = "./greptimedb_data/test/logs"
 
             [failure_detector]
             threshold = 8.0
@@ -358,7 +358,10 @@ mod tests {
         assert_eq!(vec!["127.0.0.1:2379".to_string()], options.store_addrs);
         assert_eq!(SelectorType::LeaseBased, options.selector);
         assert_eq!("debug", options.logging.level.as_ref().unwrap());
-        assert_eq!("/tmp/greptimedb/test/logs".to_string(), options.logging.dir);
+        assert_eq!(
+            "./greptimedb_data/test/logs".to_string(),
+            options.logging.dir
+        );
         assert_eq!(8.0, options.failure_detector.threshold);
         assert_eq!(
             100.0,
@@ -396,7 +399,7 @@ mod tests {
 
         let options = cmd
             .load_options(&GlobalOptions {
-                log_dir: Some("/tmp/greptimedb/test/logs".to_string()),
+                log_dir: Some("./greptimedb_data/test/logs".to_string()),
                 log_level: Some("debug".to_string()),
 
                 #[cfg(feature = "tokio-console")]
@@ -406,7 +409,7 @@ mod tests {
             .component;
 
         let logging_opt = options.logging;
-        assert_eq!("/tmp/greptimedb/test/logs", logging_opt.dir);
+        assert_eq!("./greptimedb_data/test/logs", logging_opt.dir);
         assert_eq!("debug", logging_opt.level.as_ref().unwrap());
     }
 
@@ -424,7 +427,7 @@ mod tests {
 
             [logging]
             level = "debug"
-            dir = "/tmp/greptimedb/test/logs"
+            dir = "./greptimedb_data/test/logs"
         "#;
         write!(file, "{}", toml_str).unwrap();
 

src/cmd/src/standalone.rs

@@ -42,6 +42,7 @@ use common_meta::kv_backend::KvBackendRef;
 use common_meta::node_manager::NodeManagerRef;
 use common_meta::peer::Peer;
 use common_meta::region_keeper::MemoryRegionKeeper;
+use common_meta::region_registry::LeaderRegionRegistry;
 use common_meta::sequence::SequenceBuilder;
 use common_meta::wal_options_allocator::{build_wal_options_allocator, WalOptionsAllocatorRef};
 use common_procedure::{ProcedureInfo, ProcedureManagerRef};
@@ -54,13 +55,10 @@ use datanode::config::{DatanodeOptions, ProcedureConfig, RegionEngineConfig, Sto
 use datanode::datanode::{Datanode, DatanodeBuilder};
 use datanode::region_server::RegionServer;
 use file_engine::config::EngineConfig as FileEngineConfig;
-use flow::{
-    FlowConfig, FlowWorkerManager, FlownodeBuilder, FlownodeOptions, FrontendClient,
-    FrontendInvoker,
-};
-use frontend::frontend::FrontendOptions;
+use flow::{FlowConfig, FlowWorkerManager, FlownodeBuilder, FlownodeOptions, FrontendInvoker};
+use frontend::frontend::{Frontend, FrontendOptions};
 use frontend::instance::builder::FrontendBuilder;
-use frontend::instance::{FrontendInstance, Instance as FeInstance, StandaloneDatanodeManager};
+use frontend::instance::{Instance as FeInstance, StandaloneDatanodeManager};
 use frontend::server::Services;
 use frontend::service_config::{
     InfluxdbOptions, JaegerOptions, MysqlOptions, OpentsdbOptions, PostgresOptions,
@@ -70,7 +68,7 @@ use meta_srv::metasrv::{FLOW_ID_SEQ, TABLE_ID_SEQ};
 use mito2::config::MitoConfig;
 use query::stats::StatementStatistics;
 use serde::{Deserialize, Serialize};
-use servers::export_metrics::ExportMetricsOption;
+use servers::export_metrics::{ExportMetricsOption, ExportMetricsTask};
 use servers::grpc::GrpcOptions;
 use servers::http::HttpOptions;
 use servers::tls::{TlsMode, TlsOption};
@@ -79,15 +77,9 @@ use snafu::ResultExt;
 use tokio::sync::{broadcast, RwLock};
 use tracing_appender::non_blocking::WorkerGuard;
 
-use crate::error::{
-    BuildCacheRegistrySnafu, BuildWalOptionsAllocatorSnafu, CreateDirSnafu, IllegalConfigSnafu,
-    InitDdlManagerSnafu, InitMetadataSnafu, InitTimezoneSnafu, LoadLayeredConfigSnafu, OtherSnafu,
-    Result, ShutdownDatanodeSnafu, ShutdownFlownodeSnafu, ShutdownFrontendSnafu,
-    StartDatanodeSnafu, StartFlownodeSnafu, StartFrontendSnafu, StartProcedureManagerSnafu,
-    StartWalOptionsAllocatorSnafu, StopProcedureManagerSnafu,
-};
+use crate::error::Result;
 use crate::options::{GlobalOptions, GreptimeOptions};
-use crate::{log_versions, App};
+use crate::{error, log_versions, App};
 
 pub const APP_NAME: &str = "greptime-standalone";
 
@@ -135,7 +127,6 @@ impl SubCommand {
 #[derive(Clone, Debug, Serialize, Deserialize, PartialEq)]
 #[serde(default)]
 pub struct StandaloneOptions {
-    pub mode: Mode,
     pub enable_telemetry: bool,
     pub default_timezone: Option<String>,
     pub http: HttpOptions,
@@ -165,7 +156,6 @@ pub struct StandaloneOptions {
 impl Default for StandaloneOptions {
     fn default() -> Self {
         Self {
-            mode: Mode::Standalone,
             enable_telemetry: true,
             default_timezone: None,
             http: HttpOptions::default(),
@@ -246,7 +236,6 @@ impl StandaloneOptions {
             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()
         }
     }
@@ -254,13 +243,12 @@ impl StandaloneOptions {
 
 pub struct Instance {
     datanode: Datanode,
-    frontend: FeInstance,
+    frontend: Frontend,
     // TODO(discord9): wrapped it in flownode instance instead
     flow_worker_manager: Arc<FlowWorkerManager>,
     flow_shutdown: broadcast::Sender<()>,
     procedure_manager: ProcedureManagerRef,
     wal_options_allocator: WalOptionsAllocatorRef,
-
     // Keep the logging guard to prevent the worker from being dropped.
     _guard: Vec<WorkerGuard>,
 }
@@ -284,21 +272,26 @@ impl App for Instance {
         self.procedure_manager
             .start()
             .await
-            .context(StartProcedureManagerSnafu)?;
+            .context(error::StartProcedureManagerSnafu)?;
 
         self.wal_options_allocator
             .start()
             .await
-            .context(StartWalOptionsAllocatorSnafu)?;
+            .context(error::StartWalOptionsAllocatorSnafu)?;
 
-        plugins::start_frontend_plugins(self.frontend.plugins().clone())
+        plugins::start_frontend_plugins(self.frontend.instance.plugins().clone())
             .await
-            .context(StartFrontendSnafu)?;
+            .context(error::StartFrontendSnafu)?;
+
+        self.frontend
+            .start()
+            .await
+            .context(error::StartFrontendSnafu)?;
 
-        self.frontend.start().await.context(StartFrontendSnafu)?;
         self.flow_worker_manager
             .clone()
             .run_background(Some(self.flow_shutdown.subscribe()));
+
         Ok(())
     }
 
@@ -306,17 +299,18 @@ impl App for Instance {
         self.frontend
             .shutdown()
             .await
-            .context(ShutdownFrontendSnafu)?;
+            .context(error::ShutdownFrontendSnafu)?;
 
         self.procedure_manager
             .stop()
             .await
-            .context(StopProcedureManagerSnafu)?;
+            .context(error::StopProcedureManagerSnafu)?;
 
         self.datanode
             .shutdown()
             .await
-            .context(ShutdownDatanodeSnafu)?;
+            .context(error::ShutdownDatanodeSnafu)?;
+
         self.flow_shutdown
             .send(())
             .map_err(|_e| {
@@ -325,7 +319,8 @@ impl App for Instance {
                 }
                 .build()
             })
-            .context(ShutdownFlownodeSnafu)?;
+            .context(error::ShutdownFlownodeSnafu)?;
+
         info!("Datanode instance stopped.");
 
         Ok(())
@@ -371,7 +366,7 @@ impl StartCommand {
             self.config_file.as_deref(),
             self.env_prefix.as_ref(),
         )
-        .context(LoadLayeredConfigSnafu)?;
+        .context(error::LoadLayeredConfigSnafu)?;
 
         self.merge_with_cli_options(global_options, &mut opts.component)?;
 
@@ -384,9 +379,6 @@ impl StartCommand {
         global_options: &GlobalOptions,
         opts: &mut StandaloneOptions,
     ) -> Result<()> {
-        // Should always be standalone mode.
-        opts.mode = Mode::Standalone;
-
         if let Some(dir) = &global_options.log_dir {
             opts.logging.dir.clone_from(dir);
         }
@@ -418,7 +410,7 @@ impl StartCommand {
             // frontend grpc addr conflict with datanode default grpc addr
             let datanode_grpc_addr = DatanodeOptions::default().grpc.bind_addr;
             if addr.eq(&datanode_grpc_addr) {
-                return IllegalConfigSnafu {
+                return error::IllegalConfigSnafu {
                     msg: format!(
                         "gRPC listen address conflicts with datanode reserved gRPC addr: {datanode_grpc_addr}",
                     ),
@@ -477,18 +469,19 @@ impl StartCommand {
 
         plugins::setup_frontend_plugins(&mut plugins, &plugin_opts, &fe_opts)
             .await
-            .context(StartFrontendSnafu)?;
+            .context(error::StartFrontendSnafu)?;
 
         plugins::setup_datanode_plugins(&mut plugins, &plugin_opts, &dn_opts)
             .await
-            .context(StartDatanodeSnafu)?;
+            .context(error::StartDatanodeSnafu)?;
 
-        set_default_timezone(fe_opts.default_timezone.as_deref()).context(InitTimezoneSnafu)?;
+        set_default_timezone(fe_opts.default_timezone.as_deref())
+            .context(error::InitTimezoneSnafu)?;
 
         let data_home = &dn_opts.storage.data_home;
         // Ensure the data_home directory exists.
         fs::create_dir_all(path::Path::new(data_home))
-            .context(CreateDirSnafu { dir: data_home })?;
+            .context(error::CreateDirSnafu { dir: data_home })?;
 
         let metadata_dir = metadata_store_dir(data_home);
         let (kv_backend, procedure_manager) = FeInstance::try_build_standalone_components(
@@ -497,7 +490,7 @@ impl StartCommand {
             opts.procedure,
         )
         .await
-        .context(StartFrontendSnafu)?;
+        .context(error::StartFrontendSnafu)?;
 
         // Builds cache registry
         let layered_cache_builder = LayeredCacheRegistryBuilder::default();
@@ -506,16 +499,16 @@ impl StartCommand {
             with_default_composite_cache_registry(
                 layered_cache_builder.add_cache_registry(fundamental_cache_registry),
             )
-            .context(BuildCacheRegistrySnafu)?
+            .context(error::BuildCacheRegistrySnafu)?
             .build(),
         );
 
-        let datanode = DatanodeBuilder::new(dn_opts, plugins.clone())
+        let datanode = DatanodeBuilder::new(dn_opts, plugins.clone(), Mode::Standalone)
             .with_kv_backend(kv_backend.clone())
             .with_cache_registry(layered_cache_registry.clone())
             .build()
             .await
-            .context(StartDatanodeSnafu)?;
+            .context(error::StartDatanodeSnafu)?;
 
         let information_extension = Arc::new(StandaloneInformationExtension::new(
             datanode.region_server(),
@@ -536,23 +529,19 @@ impl StartCommand {
             flow: opts.flow.clone(),
             ..Default::default()
         };
-
-        let fe_server_addr = fe_opts.grpc.bind_addr.clone();
-        let frontend_client = FrontendClient::from_static_grpc_addr(fe_server_addr);
         let flow_builder = FlownodeBuilder::new(
             flownode_options,
             plugins.clone(),
             table_metadata_manager.clone(),
             catalog_manager.clone(),
             flow_metadata_manager.clone(),
-            Arc::new(frontend_client),
         );
         let flownode = Arc::new(
             flow_builder
                 .build()
                 .await
                 .map_err(BoxedError::new)
-                .context(OtherSnafu)?,
+                .context(error::OtherSnafu)?,
         );
 
         // set the ref to query for the local flow state
@@ -583,7 +572,7 @@ impl StartCommand {
         let kafka_options = opts.wal.clone().into();
         let wal_options_allocator = build_wal_options_allocator(&kafka_options, kv_backend.clone())
             .await
-            .context(BuildWalOptionsAllocatorSnafu)?;
+            .context(error::BuildWalOptionsAllocatorSnafu)?;
         let wal_options_allocator = Arc::new(wal_options_allocator);
         let table_meta_allocator = Arc::new(TableMetadataAllocator::new(
             table_id_sequence,
@@ -604,8 +593,8 @@ impl StartCommand {
         )
         .await?;
 
-        let mut frontend = FrontendBuilder::new(
-            fe_opts,
+        let fe_instance = FrontendBuilder::new(
+            fe_opts.clone(),
             kv_backend.clone(),
             layered_cache_registry.clone(),
             catalog_manager.clone(),
@@ -616,7 +605,8 @@ impl StartCommand {
         .with_plugin(plugins.clone())
         .try_build()
         .await
-        .context(StartFrontendSnafu)?;
+        .context(error::StartFrontendSnafu)?;
+        let fe_instance = Arc::new(fe_instance);
 
         let flow_worker_manager = flownode.flow_worker_manager();
         // flow server need to be able to use frontend to write insert requests back
@@ -629,18 +619,25 @@ impl StartCommand {
             node_manager,
         )
         .await
-        .context(StartFlownodeSnafu)?;
+        .context(error::StartFlownodeSnafu)?;
         flow_worker_manager.set_frontend_invoker(invoker).await;
 
         let (tx, _rx) = broadcast::channel(1);
 
-        let servers = Services::new(opts, Arc::new(frontend.clone()), plugins)
+        let export_metrics_task = ExportMetricsTask::try_new(&opts.export_metrics, Some(&plugins))
+            .context(error::ServersSnafu)?;
+
+        let servers = Services::new(opts, fe_instance.clone(), plugins)
             .build()
             .await
-            .context(StartFrontendSnafu)?;
-        frontend
-            .build_servers(servers)
-            .context(StartFrontendSnafu)?;
+            .context(error::StartFrontendSnafu)?;
+
+        let frontend = Frontend {
+            instance: fe_instance,
+            servers,
+            heartbeat_task: None,
+            export_metrics_task,
+        };
 
         Ok(Instance {
             datanode,
@@ -668,6 +665,7 @@ impl StartCommand {
                     node_manager,
                     cache_invalidator,
                     memory_region_keeper: Arc::new(MemoryRegionKeeper::default()),
+                    leader_region_registry: Arc::new(LeaderRegionRegistry::default()),
                     table_metadata_manager,
                     table_metadata_allocator,
                     flow_metadata_manager,
@@ -677,7 +675,7 @@ impl StartCommand {
                 procedure_manager,
                 true,
             )
-            .context(InitDdlManagerSnafu)?,
+            .context(error::InitDdlManagerSnafu)?,
         );
 
         Ok(procedure_executor)
@@ -691,7 +689,7 @@ impl StartCommand {
         table_metadata_manager
             .init()
             .await
-            .context(InitMetadataSnafu)?;
+            .context(error::InitMetadataSnafu)?;
 
         Ok(table_metadata_manager)
     }
@@ -785,6 +783,7 @@ impl InformationExtension for StandaloneInformationExtension {
                     manifest_size: region_stat.manifest_size,
                     sst_size: region_stat.sst_size,
                     index_size: region_stat.index_size,
+                    region_manifest: region_stat.manifest.into(),
                 }
             })
             .collect::<Vec<_>>();
@@ -859,7 +858,7 @@ mod tests {
 
             [wal]
             provider = "raft_engine"
-            dir = "/tmp/greptimedb/test/wal"
+            dir = "./greptimedb_data/test/wal"
             file_size = "1GB"
             purge_threshold = "50GB"
             purge_interval = "10m"
@@ -867,7 +866,7 @@ mod tests {
             sync_write = false
 
             [storage]
-            data_home = "/tmp/greptimedb/"
+            data_home = "./greptimedb_data/"
             type = "File"
 
             [[storage.providers]]
@@ -899,7 +898,7 @@ mod tests {
 
             [logging]
             level = "debug"
-            dir = "/tmp/greptimedb/test/logs"
+            dir = "./greptimedb_data/test/logs"
         "#;
         write!(file, "{}", toml_str).unwrap();
         let cmd = StartCommand {
@@ -929,7 +928,10 @@ mod tests {
         let DatanodeWalConfig::RaftEngine(raft_engine_config) = dn_opts.wal else {
             unreachable!()
         };
-        assert_eq!("/tmp/greptimedb/test/wal", raft_engine_config.dir.unwrap());
+        assert_eq!(
+            "./greptimedb_data/test/wal",
+            raft_engine_config.dir.unwrap()
+        );
 
         assert!(matches!(
             &dn_opts.storage.store,
@@ -953,7 +955,7 @@ mod tests {
         }
 
         assert_eq!("debug", logging_opts.level.as_ref().unwrap());
-        assert_eq!("/tmp/greptimedb/test/logs".to_string(), logging_opts.dir);
+        assert_eq!("./greptimedb_data/test/logs".to_string(), logging_opts.dir);
     }
 
     #[test]
@@ -965,7 +967,7 @@ mod tests {
 
         let opts = cmd
             .load_options(&GlobalOptions {
-                log_dir: Some("/tmp/greptimedb/test/logs".to_string()),
+                log_dir: Some("./greptimedb_data/test/logs".to_string()),
                 log_level: Some("debug".to_string()),
 
                 #[cfg(feature = "tokio-console")]
@@ -974,7 +976,7 @@ mod tests {
             .unwrap()
             .component;
 
-        assert_eq!("/tmp/greptimedb/test/logs", opts.logging.dir);
+        assert_eq!("./greptimedb_data/test/logs", opts.logging.dir);
         assert_eq!("debug", opts.logging.level.unwrap());
     }
 
@@ -1058,7 +1060,6 @@ mod tests {
         let options =
             StandaloneOptions::load_layered_options(None, "GREPTIMEDB_STANDALONE").unwrap();
         let default_options = StandaloneOptions::default();
-        assert_eq!(options.mode, default_options.mode);
         assert_eq!(options.enable_telemetry, default_options.enable_telemetry);
         assert_eq!(options.http, default_options.http);
         assert_eq!(options.grpc, default_options.grpc);

src/cmd/tests/load_config_test.rs

@@ -56,13 +56,13 @@ fn test_load_datanode_example_config() {
                 metadata_cache_tti: Duration::from_secs(300),
             }),
             wal: DatanodeWalConfig::RaftEngine(RaftEngineConfig {
-                dir: Some("/tmp/greptimedb/wal".to_string()),
+                dir: Some("./greptimedb_data/wal".to_string()),
                 sync_period: Some(Duration::from_secs(10)),
                 recovery_parallelism: 2,
                 ..Default::default()
             }),
             storage: StorageConfig {
-                data_home: "/tmp/greptimedb/".to_string(),
+                data_home: "./greptimedb_data/".to_string(),
                 ..Default::default()
             },
             region_engine: vec![
@@ -159,17 +159,17 @@ fn test_load_metasrv_example_config() {
     let expected = GreptimeOptions::<MetasrvOptions> {
         component: MetasrvOptions {
             selector: SelectorType::default(),
-            data_home: "/tmp/metasrv/".to_string(),
+            data_home: "./greptimedb_data/metasrv/".to_string(),
             server_addr: "127.0.0.1:3002".to_string(),
             logging: LoggingOptions {
-                dir: "/tmp/greptimedb/logs".to_string(),
+                dir: "./greptimedb_data/logs".to_string(),
                 level: Some("info".to_string()),
                 otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                 tracing_sample_ratio: Some(Default::default()),
                 slow_query: SlowQueryOptions {
                     enable: false,
-                    threshold: Some(Duration::from_secs(10)),
-                    sample_ratio: Some(1.0),
+                    threshold: None,
+                    sample_ratio: None,
                 },
                 ..Default::default()
             },
@@ -202,7 +202,7 @@ fn test_load_standalone_example_config() {
         component: StandaloneOptions {
             default_timezone: Some("UTC".to_string()),
             wal: DatanodeWalConfig::RaftEngine(RaftEngineConfig {
-                dir: Some("/tmp/greptimedb/wal".to_string()),
+                dir: Some("./greptimedb_data/wal".to_string()),
                 sync_period: Some(Duration::from_secs(10)),
                 recovery_parallelism: 2,
                 ..Default::default()
@@ -219,7 +219,7 @@ fn test_load_standalone_example_config() {
                 }),
             ],
             storage: StorageConfig {
-                data_home: "/tmp/greptimedb/".to_string(),
+                data_home: "./greptimedb_data/".to_string(),
                 ..Default::default()
             },
             logging: LoggingOptions {

src/common/catalog/src/consts.rs

@@ -135,5 +135,14 @@ pub fn is_readonly_schema(schema: &str) -> bool {
 pub const TRACE_ID_COLUMN: &str = "trace_id";
 pub const SPAN_ID_COLUMN: &str = "span_id";
 pub const SPAN_NAME_COLUMN: &str = "span_name";
+pub const SERVICE_NAME_COLUMN: &str = "service_name";
 pub const PARENT_SPAN_ID_COLUMN: &str = "parent_span_id";
+pub const TRACE_TABLE_NAME: &str = "opentelemetry_traces";
+pub const TRACE_TABLE_NAME_SESSION_KEY: &str = "trace_table_name";
+// ---- End of special table and fields ----
+
+/// Generate the trace services table name from the trace table name by adding `_services` suffix.
+pub fn trace_services_table_name(trace_table_name: &str) -> String {
+    format!("{}_services", trace_table_name)
+}
 // ---- End of special table and fields ----

src/common/config/src/config.rs

@@ -161,7 +161,7 @@ mod tests {
 
             [wal]
             provider = "raft_engine"
-            dir = "/tmp/greptimedb/wal"
+            dir = "./greptimedb_data/wal"
             file_size = "1GB"
             purge_threshold = "50GB"
             purge_interval = "10m"
@@ -170,7 +170,7 @@ mod tests {
 
             [logging]
             level = "debug"
-            dir = "/tmp/greptimedb/test/logs"
+            dir = "./greptimedb_data/test/logs"
         "#;
         write!(file, "{}", toml_str).unwrap();
 
@@ -246,7 +246,7 @@ mod tests {
                 let DatanodeWalConfig::RaftEngine(raft_engine_config) = opts.wal else {
                     unreachable!()
                 };
-                assert_eq!(raft_engine_config.dir.unwrap(), "/tmp/greptimedb/wal");
+                assert_eq!(raft_engine_config.dir.unwrap(), "./greptimedb_data/wal");
 
                 // Should be default values.
                 assert_eq!(opts.node_id, None);

src/common/function/Cargo.toml

@@ -17,6 +17,7 @@ api.workspace = true
 arc-swap = "1.0"
 async-trait.workspace = true
 bincode = "1.3"
+chrono.workspace = true
 common-base.workspace = true
 common-catalog.workspace = true
 common-error.workspace = true
@@ -38,6 +39,7 @@ geohash = { version = "0.13", optional = true }
 h3o = { version = "0.6", optional = true }
 hyperloglogplus = "0.4"
 jsonb.workspace = true
+memchr = "2.7"
 nalgebra.workspace = true
 num = "0.4"
 num-traits = "0.2"

src/common/function/src/admin.rs

@@ -12,15 +12,19 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+mod add_region_follower;
 mod flush_compact_region;
 mod flush_compact_table;
 mod migrate_region;
+mod remove_region_follower;
 
 use std::sync::Arc;
 
+use add_region_follower::AddRegionFollowerFunction;
 use flush_compact_region::{CompactRegionFunction, FlushRegionFunction};
 use flush_compact_table::{CompactTableFunction, FlushTableFunction};
 use migrate_region::MigrateRegionFunction;
+use remove_region_follower::RemoveRegionFollowerFunction;
 
 use crate::flush_flow::FlushFlowFunction;
 use crate::function_registry::FunctionRegistry;
@@ -32,6 +36,8 @@ impl AdminFunction {
     /// Register all table functions to [`FunctionRegistry`].
     pub fn register(registry: &FunctionRegistry) {
         registry.register_async(Arc::new(MigrateRegionFunction));
+        registry.register_async(Arc::new(AddRegionFollowerFunction));
+        registry.register_async(Arc::new(RemoveRegionFollowerFunction));
         registry.register_async(Arc::new(FlushRegionFunction));
         registry.register_async(Arc::new(CompactRegionFunction));
         registry.register_async(Arc::new(FlushTableFunction));

src/common/function/src/admin/add_region_follower.rs

@@ -0,0 +1,129 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use common_macro::admin_fn;
+use common_meta::rpc::procedure::AddRegionFollowerRequest;
+use common_query::error::{
+    InvalidFuncArgsSnafu, MissingProcedureServiceHandlerSnafu, Result,
+    UnsupportedInputDataTypeSnafu,
+};
+use common_query::prelude::{Signature, TypeSignature, Volatility};
+use datatypes::prelude::ConcreteDataType;
+use datatypes::value::{Value, ValueRef};
+use session::context::QueryContextRef;
+use snafu::ensure;
+
+use crate::handlers::ProcedureServiceHandlerRef;
+use crate::helper::cast_u64;
+
+/// A function to add a follower to a region.
+/// Only available in cluster mode.
+///
+/// - `add_region_follower(region_id, peer_id)`.
+///
+/// The parameters:
+/// - `region_id`:  the region id
+/// - `peer_id`:  the peer id
+#[admin_fn(
+    name = AddRegionFollowerFunction,
+    display_name = add_region_follower,
+    sig_fn = signature,
+    ret = uint64
+)]
+pub(crate) async fn add_region_follower(
+    procedure_service_handler: &ProcedureServiceHandlerRef,
+    _ctx: &QueryContextRef,
+    params: &[ValueRef<'_>],
+) -> Result<Value> {
+    ensure!(
+        params.len() == 2,
+        InvalidFuncArgsSnafu {
+            err_msg: format!(
+                "The length of the args is not correct, expect exactly 2, have: {}",
+                params.len()
+            ),
+        }
+    );
+
+    let Some(region_id) = cast_u64(&params[0])? else {
+        return UnsupportedInputDataTypeSnafu {
+            function: "add_region_follower",
+            datatypes: params.iter().map(|v| v.data_type()).collect::<Vec<_>>(),
+        }
+        .fail();
+    };
+    let Some(peer_id) = cast_u64(&params[1])? else {
+        return UnsupportedInputDataTypeSnafu {
+            function: "add_region_follower",
+            datatypes: params.iter().map(|v| v.data_type()).collect::<Vec<_>>(),
+        }
+        .fail();
+    };
+
+    procedure_service_handler
+        .add_region_follower(AddRegionFollowerRequest { region_id, peer_id })
+        .await?;
+
+    Ok(Value::from(0u64))
+}
+
+fn signature() -> Signature {
+    Signature::one_of(
+        vec![
+            // add_region_follower(region_id, peer)
+            TypeSignature::Uniform(2, ConcreteDataType::numerics()),
+        ],
+        Volatility::Immutable,
+    )
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::Arc;
+
+    use common_query::prelude::TypeSignature;
+    use datatypes::vectors::{UInt64Vector, VectorRef};
+
+    use super::*;
+    use crate::function::{AsyncFunction, FunctionContext};
+
+    #[test]
+    fn test_add_region_follower_misc() {
+        let f = AddRegionFollowerFunction;
+        assert_eq!("add_region_follower", f.name());
+        assert_eq!(
+            ConcreteDataType::uint64_datatype(),
+            f.return_type(&[]).unwrap()
+        );
+        assert!(matches!(f.signature(),
+                         Signature {
+                             type_signature: TypeSignature::OneOf(sigs),
+                             volatility: Volatility::Immutable
+                         } if sigs.len() == 1));
+    }
+
+    #[tokio::test]
+    async fn test_add_region_follower() {
+        let f = AddRegionFollowerFunction;
+        let args = vec![1, 1];
+        let args = args
+            .into_iter()
+            .map(|arg| Arc::new(UInt64Vector::from_slice([arg])) as _)
+            .collect::<Vec<_>>();
+
+        let result = f.eval(FunctionContext::mock(), &args).await.unwrap();
+        let expect: VectorRef = Arc::new(UInt64Vector::from_slice([0u64]));
+        assert_eq!(result, expect);
+    }
+}

src/common/function/src/admin/remove_region_follower.rs

@@ -0,0 +1,129 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use common_macro::admin_fn;
+use common_meta::rpc::procedure::RemoveRegionFollowerRequest;
+use common_query::error::{
+    InvalidFuncArgsSnafu, MissingProcedureServiceHandlerSnafu, Result,
+    UnsupportedInputDataTypeSnafu,
+};
+use common_query::prelude::{Signature, TypeSignature, Volatility};
+use datatypes::prelude::ConcreteDataType;
+use datatypes::value::{Value, ValueRef};
+use session::context::QueryContextRef;
+use snafu::ensure;
+
+use crate::handlers::ProcedureServiceHandlerRef;
+use crate::helper::cast_u64;
+
+/// A function to remove a follower from a region.
+//// Only available in cluster mode.
+///
+/// - `remove_region_follower(region_id, peer_id)`.
+///
+/// The parameters:
+/// - `region_id`:  the region id
+/// - `peer_id`:  the peer id
+#[admin_fn(
+    name = RemoveRegionFollowerFunction,
+    display_name = remove_region_follower,
+    sig_fn = signature,
+    ret = uint64
+)]
+pub(crate) async fn remove_region_follower(
+    procedure_service_handler: &ProcedureServiceHandlerRef,
+    _ctx: &QueryContextRef,
+    params: &[ValueRef<'_>],
+) -> Result<Value> {
+    ensure!(
+        params.len() == 2,
+        InvalidFuncArgsSnafu {
+            err_msg: format!(
+                "The length of the args is not correct, expect exactly 2, have: {}",
+                params.len()
+            ),
+        }
+    );
+
+    let Some(region_id) = cast_u64(&params[0])? else {
+        return UnsupportedInputDataTypeSnafu {
+            function: "add_region_follower",
+            datatypes: params.iter().map(|v| v.data_type()).collect::<Vec<_>>(),
+        }
+        .fail();
+    };
+    let Some(peer_id) = cast_u64(&params[1])? else {
+        return UnsupportedInputDataTypeSnafu {
+            function: "add_region_follower",
+            datatypes: params.iter().map(|v| v.data_type()).collect::<Vec<_>>(),
+        }
+        .fail();
+    };
+
+    procedure_service_handler
+        .remove_region_follower(RemoveRegionFollowerRequest { region_id, peer_id })
+        .await?;
+
+    Ok(Value::from(0u64))
+}
+
+fn signature() -> Signature {
+    Signature::one_of(
+        vec![
+            // remove_region_follower(region_id, peer_id)
+            TypeSignature::Uniform(2, ConcreteDataType::numerics()),
+        ],
+        Volatility::Immutable,
+    )
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::Arc;
+
+    use common_query::prelude::TypeSignature;
+    use datatypes::vectors::{UInt64Vector, VectorRef};
+
+    use super::*;
+    use crate::function::{AsyncFunction, FunctionContext};
+
+    #[test]
+    fn test_remove_region_follower_misc() {
+        let f = RemoveRegionFollowerFunction;
+        assert_eq!("remove_region_follower", f.name());
+        assert_eq!(
+            ConcreteDataType::uint64_datatype(),
+            f.return_type(&[]).unwrap()
+        );
+        assert!(matches!(f.signature(),
+                         Signature {
+                             type_signature: TypeSignature::OneOf(sigs),
+                             volatility: Volatility::Immutable
+                         } if sigs.len() == 1));
+    }
+
+    #[tokio::test]
+    async fn test_remove_region_follower() {
+        let f = RemoveRegionFollowerFunction;
+        let args = vec![1, 1];
+        let args = args
+            .into_iter()
+            .map(|arg| Arc::new(UInt64Vector::from_slice([arg])) as _)
+            .collect::<Vec<_>>();
+
+        let result = f.eval(FunctionContext::mock(), &args).await.unwrap();
+        let expect: VectorRef = Arc::new(UInt64Vector::from_slice([0u64]));
+        assert_eq!(result, expect);
+    }
+}

src/common/function/src/function_registry.rs

@@ -27,6 +27,7 @@ use crate::scalars::hll_count::HllCalcFunction;
 use crate::scalars::ip::IpFunctions;
 use crate::scalars::json::JsonFunction;
 use crate::scalars::matches::MatchesFunction;
+use crate::scalars::matches_term::MatchesTermFunction;
 use crate::scalars::math::MathFunction;
 use crate::scalars::timestamp::TimestampFunction;
 use crate::scalars::uddsketch_calc::UddSketchCalcFunction;
@@ -116,6 +117,7 @@ pub static FUNCTION_REGISTRY: Lazy<Arc<FunctionRegistry>> = Lazy::new(|| {
 
     // Full text search function
     MatchesFunction::register(&function_registry);
+    MatchesTermFunction::register(&function_registry);
 
     // System and administration functions
     SystemFunction::register(&function_registry);

src/common/function/src/handlers.rs

@@ -16,7 +16,10 @@ use std::sync::Arc;
 
 use async_trait::async_trait;
 use common_base::AffectedRows;
-use common_meta::rpc::procedure::{MigrateRegionRequest, ProcedureStateResponse};
+use common_meta::rpc::procedure::{
+    AddRegionFollowerRequest, MigrateRegionRequest, ProcedureStateResponse,
+    RemoveRegionFollowerRequest,
+};
 use common_query::error::Result;
 use common_query::Output;
 use session::context::QueryContextRef;
@@ -63,6 +66,12 @@ pub trait ProcedureServiceHandler: Send + Sync {
 
     /// Query the procedure' state by its id
     async fn query_procedure_state(&self, pid: &str) -> Result<ProcedureStateResponse>;
+
+    /// Add a region follower to a region.
+    async fn add_region_follower(&self, request: AddRegionFollowerRequest) -> Result<()>;
+
+    /// Remove a region follower from a region.
+    async fn remove_region_follower(&self, request: RemoveRegionFollowerRequest) -> Result<()>;
 }
 
 /// This flow service handler is only use for flush flow for now.

src/common/function/src/scalars.rs

@@ -19,6 +19,7 @@ pub mod expression;
 pub mod geo;
 pub mod json;
 pub mod matches;
+pub mod matches_term;
 pub mod math;
 pub mod vector;
 

src/common/function/src/scalars/date/date_format.rs

@@ -43,7 +43,6 @@ impl Function for DateFormatFunction {
         helper::one_of_sigs2(
             vec![
                 ConcreteDataType::date_datatype(),
-                ConcreteDataType::datetime_datatype(),
                 ConcreteDataType::timestamp_second_datatype(),
                 ConcreteDataType::timestamp_millisecond_datatype(),
                 ConcreteDataType::timestamp_microsecond_datatype(),
@@ -105,22 +104,6 @@ impl Function for DateFormatFunction {
                     results.push(result.as_deref());
                 }
             }
-            ConcreteDataType::DateTime(_) => {
-                for i in 0..size {
-                    let datetime = left.get(i).as_datetime();
-                    let format = formats.get(i).as_string();
-
-                    let result = match (datetime, format) {
-                        (Some(datetime), Some(fmt)) => datetime
-                            .as_formatted_string(&fmt, Some(&func_ctx.query_ctx.timezone()))
-                            .map_err(BoxedError::new)
-                            .context(error::ExecuteSnafu)?,
-                        _ => None,
-                    };
-
-                    results.push(result.as_deref());
-                }
-            }
             _ => {
                 return UnsupportedInputDataTypeSnafu {
                     function: NAME,
@@ -147,7 +130,7 @@ mod tests {
     use common_query::prelude::{TypeSignature, Volatility};
     use datatypes::prelude::{ConcreteDataType, ScalarVector};
     use datatypes::value::Value;
-    use datatypes::vectors::{DateTimeVector, DateVector, StringVector, TimestampSecondVector};
+    use datatypes::vectors::{DateVector, StringVector, TimestampSecondVector};
 
     use super::{DateFormatFunction, *};
 
@@ -169,16 +152,11 @@ mod tests {
             ConcreteDataType::string_datatype(),
             f.return_type(&[ConcreteDataType::date_datatype()]).unwrap()
         );
-        assert_eq!(
-            ConcreteDataType::string_datatype(),
-            f.return_type(&[ConcreteDataType::datetime_datatype()])
-                .unwrap()
-        );
         assert!(matches!(f.signature(),
                          Signature {
                              type_signature: TypeSignature::OneOf(sigs),
                              volatility: Volatility::Immutable
-                         } if  sigs.len() == 6));
+                         } if  sigs.len() == 5));
     }
 
     #[test]
@@ -262,45 +240,4 @@ mod tests {
             }
         }
     }
-
-    #[test]
-    fn test_datetime_date_format() {
-        let f = DateFormatFunction;
-
-        let dates = vec![Some(123), None, Some(42), None];
-        let formats = vec![
-            "%Y-%m-%d %T.%3f",
-            "%Y-%m-%d %T.%3f",
-            "%Y-%m-%d %T.%3f",
-            "%Y-%m-%d %T.%3f",
-        ];
-        let results = [
-            Some("1970-01-01 00:00:00.123"),
-            None,
-            Some("1970-01-01 00:00:00.042"),
-            None,
-        ];
-
-        let date_vector = DateTimeVector::from(dates.clone());
-        let interval_vector = StringVector::from_vec(formats);
-        let args: Vec<VectorRef> = vec![Arc::new(date_vector), Arc::new(interval_vector)];
-        let vector = f.eval(&FunctionContext::default(), &args).unwrap();
-
-        assert_eq!(4, vector.len());
-        for (i, _t) in dates.iter().enumerate() {
-            let v = vector.get(i);
-            let result = results.get(i).unwrap();
-
-            if result.is_none() {
-                assert_eq!(Value::Null, v);
-                continue;
-            }
-            match v {
-                Value::String(s) => {
-                    assert_eq!(s.as_utf8(), result.unwrap());
-                }
-                _ => unreachable!(),
-            }
-        }
-    }
 }

src/common/function/src/scalars/date/date_sub.rs

@@ -118,11 +118,6 @@ mod tests {
             ConcreteDataType::date_datatype(),
             f.return_type(&[ConcreteDataType::date_datatype()]).unwrap()
         );
-        assert_eq!(
-            ConcreteDataType::datetime_datatype(),
-            f.return_type(&[ConcreteDataType::datetime_datatype()])
-                .unwrap()
-        );
         assert!(
             matches!(f.signature(),
                          Signature {

src/common/function/src/scalars/matches_term.rs

@@ -0,0 +1,375 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::sync::Arc;
+use std::{fmt, iter};
+
+use common_query::error::{InvalidFuncArgsSnafu, Result};
+use common_query::prelude::Volatility;
+use datatypes::prelude::ConcreteDataType;
+use datatypes::scalars::ScalarVectorBuilder;
+use datatypes::vectors::{BooleanVector, BooleanVectorBuilder, MutableVector, VectorRef};
+use memchr::memmem;
+use snafu::ensure;
+
+use crate::function::{Function, FunctionContext};
+use crate::function_registry::FunctionRegistry;
+
+/// Exact term/phrase matching function for text columns.
+///
+/// This function checks if a text column contains exact term/phrase matches
+/// with non-alphanumeric boundaries. Designed for:
+/// - Whole-word matching (e.g. "cat" in "cat!" but not in "category")
+/// - Phrase matching (e.g. "hello world" in "note:hello world!")
+///
+/// # Signature
+/// `matches_term(text: String, term: String) -> Boolean`
+///
+/// # Arguments
+/// * `text` - String column to search
+/// * `term` - Search term/phrase
+///
+/// # Returns
+/// BooleanVector where each element indicates if the corresponding text
+/// contains an exact match of the term, following these rules:
+/// 1. Exact substring match found (case-sensitive)
+/// 2. Match boundaries are either:
+///    - Start/end of text
+///    - Any non-alphanumeric character (including spaces, hyphens, punctuation, etc.)
+///
+/// # Examples
+/// ```
+/// -- SQL examples --
+/// -- Match phrase with space --
+/// SELECT matches_term(column, 'hello world') FROM table;
+/// -- Text: "warning:hello world!" => true
+/// -- Text: "hello-world"          => false (hyphen instead of space)
+/// -- Text: "hello world2023"      => false (ending with numbers)
+///
+/// -- Match multiple words with boundaries --
+/// SELECT matches_term(column, 'critical error') FROM logs;
+/// -- Match in: "ERROR:critical error!"
+/// -- No match: "critical_errors"
+///
+/// -- Empty string handling --
+/// SELECT matches_term(column, '') FROM table;
+/// -- Text: "" => true
+/// -- Text: "any" => false
+///
+/// -- Case sensitivity --
+/// SELECT matches_term(column, 'Cat') FROM table;
+/// -- Text: "Cat" => true
+/// -- Text: "cat" => false
+/// ```
+pub struct MatchesTermFunction;
+
+impl MatchesTermFunction {
+    pub fn register(registry: &FunctionRegistry) {
+        registry.register(Arc::new(MatchesTermFunction));
+    }
+}
+
+impl fmt::Display for MatchesTermFunction {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "MATCHES_TERM")
+    }
+}
+
+impl Function for MatchesTermFunction {
+    fn name(&self) -> &str {
+        "matches_term"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::boolean_datatype())
+    }
+
+    fn signature(&self) -> common_query::prelude::Signature {
+        common_query::prelude::Signature::exact(
+            vec![
+                ConcreteDataType::string_datatype(),
+                ConcreteDataType::string_datatype(),
+            ],
+            Volatility::Immutable,
+        )
+    }
+
+    fn eval(&self, _func_ctx: &FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure!(
+            columns.len() == 2,
+            InvalidFuncArgsSnafu {
+                err_msg: format!(
+                    "The length of the args is not correct, expect exactly 2, have: {}",
+                    columns.len()
+                ),
+            }
+        );
+
+        let text_column = &columns[0];
+        if text_column.is_empty() {
+            return Ok(Arc::new(BooleanVector::from(Vec::<bool>::with_capacity(0))));
+        }
+
+        let term_column = &columns[1];
+        let compiled_finder = if term_column.is_const() {
+            let term = term_column.get_ref(0).as_string().unwrap();
+            match term {
+                None => {
+                    return Ok(Arc::new(BooleanVector::from_iter(
+                        iter::repeat(None).take(text_column.len()),
+                    )));
+                }
+                Some(term) => Some(MatchesTermFinder::new(term)),
+            }
+        } else {
+            None
+        };
+
+        let len = text_column.len();
+        let mut result = BooleanVectorBuilder::with_capacity(len);
+        for i in 0..len {
+            let text = text_column.get_ref(i).as_string().unwrap();
+            let Some(text) = text else {
+                result.push_null();
+                continue;
+            };
+
+            let contains = match &compiled_finder {
+                Some(finder) => finder.find(text),
+                None => {
+                    let term = match term_column.get_ref(i).as_string().unwrap() {
+                        None => {
+                            result.push_null();
+                            continue;
+                        }
+                        Some(term) => term,
+                    };
+                    MatchesTermFinder::new(term).find(text)
+                }
+            };
+            result.push(Some(contains));
+        }
+
+        Ok(result.to_vector())
+    }
+}
+
+/// A compiled finder for `matches_term` function that holds the compiled term
+/// and its metadata for efficient matching.
+///
+/// A term is considered matched when:
+/// 1. The exact sequence appears in the text
+/// 2. It is either:
+///    - At the start/end of text with adjacent non-alphanumeric character
+///    - Surrounded by non-alphanumeric characters
+///
+/// # Examples
+/// ```
+/// let finder = MatchesTermFinder::new("cat");
+/// assert!(finder.find("cat!"));      // Term at end with punctuation
+/// assert!(finder.find("dog,cat"));   // Term preceded by comma
+/// assert!(!finder.find("category")); // Partial match rejected
+///
+/// let finder = MatchesTermFinder::new("world");
+/// assert!(finder.find("hello-world")); // Hyphen boundary
+/// ```
+#[derive(Clone, Debug)]
+pub struct MatchesTermFinder {
+    finder: memmem::Finder<'static>,
+    term: String,
+    starts_with_non_alnum: bool,
+    ends_with_non_alnum: bool,
+}
+
+impl MatchesTermFinder {
+    /// Create a new `MatchesTermFinder` for the given term.
+    pub fn new(term: &str) -> Self {
+        let starts_with_non_alnum = term.chars().next().is_some_and(|c| !c.is_alphanumeric());
+        let ends_with_non_alnum = term.chars().last().is_some_and(|c| !c.is_alphanumeric());
+
+        Self {
+            finder: memmem::Finder::new(term).into_owned(),
+            term: term.to_string(),
+            starts_with_non_alnum,
+            ends_with_non_alnum,
+        }
+    }
+
+    /// Find the term in the text.
+    pub fn find(&self, text: &str) -> bool {
+        if self.term.is_empty() {
+            return text.is_empty();
+        }
+
+        if text.len() < self.term.len() {
+            return false;
+        }
+
+        let mut pos = 0;
+        while let Some(found_pos) = self.finder.find(text[pos..].as_bytes()) {
+            let actual_pos = pos + found_pos;
+
+            let prev_ok = self.starts_with_non_alnum
+                || text[..actual_pos]
+                    .chars()
+                    .last()
+                    .map(|c| !c.is_alphanumeric())
+                    .unwrap_or(true);
+
+            if prev_ok {
+                let next_pos = actual_pos + self.finder.needle().len();
+                let next_ok = self.ends_with_non_alnum
+                    || text[next_pos..]
+                        .chars()
+                        .next()
+                        .map(|c| !c.is_alphanumeric())
+                        .unwrap_or(true);
+
+                if next_ok {
+                    return true;
+                }
+            }
+
+            if let Some(next_char) = text[actual_pos..].chars().next() {
+                pos = actual_pos + next_char.len_utf8();
+            } else {
+                break;
+            }
+        }
+
+        false
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn matches_term_example() {
+        let finder = MatchesTermFinder::new("hello world");
+        assert!(finder.find("warning:hello world!"));
+        assert!(!finder.find("hello-world"));
+        assert!(!finder.find("hello world2023"));
+
+        let finder = MatchesTermFinder::new("critical error");
+        assert!(finder.find("ERROR:critical error!"));
+        assert!(!finder.find("critical_errors"));
+
+        let finder = MatchesTermFinder::new("");
+        assert!(finder.find(""));
+        assert!(!finder.find("any"));
+
+        let finder = MatchesTermFinder::new("Cat");
+        assert!(finder.find("Cat"));
+        assert!(!finder.find("cat"));
+    }
+
+    #[test]
+    fn matches_term_with_punctuation() {
+        assert!(MatchesTermFinder::new("cat").find("cat!"));
+        assert!(MatchesTermFinder::new("dog").find("!dog"));
+    }
+
+    #[test]
+    fn matches_phrase_with_boundaries() {
+        assert!(MatchesTermFinder::new("hello-world").find("hello-world"));
+        assert!(MatchesTermFinder::new("'foo bar'").find("test: 'foo bar'"));
+    }
+
+    #[test]
+    fn matches_at_text_boundaries() {
+        assert!(MatchesTermFinder::new("start").find("start..."));
+        assert!(MatchesTermFinder::new("end").find("...end"));
+    }
+
+    // Negative cases
+    #[test]
+    fn rejects_partial_matches() {
+        assert!(!MatchesTermFinder::new("cat").find("category"));
+        assert!(!MatchesTermFinder::new("boot").find("rebooted"));
+    }
+
+    #[test]
+    fn rejects_missing_term() {
+        assert!(!MatchesTermFinder::new("foo").find("hello world"));
+    }
+
+    // Edge cases
+    #[test]
+    fn handles_empty_inputs() {
+        assert!(!MatchesTermFinder::new("test").find(""));
+        assert!(!MatchesTermFinder::new("").find("text"));
+    }
+
+    #[test]
+    fn different_unicode_boundaries() {
+        assert!(MatchesTermFinder::new("café").find("café>"));
+        assert!(!MatchesTermFinder::new("café").find("口café>"));
+        assert!(!MatchesTermFinder::new("café").find("café口"));
+        assert!(!MatchesTermFinder::new("café").find("cafémore"));
+        assert!(MatchesTermFinder::new("русский").find("русский!"));
+        assert!(MatchesTermFinder::new("русский").find("русский！"));
+    }
+
+    #[test]
+    fn case_sensitive_matching() {
+        assert!(!MatchesTermFinder::new("cat").find("Cat"));
+        assert!(MatchesTermFinder::new("CaT").find("CaT"));
+    }
+
+    #[test]
+    fn numbers_in_term() {
+        assert!(MatchesTermFinder::new("v1.0").find("v1.0!"));
+        assert!(!MatchesTermFinder::new("v1.0").find("v1.0a"));
+    }
+
+    #[test]
+    fn adjacent_alphanumeric_fails() {
+        assert!(!MatchesTermFinder::new("cat").find("cat5"));
+        assert!(!MatchesTermFinder::new("dog").find("dogcat"));
+    }
+
+    #[test]
+    fn empty_term_text() {
+        assert!(!MatchesTermFinder::new("").find("text"));
+        assert!(MatchesTermFinder::new("").find(""));
+        assert!(!MatchesTermFinder::new("text").find(""));
+    }
+
+    #[test]
+    fn leading_non_alphanumeric() {
+        assert!(MatchesTermFinder::new("/cat").find("dog/cat"));
+        assert!(MatchesTermFinder::new("dog/").find("dog/cat"));
+        assert!(MatchesTermFinder::new("dog/cat").find("dog/cat"));
+    }
+
+    #[test]
+    fn continues_searching_after_boundary_mismatch() {
+        assert!(!MatchesTermFinder::new("log").find("bloglog!"));
+        assert!(MatchesTermFinder::new("log").find("bloglog log"));
+        assert!(MatchesTermFinder::new("log").find("alogblog_log!"));
+
+        assert!(MatchesTermFinder::new("error").find("errorlog_error_case"));
+        assert!(MatchesTermFinder::new("test").find("atestbtestc_test_end"));
+        assert!(MatchesTermFinder::new("data").find("database_data_store"));
+        assert!(!MatchesTermFinder::new("data").find("database_datastore"));
+        assert!(MatchesTermFinder::new("log.txt").find("catalog.txt_log.txt!"));
+        assert!(!MatchesTermFinder::new("log.txt").find("catalog.txtlog.txt!"));
+        assert!(MatchesTermFinder::new("data-set").find("bigdata-set_data-set!"));
+
+        assert!(MatchesTermFinder::new("中文").find("这是中文测试，中文！"));
+        assert!(MatchesTermFinder::new("error").find("错误errorerror日志_error!"));
+    }
+}

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

@@ -23,7 +23,7 @@ use datatypes::arrow::array::AsArray;
 use datatypes::arrow::compute::cast;
 use datatypes::arrow::compute::kernels::zip;
 use datatypes::arrow::datatypes::{
-    DataType as ArrowDataType, Date32Type, Date64Type, TimestampMicrosecondType,
+    DataType as ArrowDataType, Date32Type, TimeUnit, TimestampMicrosecondType,
     TimestampMillisecondType, TimestampNanosecondType, TimestampSecondType,
 };
 use datatypes::prelude::ConcreteDataType;
@@ -69,9 +69,8 @@ impl Function for GreatestFunction {
         );
 
         match &input_types[0] {
-            ConcreteDataType::String(_) => Ok(ConcreteDataType::datetime_datatype()),
+            ConcreteDataType::String(_) => Ok(ConcreteDataType::timestamp_millisecond_datatype()),
             ConcreteDataType::Date(_) => Ok(ConcreteDataType::date_datatype()),
-            ConcreteDataType::DateTime(_) => Ok(ConcreteDataType::datetime_datatype()),
             ConcreteDataType::Timestamp(ts_type) => Ok(ConcreteDataType::Timestamp(*ts_type)),
             _ => UnsupportedInputDataTypeSnafu {
                 function: NAME,
@@ -87,7 +86,6 @@ impl Function for GreatestFunction {
             vec![
                 ConcreteDataType::string_datatype(),
                 ConcreteDataType::date_datatype(),
-                ConcreteDataType::datetime_datatype(),
                 ConcreteDataType::timestamp_nanosecond_datatype(),
                 ConcreteDataType::timestamp_microsecond_datatype(),
                 ConcreteDataType::timestamp_millisecond_datatype(),
@@ -109,20 +107,24 @@ impl Function for GreatestFunction {
         );
         match columns[0].data_type() {
             ConcreteDataType::String(_) => {
-                // Treats string as `DateTime` type.
-                let column1 = cast(&columns[0].to_arrow_array(), &ArrowDataType::Date64)
-                    .context(ArrowComputeSnafu)?;
-                let column1 = column1.as_primitive::<Date64Type>();
-                let column2 = cast(&columns[1].to_arrow_array(), &ArrowDataType::Date64)
-                    .context(ArrowComputeSnafu)?;
-                let column2 = column2.as_primitive::<Date64Type>();
+                let column1 = cast(
+                    &columns[0].to_arrow_array(),
+                    &ArrowDataType::Timestamp(TimeUnit::Millisecond, None),
+                )
+                .context(ArrowComputeSnafu)?;
+                let column1 = column1.as_primitive::<TimestampMillisecondType>();
+                let column2 = cast(
+                    &columns[1].to_arrow_array(),
+                    &ArrowDataType::Timestamp(TimeUnit::Millisecond, None),
+                )
+                .context(ArrowComputeSnafu)?;
+                let column2 = column2.as_primitive::<TimestampMillisecondType>();
                 let boolean_array = gt(&column1, &column2).context(ArrowComputeSnafu)?;
                 let result =
                     zip::zip(&boolean_array, &column1, &column2).context(ArrowComputeSnafu)?;
                 Ok(Helper::try_into_vector(&result).context(error::FromArrowArraySnafu)?)
             }
             ConcreteDataType::Date(_) => gt_time_types!(Date32Type, columns),
-            ConcreteDataType::DateTime(_) => gt_time_types!(Date64Type, columns),
             ConcreteDataType::Timestamp(ts_type) => match ts_type {
                 TimestampType::Second(_) => gt_time_types!(TimestampSecondType, columns),
                 TimestampType::Millisecond(_) => {
@@ -155,15 +157,15 @@ mod tests {
     use std::sync::Arc;
 
     use common_time::timestamp::TimeUnit;
-    use common_time::{Date, DateTime, Timestamp};
+    use common_time::{Date, Timestamp};
     use datatypes::types::{
-        DateTimeType, DateType, TimestampMicrosecondType, TimestampMillisecondType,
-        TimestampNanosecondType, TimestampSecondType,
+        DateType, TimestampMicrosecondType, TimestampMillisecondType, TimestampNanosecondType,
+        TimestampSecondType,
     };
     use datatypes::value::Value;
     use datatypes::vectors::{
-        DateTimeVector, DateVector, StringVector, TimestampMicrosecondVector,
-        TimestampMillisecondVector, TimestampNanosecondVector, TimestampSecondVector, Vector,
+        DateVector, StringVector, TimestampMicrosecondVector, TimestampMillisecondVector,
+        TimestampNanosecondVector, TimestampSecondVector, Vector,
     };
     use paste::paste;
 
@@ -178,7 +180,7 @@ mod tests {
                     ConcreteDataType::string_datatype()
                 ])
                 .unwrap(),
-            ConcreteDataType::DateTime(DateTimeType)
+            ConcreteDataType::timestamp_millisecond_datatype()
         );
         let columns = vec![
             Arc::new(StringVector::from(vec![
@@ -194,15 +196,18 @@ mod tests {
         let result = function
             .eval(&FunctionContext::default(), &columns)
             .unwrap();
-        let result = result.as_any().downcast_ref::<DateTimeVector>().unwrap();
+        let result = result
+            .as_any()
+            .downcast_ref::<TimestampMillisecondVector>()
+            .unwrap();
         assert_eq!(result.len(), 2);
         assert_eq!(
             result.get(0),
-            Value::DateTime(DateTime::from_str("2001-02-01 00:00:00", None).unwrap())
+            Value::Timestamp(Timestamp::from_str("2001-02-01 00:00:00", None).unwrap())
         );
         assert_eq!(
             result.get(1),
-            Value::DateTime(DateTime::from_str("2012-12-23 00:00:00", None).unwrap())
+            Value::Timestamp(Timestamp::from_str("2012-12-23 00:00:00", None).unwrap())
         );
     }
 
@@ -245,30 +250,33 @@ mod tests {
         assert_eq!(
             function
                 .return_type(&[
-                    ConcreteDataType::datetime_datatype(),
-                    ConcreteDataType::datetime_datatype()
+                    ConcreteDataType::timestamp_millisecond_datatype(),
+                    ConcreteDataType::timestamp_millisecond_datatype()
                 ])
                 .unwrap(),
-            ConcreteDataType::DateTime(DateTimeType)
+            ConcreteDataType::timestamp_millisecond_datatype()
         );
 
         let columns = vec![
-            Arc::new(DateTimeVector::from_slice(vec![-1, 2])) as _,
-            Arc::new(DateTimeVector::from_slice(vec![0, 1])) as _,
+            Arc::new(TimestampMillisecondVector::from_slice(vec![-1, 2])) as _,
+            Arc::new(TimestampMillisecondVector::from_slice(vec![0, 1])) as _,
         ];
 
         let result = function
             .eval(&FunctionContext::default(), &columns)
             .unwrap();
-        let result = result.as_any().downcast_ref::<DateTimeVector>().unwrap();
+        let result = result
+            .as_any()
+            .downcast_ref::<TimestampMillisecondVector>()
+            .unwrap();
         assert_eq!(result.len(), 2);
         assert_eq!(
             result.get(0),
-            Value::DateTime(DateTime::from_str("1970-01-01 00:00:00", None).unwrap())
+            Value::Timestamp(Timestamp::from_str("1970-01-01 00:00:00", None).unwrap())
         );
         assert_eq!(
             result.get(1),
-            Value::DateTime(DateTime::from_str("1970-01-01 00:00:00.002", None).unwrap())
+            Value::Timestamp(Timestamp::from_str("1970-01-01 00:00:00.002", None).unwrap())
         );
     }
 

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

@@ -17,7 +17,7 @@ use std::sync::Arc;
 
 use common_query::error::{InvalidFuncArgsSnafu, Result, UnsupportedInputDataTypeSnafu};
 use common_query::prelude::{Signature, Volatility};
-use common_time::{Date, DateTime, Timestamp};
+use common_time::{Date, Timestamp};
 use datatypes::prelude::ConcreteDataType;
 use datatypes::vectors::{Int64Vector, VectorRef};
 use snafu::ensure;
@@ -32,10 +32,6 @@ const NAME: &str = "to_unixtime";
 
 fn convert_to_seconds(arg: &str, func_ctx: &FunctionContext) -> Option<i64> {
     let timezone = &func_ctx.query_ctx.timezone();
-    if let Ok(dt) = DateTime::from_str(arg, Some(timezone)) {
-        return Some(dt.val() / 1000);
-    }
-
     if let Ok(ts) = Timestamp::from_str(arg, Some(timezone)) {
         return Some(ts.split().0);
     }
@@ -59,12 +55,6 @@ fn convert_dates_to_seconds(vector: &VectorRef) -> Vec<Option<i64>> {
         .collect::<Vec<Option<i64>>>()
 }
 
-fn convert_datetimes_to_seconds(vector: &VectorRef) -> Vec<Option<i64>> {
-    (0..vector.len())
-        .map(|i| vector.get(i).as_datetime().map(|dt| dt.val() / 1000))
-        .collect::<Vec<Option<i64>>>()
-}
-
 impl Function for ToUnixtimeFunction {
     fn name(&self) -> &str {
         NAME
@@ -82,7 +72,6 @@ impl Function for ToUnixtimeFunction {
                 ConcreteDataType::int32_datatype(),
                 ConcreteDataType::int64_datatype(),
                 ConcreteDataType::date_datatype(),
-                ConcreteDataType::datetime_datatype(),
                 ConcreteDataType::timestamp_second_datatype(),
                 ConcreteDataType::timestamp_millisecond_datatype(),
                 ConcreteDataType::timestamp_microsecond_datatype(),
@@ -119,10 +108,6 @@ impl Function for ToUnixtimeFunction {
                 let seconds = convert_dates_to_seconds(vector);
                 Ok(Arc::new(Int64Vector::from(seconds)))
             }
-            ConcreteDataType::DateTime(_) => {
-                let seconds = convert_datetimes_to_seconds(vector);
-                Ok(Arc::new(Int64Vector::from(seconds)))
-            }
             ConcreteDataType::Timestamp(_) => {
                 let seconds = convert_timestamps_to_seconds(vector);
                 Ok(Arc::new(Int64Vector::from(seconds)))
@@ -148,7 +133,7 @@ mod tests {
     use datatypes::prelude::ConcreteDataType;
     use datatypes::value::Value;
     use datatypes::vectors::{
-        DateTimeVector, DateVector, StringVector, TimestampMillisecondVector, TimestampSecondVector,
+        DateVector, StringVector, TimestampMillisecondVector, TimestampSecondVector,
     };
 
     use super::{ToUnixtimeFunction, *};
@@ -171,7 +156,6 @@ mod tests {
                              ConcreteDataType::int32_datatype(),
                              ConcreteDataType::int64_datatype(),
                              ConcreteDataType::date_datatype(),
-                             ConcreteDataType::datetime_datatype(),
                              ConcreteDataType::timestamp_second_datatype(),
                              ConcreteDataType::timestamp_millisecond_datatype(),
                              ConcreteDataType::timestamp_microsecond_datatype(),
@@ -253,31 +237,6 @@ mod tests {
         }
     }
 
-    #[test]
-    fn test_datetime_to_unixtime() {
-        let f = ToUnixtimeFunction;
-
-        let times = vec![Some(123000), None, Some(42000), None];
-        let results = [Some(123), None, Some(42), None];
-        let date_vector = DateTimeVector::from(times.clone());
-        let args: Vec<VectorRef> = vec![Arc::new(date_vector)];
-        let vector = f.eval(&FunctionContext::default(), &args).unwrap();
-        assert_eq!(4, vector.len());
-        for (i, _t) in times.iter().enumerate() {
-            let v = vector.get(i);
-            if i == 1 || i == 3 {
-                assert_eq!(Value::Null, v);
-                continue;
-            }
-            match v {
-                Value::Int64(ts) => {
-                    assert_eq!(ts, (*results.get(i).unwrap()).unwrap());
-                }
-                _ => unreachable!(),
-            }
-        }
-    }
-
     #[test]
     fn test_timestamp_to_unixtime() {
         let f = ToUnixtimeFunction;

src/common/function/src/scalars/vector.rs

@@ -24,9 +24,11 @@ pub(crate) mod sum;
 mod vector_add;
 mod vector_dim;
 mod vector_div;
+mod vector_kth_elem;
 mod vector_mul;
 mod vector_norm;
 mod vector_sub;
+mod vector_subvector;
 
 use std::sync::Arc;
 
@@ -56,6 +58,8 @@ impl VectorFunction {
         registry.register(Arc::new(vector_div::VectorDivFunction));
         registry.register(Arc::new(vector_norm::VectorNormFunction));
         registry.register(Arc::new(vector_dim::VectorDimFunction));
+        registry.register(Arc::new(vector_kth_elem::VectorKthElemFunction));
+        registry.register(Arc::new(vector_subvector::VectorSubvectorFunction));
         registry.register(Arc::new(elem_sum::ElemSumFunction));
         registry.register(Arc::new(elem_product::ElemProductFunction));
     }

src/common/function/src/scalars/vector/vector_kth_elem.rs

@@ -0,0 +1,211 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::borrow::Cow;
+use std::fmt::Display;
+
+use common_query::error::{InvalidFuncArgsSnafu, Result};
+use common_query::prelude::Signature;
+use datatypes::prelude::ConcreteDataType;
+use datatypes::scalars::ScalarVectorBuilder;
+use datatypes::vectors::{Float32VectorBuilder, MutableVector, VectorRef};
+use snafu::ensure;
+
+use crate::function::{Function, FunctionContext};
+use crate::helper;
+use crate::scalars::vector::impl_conv::{as_veclit, as_veclit_if_const};
+
+const NAME: &str = "vec_kth_elem";
+
+/// Returns the k-th(0-based index) element of the vector.
+///
+/// # Example
+///
+/// ```sql
+/// SELECT vec_kth_elem("[2, 4, 6]",1) as result;
+///
+/// +---------+
+/// | result  |
+/// +---------+
+/// | 4 |
+/// +---------+
+///
+/// ```
+///
+
+#[derive(Debug, Clone, Default)]
+pub struct VectorKthElemFunction;
+
+impl Function for VectorKthElemFunction {
+    fn name(&self) -> &str {
+        NAME
+    }
+
+    fn return_type(
+        &self,
+        _input_types: &[ConcreteDataType],
+    ) -> common_query::error::Result<ConcreteDataType> {
+        Ok(ConcreteDataType::float32_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        helper::one_of_sigs2(
+            vec![
+                ConcreteDataType::string_datatype(),
+                ConcreteDataType::binary_datatype(),
+            ],
+            vec![ConcreteDataType::int64_datatype()],
+        )
+    }
+
+    fn eval(&self, _func_ctx: &FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure!(
+            columns.len() == 2,
+            InvalidFuncArgsSnafu {
+                err_msg: format!(
+                    "The length of the args is not correct, expect exactly two, have: {}",
+                    columns.len()
+                ),
+            }
+        );
+
+        let arg0 = &columns[0];
+        let arg1 = &columns[1];
+
+        let len = arg0.len();
+        let mut result = Float32VectorBuilder::with_capacity(len);
+        if len == 0 {
+            return Ok(result.to_vector());
+        };
+
+        let arg0_const = as_veclit_if_const(arg0)?;
+
+        for i in 0..len {
+            let arg0 = match arg0_const.as_ref() {
+                Some(arg0) => Some(Cow::Borrowed(arg0.as_ref())),
+                None => as_veclit(arg0.get_ref(i))?,
+            };
+            let Some(arg0) = arg0 else {
+                result.push_null();
+                continue;
+            };
+
+            let arg1 = arg1.get(i).as_f64_lossy();
+            let Some(arg1) = arg1 else {
+                result.push_null();
+                continue;
+            };
+
+            ensure!(
+                arg1 >= 0.0 && arg1.fract() == 0.0,
+                InvalidFuncArgsSnafu {
+                    err_msg: format!(
+                        "Invalid argument: k must be a non-negative integer, but got k = {}.",
+                        arg1
+                    ),
+                }
+            );
+
+            let k = arg1 as usize;
+
+            ensure!(
+                k < arg0.len(),
+                InvalidFuncArgsSnafu {
+                    err_msg: format!(
+                        "Out of range: k must be in the range [0, {}], but got k = {}.",
+                        arg0.len() - 1,
+                        k
+                    ),
+                }
+            );
+
+            let value = arg0[k];
+
+            result.push(Some(value));
+        }
+        Ok(result.to_vector())
+    }
+}
+
+impl Display for VectorKthElemFunction {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{}", NAME.to_ascii_uppercase())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::Arc;
+
+    use common_query::error;
+    use datatypes::vectors::{Int64Vector, StringVector};
+
+    use super::*;
+
+    #[test]
+    fn test_vec_kth_elem() {
+        let func = VectorKthElemFunction;
+
+        let input0 = Arc::new(StringVector::from(vec![
+            Some("[1.0,2.0,3.0]".to_string()),
+            Some("[4.0,5.0,6.0]".to_string()),
+            Some("[7.0,8.0,9.0]".to_string()),
+            None,
+        ]));
+        let input1 = Arc::new(Int64Vector::from(vec![Some(0), Some(2), None, Some(1)]));
+
+        let result = func
+            .eval(&FunctionContext::default(), &[input0, input1])
+            .unwrap();
+
+        let result = result.as_ref();
+        assert_eq!(result.len(), 4);
+        assert_eq!(result.get_ref(0).as_f32().unwrap(), Some(1.0));
+        assert_eq!(result.get_ref(1).as_f32().unwrap(), Some(6.0));
+        assert!(result.get_ref(2).is_null());
+        assert!(result.get_ref(3).is_null());
+
+        let input0 = Arc::new(StringVector::from(vec![Some("[1.0,2.0,3.0]".to_string())]));
+        let input1 = Arc::new(Int64Vector::from(vec![Some(3)]));
+
+        let err = func
+            .eval(&FunctionContext::default(), &[input0, input1])
+            .unwrap_err();
+        match err {
+            error::Error::InvalidFuncArgs { err_msg, .. } => {
+                assert_eq!(
+                    err_msg,
+                    format!("Out of range: k must be in the range [0, 2], but got k = 3.")
+                )
+            }
+            _ => unreachable!(),
+        }
+
+        let input0 = Arc::new(StringVector::from(vec![Some("[1.0,2.0,3.0]".to_string())]));
+        let input1 = Arc::new(Int64Vector::from(vec![Some(-1)]));
+
+        let err = func
+            .eval(&FunctionContext::default(), &[input0, input1])
+            .unwrap_err();
+        match err {
+            error::Error::InvalidFuncArgs { err_msg, .. } => {
+                assert_eq!(
+                    err_msg,
+                    format!("Invalid argument: k must be a non-negative integer, but got k = -1.")
+                )
+            }
+            _ => unreachable!(),
+        }
+    }
+}

src/common/function/src/scalars/vector/vector_subvector.rs

@@ -0,0 +1,240 @@
+// 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::borrow::Cow;
+use std::fmt::Display;
+
+use common_query::error::{InvalidFuncArgsSnafu, Result};
+use common_query::prelude::{Signature, TypeSignature};
+use datafusion_expr::Volatility;
+use datatypes::prelude::ConcreteDataType;
+use datatypes::scalars::ScalarVectorBuilder;
+use datatypes::vectors::{BinaryVectorBuilder, MutableVector, VectorRef};
+use snafu::ensure;
+
+use crate::function::{Function, FunctionContext};
+use crate::scalars::vector::impl_conv::{as_veclit, as_veclit_if_const, veclit_to_binlit};
+
+const NAME: &str = "vec_subvector";
+
+/// Returns a subvector from start(included) to end(excluded) index.
+///
+/// # Example
+///
+/// ```sql
+/// SELECT vec_to_string(vec_subvector("[1, 2, 3, 4, 5]", 1, 3)) as result;
+///
+/// +---------+
+/// | result  |
+/// +---------+
+/// | [2, 3]  |
+/// +---------+
+///
+/// ```
+///
+
+#[derive(Debug, Clone, Default)]
+pub struct VectorSubvectorFunction;
+
+impl Function for VectorSubvectorFunction {
+    fn name(&self) -> &str {
+        NAME
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::binary_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        Signature::one_of(
+            vec![
+                TypeSignature::Exact(vec![
+                    ConcreteDataType::string_datatype(),
+                    ConcreteDataType::int64_datatype(),
+                    ConcreteDataType::int64_datatype(),
+                ]),
+                TypeSignature::Exact(vec![
+                    ConcreteDataType::binary_datatype(),
+                    ConcreteDataType::int64_datatype(),
+                    ConcreteDataType::int64_datatype(),
+                ]),
+            ],
+            Volatility::Immutable,
+        )
+    }
+
+    fn eval(&self, _func_ctx: &FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure!(
+            columns.len() == 3,
+            InvalidFuncArgsSnafu {
+                err_msg: format!(
+                    "The length of the args is not correct, expect exactly three, have: {}",
+                    columns.len()
+                )
+            }
+        );
+
+        let arg0 = &columns[0];
+        let arg1 = &columns[1];
+        let arg2 = &columns[2];
+
+        ensure!(
+            arg0.len() == arg1.len() && arg1.len() == arg2.len(),
+            InvalidFuncArgsSnafu {
+                err_msg: format!(
+                    "The lengths of the vector are not aligned, args 0: {}, args 1: {}, args 2: {}",
+                    arg0.len(),
+                    arg1.len(),
+                    arg2.len()
+                )
+            }
+        );
+
+        let len = arg0.len();
+        let mut result = BinaryVectorBuilder::with_capacity(len);
+        if len == 0 {
+            return Ok(result.to_vector());
+        }
+
+        let arg0_const = as_veclit_if_const(arg0)?;
+
+        for i in 0..len {
+            let arg0 = match arg0_const.as_ref() {
+                Some(arg0) => Some(Cow::Borrowed(arg0.as_ref())),
+                None => as_veclit(arg0.get_ref(i))?,
+            };
+            let arg1 = arg1.get(i).as_i64();
+            let arg2 = arg2.get(i).as_i64();
+            let (Some(arg0), Some(arg1), Some(arg2)) = (arg0, arg1, arg2) else {
+                result.push_null();
+                continue;
+            };
+
+            ensure!(
+                0 <= arg1 && arg1 <= arg2 && arg2 as usize <= arg0.len(),
+                InvalidFuncArgsSnafu {
+                    err_msg: format!(
+                        "Invalid start and end indices: start={}, end={}, vec_len={}",
+                        arg1,
+                        arg2,
+                        arg0.len()
+                    )
+                }
+            );
+
+            let subvector = &arg0[arg1 as usize..arg2 as usize];
+            let binlit = veclit_to_binlit(subvector);
+            result.push(Some(&binlit));
+        }
+
+        Ok(result.to_vector())
+    }
+}
+
+impl Display for VectorSubvectorFunction {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{}", NAME.to_ascii_uppercase())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::Arc;
+
+    use common_query::error::Error;
+    use datatypes::vectors::{Int64Vector, StringVector};
+
+    use super::*;
+    use crate::function::FunctionContext;
+    #[test]
+    fn test_subvector() {
+        let func = VectorSubvectorFunction;
+
+        let input0 = Arc::new(StringVector::from(vec![
+            Some("[1.0, 2.0, 3.0, 4.0, 5.0]".to_string()),
+            Some("[6.0, 7.0, 8.0, 9.0, 10.0]".to_string()),
+            None,
+            Some("[11.0, 12.0, 13.0]".to_string()),
+        ]));
+        let input1 = Arc::new(Int64Vector::from(vec![Some(1), Some(0), Some(0), Some(1)]));
+        let input2 = Arc::new(Int64Vector::from(vec![Some(3), Some(5), Some(2), Some(3)]));
+
+        let result = func
+            .eval(&FunctionContext::default(), &[input0, input1, input2])
+            .unwrap();
+
+        let result = result.as_ref();
+        assert_eq!(result.len(), 4);
+        assert_eq!(
+            result.get_ref(0).as_binary().unwrap(),
+            Some(veclit_to_binlit(&[2.0, 3.0]).as_slice())
+        );
+        assert_eq!(
+            result.get_ref(1).as_binary().unwrap(),
+            Some(veclit_to_binlit(&[6.0, 7.0, 8.0, 9.0, 10.0]).as_slice())
+        );
+        assert!(result.get_ref(2).is_null());
+        assert_eq!(
+            result.get_ref(3).as_binary().unwrap(),
+            Some(veclit_to_binlit(&[12.0, 13.0]).as_slice())
+        );
+    }
+    #[test]
+    fn test_subvector_error() {
+        let func = VectorSubvectorFunction;
+
+        let input0 = Arc::new(StringVector::from(vec![
+            Some("[1.0, 2.0, 3.0]".to_string()),
+            Some("[4.0, 5.0, 6.0]".to_string()),
+        ]));
+        let input1 = Arc::new(Int64Vector::from(vec![Some(1), Some(2)]));
+        let input2 = Arc::new(Int64Vector::from(vec![Some(3)]));
+
+        let result = func.eval(&FunctionContext::default(), &[input0, input1, input2]);
+
+        match result {
+            Err(Error::InvalidFuncArgs { err_msg, .. }) => {
+                assert_eq!(
+                    err_msg,
+                    "The lengths of the vector are not aligned, args 0: 2, args 1: 2, args 2: 1"
+                )
+            }
+            _ => unreachable!(),
+        }
+    }
+
+    #[test]
+    fn test_subvector_invalid_indices() {
+        let func = VectorSubvectorFunction;
+
+        let input0 = Arc::new(StringVector::from(vec![
+            Some("[1.0, 2.0, 3.0]".to_string()),
+            Some("[4.0, 5.0, 6.0]".to_string()),
+        ]));
+        let input1 = Arc::new(Int64Vector::from(vec![Some(1), Some(3)]));
+        let input2 = Arc::new(Int64Vector::from(vec![Some(3), Some(4)]));
+
+        let result = func.eval(&FunctionContext::default(), &[input0, input1, input2]);
+
+        match result {
+            Err(Error::InvalidFuncArgs { err_msg, .. }) => {
+                assert_eq!(
+                    err_msg,
+                    "Invalid start and end indices: start=3, end=4, vec_len=3"
+                )
+            }
+            _ => unreachable!(),
+        }
+    }
+}

src/common/function/src/state.rs

@@ -35,7 +35,10 @@ impl FunctionState {
         use api::v1::meta::ProcedureStatus;
         use async_trait::async_trait;
         use common_base::AffectedRows;
-        use common_meta::rpc::procedure::{MigrateRegionRequest, ProcedureStateResponse};
+        use common_meta::rpc::procedure::{
+            AddRegionFollowerRequest, MigrateRegionRequest, ProcedureStateResponse,
+            RemoveRegionFollowerRequest,
+        };
         use common_query::error::Result;
         use common_query::Output;
         use session::context::QueryContextRef;
@@ -66,6 +69,17 @@ impl FunctionState {
                     ..Default::default()
                 })
             }
+
+            async fn add_region_follower(&self, _request: AddRegionFollowerRequest) -> Result<()> {
+                Ok(())
+            }
+
+            async fn remove_region_follower(
+                &self,
+                _request: RemoveRegionFollowerRequest,
+            ) -> Result<()> {
+                Ok(())
+            }
         }
 
         #[async_trait]

src/common/function/src/system.rs

@@ -22,7 +22,9 @@ mod version;
 use std::sync::Arc;
 
 use build::BuildFunction;
-use database::{CurrentSchemaFunction, DatabaseFunction, SessionUserFunction};
+use database::{
+    CurrentSchemaFunction, DatabaseFunction, ReadPreferenceFunction, SessionUserFunction,
+};
 use pg_catalog::PGCatalogFunction;
 use procedure_state::ProcedureStateFunction;
 use timezone::TimezoneFunction;
@@ -39,6 +41,7 @@ impl SystemFunction {
         registry.register(Arc::new(CurrentSchemaFunction));
         registry.register(Arc::new(DatabaseFunction));
         registry.register(Arc::new(SessionUserFunction));
+        registry.register(Arc::new(ReadPreferenceFunction));
         registry.register(Arc::new(TimezoneFunction));
         registry.register_async(Arc::new(ProcedureStateFunction));
         PGCatalogFunction::register(registry);

src/common/function/src/system/database.rs

@@ -30,9 +30,12 @@ pub struct DatabaseFunction;
 pub struct CurrentSchemaFunction;
 pub struct SessionUserFunction;
 
+pub struct ReadPreferenceFunction;
+
 const DATABASE_FUNCTION_NAME: &str = "database";
 const CURRENT_SCHEMA_FUNCTION_NAME: &str = "current_schema";
 const SESSION_USER_FUNCTION_NAME: &str = "session_user";
+const READ_PREFERENCE_FUNCTION_NAME: &str = "read_preference";
 
 impl Function for DatabaseFunction {
     fn name(&self) -> &str {
@@ -94,6 +97,26 @@ impl Function for SessionUserFunction {
     }
 }
 
+impl Function for ReadPreferenceFunction {
+    fn name(&self) -> &str {
+        READ_PREFERENCE_FUNCTION_NAME
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::string_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        Signature::nullary(Volatility::Immutable)
+    }
+
+    fn eval(&self, func_ctx: &FunctionContext, _columns: &[VectorRef]) -> Result<VectorRef> {
+        let read_preference = func_ctx.query_ctx.read_preference();
+
+        Ok(Arc::new(StringVector::from_slice(&[read_preference.as_ref()])) as _)
+    }
+}
+
 impl fmt::Display for DatabaseFunction {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         write!(f, "DATABASE")
@@ -112,6 +135,12 @@ impl fmt::Display for SessionUserFunction {
     }
 }
 
+impl fmt::Display for ReadPreferenceFunction {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "READ_PREFERENCE")
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use std::sync::Arc;

src/common/grpc-expr/src/alter.rs

@@ -15,11 +15,13 @@
 use api::helper::ColumnDataTypeWrapper;
 use api::v1::add_column_location::LocationType;
 use api::v1::alter_table_expr::Kind;
-use api::v1::column_def::{as_fulltext_option, as_skipping_index_type};
+use api::v1::column_def::{
+    as_fulltext_option_analyzer, as_fulltext_option_backend, as_skipping_index_type,
+};
 use api::v1::{
     column_def, AddColumnLocation as Location, AlterTableExpr, Analyzer, CreateTableExpr,
-    DropColumns, ModifyColumnTypes, RenameTable, SemanticType,
-    SkippingIndexType as PbSkippingIndexType,
+    DropColumns, FulltextBackend as PbFulltextBackend, ModifyColumnTypes, RenameTable,
+    SemanticType, SkippingIndexType as PbSkippingIndexType,
 };
 use common_query::AddColumnLocation;
 use datatypes::schema::{ColumnSchema, FulltextOptions, RawSchema, SkippingIndexOptions};
@@ -126,11 +128,15 @@ pub fn alter_expr_to_request(table_id: TableId, expr: AlterTableExpr) -> Result<
                         column_name: f.column_name.clone(),
                         options: FulltextOptions {
                             enable: f.enable,
-                            analyzer: as_fulltext_option(
+                            analyzer: as_fulltext_option_analyzer(
                                 Analyzer::try_from(f.analyzer)
                                     .context(InvalidSetFulltextOptionRequestSnafu)?,
                             ),
                             case_sensitive: f.case_sensitive,
+                            backend: as_fulltext_option_backend(
+                                PbFulltextBackend::try_from(f.backend)
+                                    .context(InvalidSetFulltextOptionRequestSnafu)?,
+                            ),
                         },
                     },
                 },

src/common/grpc/benches/channel_manager.rs

@@ -25,7 +25,7 @@ async fn do_bench_channel_manager() {
         let m_clone = m.clone();
         let join = tokio::spawn(async move {
             for _ in 0..10000 {
-                let idx = rand::random::<usize>() % 100;
+                let idx = rand::random::<u32>() % 100;
                 let ret = m_clone.get(format!("{idx}"));
                 let _ = ret.unwrap();
             }

src/common/grpc/src/channel_manager.rs

@@ -445,20 +445,10 @@ impl Pool {
 
 async fn recycle_channel_in_loop(pool: Arc<Pool>, interval_secs: u64) {
     let mut interval = tokio::time::interval(Duration::from_secs(interval_secs));
-    // use weak ref here to prevent pool being leaked
-    let pool_weak = {
-        let weak = Arc::downgrade(&pool);
-        drop(pool);
-        weak
-    };
+
     loop {
         let _ = interval.tick().await;
-        if let Some(pool) = pool_weak.upgrade() {
-            pool.retain_channel(|_, c| c.access.swap(0, Ordering::Relaxed) != 0)
-        } else {
-            // no one is using this pool, so we can also let go
-            break;
-        }
+        pool.retain_channel(|_, c| c.access.swap(0, Ordering::Relaxed) != 0)
     }
 }
 

src/common/grpc/src/select.rs

@@ -17,8 +17,8 @@ use api::v1::column::Values;
 use common_base::BitVec;
 use datatypes::types::{IntervalType, TimeType, TimestampType, WrapperType};
 use datatypes::vectors::{
-    BinaryVector, BooleanVector, DateTimeVector, DateVector, Decimal128Vector, Float32Vector,
-    Float64Vector, Int16Vector, Int32Vector, Int64Vector, Int8Vector, IntervalDayTimeVector,
+    BinaryVector, BooleanVector, DateVector, Decimal128Vector, Float32Vector, Float64Vector,
+    Int16Vector, Int32Vector, Int64Vector, Int8Vector, IntervalDayTimeVector,
     IntervalMonthDayNanoVector, IntervalYearMonthVector, StringVector, TimeMicrosecondVector,
     TimeMillisecondVector, TimeNanosecondVector, TimeSecondVector, TimestampMicrosecondVector,
     TimestampMillisecondVector, TimestampNanosecondVector, TimestampSecondVector, UInt16Vector,
@@ -141,12 +141,6 @@ pub fn values(arrays: &[VectorRef]) -> Result<Values> {
         (ConcreteDataType::Date(_), DateVector, date_values, |x| {
             x.val()
         }),
-        (
-            ConcreteDataType::DateTime(_),
-            DateTimeVector,
-            datetime_values,
-            |x| { x.val() }
-        ),
         (
             ConcreteDataType::Timestamp(TimestampType::Second(_)),
             TimestampSecondVector,

src/common/macro/src/lib.rs

@@ -18,11 +18,13 @@ mod print_caller;
 mod range_fn;
 mod stack_trace_debug;
 mod utils;
+
 use aggr_func::{impl_aggr_func_type_store, impl_as_aggr_func_creator};
 use print_caller::process_print_caller;
 use proc_macro::TokenStream;
+use quote::quote;
 use range_fn::process_range_fn;
-use syn::{parse_macro_input, DeriveInput};
+use syn::{parse_macro_input, Data, DeriveInput, Fields};
 
 use crate::admin_fn::process_admin_fn;
 
@@ -136,3 +138,51 @@ pub fn print_caller(args: TokenStream, input: TokenStream) -> TokenStream {
 pub fn stack_trace_debug(args: TokenStream, input: TokenStream) -> TokenStream {
     stack_trace_debug::stack_trace_style_impl(args.into(), input.into()).into()
 }
+
+/// Generates implementation for `From<&TableMeta> for TableMetaBuilder`
+#[proc_macro_derive(ToMetaBuilder)]
+pub fn derive_meta_builder(input: TokenStream) -> TokenStream {
+    let input = parse_macro_input!(input as DeriveInput);
+
+    let Data::Struct(data_struct) = input.data else {
+        panic!("ToMetaBuilder can only be derived for structs");
+    };
+
+    let Fields::Named(fields) = data_struct.fields else {
+        panic!("ToMetaBuilder can only be derived for structs with named fields");
+    };
+
+    // Check that this is being applied to TableMeta struct
+    if input.ident != "TableMeta" {
+        panic!("ToMetaBuilder can only be derived for TableMeta struct");
+    }
+
+    let field_init = fields.named.iter().map(|field| {
+        let field_name = field.ident.as_ref().unwrap();
+        quote! {
+            #field_name: Default::default(),
+        }
+    });
+
+    let field_assignments = fields.named.iter().map(|field| {
+        let field_name = field.ident.as_ref().unwrap();
+        quote! {
+            builder.#field_name(meta.#field_name.clone());
+        }
+    });
+
+    let gen = quote! {
+        impl From<&TableMeta> for TableMetaBuilder {
+            fn from(meta: &TableMeta) -> Self {
+                let mut builder = Self {
+                    #(#field_init)*
+                };
+
+                #(#field_assignments)*
+                builder
+            }
+        }
+    };
+
+    gen.into()
+}

src/common/meta/Cargo.toml

@@ -7,6 +7,7 @@ license.workspace = true
 [features]
 testing = []
 pg_kvbackend = ["dep:tokio-postgres", "dep:backon", "dep:deadpool-postgres", "dep:deadpool"]
+mysql_kvbackend = ["dep:sqlx", "dep:backon"]
 
 [lints]
 workspace = true
@@ -57,9 +58,10 @@ serde_json.workspace = true
 serde_with.workspace = true
 session.workspace = true
 snafu.workspace = true
+sqlx = { workspace = true, optional = true }
 store-api.workspace = true
 strum.workspace = true
-table.workspace = true
+table = { workspace = true, features = ["testing"] }
 tokio.workspace = true
 tokio-postgres = { workspace = true, optional = true }
 tonic.workspace = true

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

@@ -192,6 +192,8 @@ mod tests {
                     expire_after: Some(300),
                     comment: "comment".to_string(),
                     options: Default::default(),
+                    created_time: chrono::Utc::now(),
+                    updated_time: chrono::Utc::now(),
                 },
                 (1..=3)
                     .map(|i| {

src/common/meta/src/cluster.rs

@@ -27,6 +27,7 @@ use crate::error::{
     DecodeJsonSnafu, EncodeJsonSnafu, Error, FromUtf8Snafu, InvalidNodeInfoKeySnafu,
     InvalidRoleSnafu, ParseNumSnafu, Result,
 };
+use crate::key::flow::flow_state::FlowStat;
 use crate::peer::Peer;
 
 const CLUSTER_NODE_INFO_PREFIX: &str = "__meta_cluster_node_info";
@@ -52,6 +53,9 @@ pub trait ClusterInfo {
     /// List all region stats in the cluster.
     async fn list_region_stats(&self) -> std::result::Result<Vec<RegionStat>, Self::Error>;
 
+    /// List all flow stats in the cluster.
+    async fn list_flow_stats(&self) -> std::result::Result<Option<FlowStat>, Self::Error>;
+
     // TODO(jeremy): Other info, like region status, etc.
 }
 

src/common/meta/src/datanode.rs

@@ -92,6 +92,22 @@ pub struct RegionStat {
     pub sst_size: u64,
     /// The size of the SST index files in bytes.
     pub index_size: u64,
+    /// The manifest infoof the region.
+    pub region_manifest: RegionManifestInfo,
+}
+
+#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
+pub enum RegionManifestInfo {
+    Mito {
+        manifest_version: u64,
+        flushed_entry_id: u64,
+    },
+    Metric {
+        data_manifest_version: u64,
+        data_flushed_entry_id: u64,
+        metadata_manifest_version: u64,
+        metadata_flushed_entry_id: u64,
+    },
 }
 
 impl Stat {
@@ -165,6 +181,31 @@ impl TryFrom<&HeartbeatRequest> for Stat {
     }
 }
 
+impl From<store_api::region_engine::RegionManifestInfo> for RegionManifestInfo {
+    fn from(value: store_api::region_engine::RegionManifestInfo) -> Self {
+        match value {
+            store_api::region_engine::RegionManifestInfo::Mito {
+                manifest_version,
+                flushed_entry_id,
+            } => RegionManifestInfo::Mito {
+                manifest_version,
+                flushed_entry_id,
+            },
+            store_api::region_engine::RegionManifestInfo::Metric {
+                data_manifest_version,
+                data_flushed_entry_id,
+                metadata_manifest_version,
+                metadata_flushed_entry_id,
+            } => RegionManifestInfo::Metric {
+                data_manifest_version,
+                data_flushed_entry_id,
+                metadata_manifest_version,
+                metadata_flushed_entry_id,
+            },
+        }
+    }
+}
+
 impl From<&api::v1::meta::RegionStat> for RegionStat {
     fn from(value: &api::v1::meta::RegionStat) -> Self {
         let region_stat = value
@@ -185,6 +226,7 @@ impl From<&api::v1::meta::RegionStat> for RegionStat {
             manifest_size: region_stat.manifest_size,
             sst_size: region_stat.sst_size,
             index_size: region_stat.index_size,
+            region_manifest: region_stat.manifest.into(),
         }
     }
 }

src/common/meta/src/ddl.rs

@@ -22,14 +22,18 @@ use store_api::storage::{RegionId, RegionNumber, TableId};
 use crate::cache_invalidator::CacheInvalidatorRef;
 use crate::ddl::flow_meta::FlowMetadataAllocatorRef;
 use crate::ddl::table_meta::TableMetadataAllocatorRef;
-use crate::error::Result;
+use crate::error::{Result, UnsupportedSnafu};
 use crate::key::flow::FlowMetadataManagerRef;
 use crate::key::table_route::PhysicalTableRouteValue;
 use crate::key::TableMetadataManagerRef;
 use crate::node_manager::NodeManagerRef;
 use crate::region_keeper::MemoryRegionKeeperRef;
+use crate::region_registry::LeaderRegionRegistryRef;
 use crate::rpc::ddl::{SubmitDdlTaskRequest, SubmitDdlTaskResponse};
-use crate::rpc::procedure::{MigrateRegionRequest, MigrateRegionResponse, ProcedureStateResponse};
+use crate::rpc::procedure::{
+    AddRegionFollowerRequest, MigrateRegionRequest, MigrateRegionResponse, ProcedureStateResponse,
+    RemoveRegionFollowerRequest,
+};
 use crate::DatanodeId;
 
 pub mod alter_database;
@@ -70,6 +74,30 @@ pub trait ProcedureExecutor: Send + Sync {
         request: SubmitDdlTaskRequest,
     ) -> Result<SubmitDdlTaskResponse>;
 
+    /// Add a region follower
+    async fn add_region_follower(
+        &self,
+        _ctx: &ExecutorContext,
+        _request: AddRegionFollowerRequest,
+    ) -> Result<()> {
+        UnsupportedSnafu {
+            operation: "add_region_follower",
+        }
+        .fail()
+    }
+
+    /// Remove a region follower
+    async fn remove_region_follower(
+        &self,
+        _ctx: &ExecutorContext,
+        _request: RemoveRegionFollowerRequest,
+    ) -> Result<()> {
+        UnsupportedSnafu {
+            operation: "remove_region_follower",
+        }
+        .fail()
+    }
+
     /// Submit a region migration task
     async fn migrate_region(
         &self,
@@ -137,6 +165,8 @@ pub struct DdlContext {
     pub cache_invalidator: CacheInvalidatorRef,
     /// Keep tracking operating regions.
     pub memory_region_keeper: MemoryRegionKeeperRef,
+    /// The leader region registry.
+    pub leader_region_registry: LeaderRegionRegistryRef,
     /// Table metadata manager.
     pub table_metadata_manager: TableMetadataManagerRef,
     /// Allocator for table metadata.

src/common/meta/src/ddl/alter_table.rs

@@ -22,30 +22,31 @@ use std::vec;
 use api::v1::alter_table_expr::Kind;
 use api::v1::RenameTable;
 use async_trait::async_trait;
-use common_error::ext::ErrorExt;
-use common_error::status_code::StatusCode;
+use common_error::ext::BoxedError;
 use common_procedure::error::{FromJsonSnafu, Result as ProcedureResult, ToJsonSnafu};
 use common_procedure::{
-    Context as ProcedureContext, Error as ProcedureError, LockKey, Procedure, Status, StringKey,
+    Context as ProcedureContext, ContextProvider, Error as ProcedureError, LockKey, PoisonKey,
+    PoisonKeys, Procedure, ProcedureId, Status, StringKey,
 };
 use common_telemetry::{debug, error, info};
 use futures::future;
 use serde::{Deserialize, Serialize};
-use snafu::ResultExt;
+use snafu::{ensure, ResultExt};
 use store_api::storage::RegionId;
 use strum::AsRefStr;
 use table::metadata::{RawTableInfo, TableId, TableInfo};
 use table::table_reference::TableReference;
 
 use crate::cache_invalidator::Context;
-use crate::ddl::utils::add_peer_context_if_needed;
+use crate::ddl::utils::{add_peer_context_if_needed, handle_multiple_results, MultipleResults};
 use crate::ddl::DdlContext;
-use crate::error::{Error, Result};
+use crate::error::{AbortProcedureSnafu, Error, NoLeaderSnafu, PutPoisonSnafu, Result};
 use crate::instruction::CacheIdent;
 use crate::key::table_info::TableInfoValue;
 use crate::key::{DeserializedValueWithBytes, RegionDistribution};
 use crate::lock_key::{CatalogLock, SchemaLock, TableLock, TableNameLock};
 use crate::metrics;
+use crate::poison_key::table_poison_key;
 use crate::rpc::ddl::AlterTableTask;
 use crate::rpc::router::{find_leader_regions, find_leaders, region_distribution};
 
@@ -104,7 +105,27 @@ impl AlterTableProcedure {
         Ok(Status::executing(true))
     }
 
-    pub async fn submit_alter_region_requests(&mut self) -> Result<Status> {
+    fn table_poison_key(&self) -> PoisonKey {
+        table_poison_key(self.data.table_id())
+    }
+
+    async fn put_poison(
+        &self,
+        ctx_provider: &dyn ContextProvider,
+        procedure_id: ProcedureId,
+    ) -> Result<()> {
+        let poison_key = self.table_poison_key();
+        ctx_provider
+            .try_put_poison(&poison_key, procedure_id)
+            .await
+            .context(PutPoisonSnafu)
+    }
+
+    pub async fn submit_alter_region_requests(
+        &mut self,
+        procedure_id: ProcedureId,
+        ctx_provider: &dyn ContextProvider,
+    ) -> Result<Status> {
         let table_id = self.data.table_id();
         let (_, physical_table_route) = self
             .context
@@ -127,6 +148,9 @@ impl AlterTableProcedure {
             alter_kind,
         );
 
+        ensure!(!leaders.is_empty(), NoLeaderSnafu { table_id });
+        // Puts the poison before submitting alter region requests to datanodes.
+        self.put_poison(ctx_provider, procedure_id).await?;
         for datanode in leaders {
             let requester = self.context.node_manager.datanode(&datanode).await;
             let regions = find_leader_regions(&physical_table_route.region_routes, &datanode);
@@ -140,28 +164,51 @@ impl AlterTableProcedure {
                 let requester = requester.clone();
 
                 alter_region_tasks.push(async move {
-                    if let Err(err) = requester.handle(request).await {
-                        if err.status_code() != StatusCode::RequestOutdated {
-                            // Treat request outdated as success.
-                            // The engine will throw this code when the schema version not match.
-                            // As this procedure has locked the table, the only reason for this error
-                            // is procedure is succeeded before and is retrying.
-                            return Err(add_peer_context_if_needed(datanode)(err));
-                        }
-                    }
-                    Ok(())
+                    requester
+                        .handle(request)
+                        .await
+                        .map_err(add_peer_context_if_needed(datanode))
                 });
             }
         }
 
-        future::join_all(alter_region_tasks)
+        let results = future::join_all(alter_region_tasks)
             .await
             .into_iter()
-            .collect::<Result<Vec<_>>>()?;
+            .collect::<Vec<_>>();
 
-        self.data.state = AlterTableState::UpdateMetadata;
+        match handle_multiple_results(results) {
+            MultipleResults::PartialRetryable(error) => {
+                // Just returns the error, and wait for the next try.
+                Err(error)
+            }
+            MultipleResults::PartialNonRetryable(error) => {
+                error!(error; "Partial non-retryable errors occurred during alter table, table {}, table_id: {}", self.data.table_ref(), self.data.table_id());
+                // No retry will be done.
+                Ok(Status::poisoned(
+                    Some(self.table_poison_key()),
+                    ProcedureError::external(error),
+                ))
+            }
+            MultipleResults::AllRetryable(error) => {
+                // Just returns the error, and wait for the next try.
+                Err(error)
+            }
+            MultipleResults::Ok => {
+                self.data.state = AlterTableState::UpdateMetadata;
+                Ok(Status::executing_with_clean_poisons(true))
+            }
+            MultipleResults::AllNonRetryable(error) => {
+                error!(error; "All alter requests returned non-retryable errors for table {}, table_id: {}", self.data.table_ref(), self.data.table_id());
+                // It assumes the metadata on datanode is not changed.
+                // Case: The alter region request is sent but not applied. (e.g., InvalidArgument)
 
-        Ok(Status::executing(true))
+                let err = BoxedError::new(error);
+                Err(err).context(AbortProcedureSnafu {
+                    clean_poisons: true,
+                })
+            }
+        }
     }
 
     /// Update table metadata.
@@ -250,10 +297,12 @@ impl Procedure for AlterTableProcedure {
         Self::TYPE_NAME
     }
 
-    async fn execute(&mut self, _ctx: &ProcedureContext) -> ProcedureResult<Status> {
+    async fn execute(&mut self, ctx: &ProcedureContext) -> ProcedureResult<Status> {
         let error_handler = |e: Error| {
             if e.is_retry_later() {
                 ProcedureError::retry_later(e)
+            } else if e.need_clean_poisons() {
+                ProcedureError::external_and_clean_poisons(e)
             } else {
                 ProcedureError::external(e)
             }
@@ -269,7 +318,10 @@ impl Procedure for AlterTableProcedure {
 
         match state {
             AlterTableState::Prepare => self.on_prepare().await,
-            AlterTableState::SubmitAlterRegionRequests => self.submit_alter_region_requests().await,
+            AlterTableState::SubmitAlterRegionRequests => {
+                self.submit_alter_region_requests(ctx.procedure_id, ctx.provider.as_ref())
+                    .await
+            }
             AlterTableState::UpdateMetadata => self.on_update_metadata().await,
             AlterTableState::InvalidateTableCache => self.on_broadcast().await,
         }
@@ -285,6 +337,10 @@ impl Procedure for AlterTableProcedure {
 
         LockKey::new(key)
     }
+
+    fn poison_keys(&self) -> PoisonKeys {
+        PoisonKeys::new(vec![self.table_poison_key()])
+    }
 }
 
 #[derive(Debug, Serialize, Deserialize, AsRefStr)]

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

@@ -337,7 +337,6 @@ pub enum FlowType {
 impl FlowType {
     pub const RECORDING_RULE: &str = "recording_rule";
     pub const STREAMING: &str = "streaming";
-    pub const FLOW_TYPE_KEY: &str = "flow_type";
 }
 
 impl Default for FlowType {
@@ -392,8 +391,7 @@ impl From<&CreateFlowData> for CreateRequest {
         };
 
         let flow_type = value.flow_type.unwrap_or_default().to_string();
-        req.flow_options
-            .insert(FlowType::FLOW_TYPE_KEY.to_string(), flow_type);
+        req.flow_options.insert("flow_type".to_string(), flow_type);
         req
     }
 }
@@ -425,9 +423,16 @@ impl From<&CreateFlowData> for (FlowInfoValue, Vec<(FlowPartitionId, FlowRouteVa
             .collect::<Vec<_>>();
 
         let flow_type = value.flow_type.unwrap_or_default().to_string();
-        options.insert(FlowType::FLOW_TYPE_KEY.to_string(), flow_type);
+        options.insert("flow_type".to_string(), flow_type);
 
-        let flow_info = FlowInfoValue {
+        let mut create_time = chrono::Utc::now();
+        if let Some(prev_flow_value) = value.prev_flow_info_value.as_ref()
+            && value.task.or_replace
+        {
+            create_time = prev_flow_value.get_inner_ref().created_time;
+        }
+
+        let flow_info: FlowInfoValue = FlowInfoValue {
             source_table_ids: value.source_table_ids.clone(),
             sink_table_name,
             flownode_ids,
@@ -437,6 +442,8 @@ impl From<&CreateFlowData> for (FlowInfoValue, Vec<(FlowPartitionId, FlowRouteVa
             expire_after,
             comment,
             options,
+            created_time: create_time,
+            updated_time: chrono::Utc::now(),
         };
 
         (flow_info, flow_routes)

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

@@ -299,7 +299,9 @@ impl Procedure for CreateTableProcedure {
                 .creator
                 .register_opening_regions(&self.context, &x.region_routes)
                 .map_err(BoxedError::new)
-                .context(ExternalSnafu)?;
+                .context(ExternalSnafu {
+                    clean_poisons: false,
+                })?;
         }
 
         Ok(())

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

@@ -130,7 +130,9 @@ impl Procedure for DropDatabaseProcedure {
         self.state
             .recover(&self.runtime_context)
             .map_err(BoxedError::new)
-            .context(ExternalSnafu)
+            .context(ExternalSnafu {
+                clean_poisons: false,
+            })
     }
 
     async fn execute(&mut self, _ctx: &ProcedureContext) -> ProcedureResult<Status> {

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

@@ -200,7 +200,9 @@ impl Procedure for DropTableProcedure {
         if register_operating_regions {
             self.register_dropping_regions()
                 .map_err(BoxedError::new)
-                .context(ExternalSnafu)?;
+                .context(ExternalSnafu {
+                    clean_poisons: false,
+                })?;
         }
 
         Ok(())

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

@@ -35,7 +35,9 @@ use crate::error::{self, Result};
 use crate::instruction::CacheIdent;
 use crate::key::table_name::TableNameKey;
 use crate::key::table_route::TableRouteValue;
-use crate::rpc::router::{find_leader_regions, find_leaders, RegionRoute};
+use crate::rpc::router::{
+    find_leader_regions, find_leaders, operating_leader_regions, RegionRoute,
+};
 
 /// [Control] indicated to the caller whether to go to the next step.
 #[derive(Debug)]
@@ -250,6 +252,11 @@ impl DropTableExecutor {
             .into_iter()
             .collect::<Result<Vec<_>>>()?;
 
+        // Deletes the leader region from registry.
+        let region_ids = operating_leader_regions(region_routes);
+        ctx.leader_region_registry
+            .batch_delete(region_ids.into_iter().map(|(region_id, _)| region_id));
+
         Ok(())
     }
 }

src/common/meta/src/ddl/table_meta.rs

@@ -98,13 +98,14 @@ impl TableMetadataAllocator {
     fn create_wal_options(
         &self,
         table_route: &PhysicalTableRouteValue,
+        skip_wal: bool,
     ) -> Result<HashMap<RegionNumber, String>> {
         let region_numbers = table_route
             .region_routes
             .iter()
             .map(|route| route.region.id.region_number())
             .collect();
-        allocate_region_wal_options(region_numbers, &self.wal_options_allocator)
+        allocate_region_wal_options(region_numbers, &self.wal_options_allocator, skip_wal)
     }
 
     async fn create_table_route(
@@ -158,7 +159,9 @@ impl TableMetadataAllocator {
     pub async fn create(&self, task: &CreateTableTask) -> Result<TableMetadata> {
         let table_id = self.allocate_table_id(&task.create_table.table_id).await?;
         let table_route = self.create_table_route(table_id, task).await?;
-        let region_wal_options = self.create_wal_options(&table_route)?;
+
+        let region_wal_options =
+            self.create_wal_options(&table_route, task.table_info.meta.options.skip_wal)?;
 
         debug!(
             "Allocated region wal options {:?} for table {}",

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

@@ -80,7 +80,13 @@ pub async fn create_logical_table(
     let tasks = vec![test_create_logical_table_task(table_name)];
     let mut procedure = CreateLogicalTablesProcedure::new(tasks, physical_table_id, ddl_context);
     let status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
     let status = procedure.on_create_metadata().await.unwrap();
     assert_matches!(status, Status::Done { .. });
 

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

@@ -171,3 +171,74 @@ impl MockDatanodeHandler for NaiveDatanodeHandler {
         unreachable!()
     }
 }
+
+#[derive(Clone)]
+pub struct PartialSuccessDatanodeHandler {
+    pub retryable: bool,
+}
+
+#[async_trait::async_trait]
+impl MockDatanodeHandler for PartialSuccessDatanodeHandler {
+    async fn handle(&self, peer: &Peer, _request: RegionRequest) -> Result<RegionResponse> {
+        let success = peer.id % 2 == 0;
+        if success {
+            Ok(RegionResponse::new(0))
+        } else if self.retryable {
+            Err(Error::RetryLater {
+                source: BoxedError::new(
+                    error::UnexpectedSnafu {
+                        err_msg: "retry later",
+                    }
+                    .build(),
+                ),
+            })
+        } else {
+            error::UnexpectedSnafu {
+                err_msg: "mock error",
+            }
+            .fail()
+        }
+    }
+
+    async fn handle_query(
+        &self,
+        _peer: &Peer,
+        _request: QueryRequest,
+    ) -> Result<SendableRecordBatchStream> {
+        unreachable!()
+    }
+}
+
+#[derive(Clone)]
+pub struct AllFailureDatanodeHandler {
+    pub retryable: bool,
+}
+
+#[async_trait::async_trait]
+impl MockDatanodeHandler for AllFailureDatanodeHandler {
+    async fn handle(&self, _peer: &Peer, _request: RegionRequest) -> Result<RegionResponse> {
+        if self.retryable {
+            Err(Error::RetryLater {
+                source: BoxedError::new(
+                    error::UnexpectedSnafu {
+                        err_msg: "retry later",
+                    }
+                    .build(),
+                ),
+            })
+        } else {
+            error::UnexpectedSnafu {
+                err_msg: "mock error",
+            }
+            .fail()
+        }
+    }
+
+    async fn handle_query(
+        &self,
+        _peer: &Peer,
+        _request: QueryRequest,
+    ) -> Result<SendableRecordBatchStream> {
+        unreachable!()
+    }
+}

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

@@ -180,7 +180,13 @@ async fn test_on_prepare() {
 
     let mut procedure = AlterLogicalTablesProcedure::new(tasks, phy_id, ddl_context);
     let result = procedure.on_prepare().await;
-    assert_matches!(result, Ok(Status::Executing { persist: true }));
+    assert_matches!(
+        result,
+        Ok(Status::Executing {
+            persist: true,
+            clean_poisons: false
+        })
+    );
 }
 
 #[tokio::test]
@@ -205,7 +211,13 @@ async fn test_on_update_metadata() {
 
     let mut procedure = AlterLogicalTablesProcedure::new(tasks, phy_id, ddl_context);
     let mut status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
 
     let ctx = common_procedure::Context {
         procedure_id: ProcedureId::random(),
@@ -213,10 +225,22 @@ async fn test_on_update_metadata() {
     };
     // on_submit_alter_region_requests
     status = procedure.execute(&ctx).await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
     // on_update_metadata
     status = procedure.execute(&ctx).await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
 }
 
 #[tokio::test]
@@ -237,7 +261,13 @@ async fn test_on_part_duplicate_alter_request() {
 
     let mut procedure = AlterLogicalTablesProcedure::new(tasks, phy_id, ddl_context.clone());
     let mut status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
 
     let ctx = common_procedure::Context {
         procedure_id: ProcedureId::random(),
@@ -245,10 +275,22 @@ async fn test_on_part_duplicate_alter_request() {
     };
     // on_submit_alter_region_requests
     status = procedure.execute(&ctx).await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
     // on_update_metadata
     status = procedure.execute(&ctx).await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
 
     // re-alter
     let tasks = vec![
@@ -270,7 +312,13 @@ async fn test_on_part_duplicate_alter_request() {
 
     let mut procedure = AlterLogicalTablesProcedure::new(tasks, phy_id, ddl_context.clone());
     let mut status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
 
     let ctx = common_procedure::Context {
         procedure_id: ProcedureId::random(),
@@ -278,10 +326,22 @@ async fn test_on_part_duplicate_alter_request() {
     };
     // on_submit_alter_region_requests
     status = procedure.execute(&ctx).await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
     // on_update_metadata
     status = procedure.execute(&ctx).await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
 
     let table_name_keys = vec![
         TableNameKey::new(DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME, "table1"),

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

@@ -25,6 +25,9 @@ use api::v1::{
 use common_catalog::consts::{DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
 use common_error::ext::ErrorExt;
 use common_error::status_code::StatusCode;
+use common_procedure::store::poison_store::PoisonStore;
+use common_procedure::{ProcedureId, Status};
+use common_procedure_test::MockContextProvider;
 use store_api::storage::RegionId;
 use table::requests::TTL_KEY;
 use tokio::sync::mpsc::{self};
@@ -33,16 +36,46 @@ use crate::ddl::alter_table::AlterTableProcedure;
 use crate::ddl::test_util::alter_table::TestAlterTableExprBuilder;
 use crate::ddl::test_util::create_table::test_create_table_task;
 use crate::ddl::test_util::datanode_handler::{
-    DatanodeWatcher, RequestOutdatedErrorDatanodeHandler,
+    AllFailureDatanodeHandler, DatanodeWatcher, PartialSuccessDatanodeHandler,
+    RequestOutdatedErrorDatanodeHandler,
 };
+use crate::error::Error;
 use crate::key::datanode_table::DatanodeTableKey;
 use crate::key::table_name::TableNameKey;
 use crate::key::table_route::TableRouteValue;
+use crate::node_manager::NodeManagerRef;
 use crate::peer::Peer;
+use crate::poison_key::table_poison_key;
 use crate::rpc::ddl::AlterTableTask;
 use crate::rpc::router::{Region, RegionRoute};
 use crate::test_util::{new_ddl_context, MockDatanodeManager};
 
+fn prepare_table_route(table_id: u32) -> TableRouteValue {
+    TableRouteValue::physical(vec![
+        RegionRoute {
+            region: Region::new_test(RegionId::new(table_id, 1)),
+            leader_peer: Some(Peer::empty(1)),
+            follower_peers: vec![Peer::empty(5)],
+            leader_state: None,
+            leader_down_since: None,
+        },
+        RegionRoute {
+            region: Region::new_test(RegionId::new(table_id, 2)),
+            leader_peer: Some(Peer::empty(2)),
+            follower_peers: vec![Peer::empty(4)],
+            leader_state: None,
+            leader_down_since: None,
+        },
+        RegionRoute {
+            region: Region::new_test(RegionId::new(table_id, 3)),
+            leader_peer: Some(Peer::empty(3)),
+            follower_peers: vec![],
+            leader_state: None,
+            leader_down_since: None,
+        },
+    ])
+}
+
 fn test_rename_alter_table_task(table_name: &str, new_table_name: &str) -> AlterTableTask {
     let builder = TestAlterTableExprBuilder::default()
         .table_name(table_name)
@@ -101,29 +134,7 @@ async fn test_on_submit_alter_request() {
         .table_metadata_manager
         .create_table_metadata(
             task.table_info.clone(),
-            TableRouteValue::physical(vec![
-                RegionRoute {
-                    region: Region::new_test(RegionId::new(table_id, 1)),
-                    leader_peer: Some(Peer::empty(1)),
-                    follower_peers: vec![Peer::empty(5)],
-                    leader_state: None,
-                    leader_down_since: None,
-                },
-                RegionRoute {
-                    region: Region::new_test(RegionId::new(table_id, 2)),
-                    leader_peer: Some(Peer::empty(2)),
-                    follower_peers: vec![Peer::empty(4)],
-                    leader_state: None,
-                    leader_down_since: None,
-                },
-                RegionRoute {
-                    region: Region::new_test(RegionId::new(table_id, 3)),
-                    leader_peer: Some(Peer::empty(3)),
-                    follower_peers: vec![],
-                    leader_state: None,
-                    leader_down_since: None,
-                },
-            ]),
+            prepare_table_route(table_id),
             HashMap::new(),
         )
         .await
@@ -141,9 +152,15 @@ async fn test_on_submit_alter_request() {
             })),
         },
     };
-    let mut procedure = AlterTableProcedure::new(table_id, alter_table_task, ddl_context).unwrap();
+    let procedure_id = ProcedureId::random();
+    let provider = Arc::new(MockContextProvider::default());
+    let mut procedure =
+        AlterTableProcedure::new(table_id, alter_table_task, ddl_context.clone()).unwrap();
     procedure.on_prepare().await.unwrap();
-    procedure.submit_alter_region_requests().await.unwrap();
+    procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap();
 
     let check = |peer: Peer,
                  request: RegionRequest,
@@ -185,29 +202,7 @@ async fn test_on_submit_alter_request_with_outdated_request() {
         .table_metadata_manager
         .create_table_metadata(
             task.table_info.clone(),
-            TableRouteValue::physical(vec![
-                RegionRoute {
-                    region: Region::new_test(RegionId::new(table_id, 1)),
-                    leader_peer: Some(Peer::empty(1)),
-                    follower_peers: vec![Peer::empty(5)],
-                    leader_state: None,
-                    leader_down_since: None,
-                },
-                RegionRoute {
-                    region: Region::new_test(RegionId::new(table_id, 2)),
-                    leader_peer: Some(Peer::empty(2)),
-                    follower_peers: vec![Peer::empty(4)],
-                    leader_state: None,
-                    leader_down_since: None,
-                },
-                RegionRoute {
-                    region: Region::new_test(RegionId::new(table_id, 3)),
-                    leader_peer: Some(Peer::empty(3)),
-                    follower_peers: vec![],
-                    leader_state: None,
-                    leader_down_since: None,
-                },
-            ]),
+            prepare_table_route(table_id),
             HashMap::new(),
         )
         .await
@@ -225,9 +220,15 @@ async fn test_on_submit_alter_request_with_outdated_request() {
             })),
         },
     };
+    let procedure_id = ProcedureId::random();
+    let provider = Arc::new(MockContextProvider::default());
     let mut procedure = AlterTableProcedure::new(table_id, alter_table_task, ddl_context).unwrap();
     procedure.on_prepare().await.unwrap();
-    procedure.submit_alter_region_requests().await.unwrap();
+    let err = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap_err();
+    assert!(!err.is_retry_later());
 }
 
 #[tokio::test]
@@ -326,9 +327,14 @@ async fn test_on_update_metadata_add_columns() {
             })),
         },
     };
+    let procedure_id = ProcedureId::random();
+    let provider = Arc::new(MockContextProvider::default());
     let mut procedure = AlterTableProcedure::new(table_id, task, ddl_context.clone()).unwrap();
     procedure.on_prepare().await.unwrap();
-    procedure.submit_alter_region_requests().await.unwrap();
+    procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap();
     procedure.on_update_metadata().await.unwrap();
 
     let table_info = ddl_context
@@ -387,9 +393,14 @@ async fn test_on_update_table_options() {
             })),
         },
     };
+    let procedure_id = ProcedureId::random();
+    let provider = Arc::new(MockContextProvider::default());
     let mut procedure = AlterTableProcedure::new(table_id, task, ddl_context.clone()).unwrap();
     procedure.on_prepare().await.unwrap();
-    procedure.submit_alter_region_requests().await.unwrap();
+    procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap();
     procedure.on_update_metadata().await.unwrap();
 
     let table_info = ddl_context
@@ -417,3 +428,156 @@ async fn test_on_update_table_options() {
         HashMap::from(&table_info.meta.options)
     );
 }
+
+async fn prepare_alter_table_procedure(
+    node_manager: NodeManagerRef,
+) -> (AlterTableProcedure, ProcedureId) {
+    common_telemetry::init_default_ut_logging();
+    let ddl_context = new_ddl_context(node_manager);
+    let table_id = 1024;
+    let table_name = "foo";
+    let task = test_create_table_task(table_name, table_id);
+    // Puts a value to table name key.
+    ddl_context
+        .table_metadata_manager
+        .create_table_metadata(
+            task.table_info.clone(),
+            prepare_table_route(table_id),
+            HashMap::new(),
+        )
+        .await
+        .unwrap();
+
+    let alter_table_task = AlterTableTask {
+        alter_table: AlterTableExpr {
+            catalog_name: DEFAULT_CATALOG_NAME.to_string(),
+            schema_name: DEFAULT_SCHEMA_NAME.to_string(),
+            table_name: table_name.to_string(),
+            kind: Some(Kind::DropColumns(DropColumns {
+                drop_columns: vec![DropColumn {
+                    name: "cpu".to_string(),
+                }],
+            })),
+        },
+    };
+    let procedure_id = ProcedureId::random();
+    let mut procedure =
+        AlterTableProcedure::new(table_id, alter_table_task, ddl_context.clone()).unwrap();
+    procedure.on_prepare().await.unwrap();
+    (procedure, procedure_id)
+}
+
+#[tokio::test]
+async fn test_on_submit_alter_request_with_partial_success_retryable() {
+    let node_manager = Arc::new(MockDatanodeManager::new(PartialSuccessDatanodeHandler {
+        retryable: true,
+    }));
+    let provider = Arc::new(MockContextProvider::default());
+    let (mut procedure, procedure_id) = prepare_alter_table_procedure(node_manager).await;
+    let result = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap_err();
+    assert!(result.is_retry_later());
+
+    // Submits again
+    let result = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap_err();
+    assert!(result.is_retry_later());
+}
+
+#[tokio::test]
+async fn test_on_submit_alter_request_with_partial_success_non_retryable() {
+    let node_manager = Arc::new(MockDatanodeManager::new(PartialSuccessDatanodeHandler {
+        retryable: false,
+    }));
+    let provider = Arc::new(MockContextProvider::default());
+    let (mut procedure, procedure_id) = prepare_alter_table_procedure(node_manager).await;
+    let result = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap();
+    assert_matches!(result, Status::Poisoned { .. });
+
+    // submits again
+    let result = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap();
+    assert_matches!(result, Status::Poisoned { .. });
+}
+
+#[tokio::test]
+async fn test_on_submit_alter_request_with_all_failure_retrybale() {
+    common_telemetry::init_default_ut_logging();
+    let node_manager = Arc::new(MockDatanodeManager::new(AllFailureDatanodeHandler {
+        retryable: true,
+    }));
+    let provider = Arc::new(MockContextProvider::default());
+    let (mut procedure, procedure_id) = prepare_alter_table_procedure(node_manager).await;
+    let err = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap_err();
+    assert!(err.is_retry_later());
+    // submits again
+    let err = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap_err();
+    assert!(err.is_retry_later());
+}
+
+#[tokio::test]
+async fn test_on_submit_alter_request_with_all_failure_non_retrybale() {
+    common_telemetry::init_default_ut_logging();
+    let node_manager = Arc::new(MockDatanodeManager::new(AllFailureDatanodeHandler {
+        retryable: false,
+    }));
+    let provider = Arc::new(MockContextProvider::default());
+    let (mut procedure, procedure_id) = prepare_alter_table_procedure(node_manager).await;
+    let err = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap_err();
+    assert_matches!(err, Error::AbortProcedure { .. });
+    assert!(!err.is_retry_later());
+    assert!(err.need_clean_poisons());
+
+    // submits again
+    let err = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap_err();
+    assert_matches!(err, Error::AbortProcedure { .. });
+    assert!(!err.is_retry_later());
+    assert!(err.need_clean_poisons());
+}
+
+#[tokio::test]
+async fn test_on_submit_alter_request_with_exist_poison() {
+    common_telemetry::init_default_ut_logging();
+    let node_manager = Arc::new(MockDatanodeManager::new(AllFailureDatanodeHandler {
+        retryable: false,
+    }));
+    let provider = Arc::new(MockContextProvider::default());
+    let (mut procedure, procedure_id) = prepare_alter_table_procedure(node_manager).await;
+
+    let table_id = 1024;
+    let key = table_poison_key(table_id).to_string();
+    let another_procedure_id = ProcedureId::random();
+    provider
+        .poison_manager()
+        .try_put_poison(key, another_procedure_id.to_string())
+        .await
+        .unwrap();
+
+    procedure.on_prepare().await.unwrap();
+    let err = procedure
+        .submit_alter_region_requests(procedure_id, provider.as_ref())
+        .await
+        .unwrap_err();
+    assert_matches!(err, Error::PutPoison { .. });
+}

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

@@ -69,7 +69,13 @@ async fn test_on_prepare() {
     let physical_table_id = table_id;
     let mut procedure = CreateLogicalTablesProcedure::new(tasks, physical_table_id, ddl_context);
     let status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
 }
 
 #[tokio::test]
@@ -202,7 +208,13 @@ async fn test_on_prepare_part_logical_tables_exist() {
         ddl_context,
     );
     let status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
 }
 
 #[tokio::test]
@@ -238,7 +250,13 @@ async fn test_on_create_metadata() {
         ddl_context,
     );
     let status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
     let ctx = ProcedureContext {
         procedure_id: ProcedureId::random(),
         provider: Arc::new(MockContextProvider::default()),
@@ -294,7 +312,13 @@ async fn test_on_create_metadata_part_logical_tables_exist() {
         ddl_context,
     );
     let status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
     let ctx = ProcedureContext {
         procedure_id: ProcedureId::random(),
         provider: Arc::new(MockContextProvider::default()),
@@ -339,7 +363,13 @@ async fn test_on_create_metadata_err() {
         ddl_context.clone(),
     );
     let status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
     let ctx = ProcedureContext {
         procedure_id: ProcedureId::random(),
         provider: Arc::new(MockContextProvider::default()),

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

@@ -137,7 +137,13 @@ async fn test_on_prepare_without_create_if_table_exists() {
     task.create_table.create_if_not_exists = true;
     let mut procedure = CreateTableProcedure::new(task, ddl_context);
     let status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
     assert_eq!(procedure.table_id(), 1024);
 }
 

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

@@ -153,7 +153,13 @@ async fn test_on_prepare_without_create_if_table_exists() {
     task.create_view.create_if_not_exists = true;
     let mut procedure = CreateViewProcedure::new(task, ddl_context);
     let status = procedure.on_prepare().await.unwrap();
-    assert_matches!(status, Status::Executing { persist: true });
+    assert_matches!(
+        status,
+        Status::Executing {
+            persist: true,
+            clean_poisons: false
+        }
+    );
     assert_eq!(procedure.view_id(), 1024);
 }
 

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

@@ -13,10 +13,12 @@
 // limitations under the License.
 
 use std::collections::HashMap;
+use std::fmt::Debug;
 
 use common_catalog::consts::METRIC_ENGINE;
 use common_error::ext::BoxedError;
 use common_procedure::error::Error as ProcedureError;
+use common_telemetry::{error, warn};
 use common_wal::options::WalOptions;
 use snafu::{ensure, OptionExt, ResultExt};
 use store_api::metric_engine_consts::LOGICAL_TABLE_METADATA_KEY;
@@ -38,6 +40,7 @@ use crate::rpc::router::RegionRoute;
 /// Adds [Peer] context if the error is unretryable.
 pub fn add_peer_context_if_needed(datanode: Peer) -> impl FnOnce(Error) -> Error {
     move |err| {
+        error!(err; "Failed to operate datanode, peer: {}", datanode);
         if !err.is_retry_later() {
             return Err::<(), BoxedError>(BoxedError::new(err))
                 .context(OperateDatanodeSnafu { peer: datanode })
@@ -182,6 +185,85 @@ pub fn extract_region_wal_options(
     Ok(region_wal_options)
 }
 
+/// The result of multiple operations.
+///
+/// - Ok: all operations are successful.
+/// - PartialRetryable: if any operation is retryable and without non retryable error, the result is retryable.
+/// - PartialNonRetryable: if any operation is non retryable, the result is non retryable.
+/// - AllRetryable: all operations are retryable.
+/// - AllNonRetryable: all operations are not retryable.
+pub enum MultipleResults {
+    Ok,
+    PartialRetryable(Error),
+    PartialNonRetryable(Error),
+    AllRetryable(Error),
+    AllNonRetryable(Error),
+}
+
+/// Handles the results of alter region requests.
+///
+/// For partial success, we need to check if the errors are retryable.
+/// If all the errors are retryable, we return a retryable error.
+/// Otherwise, we return the first error.
+pub fn handle_multiple_results<T: Debug>(results: Vec<Result<T>>) -> MultipleResults {
+    if results.is_empty() {
+        return MultipleResults::Ok;
+    }
+    let num_results = results.len();
+    let mut retryable_results = Vec::new();
+    let mut non_retryable_results = Vec::new();
+    let mut ok_results = Vec::new();
+
+    for result in results {
+        match result {
+            Ok(_) => ok_results.push(result),
+            Err(err) => {
+                if err.is_retry_later() {
+                    retryable_results.push(err);
+                } else {
+                    non_retryable_results.push(err);
+                }
+            }
+        }
+    }
+
+    common_telemetry::debug!(
+        "retryable_results: {}, non_retryable_results: {}, ok_results: {}",
+        retryable_results.len(),
+        non_retryable_results.len(),
+        ok_results.len()
+    );
+
+    if retryable_results.len() == num_results {
+        return MultipleResults::AllRetryable(retryable_results.into_iter().next().unwrap());
+    } else if non_retryable_results.len() == num_results {
+        warn!("all non retryable results: {}", non_retryable_results.len());
+        for err in &non_retryable_results {
+            error!(err; "non retryable error");
+        }
+        return MultipleResults::AllNonRetryable(non_retryable_results.into_iter().next().unwrap());
+    } else if ok_results.len() == num_results {
+        return MultipleResults::Ok;
+    } else if !retryable_results.is_empty()
+        && !ok_results.is_empty()
+        && non_retryable_results.is_empty()
+    {
+        return MultipleResults::PartialRetryable(retryable_results.into_iter().next().unwrap());
+    }
+
+    warn!(
+        "partial non retryable results: {}, retryable results: {}, ok results: {}",
+        non_retryable_results.len(),
+        retryable_results.len(),
+        ok_results.len()
+    );
+    for err in &non_retryable_results {
+        error!(err; "non retryable error");
+    }
+    // non_retryable_results.len() > 0
+    MultipleResults::PartialNonRetryable(non_retryable_results.into_iter().next().unwrap())
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

src/common/meta/src/ddl_manager.rs

@@ -834,6 +834,7 @@ mod tests {
     use std::sync::Arc;
 
     use common_procedure::local::LocalManager;
+    use common_procedure::test_util::InMemoryPoisonStore;
 
     use super::DdlManager;
     use crate::cache_invalidator::DummyCacheInvalidator;
@@ -850,6 +851,7 @@ mod tests {
     use crate::node_manager::{DatanodeRef, FlownodeRef, NodeManager};
     use crate::peer::Peer;
     use crate::region_keeper::MemoryRegionKeeper;
+    use crate::region_registry::LeaderRegionRegistry;
     use crate::sequence::SequenceBuilder;
     use crate::state_store::KvStateStore;
     use crate::wal_options_allocator::WalOptionsAllocator;
@@ -882,7 +884,12 @@ mod tests {
         ));
 
         let state_store = Arc::new(KvStateStore::new(kv_backend.clone()));
-        let procedure_manager = Arc::new(LocalManager::new(Default::default(), state_store));
+        let poison_manager = Arc::new(InMemoryPoisonStore::default());
+        let procedure_manager = Arc::new(LocalManager::new(
+            Default::default(),
+            state_store,
+            poison_manager,
+        ));
 
         let _ = DdlManager::try_new(
             DdlContext {
@@ -893,6 +900,7 @@ mod tests {
                 flow_metadata_manager,
                 flow_metadata_allocator,
                 memory_region_keeper: Arc::new(MemoryRegionKeeper::default()),
+                leader_region_registry: Arc::new(LeaderRegionRegistry::default()),
                 region_failure_detector_controller: Arc::new(NoopRegionFailureDetectorControl),
             },
             procedure_manager.clone(),

src/common/meta/src/error.rs

@@ -449,6 +449,14 @@ pub enum Error {
     #[snafu(display("Retry later"))]
     RetryLater { source: BoxedError },
 
+    #[snafu(display("Abort procedure"))]
+    AbortProcedure {
+        #[snafu(implicit)]
+        location: Location,
+        source: BoxedError,
+        clean_poisons: bool,
+    },
+
     #[snafu(display(
         "Failed to encode a wal options to json string, wal_options: {:?}",
         wal_options
@@ -685,7 +693,36 @@ pub enum Error {
         operation: String,
     },
 
-    #[cfg(feature = "pg_kvbackend")]
+    #[cfg(feature = "mysql_kvbackend")]
+    #[snafu(display("Failed to execute via MySql, sql: {}", sql))]
+    MySqlExecution {
+        sql: String,
+        #[snafu(source)]
+        error: sqlx::Error,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
+    #[cfg(feature = "mysql_kvbackend")]
+    #[snafu(display("Failed to create connection pool for MySql"))]
+    CreateMySqlPool {
+        #[snafu(source)]
+        error: sqlx::Error,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
+    #[cfg(feature = "mysql_kvbackend")]
+    #[snafu(display("Failed to {} MySql transaction", operation))]
+    MySqlTransaction {
+        #[snafu(source)]
+        error: sqlx::Error,
+        #[snafu(implicit)]
+        location: Location,
+        operation: String,
+    },
+
+    #[cfg(any(feature = "pg_kvbackend", feature = "mysql_kvbackend"))]
     #[snafu(display("Rds transaction retry failed"))]
     RdsTransactionRetryFailed {
         #[snafu(implicit)]
@@ -719,6 +756,33 @@ pub enum Error {
         #[snafu(source)]
         error: serde_json::Error,
     },
+
+    #[snafu(display("No leader found for table_id: {}", table_id))]
+    NoLeader {
+        table_id: TableId,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
+    #[snafu(display(
+        "Procedure poison key already exists with a different value, key: {}, value: {}",
+        key,
+        value
+    ))]
+    ProcedurePoisonConflict {
+        key: String,
+        value: String,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
+    #[snafu(display("Failed to put poison, table metadata may be corrupted"))]
+    PutPoison {
+        #[snafu(implicit)]
+        location: Location,
+        #[snafu(source)]
+        source: common_procedure::error::Error,
+    },
 }
 
 pub type Result<T> = std::result::Result<T, Error>;
@@ -737,7 +801,8 @@ impl ErrorExt for Error {
             | SerializeToJson { .. }
             | DeserializeFromJson { .. } => StatusCode::Internal,
 
-            ValueNotExist { .. } => StatusCode::Unexpected,
+            NoLeader { .. } => StatusCode::TableUnavailable,
+            ValueNotExist { .. } | ProcedurePoisonConflict { .. } => StatusCode::Unexpected,
 
             Unsupported { .. } => StatusCode::Unsupported,
 
@@ -808,7 +873,9 @@ impl ErrorExt for Error {
             OperateDatanode { source, .. } => source.status_code(),
             Table { source, .. } => source.status_code(),
             RetryLater { source, .. } => source.status_code(),
+            AbortProcedure { source, .. } => source.status_code(),
             ConvertAlterTableRequest { source, .. } => source.status_code(),
+            PutPoison { source, .. } => source.status_code(),
 
             ParseProcedureId { .. }
             | InvalidNumTopics { .. }
@@ -823,8 +890,13 @@ impl ErrorExt for Error {
             PostgresExecution { .. }
             | CreatePostgresPool { .. }
             | GetPostgresConnection { .. }
-            | PostgresTransaction { .. }
-            | RdsTransactionRetryFailed { .. } => StatusCode::Internal,
+            | PostgresTransaction { .. } => StatusCode::Internal,
+            #[cfg(feature = "mysql_kvbackend")]
+            MySqlExecution { .. } | CreateMySqlPool { .. } | MySqlTransaction { .. } => {
+                StatusCode::Internal
+            }
+            #[cfg(any(feature = "pg_kvbackend", feature = "mysql_kvbackend"))]
+            RdsTransactionRetryFailed { .. } => StatusCode::Internal,
             Error::DatanodeTableInfoNotFound { .. } => StatusCode::Internal,
         }
     }
@@ -835,16 +907,29 @@ impl ErrorExt for Error {
 }
 
 impl Error {
-    #[cfg(feature = "pg_kvbackend")]
+    #[cfg(any(feature = "pg_kvbackend", feature = "mysql_kvbackend"))]
     /// Check if the error is a serialization error.
     pub fn is_serialization_error(&self) -> bool {
         match self {
+            #[cfg(feature = "pg_kvbackend")]
             Error::PostgresTransaction { error, .. } => {
                 error.code() == Some(&tokio_postgres::error::SqlState::T_R_SERIALIZATION_FAILURE)
             }
+            #[cfg(feature = "pg_kvbackend")]
             Error::PostgresExecution { error, .. } => {
                 error.code() == Some(&tokio_postgres::error::SqlState::T_R_SERIALIZATION_FAILURE)
             }
+            #[cfg(feature = "mysql_kvbackend")]
+            Error::MySqlExecution {
+                error: sqlx::Error::Database(database_error),
+                ..
+            } => {
+                matches!(
+                    database_error.message(),
+                    "Deadlock found when trying to get lock; try restarting transaction"
+                        | "can't serialize access for this transaction"
+                )
+            }
             _ => false,
         }
     }
@@ -861,6 +946,11 @@ impl Error {
         matches!(self, Error::RetryLater { .. })
     }
 
+    /// Determine whether it needs to clean poisons.
+    pub fn need_clean_poisons(&self) -> bool {
+        matches!(self, Error::AbortProcedure { clean_poisons, .. } if *clean_poisons)
+    }
+
     /// Returns true if the response exceeds the size limit.
     pub fn is_exceeded_size_limit(&self) -> bool {
         match self {

src/common/meta/src/instruction.rs

@@ -192,6 +192,12 @@ pub struct DropFlow {
     pub flownode_ids: Vec<FlownodeId>,
 }
 
+/// Flushes a batch of regions.
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
+pub struct FlushRegions {
+    pub region_ids: Vec<RegionId>,
+}
+
 #[derive(Debug, Clone, Serialize, Deserialize, Display, PartialEq)]
 pub enum Instruction {
     /// Opens a region.
@@ -208,6 +214,8 @@ pub enum Instruction {
     DowngradeRegion(DowngradeRegion),
     /// Invalidates batch cache.
     InvalidateCaches(Vec<CacheIdent>),
+    /// Flushes regions.
+    FlushRegion(FlushRegions),
 }
 
 /// The reply of [UpgradeRegion].

src/common/meta/src/key.rs

@@ -57,7 +57,10 @@
 //!     - This key is mainly used in constructing the view in Datanode and Frontend.
 //!
 //! 12. Kafka topic key: `__topic_name/kafka/{topic_name}`
-//!     - The key is used to mark existing topics in kafka for WAL.
+//!     - The key is used to track existing topics in Kafka.
+//!     - The value is a [TopicNameValue](crate::key::topic_name::TopicNameValue) struct; it contains the `pruned_entry_id` which represents
+//!       the highest entry id that has been pruned from the remote WAL.
+//!     - When a region uses this topic, it should start replaying entries from `pruned_entry_id + 1` (minimum available entry id).
 //!
 //! 13. Topic name to region map key `__topic_region/{topic_name}/{region_id}`
 //!     - Mapping {topic_name} to {region_id}
@@ -137,6 +140,7 @@ use table::metadata::{RawTableInfo, TableId};
 use table::table_name::TableName;
 use table_info::{TableInfoKey, TableInfoManager, TableInfoValue};
 use table_name::{TableNameKey, TableNameManager, TableNameValue};
+use topic_name::TopicNameManager;
 use topic_region::{TopicRegionKey, TopicRegionManager};
 use view_info::{ViewInfoKey, ViewInfoManager, ViewInfoValue};
 
@@ -156,6 +160,7 @@ use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
 use crate::rpc::router::{region_distribution, LeaderState, RegionRoute};
 use crate::rpc::store::BatchDeleteRequest;
+use crate::state_store::PoisonValue;
 use crate::DatanodeId;
 
 pub const NAME_PATTERN: &str = r"[a-zA-Z_:-][a-zA-Z0-9_:\-\.@#]*";
@@ -308,6 +313,7 @@ pub struct TableMetadataManager {
     schema_manager: SchemaManager,
     table_route_manager: TableRouteManager,
     tombstone_manager: TombstoneManager,
+    topic_name_manager: TopicNameManager,
     topic_region_manager: TopicRegionManager,
     kv_backend: KvBackendRef,
 }
@@ -459,6 +465,7 @@ impl TableMetadataManager {
             schema_manager: SchemaManager::new(kv_backend.clone()),
             table_route_manager: TableRouteManager::new(kv_backend.clone()),
             tombstone_manager: TombstoneManager::new(kv_backend.clone()),
+            topic_name_manager: TopicNameManager::new(kv_backend.clone()),
             topic_region_manager: TopicRegionManager::new(kv_backend.clone()),
             kv_backend,
         }
@@ -512,6 +519,14 @@ impl TableMetadataManager {
         &self.table_route_manager
     }
 
+    pub fn topic_name_manager(&self) -> &TopicNameManager {
+        &self.topic_name_manager
+    }
+
+    pub fn topic_region_manager(&self) -> &TopicRegionManager {
+        &self.topic_region_manager
+    }
+
     #[cfg(feature = "testing")]
     pub fn kv_backend(&self) -> &KvBackendRef {
         &self.kv_backend
@@ -1320,7 +1335,8 @@ impl_metadata_value! {
     TableFlowValue,
     NodeAddressValue,
     SchemaNameValue,
-    FlowStateValue
+    FlowStateValue,
+    PoisonValue
 }
 
 impl_optional_metadata_value! {
@@ -1471,7 +1487,8 @@ mod tests {
             new_test_table_info(region_routes.iter().map(|r| r.region.id.region_number())).into();
         let wal_allocator = WalOptionsAllocator::RaftEngine;
         let regions = (0..16).collect();
-        let region_wal_options = allocate_region_wal_options(regions, &wal_allocator).unwrap();
+        let region_wal_options =
+            allocate_region_wal_options(regions, &wal_allocator, false).unwrap();
         create_physical_table_metadata(
             &table_metadata_manager,
             table_info.clone(),

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

@@ -461,6 +461,8 @@ mod tests {
             expire_after: Some(300),
             comment: "hi".to_string(),
             options: Default::default(),
+            created_time: chrono::Utc::now(),
+            updated_time: chrono::Utc::now(),
         }
     }
 
@@ -632,6 +634,8 @@ mod tests {
             expire_after: Some(300),
             comment: "hi".to_string(),
             options: Default::default(),
+            created_time: chrono::Utc::now(),
+            updated_time: chrono::Utc::now(),
         };
         let err = flow_metadata_manager
             .create_flow_metadata(flow_id, flow_value, flow_routes.clone())
@@ -869,6 +873,8 @@ mod tests {
             expire_after: Some(300),
             comment: "hi".to_string(),
             options: Default::default(),
+            created_time: chrono::Utc::now(),
+            updated_time: chrono::Utc::now(),
         };
         let err = flow_metadata_manager
             .update_flow_metadata(

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

@@ -15,6 +15,7 @@
 use std::collections::{BTreeMap, HashMap};
 use std::sync::Arc;
 
+use chrono::{DateTime, Utc};
 use lazy_static::lazy_static;
 use regex::Regex;
 use serde::{Deserialize, Serialize};
@@ -131,6 +132,12 @@ pub struct FlowInfoValue {
     pub(crate) comment: String,
     /// The options.
     pub(crate) options: HashMap<String, String>,
+    /// The created time
+    #[serde(default)]
+    pub(crate) created_time: DateTime<Utc>,
+    /// The updated time.
+    #[serde(default)]
+    pub(crate) updated_time: DateTime<Utc>,
 }
 
 impl FlowInfoValue {
@@ -171,6 +178,14 @@ impl FlowInfoValue {
     pub fn options(&self) -> &HashMap<String, String> {
         &self.options
     }
+
+    pub fn created_time(&self) -> &DateTime<Utc> {
+        &self.created_time
+    }
+
+    pub fn updated_time(&self) -> &DateTime<Utc> {
+        &self.updated_time
+    }
 }
 
 pub type FlowInfoManagerRef = Arc<FlowInfoManager>;

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

@@ -97,11 +97,19 @@ impl<'a> MetadataKey<'a, FlowStateKey> for FlowStateKey {
 pub struct FlowStateValue {
     /// For each key, the bytes of the state in memory
     pub state_size: BTreeMap<FlowId, usize>,
+    /// For each key, the last execution time of flow in unix timestamp milliseconds.
+    pub last_exec_time_map: BTreeMap<FlowId, i64>,
 }
 
 impl FlowStateValue {
-    pub fn new(state_size: BTreeMap<FlowId, usize>) -> Self {
-        Self { state_size }
+    pub fn new(
+        state_size: BTreeMap<FlowId, usize>,
+        last_exec_time_map: BTreeMap<FlowId, i64>,
+    ) -> Self {
+        Self {
+            state_size,
+            last_exec_time_map,
+        }
     }
 }
 
@@ -143,12 +151,15 @@ impl FlowStateManager {
 pub struct FlowStat {
     /// For each key, the bytes of the state in memory
     pub state_size: BTreeMap<u32, usize>,
+    /// For each key, the last execution time of flow in unix timestamp milliseconds.
+    pub last_exec_time_map: BTreeMap<FlowId, i64>,
 }
 
 impl From<FlowStateValue> for FlowStat {
     fn from(value: FlowStateValue) -> Self {
         Self {
             state_size: value.state_size,
+            last_exec_time_map: value.last_exec_time_map,
         }
     }
 }
@@ -157,6 +168,7 @@ impl From<FlowStat> for FlowStateValue {
     fn from(value: FlowStat) -> Self {
         Self {
             state_size: value.state_size,
+            last_exec_time_map: value.last_exec_time_map,
         }
     }
 }

src/common/meta/src/key/test_utils.rs

@@ -40,7 +40,7 @@ pub fn new_test_table_info_with_name<I: IntoIterator<Item = u32>>(
         .build()
         .unwrap();
 
-    let meta = TableMetaBuilder::default()
+    let meta = TableMetaBuilder::empty()
         .schema(Arc::new(schema))
         .primary_key_indices(vec![0])
         .engine("engine")

src/common/meta/src/key/topic_name.rs

@@ -15,11 +15,14 @@
 use std::fmt::{self, Display};
 
 use serde::{Deserialize, Serialize};
-use snafu::{OptionExt, ResultExt};
+use snafu::{ensure, OptionExt, ResultExt};
 
-use crate::error::{DecodeJsonSnafu, Error, InvalidMetadataSnafu, Result};
+use crate::ensure_values;
+use crate::error::{self, DecodeJsonSnafu, Error, InvalidMetadataSnafu, Result, UnexpectedSnafu};
+use crate::key::txn_helper::TxnOpGetResponseSet;
 use crate::key::{
-    MetadataKey, KAFKA_TOPIC_KEY_PATTERN, KAFKA_TOPIC_KEY_PREFIX, LEGACY_TOPIC_KEY_PREFIX,
+    DeserializedValueWithBytes, MetadataKey, MetadataValue, KAFKA_TOPIC_KEY_PATTERN,
+    KAFKA_TOPIC_KEY_PREFIX, LEGACY_TOPIC_KEY_PREFIX,
 };
 use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
@@ -31,8 +34,32 @@ pub struct TopicNameKey<'a> {
     pub topic: &'a str,
 }
 
-#[derive(Debug, Serialize, Deserialize)]
-pub struct TopicNameValue;
+/// The value associated with a topic name key.
+///
+/// The `pruned_entry_id` is the highest entry id that has been pruned from the remote WAL.
+/// When a region uses this topic, it should start replaying entries from `pruned_entry_id + 1` (minimal available entry id).
+#[derive(Debug, Serialize, Deserialize, Default, Clone)]
+pub struct TopicNameValue {
+    pub pruned_entry_id: u64,
+}
+
+impl TopicNameValue {
+    pub fn new(pruned_entry_id: u64) -> Self {
+        Self { pruned_entry_id }
+    }
+}
+
+impl MetadataValue for TopicNameValue {
+    fn try_from_raw_value(raw_value: &[u8]) -> Result<Self> {
+        let value = serde_json::from_slice::<TopicNameValue>(raw_value).context(DecodeJsonSnafu)?;
+        Ok(value)
+    }
+
+    fn try_as_raw_value(&self) -> Result<Vec<u8>> {
+        let raw_value = serde_json::to_vec(self).context(DecodeJsonSnafu)?;
+        Ok(raw_value)
+    }
+}
 
 impl<'a> TopicNameKey<'a> {
     pub fn new(topic: &'a str) -> Self {
@@ -114,13 +141,16 @@ impl TopicNameManager {
         {
             let topics =
                 serde_json::from_slice::<Vec<String>>(&kv.value).context(DecodeJsonSnafu)?;
-            let mut reqs = topics
-                .iter()
-                .map(|topic| {
-                    let key = TopicNameKey::new(topic);
-                    TxnOp::Put(key.to_bytes(), vec![])
-                })
-                .collect::<Vec<_>>();
+            let mut reqs = Vec::with_capacity(topics.len() + 1);
+            for topic in topics {
+                let topic_name_key = TopicNameKey::new(&topic);
+                let topic_name_value = TopicNameValue::new(0);
+                let put_req = TxnOp::Put(
+                    topic_name_key.to_bytes(),
+                    topic_name_value.try_as_raw_value()?,
+                );
+                reqs.push(put_req);
+            }
             let delete_req = TxnOp::Delete(LEGACY_TOPIC_KEY_PREFIX.as_bytes().to_vec());
             reqs.push(delete_req);
             let txn = Txn::new().and_then(reqs);
@@ -129,7 +159,7 @@ impl TopicNameManager {
         Ok(())
     }
 
-    /// Range query for topics.
+    /// Range query for topics. Only the keys are returned.
     /// Caution: this method returns keys as String instead of values of range query since the topics are stored in keys.
     pub async fn range(&self) -> Result<Vec<String>> {
         let prefix = TopicNameKey::range_start_key();
@@ -142,25 +172,72 @@ impl TopicNameManager {
             .collect::<Result<Vec<String>>>()
     }
 
-    /// Put topics into kvbackend.
+    /// Put topics into kvbackend. The value is set to 0 by default.
     pub async fn batch_put(&self, topic_name_keys: Vec<TopicNameKey<'_>>) -> Result<()> {
+        let mut kvs = Vec::with_capacity(topic_name_keys.len());
+        let topic_name_value = TopicNameValue::new(0);
+        for topic_name_key in &topic_name_keys {
+            let kv = KeyValue {
+                key: topic_name_key.to_bytes(),
+                value: topic_name_value.clone().try_as_raw_value()?,
+            };
+            kvs.push(kv);
+        }
         let req = BatchPutRequest {
-            kvs: topic_name_keys
-                .iter()
-                .map(|key| KeyValue {
-                    key: key.to_bytes(),
-                    value: vec![],
-                })
-                .collect(),
+            kvs,
             prev_kv: false,
         };
         self.kv_backend.batch_put(req).await?;
         Ok(())
     }
+
+    /// Get value for a specific topic.
+    pub async fn get(
+        &self,
+        topic: &str,
+    ) -> Result<Option<DeserializedValueWithBytes<TopicNameValue>>> {
+        let key = TopicNameKey::new(topic);
+        let raw_key = key.to_bytes();
+        self.kv_backend
+            .get(&raw_key)
+            .await?
+            .map(|x| DeserializedValueWithBytes::from_inner_slice(&x.value))
+            .transpose()
+    }
+
+    /// Update the topic name key and value in the kv backend.
+    pub async fn update(
+        &self,
+        topic: &str,
+        pruned_entry_id: u64,
+        prev: Option<DeserializedValueWithBytes<TopicNameValue>>,
+    ) -> Result<()> {
+        let key = TopicNameKey::new(topic);
+        let raw_key = key.to_bytes();
+        let value = TopicNameValue::new(pruned_entry_id);
+        let new_raw_value = value.try_as_raw_value()?;
+        let raw_value = prev.map(|v| v.get_raw_bytes()).unwrap_or_default();
+
+        let txn = Txn::compare_and_put(raw_key.clone(), raw_value, new_raw_value.clone());
+        let mut r = self.kv_backend.txn(txn).await?;
+
+        if !r.succeeded {
+            let mut set = TxnOpGetResponseSet::from(&mut r.responses);
+            let raw_value = TxnOpGetResponseSet::filter(raw_key)(&mut set)
+                .context(UnexpectedSnafu {
+                    err_msg: "Reads the empty topic name value in comparing operation while updating TopicNameValue",
+                })?;
+
+            let op_name = "updating TopicNameValue";
+            ensure_values!(raw_value, new_raw_value, op_name);
+        }
+        Ok(())
+    }
 }
 
 #[cfg(test)]
 mod tests {
+    use std::assert_matches::assert_matches;
     use std::sync::Arc;
 
     use super::*;
@@ -204,7 +281,19 @@ mod tests {
         let topics = manager.range().await.unwrap();
         assert_eq!(topics, all_topics);
 
-        let topics = manager.range().await.unwrap();
-        assert_eq!(topics, all_topics);
+        for topic in &topics {
+            let value = manager.get(topic).await.unwrap().unwrap();
+            assert_eq!(value.pruned_entry_id, 0);
+            manager.update(topic, 1, Some(value.clone())).await.unwrap();
+            let new_value = manager.get(topic).await.unwrap().unwrap();
+            assert_eq!(new_value.pruned_entry_id, 1);
+            // Update twice, nothing changed
+            manager.update(topic, 1, Some(value.clone())).await.unwrap();
+            let new_value = manager.get(topic).await.unwrap().unwrap();
+            assert_eq!(new_value.pruned_entry_id, 1);
+            // Bad cas, emit error
+            let err = manager.update(topic, 3, Some(value)).await.unwrap_err();
+            assert_matches!(err, error::Error::Unexpected { .. });
+        }
     }
 }

src/common/meta/src/key/topic_region.rs

@@ -224,6 +224,7 @@ impl TopicRegionManager {
                         Some((region_id, kafka.topic.as_str()))
                     }
                     Some(WalOptions::RaftEngine) => None,
+                    Some(WalOptions::Noop) => None,
                     None => None,
                 },
             )

src/common/meta/src/kv_backend.rs

@@ -31,7 +31,7 @@ use crate::rpc::KeyValue;
 pub mod chroot;
 pub mod etcd;
 pub mod memory;
-#[cfg(feature = "pg_kvbackend")]
+#[cfg(any(feature = "mysql_kvbackend", feature = "pg_kvbackend"))]
 pub mod rds;
 pub mod test;
 pub mod txn;