fix: close issue #3902 since upstream fixed (#5801)

Signed-off-by: yihong0618 <zouzou0208@gmail.com>

.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/develop.yml

@@ -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

.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

@@ -67,7 +67,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "0.13.0"
+version = "0.14.0"
 edition = "2021"
 license = "Apache-2.0"
 
@@ -88,7 +88,7 @@ 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 +99,9 @@ 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 +111,7 @@ 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" }
@@ -121,32 +121,31 @@ 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"
-flate2 = { version = "1.1.0", default-features = false, features = ["zlib-rs"] }
 fst = "0.4.7"
 futures = "0.3"
 futures-util = "0.3"
-greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "c5419bbd20cb42e568ec325a4d71a3c94cc327e1" }
+greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "97e298d119fdb9499bc6ba9e03f375cfa7cdf130" }
 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 = [
@@ -164,8 +163,8 @@ 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 = [
@@ -177,7 +176,7 @@ 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"
@@ -185,21 +184,23 @@ rustls = { version = "0.23.20", default-features = false } # override by patch,
 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"
 similar-asserts = "1.6.0"
 smallvec = { version = "1", features = ["serde"] }
 snafu = "0.8"
 sqlx = { version = "0.8", features = [
     "runtime-tokio-rustls",
     "mysql",
+    "postgres",
+    "chrono",
 ] }
-sysinfo = "0.30"
+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"

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 archives 1 billion cold run #1 in JSONBench!](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

@@ -24,7 +24,7 @@
 | `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. |
@@ -101,7 +101,7 @@
 | `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 +181,7 @@
 | `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 +222,7 @@
 | `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 +279,7 @@
 | `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 +308,7 @@
 
 | 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" |
@@ -352,7 +352,7 @@
 | `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. |
@@ -390,7 +390,7 @@
 | `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. |
@@ -442,7 +442,7 @@
 | `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 +522,7 @@
 | `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. |
@@ -563,7 +563,7 @@
 | `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 +579,7 @@
 | `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

@@ -27,7 +27,7 @@ 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 +119,7 @@ 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`**.
@@ -265,7 +265,7 @@ 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.
@@ -618,7 +618,7 @@ 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

@@ -30,7 +30,7 @@ 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 +76,7 @@ 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 +121,3 @@ 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"
@@ -177,7 +177,7 @@ 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

@@ -34,7 +34,7 @@ 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 +164,7 @@ 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`**.
@@ -352,7 +352,7 @@ 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.
@@ -705,7 +705,7 @@ 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/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";

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/cli/src/bench.rs

@@ -177,7 +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 {
@@ -188,7 +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/client/src/lib.rs

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

src/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

@@ -295,10 +295,13 @@ 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(),
@@ -406,7 +409,7 @@ mod tests {
             sync_write = false
 
             [storage]
-            data_home = "/tmp/greptimedb/"
+            data_home = "./greptimedb_data/"
             type = "File"
 
             [[storage.providers]]
@@ -420,7 +423,7 @@ mod tests {
 
             [logging]
             level = "debug"
-            dir = "/tmp/greptimedb/test/logs"
+            dir = "./greptimedb_data/test/logs"
         "#;
         write!(file, "{}", toml_str).unwrap();
 
@@ -467,7 +470,7 @@ 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 { .. })
@@ -483,7 +486,10 @@ 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]
@@ -526,7 +532,7 @@ 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")]
@@ -536,7 +542,7 @@ 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 +571,11 @@ 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

@@ -249,10 +249,13 @@ 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;

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

@@ -55,9 +55,9 @@ use datanode::datanode::{Datanode, DatanodeBuilder};
 use datanode::region_server::RegionServer;
 use file_engine::config::EngineConfig as FileEngineConfig;
 use flow::{FlowConfig, FlowWorkerManager, FlownodeBuilder, FlownodeOptions, FrontendInvoker};
-use frontend::frontend::FrontendOptions;
+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,
@@ -67,7 +67,7 @@ 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};
@@ -76,15 +76,9 @@ 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";
 
@@ -251,13 +245,12 @@ 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>,
 }
@@ -281,21 +274,26 @@ 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(())
     }
 
@@ -303,17 +301,18 @@ 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| {
@@ -322,7 +321,8 @@ impl App for Instance {
                 }
                 .build()
             })
-            .context(ShutdownFlownodeSnafu)?;
+            .context(error::ShutdownFlownodeSnafu)?;
+
         info!("Datanode instance stopped.");
 
         Ok(())
@@ -368,7 +368,7 @@ 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)?;
 
@@ -415,7 +415,7 @@ 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}",
                     ),
@@ -474,18 +474,19 @@ 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(
@@ -494,7 +495,7 @@ impl StartCommand {
             opts.procedure,
         )
         .await
-        .context(StartFrontendSnafu)?;
+        .context(error::StartFrontendSnafu)?;
 
         // Builds cache registry
         let layered_cache_builder = LayeredCacheRegistryBuilder::default();
@@ -503,7 +504,7 @@ impl StartCommand {
             with_default_composite_cache_registry(
                 layered_cache_builder.add_cache_registry(fundamental_cache_registry),
             )
-            .context(BuildCacheRegistrySnafu)?
+            .context(error::BuildCacheRegistrySnafu)?
             .build(),
         );
 
@@ -512,7 +513,7 @@ impl StartCommand {
             .with_cache_registry(layered_cache_registry.clone())
             .build()
             .await
-            .context(StartDatanodeSnafu)?;
+            .context(error::StartDatanodeSnafu)?;
 
         let information_extension = Arc::new(StandaloneInformationExtension::new(
             datanode.region_server(),
@@ -545,7 +546,7 @@ impl StartCommand {
                 .build()
                 .await
                 .map_err(BoxedError::new)
-                .context(OtherSnafu)?,
+                .context(error::OtherSnafu)?,
         );
 
         // set the ref to query for the local flow state
@@ -576,7 +577,7 @@ 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,
@@ -597,8 +598,8 @@ 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(),
@@ -609,7 +610,8 @@ 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
@@ -622,18 +624,25 @@ 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,
@@ -670,7 +679,7 @@ impl StartCommand {
                 procedure_manager,
                 true,
             )
-            .context(InitDdlManagerSnafu)?,
+            .context(error::InitDdlManagerSnafu)?,
         );
 
         Ok(procedure_executor)
@@ -684,7 +693,7 @@ impl StartCommand {
         table_metadata_manager
             .init()
             .await
-            .context(InitMetadataSnafu)?;
+            .context(error::InitMetadataSnafu)?;
 
         Ok(table_metadata_manager)
     }
@@ -852,7 +861,7 @@ 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"
@@ -860,7 +869,7 @@ mod tests {
             sync_write = false
 
             [storage]
-            data_home = "/tmp/greptimedb/"
+            data_home = "./greptimedb_data/"
             type = "File"
 
             [[storage.providers]]
@@ -892,7 +901,7 @@ mod tests {
 
             [logging]
             level = "debug"
-            dir = "/tmp/greptimedb/test/logs"
+            dir = "./greptimedb_data/test/logs"
         "#;
         write!(file, "{}", toml_str).unwrap();
         let cmd = StartCommand {
@@ -922,7 +931,10 @@ 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,
@@ -946,7 +958,7 @@ 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]
@@ -958,7 +970,7 @@ 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")]
@@ -967,7 +979,7 @@ 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());
     }
 

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![
@@ -159,10 +159,10 @@ 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()),
@@ -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 +135,6 @@ 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";
 // ---- 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/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/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/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/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/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/ddl.rs

@@ -22,14 +22,17 @@ 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::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 +73,30 @@ 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,

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/key.rs

@@ -1471,7 +1471,8 @@ 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/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/rds/mysql.rs

@@ -155,21 +155,21 @@ impl<'a> MySqlTemplateFactory<'a> {
             table_name: table_name.to_string(),
             create_table_statement: format!(
                 // Cannot be more than 3072 bytes in PRIMARY KEY
-                "CREATE TABLE IF NOT EXISTS {table_name}(k VARBINARY(3072) PRIMARY KEY, v BLOB);",
+                "CREATE TABLE IF NOT EXISTS `{table_name}`(k VARBINARY(3072) PRIMARY KEY, v BLOB);",
             ),
             range_template: RangeTemplate {
-                point: format!("SELECT k, v FROM {table_name} WHERE k = ?"),
-                range: format!("SELECT k, v FROM {table_name} WHERE k >= ? AND k < ? ORDER BY k"),
-                full: format!("SELECT k, v FROM {table_name} ? ORDER BY k"),
-                left_bounded: format!("SELECT k, v FROM {table_name} WHERE k >= ? ORDER BY k"),
-                prefix: format!("SELECT k, v FROM {table_name} WHERE k LIKE ? ORDER BY k"),
+                point: format!("SELECT k, v FROM `{table_name}` WHERE k = ?"),
+                range: format!("SELECT k, v FROM `{table_name}` WHERE k >= ? AND k < ? ORDER BY k"),
+                full: format!("SELECT k, v FROM `{table_name}` ? ORDER BY k"),
+                left_bounded: format!("SELECT k, v FROM `{table_name}` WHERE k >= ? ORDER BY k"),
+                prefix: format!("SELECT k, v FROM `{table_name}` WHERE k LIKE ? ORDER BY k"),
             },
             delete_template: RangeTemplate {
-                point: format!("DELETE FROM {table_name} WHERE k = ?;"),
-                range: format!("DELETE FROM {table_name} WHERE k >= ? AND k < ?;"),
-                full: format!("DELETE FROM {table_name}"),
-                left_bounded: format!("DELETE FROM {table_name} WHERE k >= ?;"),
-                prefix: format!("DELETE FROM {table_name} WHERE k LIKE ?;"),
+                point: format!("DELETE FROM `{table_name}` WHERE k = ?;"),
+                range: format!("DELETE FROM `{table_name}` WHERE k >= ? AND k < ?;"),
+                full: format!("DELETE FROM `{table_name}`"),
+                left_bounded: format!("DELETE FROM `{table_name}` WHERE k >= ?;"),
+                prefix: format!("DELETE FROM `{table_name}` WHERE k LIKE ?;"),
             },
         }
     }
@@ -189,14 +189,17 @@ impl MySqlTemplateSet {
     fn generate_batch_get_query(&self, key_len: usize) -> String {
         let table_name = &self.table_name;
         let in_clause = mysql_generate_in_placeholders(1, key_len).join(", ");
-        format!("SELECT k, v FROM {table_name} WHERE k in ({});", in_clause)
+        format!(
+            "SELECT k, v FROM `{table_name}` WHERE k in ({});",
+            in_clause
+        )
     }
 
     /// Generates the sql for batch delete.
     fn generate_batch_delete_query(&self, key_len: usize) -> String {
         let table_name = &self.table_name;
         let in_clause = mysql_generate_in_placeholders(1, key_len).join(", ");
-        format!("DELETE FROM {table_name} WHERE k in ({});", in_clause)
+        format!("DELETE FROM `{table_name}` WHERE k in ({});", in_clause)
     }
 
     /// Generates the sql for batch upsert.
@@ -212,9 +215,9 @@ impl MySqlTemplateSet {
         let values_clause = values_placeholders.join(", ");
 
         (
-            format!(r#"SELECT k, v FROM {table_name} WHERE k IN ({in_clause})"#,),
+            format!(r#"SELECT k, v FROM `{table_name}` WHERE k IN ({in_clause})"#,),
             format!(
-                r#"INSERT INTO {table_name} (k, v) VALUES {values_clause} ON DUPLICATE KEY UPDATE v = VALUES(v);"#,
+                r#"INSERT INTO `{table_name}` (k, v) VALUES {values_clause} ON DUPLICATE KEY UPDATE v = VALUES(v);"#,
             ),
         )
     }

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

@@ -157,21 +157,25 @@ impl<'a> PgSqlTemplateFactory<'a> {
         PgSqlTemplateSet {
             table_name: table_name.to_string(),
             create_table_statement: format!(
-                "CREATE TABLE IF NOT EXISTS {table_name}(k bytea PRIMARY KEY, v bytea)",
+                "CREATE TABLE IF NOT EXISTS \"{table_name}\"(k bytea PRIMARY KEY, v bytea)",
             ),
             range_template: RangeTemplate {
-                point: format!("SELECT k, v FROM {table_name} WHERE k = $1"),
-                range: format!("SELECT k, v FROM {table_name} WHERE k >= $1 AND k < $2 ORDER BY k"),
-                full: format!("SELECT k, v FROM {table_name} $1 ORDER BY k"),
-                left_bounded: format!("SELECT k, v FROM {table_name} WHERE k >= $1 ORDER BY k"),
-                prefix: format!("SELECT k, v FROM {table_name} WHERE k LIKE $1 ORDER BY k"),
+                point: format!("SELECT k, v FROM \"{table_name}\" WHERE k = $1"),
+                range: format!(
+                    "SELECT k, v FROM \"{table_name}\" WHERE k >= $1 AND k < $2 ORDER BY k"
+                ),
+                full: format!("SELECT k, v FROM \"{table_name}\" $1 ORDER BY k"),
+                left_bounded: format!("SELECT k, v FROM \"{table_name}\" WHERE k >= $1 ORDER BY k"),
+                prefix: format!("SELECT k, v FROM \"{table_name}\" WHERE k LIKE $1 ORDER BY k"),
             },
             delete_template: RangeTemplate {
-                point: format!("DELETE FROM {table_name} WHERE k = $1 RETURNING k,v;"),
-                range: format!("DELETE FROM {table_name} WHERE k >= $1 AND k < $2 RETURNING k,v;"),
-                full: format!("DELETE FROM {table_name} RETURNING k,v"),
-                left_bounded: format!("DELETE FROM {table_name} WHERE k >= $1 RETURNING k,v;"),
-                prefix: format!("DELETE FROM {table_name} WHERE k LIKE $1 RETURNING k,v;"),
+                point: format!("DELETE FROM \"{table_name}\" WHERE k = $1 RETURNING k,v;"),
+                range: format!(
+                    "DELETE FROM \"{table_name}\" WHERE k >= $1 AND k < $2 RETURNING k,v;"
+                ),
+                full: format!("DELETE FROM \"{table_name}\" RETURNING k,v"),
+                left_bounded: format!("DELETE FROM \"{table_name}\" WHERE k >= $1 RETURNING k,v;"),
+                prefix: format!("DELETE FROM \"{table_name}\" WHERE k LIKE $1 RETURNING k,v;"),
             },
         }
     }
@@ -191,7 +195,10 @@ impl PgSqlTemplateSet {
     fn generate_batch_get_query(&self, key_len: usize) -> String {
         let table_name = &self.table_name;
         let in_clause = pg_generate_in_placeholders(1, key_len).join(", ");
-        format!("SELECT k, v FROM {table_name} WHERE k in ({});", in_clause)
+        format!(
+            "SELECT k, v FROM \"{table_name}\" WHERE k in ({});",
+            in_clause
+        )
     }
 
     /// Generates the sql for batch delete.
@@ -199,7 +206,7 @@ impl PgSqlTemplateSet {
         let table_name = &self.table_name;
         let in_clause = pg_generate_in_placeholders(1, key_len).join(", ");
         format!(
-            "DELETE FROM {table_name} WHERE k in ({}) RETURNING k,v;",
+            "DELETE FROM \"{table_name}\" WHERE k in ({}) RETURNING k,v;",
             in_clause
         )
     }
@@ -220,9 +227,9 @@ impl PgSqlTemplateSet {
         format!(
             r#"
     WITH prev AS (
-        SELECT k,v FROM {table_name} WHERE k IN ({in_clause})
+        SELECT k,v FROM "{table_name}" WHERE k IN ({in_clause})
     ), update AS (
-    INSERT INTO {table_name} (k, v) VALUES
+    INSERT INTO "{table_name}" (k, v) VALUES
         {values_clause}
     ON CONFLICT (
         k

src/common/meta/src/peer.rs

@@ -12,63 +12,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::fmt::{Display, Formatter};
 use std::sync::Arc;
 
-use api::v1::meta::Peer as PbPeer;
-use serde::{Deserialize, Serialize};
+pub use api::v1::meta::Peer;
 
 use crate::error::Error;
 use crate::{DatanodeId, FlownodeId};
 
-#[derive(Debug, Default, Clone, Hash, Eq, PartialEq, Deserialize, Serialize)]
-pub struct Peer {
-    /// Node identifier. Unique in a cluster.
-    pub id: u64,
-    pub addr: String,
-}
-
-impl From<PbPeer> for Peer {
-    fn from(p: PbPeer) -> Self {
-        Self {
-            id: p.id,
-            addr: p.addr,
-        }
-    }
-}
-
-impl From<Peer> for PbPeer {
-    fn from(p: Peer) -> Self {
-        Self {
-            id: p.id,
-            addr: p.addr,
-        }
-    }
-}
-
-impl Peer {
-    pub fn new(id: u64, addr: impl Into<String>) -> Self {
-        Self {
-            id,
-            addr: addr.into(),
-        }
-    }
-
-    #[cfg(any(test, feature = "testing"))]
-    pub fn empty(id: u64) -> Self {
-        Self {
-            id,
-            addr: String::new(),
-        }
-    }
-}
-
-impl Display for Peer {
-    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
-        write!(f, "peer-{}({})", self.id, self.addr)
-    }
-}
-
 /// can query peer given a node id
 #[async_trait::async_trait]
 pub trait PeerLookupService {

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

@@ -1240,6 +1240,7 @@ impl From<QueryContext> for PbQueryContext {
             extensions,
             channel: channel as u32,
             snapshot_seqs: None,
+            explain: None,
         }
     }
 }

src/common/meta/src/state_store.rs

@@ -290,13 +290,13 @@ mod tests {
         num_per_range: u32,
         max_bytes: u32,
     ) {
-        let num_cases = rand::thread_rng().gen_range(1..=8);
+        let num_cases = rand::rng().random_range(1..=8);
         common_telemetry::info!("num_cases: {}", num_cases);
         let mut cases = Vec::with_capacity(num_cases);
         for i in 0..num_cases {
-            let size = rand::thread_rng().gen_range(size_limit..=max_bytes);
+            let size = rand::rng().random_range(size_limit..=max_bytes);
             let mut large_value = vec![0u8; size as usize];
-            rand::thread_rng().fill_bytes(&mut large_value);
+            rand::rng().fill_bytes(&mut large_value);
 
             // Starts from `a`.
             let prefix = format!("{}/", std::char::from_u32(97 + i as u32).unwrap());
@@ -354,8 +354,8 @@ mod tests {
 
     #[tokio::test]
     async fn test_meta_state_store_split_value() {
-        let size_limit = rand::thread_rng().gen_range(128..=512);
-        let page_size = rand::thread_rng().gen_range(1..10);
+        let size_limit = rand::rng().random_range(128..=512);
+        let page_size = rand::rng().random_range(1..10);
         let kv_backend = Arc::new(MemoryKvBackend::new());
         test_meta_state_store_split_value_with_size_limit(kv_backend, size_limit, page_size, 8192)
             .await;
@@ -388,7 +388,7 @@ mod tests {
         // However, some KvBackends, the `ChrootKvBackend`, will add the prefix to `key`;
         // we don't know the exact size of the key.
         let size_limit = 1536 * 1024 - key_size;
-        let page_size = rand::thread_rng().gen_range(1..10);
+        let page_size = rand::rng().random_range(1..10);
         test_meta_state_store_split_value_with_size_limit(
             kv_backend,
             size_limit,

src/common/meta/src/wal_options_allocator.rs

@@ -53,21 +53,12 @@ impl WalOptionsAllocator {
         }
     }
 
-    /// Allocates a wal options for a region.
-    pub fn alloc(&self) -> Result<WalOptions> {
-        match self {
-            Self::RaftEngine => Ok(WalOptions::RaftEngine),
-            Self::Kafka(topic_manager) => {
-                let topic = topic_manager.select()?;
-                Ok(WalOptions::Kafka(KafkaWalOptions {
-                    topic: topic.clone(),
-                }))
-            }
-        }
-    }
-
     /// Allocates a batch of wal options where each wal options goes to a region.
-    pub fn alloc_batch(&self, num_regions: usize) -> Result<Vec<WalOptions>> {
+    /// If skip_wal is true, the wal options will be set to Noop regardless of the allocator type.
+    pub fn alloc_batch(&self, num_regions: usize, skip_wal: bool) -> Result<Vec<WalOptions>> {
+        if skip_wal {
+            return Ok(vec![WalOptions::Noop; num_regions]);
+        }
         match self {
             WalOptionsAllocator::RaftEngine => Ok(vec![WalOptions::RaftEngine; num_regions]),
             WalOptionsAllocator::Kafka(topic_manager) => {
@@ -130,9 +121,10 @@ pub async fn build_wal_options_allocator(
 pub fn allocate_region_wal_options(
     regions: Vec<RegionNumber>,
     wal_options_allocator: &WalOptionsAllocator,
+    skip_wal: bool,
 ) -> Result<HashMap<RegionNumber, String>> {
     let wal_options = wal_options_allocator
-        .alloc_batch(regions.len())?
+        .alloc_batch(regions.len(), skip_wal)?
         .into_iter()
         .map(|wal_options| {
             serde_json::to_string(&wal_options).context(EncodeWalOptionsSnafu { wal_options })
@@ -177,7 +169,7 @@ mod tests {
 
         let num_regions = 32;
         let regions = (0..num_regions).collect::<Vec<_>>();
-        let got = allocate_region_wal_options(regions.clone(), &allocator).unwrap();
+        let got = allocate_region_wal_options(regions.clone(), &allocator, false).unwrap();
 
         let encoded_wal_options = serde_json::to_string(&WalOptions::RaftEngine).unwrap();
         let expected = regions
@@ -237,7 +229,7 @@ mod tests {
 
                 let num_regions = 32;
                 let regions = (0..num_regions).collect::<Vec<_>>();
-                let got = allocate_region_wal_options(regions.clone(), &allocator).unwrap();
+                let got = allocate_region_wal_options(regions.clone(), &allocator, false).unwrap();
 
                 // Check the allocated wal options contain the expected topics.
                 let expected = (0..num_regions)
@@ -253,4 +245,18 @@ mod tests {
         })
         .await;
     }
+
+    #[tokio::test]
+    async fn test_allocator_with_skip_wal() {
+        let allocator = WalOptionsAllocator::RaftEngine;
+        allocator.start().await.unwrap();
+
+        let num_regions = 32;
+        let regions = (0..num_regions).collect::<Vec<_>>();
+        let got = allocate_region_wal_options(regions.clone(), &allocator, true).unwrap();
+        assert_eq!(got.len(), num_regions as usize);
+        for wal_options in got.values() {
+            assert_eq!(wal_options, &"{\"wal.provider\":\"noop\"}");
+        }
+    }
 }

src/common/meta/src/wal_options_allocator/selector.rs

@@ -39,7 +39,7 @@ impl RoundRobinTopicSelector {
     // The cursor in the round-robin selector is not persisted which may break the round-robin strategy cross crashes.
     // Introducing a shuffling strategy may help mitigate this issue.
     pub fn with_shuffle() -> Self {
-        let offset = rand::thread_rng().gen_range(0..64);
+        let offset = rand::rng().random_range(0..64);
         Self {
             cursor: AtomicUsize::new(offset),
         }

src/common/procedure/src/local/runner.rs

@@ -207,7 +207,7 @@ impl Runner {
                     if let Some(d) = retry.next() {
                         let millis = d.as_millis() as u64;
                         // Add random noise to the retry delay to avoid retry storms.
-                        let noise = rand::thread_rng().gen_range(0..(millis / 4) + 1);
+                        let noise = rand::rng().random_range(0..(millis / 4) + 1);
                         let d = d.add(Duration::from_millis(noise));
 
                         self.wait_on_err(d, retry_times).await;

src/common/recordbatch/src/adapter.rs

@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::fmt::Display;
+use std::fmt::{self, Display};
 use std::future::Future;
 use std::marker::PhantomData;
 use std::pin::Pin;
@@ -28,7 +28,7 @@ use datafusion::logical_expr::Expr;
 use datafusion::physical_expr::create_physical_expr;
 use datafusion::physical_plan::metrics::{BaselineMetrics, MetricValue};
 use datafusion::physical_plan::{
-    accept, displayable, ExecutionPlan, ExecutionPlanVisitor, PhysicalExpr,
+    accept, DisplayFormatType, ExecutionPlan, ExecutionPlanVisitor, PhysicalExpr,
     RecordBatchStream as DfRecordBatchStream,
 };
 use datafusion_common::arrow::error::ArrowError;
@@ -206,13 +206,16 @@ impl Stream for DfRecordBatchStreamAdapter {
 }
 
 /// DataFusion [SendableRecordBatchStream](DfSendableRecordBatchStream) -> Greptime [RecordBatchStream].
-/// The reverse one is [DfRecordBatchStreamAdapter]
+/// The reverse one is [DfRecordBatchStreamAdapter].
+/// It can collect metrics from DataFusion execution plan.
 pub struct RecordBatchStreamAdapter {
     schema: SchemaRef,
     stream: DfSendableRecordBatchStream,
     metrics: Option<BaselineMetrics>,
     /// Aggregated plan-level metrics. Resolved after an [ExecutionPlan] is finished.
     metrics_2: Metrics,
+    /// Display plan and metrics in verbose mode.
+    explain_verbose: bool,
 }
 
 /// Json encoded metrics. Contains metric from a whole plan tree.
@@ -231,6 +234,7 @@ impl RecordBatchStreamAdapter {
             stream,
             metrics: None,
             metrics_2: Metrics::Unavailable,
+            explain_verbose: false,
         })
     }
 
@@ -246,12 +250,18 @@ impl RecordBatchStreamAdapter {
             stream,
             metrics: Some(metrics),
             metrics_2: Metrics::Unresolved(df_plan),
+            explain_verbose: false,
         })
     }
 
     pub fn set_metrics2(&mut self, plan: Arc<dyn ExecutionPlan>) {
         self.metrics_2 = Metrics::Unresolved(plan)
     }
+
+    /// Set the verbose mode for displaying plan and metrics.
+    pub fn set_explain_verbose(&mut self, verbose: bool) {
+        self.explain_verbose = verbose;
+    }
 }
 
 impl RecordBatchStream for RecordBatchStreamAdapter {
@@ -296,7 +306,7 @@ impl Stream for RecordBatchStreamAdapter {
             }
             Poll::Ready(None) => {
                 if let Metrics::Unresolved(df_plan) = &self.metrics_2 {
-                    let mut metric_collector = MetricCollector::default();
+                    let mut metric_collector = MetricCollector::new(self.explain_verbose);
                     accept(df_plan.as_ref(), &mut metric_collector).unwrap();
                     self.metrics_2 = Metrics::Resolved(metric_collector.record_batch_metrics);
                 }
@@ -312,10 +322,20 @@ impl Stream for RecordBatchStreamAdapter {
 }
 
 /// An [ExecutionPlanVisitor] to collect metrics from a [ExecutionPlan].
-#[derive(Default)]
 pub struct MetricCollector {
     current_level: usize,
     pub record_batch_metrics: RecordBatchMetrics,
+    verbose: bool,
+}
+
+impl MetricCollector {
+    pub fn new(verbose: bool) -> Self {
+        Self {
+            current_level: 0,
+            record_batch_metrics: RecordBatchMetrics::default(),
+            verbose,
+        }
+    }
 }
 
 impl ExecutionPlanVisitor for MetricCollector {
@@ -339,7 +359,7 @@ impl ExecutionPlanVisitor for MetricCollector {
             .sorted_for_display()
             .timestamps_removed();
         let mut plan_metric = PlanMetrics {
-            plan: displayable(plan).one_line().to_string(),
+            plan: one_line(plan, self.verbose).to_string(),
             level: self.current_level,
             metrics: Vec::with_capacity(metric.iter().size_hint().0),
         };
@@ -371,6 +391,29 @@ impl ExecutionPlanVisitor for MetricCollector {
     }
 }
 
+/// Returns a single-line summary of the root of the plan.
+/// If the `verbose` flag is set, it will display detailed information about the plan.
+fn one_line(plan: &dyn ExecutionPlan, verbose: bool) -> impl fmt::Display + '_ {
+    struct Wrapper<'a> {
+        plan: &'a dyn ExecutionPlan,
+        format_type: DisplayFormatType,
+    }
+
+    impl fmt::Display for Wrapper<'_> {
+        fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+            self.plan.fmt_as(self.format_type, f)?;
+            writeln!(f)
+        }
+    }
+
+    let format_type = if verbose {
+        DisplayFormatType::Verbose
+    } else {
+        DisplayFormatType::Default
+    };
+    Wrapper { plan, format_type }
+}
+
 /// [`RecordBatchMetrics`] carrys metrics value
 /// from datanode to frontend through gRPC
 #[derive(serde::Serialize, serde::Deserialize, Default, Debug, Clone)]

src/common/recordbatch/src/filter.rs

@@ -22,10 +22,12 @@ use datafusion::physical_plan::PhysicalExpr;
 use datafusion_common::arrow::array::{ArrayRef, Datum, Scalar};
 use datafusion_common::arrow::buffer::BooleanBuffer;
 use datafusion_common::arrow::compute::kernels::cmp;
-use datafusion_common::cast::{as_boolean_array, as_null_array};
+use datafusion_common::cast::{as_boolean_array, as_null_array, as_string_array};
 use datafusion_common::{internal_err, DataFusionError, ScalarValue};
 use datatypes::arrow::array::{Array, BooleanArray, RecordBatch};
 use datatypes::arrow::compute::filter_record_batch;
+use datatypes::arrow::error::ArrowError;
+use datatypes::compute::kernels::regexp;
 use datatypes::compute::or_kleene;
 use datatypes::vectors::VectorRef;
 use snafu::ResultExt;
@@ -36,7 +38,8 @@ use crate::error::{ArrowComputeSnafu, Result, ToArrowScalarSnafu, UnsupportedOpe
 /// - `col` `op` `literal`
 /// - `literal` `op` `col`
 ///
-/// And the `op` is one of `=`, `!=`, `>`, `>=`, `<`, `<=`.
+/// And the `op` is one of `=`, `!=`, `>`, `>=`, `<`, `<=`,
+/// or regex operators: `~`, `~*`, `!~`, `!~*`.
 ///
 /// This struct contains normalized predicate expr. In the form of
 /// `col` `op` `literal` where the `col` is provided from input.
@@ -86,7 +89,11 @@ impl SimpleFilterEvaluator {
                     | Operator::Lt
                     | Operator::LtEq
                     | Operator::Gt
-                    | Operator::GtEq => {}
+                    | Operator::GtEq
+                    | Operator::RegexMatch
+                    | Operator::RegexIMatch
+                    | Operator::RegexNotMatch
+                    | Operator::RegexNotIMatch => {}
                     Operator::Or => {
                         let lhs = Self::try_new(&binary.left)?;
                         let rhs = Self::try_new(&binary.right)?;
@@ -172,6 +179,10 @@ impl SimpleFilterEvaluator {
             Operator::LtEq => cmp::lt_eq(input, &self.literal),
             Operator::Gt => cmp::gt(input, &self.literal),
             Operator::GtEq => cmp::gt_eq(input, &self.literal),
+            Operator::RegexMatch => self.regex_match(input, false, false),
+            Operator::RegexIMatch => self.regex_match(input, true, false),
+            Operator::RegexNotMatch => self.regex_match(input, false, true),
+            Operator::RegexNotIMatch => self.regex_match(input, true, true),
             Operator::Or => {
                 // OR operator stands for OR-chained EQs (or INLIST in other words)
                 let mut result: BooleanArray = vec![false; input_len].into();
@@ -192,6 +203,28 @@ impl SimpleFilterEvaluator {
             .context(ArrowComputeSnafu)
             .map(|array| array.values().clone())
     }
+
+    fn regex_match(
+        &self,
+        input: &impl Datum,
+        ignore_case: bool,
+        negative: bool,
+    ) -> std::result::Result<BooleanArray, ArrowError> {
+        let flag = if ignore_case { Some("i") } else { None };
+        let array = input.get().0;
+        let string_array = as_string_array(array).map_err(|_| {
+            ArrowError::CastError(format!("Cannot cast {:?} to StringArray", array))
+        })?;
+        let literal_array = self.literal.clone().into_inner();
+        let regex_array = as_string_array(&literal_array).map_err(|_| {
+            ArrowError::CastError(format!("Cannot cast {:?} to StringArray", literal_array))
+        })?;
+        let mut result = regexp::regexp_is_match_scalar(string_array, regex_array.value(0), flag)?;
+        if negative {
+            result = datatypes::compute::not(&result)?;
+        }
+        Ok(result)
+    }
 }
 
 /// Evaluate the predicate on the input [RecordBatch], and return a new [RecordBatch].

src/common/telemetry/src/logging.rs

@@ -111,7 +111,7 @@ impl Eq for LoggingOptions {}
 impl Default for LoggingOptions {
     fn default() -> Self {
         Self {
-            dir: "/tmp/greptimedb/logs".to_string(),
+            dir: "./greptimedb_data/logs".to_string(),
             level: None,
             log_format: LogFormat::Text,
             enable_otlp_tracing: false,

src/common/test-util/src/ports.rs

@@ -22,6 +22,6 @@ static PORTS: OnceCell<AtomicUsize> = OnceCell::new();
 /// Return a unique port(in runtime) for test
 pub fn get_port() -> usize {
     PORTS
-        .get_or_init(|| AtomicUsize::new(rand::thread_rng().gen_range(13000..13800)))
+        .get_or_init(|| AtomicUsize::new(rand::rng().random_range(13000..13800)))
         .fetch_add(1, Ordering::Relaxed)
 }

src/common/time/src/timestamp.rs

@@ -715,10 +715,10 @@ mod tests {
             TimeUnit::Microsecond,
             TimeUnit::Nanosecond,
         ];
-        let mut rng = rand::thread_rng();
-        let unit_idx: usize = rng.gen_range(0..4);
+        let mut rng = rand::rng();
+        let unit_idx: usize = rng.random_range(0..4);
         let unit = units[unit_idx];
-        let value: i64 = rng.gen();
+        let value: i64 = rng.random();
         Timestamp::new(value, unit)
     }
 
@@ -745,8 +745,8 @@ mod tests {
 
     /// Generate timestamp less than or equal to `threshold`
     fn gen_ts_le(threshold: &Timestamp) -> Timestamp {
-        let mut rng = rand::thread_rng();
-        let timestamp = rng.gen_range(i64::MIN..=threshold.value);
+        let mut rng = rand::rng();
+        let timestamp = rng.random_range(i64::MIN..=threshold.value);
         Timestamp::new(timestamp, threshold.unit)
     }
 

src/common/wal/src/options.rs

@@ -33,6 +33,7 @@ pub enum WalOptions {
     RaftEngine,
     #[serde(with = "kafka_prefix")]
     Kafka(KafkaWalOptions),
+    Noop,
 }
 
 with_prefix!(kafka_prefix "wal.kafka.");
@@ -62,5 +63,14 @@ mod tests {
 
         let decoded: WalOptions = serde_json::from_str(&encoded).unwrap();
         assert_eq!(decoded, wal_options);
+
+        // Test serde noop wal options.
+        let wal_options = WalOptions::Noop;
+        let encoded = serde_json::to_string(&wal_options).unwrap();
+        let expected = r#"{"wal.provider":"noop"}"#;
+        assert_eq!(&encoded, expected);
+
+        let decoded: WalOptions = serde_json::from_str(&encoded).unwrap();
+        assert_eq!(decoded, wal_options);
     }
 }

src/datanode/src/config.rs

@@ -36,7 +36,7 @@ use servers::Mode;
 pub const DEFAULT_OBJECT_STORE_CACHE_SIZE: ReadableSize = ReadableSize::gb(5);
 
 /// Default data home in file storage
-const DEFAULT_DATA_HOME: &str = "/tmp/greptimedb";
+const DEFAULT_DATA_HOME: &str = "./greptimedb_data";
 
 /// Object storage config
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]

src/datanode/src/lib.rs

@@ -25,6 +25,6 @@ pub mod heartbeat;
 pub mod metrics;
 pub mod region_server;
 pub mod service;
-mod store;
+pub mod store;
 #[cfg(any(test, feature = "testing"))]
 pub mod tests;

src/datanode/src/store.rs

@@ -15,7 +15,7 @@
 //! object storage utilities
 
 mod azblob;
-mod fs;
+pub mod fs;
 mod gcs;
 mod oss;
 mod s3;

src/datanode/src/store/fs.rs

@@ -24,7 +24,8 @@ use crate::config::FileConfig;
 use crate::error::{self, Result};
 use crate::store;
 
-pub(crate) async fn new_fs_object_store(
+/// A helper function to create a file system object store.
+pub async fn new_fs_object_store(
     data_home: &str,
     _file_config: &FileConfig,
 ) -> Result<ObjectStore> {

src/datatypes/src/value.rs

@@ -285,6 +285,20 @@ impl Value {
         }
     }
 
+    /// Cast Value to i64. Return None if value is not a valid int64 data type.
+    pub fn as_i64(&self) -> Option<i64> {
+        match self {
+            Value::Int8(v) => Some(*v as _),
+            Value::Int16(v) => Some(*v as _),
+            Value::Int32(v) => Some(*v as _),
+            Value::Int64(v) => Some(*v),
+            Value::UInt8(v) => Some(*v as _),
+            Value::UInt16(v) => Some(*v as _),
+            Value::UInt32(v) => Some(*v as _),
+            _ => None,
+        }
+    }
+
     /// Cast Value to u64. Return None if value is not a valid uint64 data type.
     pub fn as_u64(&self) -> Option<u64> {
         match self {
@@ -295,7 +309,6 @@ impl Value {
             _ => None,
         }
     }
-
     /// Cast Value to f64. Return None if it's not castable;
     pub fn as_f64_lossy(&self) -> Option<f64> {
         match self {

src/datatypes/src/vectors.rs

@@ -30,6 +30,7 @@ mod boolean;
 mod constant;
 mod date;
 mod decimal;
+mod dictionary;
 mod duration;
 mod eq;
 mod helper;
@@ -48,6 +49,7 @@ pub use boolean::{BooleanVector, BooleanVectorBuilder};
 pub use constant::ConstantVector;
 pub use date::{DateVector, DateVectorBuilder};
 pub use decimal::{Decimal128Vector, Decimal128VectorBuilder};
+pub use dictionary::{DictionaryIter, DictionaryVector};
 pub use duration::{
     DurationMicrosecondVector, DurationMicrosecondVectorBuilder, DurationMillisecondVector,
     DurationMillisecondVectorBuilder, DurationNanosecondVector, DurationNanosecondVectorBuilder,

src/datatypes/src/vectors/dictionary.rs

@@ -0,0 +1,438 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::any::Any;
+use std::sync::Arc;
+
+use arrow::array::Array;
+use arrow::datatypes::Int32Type;
+use arrow_array::{ArrayRef, DictionaryArray, Int32Array};
+use serde_json::Value as JsonValue;
+use snafu::ResultExt;
+
+use super::operations::VectorOp;
+use crate::data_type::ConcreteDataType;
+use crate::error::{self, Result};
+use crate::serialize::Serializable;
+use crate::types::DictionaryType;
+use crate::value::{Value, ValueRef};
+use crate::vectors::{self, Helper, Validity, Vector, VectorRef};
+
+/// Vector of dictionaries, basically backed by Arrow's `DictionaryArray`.
+#[derive(Debug, PartialEq)]
+pub struct DictionaryVector {
+    array: DictionaryArray<Int32Type>,
+    /// The datatype of the items in the dictionary.
+    item_type: ConcreteDataType,
+    /// The vector of items in the dictionary.
+    item_vector: VectorRef,
+}
+
+impl DictionaryVector {
+    /// Create a new instance of `DictionaryVector` from a dictionary array and item type
+    pub fn new(array: DictionaryArray<Int32Type>, item_type: ConcreteDataType) -> Result<Self> {
+        let item_vector = Helper::try_into_vector(array.values())?;
+
+        Ok(Self {
+            array,
+            item_type,
+            item_vector,
+        })
+    }
+
+    /// Returns the underlying Arrow dictionary array
+    pub fn array(&self) -> &DictionaryArray<Int32Type> {
+        &self.array
+    }
+
+    /// Returns the keys array of this dictionary
+    pub fn keys(&self) -> &arrow_array::PrimitiveArray<Int32Type> {
+        self.array.keys()
+    }
+
+    /// Returns the values array of this dictionary
+    pub fn values(&self) -> &ArrayRef {
+        self.array.values()
+    }
+
+    pub fn as_arrow(&self) -> &dyn Array {
+        &self.array
+    }
+}
+
+impl Vector for DictionaryVector {
+    fn data_type(&self) -> ConcreteDataType {
+        ConcreteDataType::Dictionary(DictionaryType::new(
+            ConcreteDataType::int32_datatype(),
+            self.item_type.clone(),
+        ))
+    }
+
+    fn vector_type_name(&self) -> String {
+        "DictionaryVector".to_string()
+    }
+
+    fn as_any(&self) -> &dyn Any {
+        self
+    }
+
+    fn len(&self) -> usize {
+        self.array.len()
+    }
+
+    fn to_arrow_array(&self) -> ArrayRef {
+        Arc::new(self.array.clone())
+    }
+
+    fn to_boxed_arrow_array(&self) -> Box<dyn Array> {
+        Box::new(self.array.clone())
+    }
+
+    fn validity(&self) -> Validity {
+        vectors::impl_validity_for_vector!(self.array)
+    }
+
+    fn memory_size(&self) -> usize {
+        self.array.get_buffer_memory_size()
+    }
+
+    fn null_count(&self) -> usize {
+        self.array.null_count()
+    }
+
+    fn is_null(&self, row: usize) -> bool {
+        self.array.is_null(row)
+    }
+
+    fn slice(&self, offset: usize, length: usize) -> VectorRef {
+        Arc::new(Self {
+            array: self.array.slice(offset, length),
+            item_type: self.item_type.clone(),
+            item_vector: self.item_vector.clone(),
+        })
+    }
+
+    fn get(&self, index: usize) -> Value {
+        if !self.array.is_valid(index) {
+            return Value::Null;
+        }
+
+        let key = self.array.keys().value(index);
+        self.item_vector.get(key as usize)
+    }
+
+    fn get_ref(&self, index: usize) -> ValueRef {
+        if !self.array.is_valid(index) {
+            return ValueRef::Null;
+        }
+
+        let key = self.array.keys().value(index);
+        self.item_vector.get_ref(key as usize)
+    }
+}
+
+impl Serializable for DictionaryVector {
+    fn serialize_to_json(&self) -> Result<Vec<JsonValue>> {
+        // Convert the dictionary array to JSON, where each element is either null or
+        // the value it refers to in the dictionary
+        let mut result = Vec::with_capacity(self.len());
+
+        for i in 0..self.len() {
+            if self.is_null(i) {
+                result.push(JsonValue::Null);
+            } else {
+                let key = self.array.keys().value(i);
+                let value = self.item_vector.get(key as usize);
+                let json_value = serde_json::to_value(value).context(error::SerializeSnafu)?;
+                result.push(json_value);
+            }
+        }
+
+        Ok(result)
+    }
+}
+
+impl TryFrom<DictionaryArray<Int32Type>> for DictionaryVector {
+    type Error = crate::error::Error;
+
+    fn try_from(array: DictionaryArray<Int32Type>) -> Result<Self> {
+        let item_type = ConcreteDataType::from_arrow_type(array.values().data_type());
+        let item_vector = Helper::try_into_vector(array.values())?;
+
+        Ok(Self {
+            array,
+            item_type,
+            item_vector,
+        })
+    }
+}
+
+pub struct DictionaryIter<'a> {
+    vector: &'a DictionaryVector,
+    idx: usize,
+}
+
+impl<'a> DictionaryIter<'a> {
+    pub fn new(vector: &'a DictionaryVector) -> DictionaryIter<'a> {
+        DictionaryIter { vector, idx: 0 }
+    }
+}
+
+impl<'a> Iterator for DictionaryIter<'a> {
+    type Item = Option<ValueRef<'a>>;
+
+    #[inline]
+    fn next(&mut self) -> Option<Self::Item> {
+        if self.idx >= self.vector.len() {
+            return None;
+        }
+
+        let idx = self.idx;
+        self.idx += 1;
+
+        if self.vector.is_null(idx) {
+            return Some(None);
+        }
+
+        Some(Some(self.vector.item_vector.get_ref(self.idx)))
+    }
+
+    #[inline]
+    fn size_hint(&self) -> (usize, Option<usize>) {
+        (
+            self.vector.len() - self.idx,
+            Some(self.vector.len() - self.idx),
+        )
+    }
+}
+
+impl VectorOp for DictionaryVector {
+    fn replicate(&self, offsets: &[usize]) -> VectorRef {
+        let keys = self.array.keys();
+        let mut replicated_keys = Vec::with_capacity(offsets.len());
+
+        let mut previous_offset = 0;
+        for (i, &offset) in offsets.iter().enumerate() {
+            let key = if i < self.len() {
+                if keys.is_valid(i) {
+                    Some(keys.value(i))
+                } else {
+                    None
+                }
+            } else {
+                None
+            };
+
+            // repeat this key (offset - previous_offset) times
+            let repeat_count = offset - previous_offset;
+            if repeat_count > 0 {
+                replicated_keys.resize(replicated_keys.len() + repeat_count, key);
+            }
+
+            previous_offset = offset;
+        }
+
+        let new_keys = Int32Array::from(replicated_keys);
+        let new_array = DictionaryArray::try_new(new_keys, self.values().clone())
+            .expect("Failed to create replicated dictionary array");
+
+        Arc::new(Self {
+            array: new_array,
+            item_type: self.item_type.clone(),
+            item_vector: self.item_vector.clone(),
+        })
+    }
+
+    fn filter(&self, filter: &vectors::BooleanVector) -> Result<VectorRef> {
+        let key_array: ArrayRef = Arc::new(self.array.keys().clone());
+        let key_vector = Helper::try_into_vector(&key_array)?;
+        let filtered_key_vector = key_vector.filter(filter)?;
+        let filtered_key_array = filtered_key_vector.to_arrow_array();
+        let filtered_key_array = filtered_key_array
+            .as_any()
+            .downcast_ref::<Int32Array>()
+            .unwrap();
+
+        let new_array = DictionaryArray::try_new(filtered_key_array.clone(), self.values().clone())
+            .expect("Failed to create filtered dictionary array");
+
+        Ok(Arc::new(Self {
+            array: new_array,
+            item_type: self.item_type.clone(),
+            item_vector: self.item_vector.clone(),
+        }))
+    }
+
+    fn cast(&self, to_type: &ConcreteDataType) -> Result<VectorRef> {
+        let new_items = self.item_vector.cast(to_type)?;
+        let new_array =
+            DictionaryArray::try_new(self.array.keys().clone(), new_items.to_arrow_array())
+                .expect("Failed to create casted dictionary array");
+        Ok(Arc::new(Self {
+            array: new_array,
+            item_type: to_type.clone(),
+            item_vector: self.item_vector.clone(),
+        }))
+    }
+
+    fn take(&self, indices: &vectors::UInt32Vector) -> Result<VectorRef> {
+        let key_array: ArrayRef = Arc::new(self.array.keys().clone());
+        let key_vector = Helper::try_into_vector(&key_array)?;
+        let new_key_vector = key_vector.take(indices)?;
+        let new_key_array = new_key_vector.to_arrow_array();
+        let new_key_array = new_key_array.as_any().downcast_ref::<Int32Array>().unwrap();
+
+        let new_array = DictionaryArray::try_new(new_key_array.clone(), self.values().clone())
+            .expect("Failed to create filtered dictionary array");
+
+        Ok(Arc::new(Self {
+            array: new_array,
+            item_type: self.item_type.clone(),
+            item_vector: self.item_vector.clone(),
+        }))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::Arc;
+
+    use arrow_array::StringArray;
+
+    use super::*;
+
+    // Helper function to create a test dictionary vector with string values
+    fn create_test_dictionary() -> DictionaryVector {
+        // Dictionary values: ["a", "b", "c", "d"]
+        // Keys: [0, 1, 2, null, 1, 3]
+        // Resulting in: ["a", "b", "c", null, "b", "d"]
+        let values = StringArray::from(vec!["a", "b", "c", "d"]);
+        let keys = Int32Array::from(vec![Some(0), Some(1), Some(2), None, Some(1), Some(3)]);
+        let dict_array = DictionaryArray::new(keys, Arc::new(values));
+        DictionaryVector::try_from(dict_array).unwrap()
+    }
+
+    #[test]
+    fn test_dictionary_vector_basics() {
+        let dict_vec = create_test_dictionary();
+
+        // Test length and null count
+        assert_eq!(dict_vec.len(), 6);
+        assert_eq!(dict_vec.null_count(), 1);
+
+        // Test data type
+        let data_type = dict_vec.data_type();
+        if let ConcreteDataType::Dictionary(dict_type) = data_type {
+            assert_eq!(*dict_type.value_type(), ConcreteDataType::string_datatype());
+        } else {
+            panic!("Expected Dictionary data type");
+        }
+
+        // Test is_null
+        assert!(!dict_vec.is_null(0));
+        assert!(dict_vec.is_null(3));
+
+        // Test get values
+        assert_eq!(dict_vec.get(0), Value::String("a".to_string().into()));
+        assert_eq!(dict_vec.get(1), Value::String("b".to_string().into()));
+        assert_eq!(dict_vec.get(3), Value::Null);
+        assert_eq!(dict_vec.get(4), Value::String("b".to_string().into()));
+    }
+
+    #[test]
+    fn test_slice() {
+        let dict_vec = create_test_dictionary();
+        let sliced = dict_vec.slice(1, 3);
+
+        assert_eq!(sliced.len(), 3);
+        assert_eq!(sliced.get(0), Value::String("b".to_string().into()));
+        assert_eq!(sliced.get(1), Value::String("c".to_string().into()));
+        assert_eq!(sliced.get(2), Value::Null);
+    }
+
+    #[test]
+    fn test_replicate() {
+        let dict_vec = create_test_dictionary();
+
+        // Replicate with offsets [0, 2, 5] - should get values at these indices
+        let offsets = vec![0, 2, 5];
+        let replicated = dict_vec.replicate(&offsets);
+        assert_eq!(replicated.len(), 5);
+        assert_eq!(replicated.get(0), Value::String("b".to_string().into()));
+        assert_eq!(replicated.get(1), Value::String("b".to_string().into()));
+        assert_eq!(replicated.get(2), Value::String("c".to_string().into()));
+        assert_eq!(replicated.get(3), Value::String("c".to_string().into()));
+        assert_eq!(replicated.get(4), Value::String("c".to_string().into()));
+    }
+
+    #[test]
+    fn test_filter() {
+        let dict_vec = create_test_dictionary();
+
+        // Keep only indices 0, 2, 4
+        let filter_values = vec![true, false, true, false, true, false];
+        let filter = vectors::BooleanVector::from(filter_values);
+
+        let filtered = dict_vec.filter(&filter).unwrap();
+        assert_eq!(filtered.len(), 3);
+
+        // Check the values
+        assert_eq!(filtered.get(0), Value::String("a".to_string().into()));
+        assert_eq!(filtered.get(1), Value::String("c".to_string().into()));
+        assert_eq!(filtered.get(2), Value::String("b".to_string().into()));
+    }
+
+    #[test]
+    fn test_cast() {
+        let dict_vec = create_test_dictionary();
+
+        // Cast to the same type should return an equivalent vector
+        let casted = dict_vec.cast(&ConcreteDataType::string_datatype()).unwrap();
+
+        // The returned vector should have string values
+        assert_eq!(
+            casted.data_type(),
+            ConcreteDataType::Dictionary(DictionaryType::new(
+                ConcreteDataType::int32_datatype(),
+                ConcreteDataType::string_datatype(),
+            ))
+        );
+        assert_eq!(casted.len(), dict_vec.len());
+
+        // Values should match the original dictionary lookups
+        assert_eq!(casted.get(0), Value::String("a".to_string().into()));
+        assert_eq!(casted.get(1), Value::String("b".to_string().into()));
+        assert_eq!(casted.get(2), Value::String("c".to_string().into()));
+        assert_eq!(casted.get(3), Value::Null);
+        assert_eq!(casted.get(4), Value::String("b".to_string().into()));
+        assert_eq!(casted.get(5), Value::String("d".to_string().into()));
+    }
+
+    #[test]
+    fn test_take() {
+        let dict_vec = create_test_dictionary();
+
+        // Take indices 2, 0, 4
+        let indices_vec = vec![Some(2u32), Some(0), Some(4)];
+        let indices = vectors::UInt32Vector::from(indices_vec);
+
+        let taken = dict_vec.take(&indices).unwrap();
+        assert_eq!(taken.len(), 3);
+
+        // Check the values
+        assert_eq!(taken.get(0), Value::String("c".to_string().into()));
+        assert_eq!(taken.get(1), Value::String("a".to_string().into()));
+        assert_eq!(taken.get(2), Value::String("b".to_string().into()));
+    }
+}

src/datatypes/src/vectors/helper.rs

@@ -20,7 +20,8 @@ use std::sync::Arc;
 use arrow::array::{Array, ArrayRef, StringArray};
 use arrow::compute;
 use arrow::compute::kernels::comparison;
-use arrow::datatypes::{DataType as ArrowDataType, TimeUnit};
+use arrow::datatypes::{DataType as ArrowDataType, Int32Type, TimeUnit};
+use arrow_array::DictionaryArray;
 use arrow_schema::IntervalUnit;
 use datafusion_common::ScalarValue;
 use snafu::{OptionExt, ResultExt};
@@ -31,7 +32,7 @@ use crate::prelude::DataType;
 use crate::scalars::{Scalar, ScalarVectorBuilder};
 use crate::value::{ListValue, ListValueRef, Value};
 use crate::vectors::{
-    BinaryVector, BooleanVector, ConstantVector, DateVector, Decimal128Vector,
+    BinaryVector, BooleanVector, ConstantVector, DateVector, Decimal128Vector, DictionaryVector,
     DurationMicrosecondVector, DurationMillisecondVector, DurationNanosecondVector,
     DurationSecondVector, Float32Vector, Float64Vector, Int16Vector, Int32Vector, Int64Vector,
     Int8Vector, IntervalDayTimeVector, IntervalMonthDayNanoVector, IntervalYearMonthVector,
@@ -347,6 +348,17 @@ impl Helper {
             ArrowDataType::Decimal128(_, _) => {
                 Arc::new(Decimal128Vector::try_from_arrow_array(array)?)
             }
+            ArrowDataType::Dictionary(key, value) if matches!(&**key, ArrowDataType::Int32) => {
+                let array = array
+                    .as_ref()
+                    .as_any()
+                    .downcast_ref::<DictionaryArray<Int32Type>>()
+                    .unwrap(); // Safety: the type is guarded by match arm condition
+                Arc::new(DictionaryVector::new(
+                    array.clone(),
+                    ConcreteDataType::try_from(value.as_ref())?,
+                )?)
+            }
             ArrowDataType::Float16
             | ArrowDataType::LargeList(_)
             | ArrowDataType::FixedSizeList(_, _)

src/datatypes/src/vectors/operations.rs

@@ -14,14 +14,11 @@
 
 mod cast;
 mod filter;
-mod find_unique;
 mod replicate;
 mod take;
 
 use std::sync::Arc;
 
-use common_base::BitVec;
-
 use crate::error::{self, Result};
 use crate::types::LogicalPrimitiveType;
 use crate::vectors::constant::ConstantVector;
@@ -40,23 +37,6 @@ pub trait VectorOp {
     /// Panics if `offsets.len() != self.len()`.
     fn replicate(&self, offsets: &[usize]) -> VectorRef;
 
-    /// Mark `i-th` bit of `selected` to `true` if the `i-th` element of `self` is unique, which
-    /// means there is no elements behind it have same value as it.
-    ///
-    /// The caller should ensure
-    /// 1. the length of `selected` bitmap is equal to `vector.len()`.
-    /// 2. `vector` and `prev_vector` are sorted.
-    ///
-    /// If there are multiple duplicate elements, this function retains the **first** element.
-    /// The first element is considered as unique if the first element of `self` is different
-    /// from its previous element, that is the last element of `prev_vector`.
-    ///
-    /// # Panics
-    /// Panics if
-    /// - `selected.len() < self.len()`.
-    /// - `prev_vector` and `self` have different data types.
-    fn find_unique(&self, selected: &mut BitVec, prev_vector: Option<&dyn Vector>);
-
     /// Filters the vector, returns elements matching the `filter` (i.e. where the values are true).
     ///
     /// Note that the nulls of `filter` are interpreted as `false` will lead to these elements being masked out.
@@ -81,11 +61,6 @@ macro_rules! impl_scalar_vector_op {
                 replicate::replicate_scalar(self, offsets)
             }
 
-            fn find_unique(&self, selected: &mut BitVec, prev_vector: Option<&dyn Vector>) {
-                let prev_vector = prev_vector.map(|pv| pv.as_any().downcast_ref::<$VectorType>().unwrap());
-                find_unique::find_unique_scalar(self, selected, prev_vector);
-            }
-
             fn filter(&self, filter: &BooleanVector) -> Result<VectorRef> {
                 filter::filter_non_constant!(self, $VectorType, filter)
             }
@@ -121,11 +96,6 @@ impl VectorOp for Decimal128Vector {
         std::sync::Arc::new(replicate::replicate_decimal128(self, offsets))
     }
 
-    fn find_unique(&self, selected: &mut BitVec, prev_vector: Option<&dyn Vector>) {
-        let prev_vector = prev_vector.and_then(|pv| pv.as_any().downcast_ref::<Decimal128Vector>());
-        find_unique::find_unique_scalar(self, selected, prev_vector);
-    }
-
     fn filter(&self, filter: &BooleanVector) -> Result<VectorRef> {
         filter::filter_non_constant!(self, Decimal128Vector, filter)
     }
@@ -144,12 +114,6 @@ impl<T: LogicalPrimitiveType> VectorOp for PrimitiveVector<T> {
         std::sync::Arc::new(replicate::replicate_primitive(self, offsets))
     }
 
-    fn find_unique(&self, selected: &mut BitVec, prev_vector: Option<&dyn Vector>) {
-        let prev_vector =
-            prev_vector.and_then(|pv| pv.as_any().downcast_ref::<PrimitiveVector<T>>());
-        find_unique::find_unique_scalar(self, selected, prev_vector);
-    }
-
     fn filter(&self, filter: &BooleanVector) -> Result<VectorRef> {
         filter::filter_non_constant!(self, PrimitiveVector<T>, filter)
     }
@@ -168,11 +132,6 @@ impl VectorOp for NullVector {
         replicate::replicate_null(self, offsets)
     }
 
-    fn find_unique(&self, selected: &mut BitVec, prev_vector: Option<&dyn Vector>) {
-        let prev_vector = prev_vector.and_then(|pv| pv.as_any().downcast_ref::<NullVector>());
-        find_unique::find_unique_null(self, selected, prev_vector);
-    }
-
     fn filter(&self, filter: &BooleanVector) -> Result<VectorRef> {
         filter::filter_non_constant!(self, NullVector, filter)
     }
@@ -195,11 +154,6 @@ impl VectorOp for ConstantVector {
         self.replicate_vector(offsets)
     }
 
-    fn find_unique(&self, selected: &mut BitVec, prev_vector: Option<&dyn Vector>) {
-        let prev_vector = prev_vector.and_then(|pv| pv.as_any().downcast_ref::<ConstantVector>());
-        find_unique::find_unique_constant(self, selected, prev_vector);
-    }
-
     fn filter(&self, filter: &BooleanVector) -> Result<VectorRef> {
         self.filter_vector(filter)
     }

src/datatypes/src/vectors/operations/find_unique.rs

@@ -1,366 +0,0 @@
-// Copyright 2023 Greptime Team
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-use common_base::BitVec;
-
-use crate::scalars::ScalarVector;
-use crate::vectors::constant::ConstantVector;
-use crate::vectors::{NullVector, Vector};
-
-// To implement `find_unique()` correctly, we need to keep in mind that always marks an element as
-// selected when it is different from the previous one, and leaves the `selected` unchanged
-// in any other case.
-pub(crate) fn find_unique_scalar<'a, T: ScalarVector>(
-    vector: &'a T,
-    selected: &'a mut BitVec,
-    prev_vector: Option<&'a T>,
-) where
-    T::RefItem<'a>: PartialEq,
-{
-    assert!(selected.len() >= vector.len());
-
-    if vector.is_empty() {
-        return;
-    }
-
-    for ((i, current), next) in vector
-        .iter_data()
-        .enumerate()
-        .zip(vector.iter_data().skip(1))
-    {
-        if current != next {
-            // If next element is a different element, we mark it as selected.
-            selected.set(i + 1, true);
-        }
-    }
-
-    // Marks first element as selected if it is different from previous element, otherwise
-    // keep selected bitmap unchanged.
-    let is_first_not_duplicate = prev_vector
-        .map(|pv| {
-            if pv.is_empty() {
-                true
-            } else {
-                let last = pv.get_data(pv.len() - 1);
-                last != vector.get_data(0)
-            }
-        })
-        .unwrap_or(true);
-    if is_first_not_duplicate {
-        selected.set(0, true);
-    }
-}
-
-pub(crate) fn find_unique_null(
-    vector: &NullVector,
-    selected: &mut BitVec,
-    prev_vector: Option<&NullVector>,
-) {
-    if vector.is_empty() {
-        return;
-    }
-
-    let is_first_not_duplicate = prev_vector.map(NullVector::is_empty).unwrap_or(true);
-    if is_first_not_duplicate {
-        selected.set(0, true);
-    }
-}
-
-pub(crate) fn find_unique_constant(
-    vector: &ConstantVector,
-    selected: &mut BitVec,
-    prev_vector: Option<&ConstantVector>,
-) {
-    if vector.is_empty() {
-        return;
-    }
-
-    let is_first_not_duplicate = prev_vector
-        .map(|pv| {
-            if pv.is_empty() {
-                true
-            } else {
-                vector.get_constant_ref() != pv.get_constant_ref()
-            }
-        })
-        .unwrap_or(true);
-
-    if is_first_not_duplicate {
-        selected.set(0, true);
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use std::sync::Arc;
-
-    use common_time::Date;
-
-    use super::*;
-    use crate::timestamp::*;
-    use crate::vectors::{Int32Vector, StringVector, Vector, VectorOp};
-
-    fn check_bitmap(expect: &[bool], selected: &BitVec) {
-        let actual = selected.iter().collect::<Vec<_>>();
-        assert_eq!(expect, actual);
-    }
-
-    fn check_find_unique_scalar(expect: &[bool], input: &[i32], prev: Option<&[i32]>) {
-        check_find_unique_scalar_opt(expect, input.iter().map(|v| Some(*v)), prev);
-    }
-
-    fn check_find_unique_scalar_opt(
-        expect: &[bool],
-        input: impl Iterator<Item = Option<i32>>,
-        prev: Option<&[i32]>,
-    ) {
-        let input = Int32Vector::from(input.collect::<Vec<_>>());
-        let prev = prev.map(Int32Vector::from_slice);
-
-        let mut selected = BitVec::repeat(false, input.len());
-        input.find_unique(&mut selected, prev.as_ref().map(|v| v as _));
-
-        check_bitmap(expect, &selected);
-    }
-
-    #[test]
-    fn test_find_unique_scalar() {
-        check_find_unique_scalar(&[], &[], None);
-        check_find_unique_scalar(&[true], &[1], None);
-        check_find_unique_scalar(&[true, false], &[1, 1], None);
-        check_find_unique_scalar(&[true, true], &[1, 2], None);
-        check_find_unique_scalar(&[true, true, true, true], &[1, 2, 3, 4], None);
-        check_find_unique_scalar(&[true, false, true, false], &[1, 1, 3, 3], None);
-        check_find_unique_scalar(&[true, false, false, false, true], &[2, 2, 2, 2, 3], None);
-
-        check_find_unique_scalar(&[true], &[5], Some(&[]));
-        check_find_unique_scalar(&[true], &[5], Some(&[3]));
-        check_find_unique_scalar(&[false], &[5], Some(&[5]));
-        check_find_unique_scalar(&[false], &[5], Some(&[4, 5]));
-        check_find_unique_scalar(&[false, true], &[5, 6], Some(&[4, 5]));
-        check_find_unique_scalar(&[false, true, false], &[5, 6, 6], Some(&[4, 5]));
-        check_find_unique_scalar(
-            &[false, true, false, true, true],
-            &[5, 6, 6, 7, 8],
-            Some(&[4, 5]),
-        );
-
-        check_find_unique_scalar_opt(
-            &[true, true, false, true, false],
-            [Some(1), Some(2), Some(2), None, None].into_iter(),
-            None,
-        );
-    }
-
-    #[test]
-    fn test_find_unique_scalar_multi_times_with_prev() {
-        let prev = Int32Vector::from_slice([1]);
-
-        let v1 = Int32Vector::from_slice([2, 3, 4]);
-        let mut selected = BitVec::repeat(false, v1.len());
-        v1.find_unique(&mut selected, Some(&prev));
-
-        // Though element in v2 are the same as prev, but we should still keep them.
-        let v2 = Int32Vector::from_slice([1, 1, 1]);
-        v2.find_unique(&mut selected, Some(&prev));
-
-        check_bitmap(&[true, true, true], &selected);
-    }
-
-    fn new_bitmap(bits: &[bool]) -> BitVec {
-        BitVec::from_iter(bits)
-    }
-
-    #[test]
-    fn test_find_unique_scalar_with_prev() {
-        let prev = Int32Vector::from_slice([1]);
-
-        let mut selected = new_bitmap(&[true, false, true, false]);
-        let v = Int32Vector::from_slice([2, 3, 4, 5]);
-        v.find_unique(&mut selected, Some(&prev));
-        // All elements are different.
-        check_bitmap(&[true, true, true, true], &selected);
-
-        let mut selected = new_bitmap(&[true, false, true, false]);
-        let v = Int32Vector::from_slice([1, 2, 3, 4]);
-        v.find_unique(&mut selected, Some(&prev));
-        // Though first element is duplicate, but we keep the flag unchanged.
-        check_bitmap(&[true, true, true, true], &selected);
-
-        // Same case as above, but now `prev` is None.
-        let mut selected = new_bitmap(&[true, false, true, false]);
-        let v = Int32Vector::from_slice([1, 2, 3, 4]);
-        v.find_unique(&mut selected, None);
-        check_bitmap(&[true, true, true, true], &selected);
-
-        // Same case as above, but now `prev` is empty.
-        let mut selected = new_bitmap(&[true, false, true, false]);
-        let v = Int32Vector::from_slice([1, 2, 3, 4]);
-        v.find_unique(&mut selected, Some(&Int32Vector::from_slice([])));
-        check_bitmap(&[true, true, true, true], &selected);
-
-        let mut selected = new_bitmap(&[false, false, false, false]);
-        let v = Int32Vector::from_slice([2, 2, 4, 5]);
-        v.find_unique(&mut selected, Some(&prev));
-        // only v[1] is duplicate.
-        check_bitmap(&[true, false, true, true], &selected);
-    }
-
-    fn check_find_unique_null(len: usize) {
-        let input = NullVector::new(len);
-        let mut selected = BitVec::repeat(false, input.len());
-        input.find_unique(&mut selected, None);
-
-        let mut expect = vec![false; len];
-        if !expect.is_empty() {
-            expect[0] = true;
-        }
-        check_bitmap(&expect, &selected);
-
-        let mut selected = BitVec::repeat(false, input.len());
-        let prev = Some(NullVector::new(1));
-        input.find_unique(&mut selected, prev.as_ref().map(|v| v as _));
-        let expect = vec![false; len];
-        check_bitmap(&expect, &selected);
-    }
-
-    #[test]
-    fn test_find_unique_null() {
-        for len in 0..5 {
-            check_find_unique_null(len);
-        }
-    }
-
-    #[test]
-    fn test_find_unique_null_with_prev() {
-        let prev = NullVector::new(1);
-
-        // Keep flags unchanged.
-        let mut selected = new_bitmap(&[true, false, true, false]);
-        let v = NullVector::new(4);
-        v.find_unique(&mut selected, Some(&prev));
-        check_bitmap(&[true, false, true, false], &selected);
-
-        // Keep flags unchanged.
-        let mut selected = new_bitmap(&[false, false, true, false]);
-        v.find_unique(&mut selected, Some(&prev));
-        check_bitmap(&[false, false, true, false], &selected);
-
-        // Prev is None, select first element.
-        let mut selected = new_bitmap(&[false, false, true, false]);
-        v.find_unique(&mut selected, None);
-        check_bitmap(&[true, false, true, false], &selected);
-
-        // Prev is empty, select first element.
-        let mut selected = new_bitmap(&[false, false, true, false]);
-        v.find_unique(&mut selected, Some(&NullVector::new(0)));
-        check_bitmap(&[true, false, true, false], &selected);
-    }
-
-    fn check_find_unique_constant(len: usize) {
-        let input = ConstantVector::new(Arc::new(Int32Vector::from_slice([8])), len);
-        let mut selected = BitVec::repeat(false, len);
-        input.find_unique(&mut selected, None);
-
-        let mut expect = vec![false; len];
-        if !expect.is_empty() {
-            expect[0] = true;
-        }
-        check_bitmap(&expect, &selected);
-
-        let mut selected = BitVec::repeat(false, len);
-        let prev = Some(ConstantVector::new(
-            Arc::new(Int32Vector::from_slice([8])),
-            1,
-        ));
-        input.find_unique(&mut selected, prev.as_ref().map(|v| v as _));
-        let expect = vec![false; len];
-        check_bitmap(&expect, &selected);
-    }
-
-    #[test]
-    fn test_find_unique_constant() {
-        for len in 0..5 {
-            check_find_unique_constant(len);
-        }
-    }
-
-    #[test]
-    fn test_find_unique_constant_with_prev() {
-        let prev = ConstantVector::new(Arc::new(Int32Vector::from_slice([1])), 1);
-
-        // Keep flags unchanged.
-        let mut selected = new_bitmap(&[true, false, true, false]);
-        let v = ConstantVector::new(Arc::new(Int32Vector::from_slice([1])), 4);
-        v.find_unique(&mut selected, Some(&prev));
-        check_bitmap(&[true, false, true, false], &selected);
-
-        // Keep flags unchanged.
-        let mut selected = new_bitmap(&[false, false, true, false]);
-        v.find_unique(&mut selected, Some(&prev));
-        check_bitmap(&[false, false, true, false], &selected);
-
-        // Prev is None, select first element.
-        let mut selected = new_bitmap(&[false, false, true, false]);
-        v.find_unique(&mut selected, None);
-        check_bitmap(&[true, false, true, false], &selected);
-
-        // Prev is empty, select first element.
-        let mut selected = new_bitmap(&[false, false, true, false]);
-        v.find_unique(
-            &mut selected,
-            Some(&ConstantVector::new(
-                Arc::new(Int32Vector::from_slice([1])),
-                0,
-            )),
-        );
-        check_bitmap(&[true, false, true, false], &selected);
-
-        // Different constant vector.
-        let mut selected = new_bitmap(&[false, false, true, false]);
-        let v = ConstantVector::new(Arc::new(Int32Vector::from_slice([2])), 4);
-        v.find_unique(&mut selected, Some(&prev));
-        check_bitmap(&[true, false, true, false], &selected);
-    }
-
-    #[test]
-    fn test_find_unique_string() {
-        let input = StringVector::from_slice(&["a", "a", "b", "c"]);
-        let mut selected = BitVec::repeat(false, 4);
-        input.find_unique(&mut selected, None);
-        let expect = vec![true, false, true, true];
-        check_bitmap(&expect, &selected);
-    }
-
-    macro_rules! impl_find_unique_date_like_test {
-        ($VectorType: ident, $ValueType: ident, $method: ident) => {{
-            use $crate::vectors::$VectorType;
-
-            let v = $VectorType::from_iterator([8, 8, 9, 10].into_iter().map($ValueType::$method));
-            let mut selected = BitVec::repeat(false, 4);
-            v.find_unique(&mut selected, None);
-            let expect = vec![true, false, true, true];
-            check_bitmap(&expect, &selected);
-        }};
-    }
-
-    #[test]
-    fn test_find_unique_date_like() {
-        impl_find_unique_date_like_test!(DateVector, Date, new);
-        impl_find_unique_date_like_test!(TimestampSecondVector, TimestampSecond, from);
-        impl_find_unique_date_like_test!(TimestampMillisecondVector, TimestampMillisecond, from);
-        impl_find_unique_date_like_test!(TimestampMicrosecondVector, TimestampMicrosecond, from);
-        impl_find_unique_date_like_test!(TimestampNanosecondVector, TimestampNanosecond, from);
-    }
-}

src/flow/src/adapter/util.rs

@@ -20,7 +20,7 @@ use api::v1::{ColumnDataType, ColumnDataTypeExtension, CreateTableExpr, Semantic
 use common_error::ext::BoxedError;
 use common_meta::key::table_info::TableInfoValue;
 use datatypes::prelude::ConcreteDataType;
-use datatypes::schema::ColumnSchema;
+use datatypes::schema::{ColumnDefaultConstraint, ColumnSchema};
 use itertools::Itertools;
 use operator::expr_helper;
 use session::context::QueryContextBuilder;
@@ -174,7 +174,15 @@ pub fn table_info_value_to_relation_desc(
     let default_values = raw_schema
         .column_schemas
         .iter()
-        .map(|c| c.default_constraint().cloned())
+        .map(|c| {
+            c.default_constraint().cloned().or_else(|| {
+                if c.is_nullable() {
+                    Some(ColumnDefaultConstraint::null_value())
+                } else {
+                    None
+                }
+            })
+        })
         .collect_vec();
 
     Ok(TableDesc::new(relation_desc, default_values))

src/flow/src/compute/render.rs

@@ -179,7 +179,7 @@ impl Context<'_, '_> {
     ) -> CollectionBundle<Batch> {
         let (send_port, recv_port) = self.df.make_edge::<_, Toff<Batch>>("constant_batch");
         let mut per_time: BTreeMap<repr::Timestamp, Vec<DiffRow>> = Default::default();
-        for (key, group) in &rows.into_iter().group_by(|(_row, ts, _diff)| *ts) {
+        for (key, group) in &rows.into_iter().chunk_by(|(_row, ts, _diff)| *ts) {
             per_time.entry(key).or_default().extend(group);
         }
 
@@ -233,7 +233,7 @@ impl Context<'_, '_> {
     pub fn render_constant(&mut self, rows: Vec<DiffRow>) -> CollectionBundle {
         let (send_port, recv_port) = self.df.make_edge::<_, Toff>("constant");
         let mut per_time: BTreeMap<repr::Timestamp, Vec<DiffRow>> = Default::default();
-        for (key, group) in &rows.into_iter().group_by(|(_row, ts, _diff)| *ts) {
+        for (key, group) in &rows.into_iter().chunk_by(|(_row, ts, _diff)| *ts) {
             per_time.entry(key).or_default().extend(group);
         }
 

src/flow/src/expr/scalar.rs

@@ -151,12 +151,12 @@ impl ScalarExpr {
 
     /// apply optimization to the expression, like flatten variadic function
     pub fn optimize(&mut self) {
-        self.flatten_varidic_fn();
+        self.flatten_variadic_fn();
     }
 
     /// Because Substrait's `And`/`Or` function is binary, but FlowPlan's
     /// `And`/`Or` function is variadic, we need to flatten the `And` function if multiple `And`/`Or` functions are nested.
-    fn flatten_varidic_fn(&mut self) {
+    fn flatten_variadic_fn(&mut self) {
         if let ScalarExpr::CallVariadic { func, exprs } = self {
             let mut new_exprs = vec![];
             for expr in std::mem::take(exprs) {
@@ -167,7 +167,7 @@ impl ScalarExpr {
                 {
                     if *func == inner_func {
                         for inner_expr in inner_exprs.iter_mut() {
-                            inner_expr.flatten_varidic_fn();
+                            inner_expr.flatten_variadic_fn();
                         }
                         new_exprs.extend(inner_exprs);
                     }

src/flow/src/lib.rs

@@ -33,6 +33,7 @@ mod expr;
 pub mod heartbeat;
 mod metrics;
 mod plan;
+mod recording_rules;
 mod repr;
 mod server;
 mod transform;

src/flow/src/recording_rules.rs

@@ -0,0 +1,24 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+//! Run flow as recording rule which is time-window-aware normal query triggered when new data arrives
+
+use std::time::Duration;
+
+mod frontend_client;
+mod utils;
+
+/// TODO(discord9): make those constants configurable
+/// The default rule engine query timeout is 10 minutes
+pub const DEFAULT_RULE_ENGINE_QUERY_TIMEOUT: Duration = Duration::from_secs(10 * 60);

src/flow/src/recording_rules/frontend_client.rs

@@ -0,0 +1,148 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+//! Frontend client to run flow as recording rule which is time-window-aware normal query triggered every tick set by user
+
+use std::sync::Arc;
+
+use client::{Client, Database, DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
+use common_error::ext::BoxedError;
+use common_grpc::channel_manager::{ChannelConfig, ChannelManager};
+use common_meta::cluster::{NodeInfo, NodeInfoKey, Role};
+use common_meta::peer::Peer;
+use common_meta::rpc::store::RangeRequest;
+use meta_client::client::MetaClient;
+use snafu::ResultExt;
+
+use crate::error::{ExternalSnafu, UnexpectedSnafu};
+use crate::recording_rules::DEFAULT_RULE_ENGINE_QUERY_TIMEOUT;
+use crate::Error;
+
+fn default_channel_mgr() -> ChannelManager {
+    let cfg = ChannelConfig::new().timeout(DEFAULT_RULE_ENGINE_QUERY_TIMEOUT);
+    ChannelManager::with_config(cfg)
+}
+
+fn client_from_urls(addrs: Vec<String>) -> Client {
+    Client::with_manager_and_urls(default_channel_mgr(), addrs)
+}
+
+/// A simple frontend client able to execute sql using grpc protocol
+#[derive(Debug)]
+pub enum FrontendClient {
+    Distributed {
+        meta_client: Arc<MetaClient>,
+    },
+    Standalone {
+        /// for the sake of simplicity still use grpc even in standalone mode
+        /// notice the client here should all be lazy, so that can wait after frontend is booted then make conn
+        /// TODO(discord9): not use grpc under standalone mode
+        database_client: DatabaseWithPeer,
+    },
+}
+
+#[derive(Debug, Clone)]
+pub struct DatabaseWithPeer {
+    pub database: Database,
+    pub peer: Peer,
+}
+
+impl DatabaseWithPeer {
+    fn new(database: Database, peer: Peer) -> Self {
+        Self { database, peer }
+    }
+}
+
+impl FrontendClient {
+    pub fn from_meta_client(meta_client: Arc<MetaClient>) -> Self {
+        Self::Distributed { meta_client }
+    }
+
+    pub fn from_static_grpc_addr(addr: String) -> Self {
+        let peer = Peer {
+            id: 0,
+            addr: addr.clone(),
+        };
+
+        let client = client_from_urls(vec![addr]);
+        let database = Database::new(DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME, client);
+        Self::Standalone {
+            database_client: DatabaseWithPeer::new(database, peer),
+        }
+    }
+}
+
+impl FrontendClient {
+    async fn scan_for_frontend(&self) -> Result<Vec<(NodeInfoKey, NodeInfo)>, Error> {
+        let Self::Distributed { meta_client, .. } = self else {
+            return Ok(vec![]);
+        };
+        let cluster_client = meta_client
+            .cluster_client()
+            .map_err(BoxedError::new)
+            .context(ExternalSnafu)?;
+
+        let prefix = NodeInfoKey::key_prefix_with_role(Role::Frontend);
+        let req = RangeRequest::new().with_prefix(prefix);
+        let resp = cluster_client
+            .range(req)
+            .await
+            .map_err(BoxedError::new)
+            .context(ExternalSnafu)?;
+        let mut res = Vec::with_capacity(resp.kvs.len());
+        for kv in resp.kvs {
+            let key = NodeInfoKey::try_from(kv.key)
+                .map_err(BoxedError::new)
+                .context(ExternalSnafu)?;
+
+            let val = NodeInfo::try_from(kv.value)
+                .map_err(BoxedError::new)
+                .context(ExternalSnafu)?;
+            res.push((key, val));
+        }
+        Ok(res)
+    }
+
+    /// Get the database with max `last_activity_ts`
+    async fn get_last_active_frontend(&self) -> Result<DatabaseWithPeer, Error> {
+        if let Self::Standalone { database_client } = self {
+            return Ok(database_client.clone());
+        }
+
+        let frontends = self.scan_for_frontend().await?;
+        let mut peer = None;
+
+        if let Some((_, val)) = frontends.iter().max_by_key(|(_, val)| val.last_activity_ts) {
+            peer = Some(val.peer.clone());
+        }
+
+        let Some(peer) = peer else {
+            UnexpectedSnafu {
+                reason: format!("No frontend available: {:?}", frontends),
+            }
+            .fail()?
+        };
+        let client = client_from_urls(vec![peer.addr.clone()]);
+        let database = Database::new(DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME, client);
+        Ok(DatabaseWithPeer::new(database, peer))
+    }
+
+    /// Get a database client, and possibly update it before returning.
+    pub async fn get_database_client(&self) -> Result<DatabaseWithPeer, Error> {
+        match self {
+            Self::Standalone { database_client } => Ok(database_client.clone()),
+            Self::Distributed { meta_client: _ } => self.get_last_active_frontend().await,
+        }
+    }
+}

src/flow/src/recording_rules/utils.rs

@@ -0,0 +1,600 @@
+// 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.
+
+//! some utils for helping with recording rule
+
+use std::collections::HashSet;
+use std::sync::Arc;
+
+use common_error::ext::BoxedError;
+use common_telemetry::{debug, info};
+use datafusion::error::Result as DfResult;
+use datafusion::logical_expr::Expr;
+use datafusion::sql::unparser::Unparser;
+use datafusion_common::tree_node::{
+    Transformed, TreeNodeRecursion, TreeNodeRewriter, TreeNodeVisitor,
+};
+use datafusion_common::DataFusionError;
+use datafusion_expr::{Distinct, LogicalPlan};
+use datatypes::schema::RawSchema;
+use query::parser::QueryLanguageParser;
+use query::QueryEngineRef;
+use session::context::QueryContextRef;
+use snafu::ResultExt;
+
+use crate::adapter::AUTO_CREATED_PLACEHOLDER_TS_COL;
+use crate::df_optimizer::apply_df_optimizer;
+use crate::error::{DatafusionSnafu, ExternalSnafu};
+use crate::Error;
+
+/// Convert sql to datafusion logical plan
+pub async fn sql_to_df_plan(
+    query_ctx: QueryContextRef,
+    engine: QueryEngineRef,
+    sql: &str,
+    optimize: bool,
+) -> Result<LogicalPlan, Error> {
+    let stmt = QueryLanguageParser::parse_sql(sql, &query_ctx)
+        .map_err(BoxedError::new)
+        .context(ExternalSnafu)?;
+    let plan = engine
+        .planner()
+        .plan(&stmt, query_ctx)
+        .await
+        .map_err(BoxedError::new)
+        .context(ExternalSnafu)?;
+    let plan = if optimize {
+        apply_df_optimizer(plan).await?
+    } else {
+        plan
+    };
+    Ok(plan)
+}
+
+pub fn df_plan_to_sql(plan: &LogicalPlan) -> Result<String, Error> {
+    /// A dialect that forces identifiers to be quoted when have uppercase
+    struct ForceQuoteIdentifiers;
+    impl datafusion::sql::unparser::dialect::Dialect for ForceQuoteIdentifiers {
+        fn identifier_quote_style(&self, identifier: &str) -> Option<char> {
+            if identifier.to_lowercase() != identifier {
+                Some('`')
+            } else {
+                None
+            }
+        }
+    }
+    let unparser = Unparser::new(&ForceQuoteIdentifiers);
+    // first make all column qualified
+    let sql = unparser
+        .plan_to_sql(plan)
+        .with_context(|_e| DatafusionSnafu {
+            context: format!("Failed to unparse logical plan {plan:?}"),
+        })?;
+    Ok(sql.to_string())
+}
+
+/// Helper to find the innermost group by expr in schema, return None if no group by expr
+#[derive(Debug, Clone, Default)]
+pub struct FindGroupByFinalName {
+    group_exprs: Option<HashSet<datafusion_expr::Expr>>,
+}
+
+impl FindGroupByFinalName {
+    pub fn get_group_expr_names(&self) -> Option<HashSet<String>> {
+        self.group_exprs
+            .as_ref()
+            .map(|exprs| exprs.iter().map(|expr| expr.qualified_name().1).collect())
+    }
+}
+
+impl TreeNodeVisitor<'_> for FindGroupByFinalName {
+    type Node = LogicalPlan;
+
+    fn f_down(&mut self, node: &Self::Node) -> datafusion_common::Result<TreeNodeRecursion> {
+        if let LogicalPlan::Aggregate(aggregate) = node {
+            self.group_exprs = Some(aggregate.group_expr.iter().cloned().collect());
+            debug!("Group by exprs: {:?}", self.group_exprs);
+        } else if let LogicalPlan::Distinct(distinct) = node {
+            debug!("Distinct: {:#?}", distinct);
+            match distinct {
+                Distinct::All(input) => {
+                    if let LogicalPlan::TableScan(table_scan) = &**input {
+                        // get column from field_qualifier, projection and projected_schema:
+                        let len = table_scan.projected_schema.fields().len();
+                        let columns = (0..len)
+                            .map(|f| {
+                                let (qualifier, field) =
+                                    table_scan.projected_schema.qualified_field(f);
+                                datafusion_common::Column::new(qualifier.cloned(), field.name())
+                            })
+                            .map(datafusion_expr::Expr::Column);
+                        self.group_exprs = Some(columns.collect());
+                    } else {
+                        self.group_exprs = Some(input.expressions().iter().cloned().collect())
+                    }
+                }
+                Distinct::On(distinct_on) => {
+                    self.group_exprs = Some(distinct_on.on_expr.iter().cloned().collect())
+                }
+            }
+            debug!("Group by exprs: {:?}", self.group_exprs);
+        }
+
+        Ok(TreeNodeRecursion::Continue)
+    }
+
+    /// deal with projection when going up with group exprs
+    fn f_up(&mut self, node: &Self::Node) -> datafusion_common::Result<TreeNodeRecursion> {
+        if let LogicalPlan::Projection(projection) = node {
+            for expr in &projection.expr {
+                let Some(group_exprs) = &mut self.group_exprs else {
+                    return Ok(TreeNodeRecursion::Continue);
+                };
+                if let datafusion_expr::Expr::Alias(alias) = expr {
+                    // if a alias exist, replace with the new alias
+                    let mut new_group_exprs = group_exprs.clone();
+                    for group_expr in group_exprs.iter() {
+                        if group_expr.name_for_alias()? == alias.expr.name_for_alias()? {
+                            new_group_exprs.remove(group_expr);
+                            new_group_exprs.insert(expr.clone());
+                            break;
+                        }
+                    }
+                    *group_exprs = new_group_exprs;
+                }
+            }
+        }
+        debug!("Aliased group by exprs: {:?}", self.group_exprs);
+        Ok(TreeNodeRecursion::Continue)
+    }
+}
+
+/// Add to the final select columns like `update_at`(which doesn't necessary need to have exact name just need to be a extra timestamp column) and `__ts_placeholder`(this column need to have exact this name and be a timestamp) with values like `now()` and `0`
+#[derive(Debug)]
+pub struct AddAutoColumnRewriter {
+    pub schema: RawSchema,
+    pub is_rewritten: bool,
+}
+
+impl AddAutoColumnRewriter {
+    pub fn new(schema: RawSchema) -> Self {
+        Self {
+            schema,
+            is_rewritten: false,
+        }
+    }
+}
+
+impl TreeNodeRewriter for AddAutoColumnRewriter {
+    type Node = LogicalPlan;
+    fn f_down(&mut self, node: Self::Node) -> DfResult<Transformed<Self::Node>> {
+        if self.is_rewritten {
+            return Ok(Transformed::no(node));
+        }
+
+        // if is distinct all, go one level down
+        if let LogicalPlan::Distinct(Distinct::All(_)) = node {
+            return Ok(Transformed::no(node));
+        }
+
+        // FIXME(discord9): just read plan.expr and do stuffs
+        let mut exprs = node.expressions();
+
+        // add columns if have different column count
+        let query_col_cnt = exprs.len();
+        let table_col_cnt = self.schema.column_schemas.len();
+        info!("query_col_cnt={query_col_cnt}, table_col_cnt={table_col_cnt}");
+        if query_col_cnt == table_col_cnt {
+            self.is_rewritten = true;
+            return Ok(Transformed::no(node));
+        } else if query_col_cnt + 1 == table_col_cnt {
+            let last_col_schema = self.schema.column_schemas.last().unwrap();
+
+            // if time index column is auto created add it
+            if last_col_schema.name == AUTO_CREATED_PLACEHOLDER_TS_COL
+                && self.schema.timestamp_index == Some(table_col_cnt - 1)
+            {
+                exprs.push(datafusion::logical_expr::lit(0));
+            } else if last_col_schema.data_type.is_timestamp() {
+                // is the update at column
+                exprs.push(datafusion::prelude::now());
+            } else {
+                // helpful error message
+                return Err(DataFusionError::Plan(format!(
+                    "Expect the last column in table to be timestamp column, found column {} with type {:?}",
+                    last_col_schema.name,
+                    last_col_schema.data_type
+                )));
+            }
+        } else if query_col_cnt + 2 == table_col_cnt {
+            let mut col_iter = self.schema.column_schemas.iter().rev();
+            let last_col_schema = col_iter.next().unwrap();
+            let second_last_col_schema = col_iter.next().unwrap();
+            if second_last_col_schema.data_type.is_timestamp() {
+                exprs.push(datafusion::prelude::now());
+            } else {
+                return Err(DataFusionError::Plan(format!(
+                    "Expect the second last column in the table to be timestamp column, found column {} with type {:?}",
+                    second_last_col_schema.name,
+                    second_last_col_schema.data_type
+                )));
+            }
+
+            if last_col_schema.name == AUTO_CREATED_PLACEHOLDER_TS_COL
+                && self.schema.timestamp_index == Some(table_col_cnt - 1)
+            {
+                exprs.push(datafusion::logical_expr::lit(0));
+            } else {
+                return Err(DataFusionError::Plan(format!(
+                    "Expect timestamp column {}, found {:?}",
+                    AUTO_CREATED_PLACEHOLDER_TS_COL, last_col_schema
+                )));
+            }
+        } else {
+            return Err(DataFusionError::Plan(format!(
+                    "Expect table have 0,1 or 2 columns more than query columns, found {} query columns {:?}, {} table columns {:?}",
+                    query_col_cnt, node.expressions(), table_col_cnt, self.schema.column_schemas
+                )));
+        }
+
+        self.is_rewritten = true;
+        let new_plan = node.with_new_exprs(exprs, node.inputs().into_iter().cloned().collect())?;
+        Ok(Transformed::yes(new_plan))
+    }
+}
+
+// TODO(discord9): a method to found out the precise time window
+
+/// Find out the `Filter` Node corresponding to innermost(deepest) `WHERE` and add a new filter expr to it
+#[derive(Debug)]
+pub struct AddFilterRewriter {
+    extra_filter: Expr,
+    is_rewritten: bool,
+}
+
+impl AddFilterRewriter {
+    fn new(filter: Expr) -> Self {
+        Self {
+            extra_filter: filter,
+            is_rewritten: false,
+        }
+    }
+}
+
+impl TreeNodeRewriter for AddFilterRewriter {
+    type Node = LogicalPlan;
+    fn f_up(&mut self, node: Self::Node) -> DfResult<Transformed<Self::Node>> {
+        if self.is_rewritten {
+            return Ok(Transformed::no(node));
+        }
+        match node {
+            LogicalPlan::Filter(mut filter) if !filter.having => {
+                filter.predicate = filter.predicate.and(self.extra_filter.clone());
+                self.is_rewritten = true;
+                Ok(Transformed::yes(LogicalPlan::Filter(filter)))
+            }
+            LogicalPlan::TableScan(_) => {
+                // add a new filter
+                let filter =
+                    datafusion_expr::Filter::try_new(self.extra_filter.clone(), Arc::new(node))?;
+                self.is_rewritten = true;
+                Ok(Transformed::yes(LogicalPlan::Filter(filter)))
+            }
+            _ => Ok(Transformed::no(node)),
+        }
+    }
+}
+
+#[cfg(test)]
+mod test {
+    use datafusion_common::tree_node::TreeNode as _;
+    use datatypes::prelude::ConcreteDataType;
+    use datatypes::schema::ColumnSchema;
+    use pretty_assertions::assert_eq;
+    use session::context::QueryContext;
+
+    use super::*;
+    use crate::test_utils::create_test_query_engine;
+
+    /// test if uppercase are handled correctly(with quote)
+    #[tokio::test]
+    async fn test_sql_plan_convert() {
+        let query_engine = create_test_query_engine();
+        let ctx = QueryContext::arc();
+        let old = r#"SELECT "NUMBER" FROM "UPPERCASE_NUMBERS_WITH_TS""#;
+        let new = sql_to_df_plan(ctx.clone(), query_engine.clone(), old, false)
+            .await
+            .unwrap();
+        let new_sql = df_plan_to_sql(&new).unwrap();
+
+        assert_eq!(
+            r#"SELECT `UPPERCASE_NUMBERS_WITH_TS`.`NUMBER` FROM `UPPERCASE_NUMBERS_WITH_TS`"#,
+            new_sql
+        );
+    }
+
+    #[tokio::test]
+    async fn test_add_filter() {
+        let testcases = vec![
+            (
+                "SELECT number FROM numbers_with_ts GROUP BY number",
+                "SELECT numbers_with_ts.number FROM numbers_with_ts WHERE (number > 4) GROUP BY numbers_with_ts.number"
+            ),
+
+            (
+                "SELECT number FROM numbers_with_ts WHERE number < 2 OR number >10",
+                "SELECT numbers_with_ts.number FROM numbers_with_ts WHERE ((numbers_with_ts.number < 2) OR (numbers_with_ts.number > 10)) AND (number > 4)"
+            ),
+
+            (
+                "SELECT date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window",
+                "SELECT date_bin('5 minutes', numbers_with_ts.ts) AS time_window FROM numbers_with_ts WHERE (number > 4) GROUP BY date_bin('5 minutes', numbers_with_ts.ts)"
+            ),
+
+            // subquery
+            (
+                "SELECT number, time_window FROM (SELECT number, date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window, number);", 
+                "SELECT numbers_with_ts.number, time_window FROM (SELECT numbers_with_ts.number, date_bin('5 minutes', numbers_with_ts.ts) AS time_window FROM numbers_with_ts WHERE (number > 4) GROUP BY date_bin('5 minutes', numbers_with_ts.ts), numbers_with_ts.number)"
+            ),
+
+            // complex subquery without alias
+            (
+                "SELECT sum(number), number, date_bin('5 minutes', ts) as time_window, bucket_name FROM (SELECT number, ts, case when number < 5 THEN 'bucket_0_5' when number >= 5 THEN 'bucket_5_inf' END as bucket_name FROM numbers_with_ts) GROUP BY number, time_window, bucket_name;",
+                "SELECT sum(numbers_with_ts.number), numbers_with_ts.number, date_bin('5 minutes', numbers_with_ts.ts) AS time_window, bucket_name FROM (SELECT numbers_with_ts.number, numbers_with_ts.ts, CASE WHEN (numbers_with_ts.number < 5) THEN 'bucket_0_5' WHEN (numbers_with_ts.number >= 5) THEN 'bucket_5_inf' END AS bucket_name FROM numbers_with_ts WHERE (number > 4)) GROUP BY numbers_with_ts.number, date_bin('5 minutes', numbers_with_ts.ts), bucket_name"
+            ),
+
+            // complex subquery alias
+            (
+                "SELECT sum(number), number, date_bin('5 minutes', ts) as time_window, bucket_name FROM (SELECT number, ts, case when number < 5 THEN 'bucket_0_5' when number >= 5 THEN 'bucket_5_inf' END as bucket_name FROM numbers_with_ts) as cte WHERE number > 1 GROUP BY number, time_window, bucket_name;",
+                "SELECT sum(cte.number), cte.number, date_bin('5 minutes', cte.ts) AS time_window, cte.bucket_name FROM (SELECT numbers_with_ts.number, numbers_with_ts.ts, CASE WHEN (numbers_with_ts.number < 5) THEN 'bucket_0_5' WHEN (numbers_with_ts.number >= 5) THEN 'bucket_5_inf' END AS bucket_name FROM numbers_with_ts WHERE (number > 4)) AS cte WHERE (cte.number > 1) GROUP BY cte.number, date_bin('5 minutes', cte.ts), cte.bucket_name"
+            )
+        ];
+        use datafusion_expr::{col, lit};
+        let query_engine = create_test_query_engine();
+        let ctx = QueryContext::arc();
+
+        for (before, after) in testcases {
+            let sql = before;
+            let plan = sql_to_df_plan(ctx.clone(), query_engine.clone(), sql, false)
+                .await
+                .unwrap();
+
+            let mut add_filter = AddFilterRewriter::new(col("number").gt(lit(4u32)));
+            let plan = plan.rewrite(&mut add_filter).unwrap().data;
+            let new_sql = df_plan_to_sql(&plan).unwrap();
+            assert_eq!(after, new_sql);
+        }
+    }
+
+    #[tokio::test]
+    async fn test_add_auto_column_rewriter() {
+        let testcases = vec![
+            // add update_at
+            (
+                "SELECT number FROM numbers_with_ts",
+                Ok("SELECT numbers_with_ts.number, now() FROM numbers_with_ts"),
+                vec![
+                    ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
+                    ColumnSchema::new(
+                        "ts",
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    )
+                    .with_time_index(true),
+                ],
+            ),
+            // add ts placeholder
+            (
+                "SELECT number FROM numbers_with_ts",
+                Ok("SELECT numbers_with_ts.number, 0 FROM numbers_with_ts"),
+                vec![
+                    ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
+                    ColumnSchema::new(
+                        AUTO_CREATED_PLACEHOLDER_TS_COL,
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    )
+                    .with_time_index(true),
+                ],
+            ),
+            // no modify
+            (
+                "SELECT number, ts FROM numbers_with_ts",
+                Ok("SELECT numbers_with_ts.number, numbers_with_ts.ts FROM numbers_with_ts"),
+                vec![
+                    ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
+                    ColumnSchema::new(
+                        "ts",
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    )
+                    .with_time_index(true),
+                ],
+            ),
+            // add update_at and ts placeholder
+            (
+                "SELECT number FROM numbers_with_ts",
+                Ok("SELECT numbers_with_ts.number, now(), 0 FROM numbers_with_ts"),
+                vec![
+                    ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
+                    ColumnSchema::new(
+                        "update_at",
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    ),
+                    ColumnSchema::new(
+                        AUTO_CREATED_PLACEHOLDER_TS_COL,
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    )
+                    .with_time_index(true),
+                ],
+            ),
+            // add ts placeholder
+            (
+                "SELECT number, ts FROM numbers_with_ts",
+                Ok("SELECT numbers_with_ts.number, numbers_with_ts.ts, 0 FROM numbers_with_ts"),
+                vec![
+                    ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
+                    ColumnSchema::new(
+                        "update_at",
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    ),
+                    ColumnSchema::new(
+                        AUTO_CREATED_PLACEHOLDER_TS_COL,
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    )
+                    .with_time_index(true),
+                ],
+            ),
+            // add update_at after time index column
+            (
+                "SELECT number, ts FROM numbers_with_ts",
+                Ok("SELECT numbers_with_ts.number, numbers_with_ts.ts, now() FROM numbers_with_ts"),
+                vec![
+                    ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
+                    ColumnSchema::new(
+                        "ts",
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    )
+                    .with_time_index(true),
+                    ColumnSchema::new(
+                        // name is irrelevant for update_at column
+                        "update_atat",
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    ),
+                ],
+            ),
+            // error datatype mismatch
+            (
+                "SELECT number, ts FROM numbers_with_ts",
+                Err("Expect the last column in table to be timestamp column, found column atat with type Int8"),
+                vec![
+                    ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
+                    ColumnSchema::new(
+                        "ts",
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    )
+                    .with_time_index(true),
+                    ColumnSchema::new(
+                        // name is irrelevant for update_at column
+                        "atat",
+                        ConcreteDataType::int8_datatype(),
+                        false,
+                    ),
+                ],
+            ),
+            // error datatype mismatch on second last column
+            (
+                "SELECT number FROM numbers_with_ts",
+                Err("Expect the second last column in the table to be timestamp column, found column ts with type Int8"),
+                vec![
+                    ColumnSchema::new("number", ConcreteDataType::int32_datatype(), true),
+                    ColumnSchema::new(
+                        "ts",
+                        ConcreteDataType::int8_datatype(),
+                        false,
+                    ),
+                    ColumnSchema::new(
+                        // name is irrelevant for update_at column
+                        "atat",
+                        ConcreteDataType::timestamp_millisecond_datatype(),
+                        false,
+                    )
+                    .with_time_index(true),
+                ],
+            ),
+        ];
+
+        let query_engine = create_test_query_engine();
+        let ctx = QueryContext::arc();
+        for (before, after, column_schemas) in testcases {
+            let raw_schema = RawSchema::new(column_schemas);
+            let mut add_auto_column_rewriter = AddAutoColumnRewriter::new(raw_schema);
+
+            let plan = sql_to_df_plan(ctx.clone(), query_engine.clone(), before, false)
+                .await
+                .unwrap();
+            let new_sql = (|| {
+                let plan = plan
+                    .rewrite(&mut add_auto_column_rewriter)
+                    .map_err(|e| e.to_string())?
+                    .data;
+                df_plan_to_sql(&plan).map_err(|e| e.to_string())
+            })();
+            match (after, new_sql.clone()) {
+                (Ok(after), Ok(new_sql)) => assert_eq!(after, new_sql),
+                (Err(expected), Err(real_err_msg)) => assert!(
+                    real_err_msg.contains(expected),
+                    "expected: {expected}, real: {real_err_msg}"
+                ),
+                _ => panic!("expected: {:?}, real: {:?}", after, new_sql),
+            }
+        }
+    }
+
+    #[tokio::test]
+    async fn test_find_group_by_exprs() {
+        let testcases = vec![
+            (
+                "SELECT arrow_cast(date_bin(INTERVAL '1 MINS', numbers_with_ts.ts), 'Timestamp(Second, None)') AS ts FROM numbers_with_ts GROUP BY ts;", 
+                vec!["ts"]
+            ),
+            (
+                "SELECT number FROM numbers_with_ts GROUP BY number",
+                vec!["number"]
+            ),
+            (
+                "SELECT date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window",
+                vec!["time_window"]
+            ),
+             // subquery
+            (
+                "SELECT number, time_window FROM (SELECT number, date_bin('5 minutes', ts) as time_window FROM numbers_with_ts GROUP BY time_window, number);", 
+                vec!["time_window", "number"]
+            ),
+            // complex subquery without alias
+            (
+                "SELECT sum(number), number, date_bin('5 minutes', ts) as time_window, bucket_name FROM (SELECT number, ts, case when number < 5 THEN 'bucket_0_5' when number >= 5 THEN 'bucket_5_inf' END as bucket_name FROM numbers_with_ts) GROUP BY number, time_window, bucket_name;",
+                vec!["number", "time_window", "bucket_name"]
+            ),
+            // complex subquery alias
+            (
+                "SELECT sum(number), number, date_bin('5 minutes', ts) as time_window, bucket_name FROM (SELECT number, ts, case when number < 5 THEN 'bucket_0_5' when number >= 5 THEN 'bucket_5_inf' END as bucket_name FROM numbers_with_ts) as cte GROUP BY number, time_window, bucket_name;",
+                vec!["number", "time_window", "bucket_name"]
+            )
+        ];
+
+        let query_engine = create_test_query_engine();
+        let ctx = QueryContext::arc();
+        for (sql, expected) in testcases {
+            // need to be unoptimize for better readiability
+            let plan = sql_to_df_plan(ctx.clone(), query_engine.clone(), sql, false)
+                .await
+                .unwrap();
+            let mut group_by_exprs = FindGroupByFinalName::default();
+            plan.visit(&mut group_by_exprs).unwrap();
+            let expected: HashSet<String> = expected.into_iter().map(|s| s.to_string()).collect();
+            assert_eq!(
+                expected,
+                group_by_exprs.get_group_expr_names().unwrap_or_default()
+            );
+        }
+    }
+}

src/flow/src/test_utils.rs

@@ -86,7 +86,8 @@ pub fn create_test_query_engine() -> Arc<dyn QueryEngine> {
 
     let schema = vec![
         datatypes::schema::ColumnSchema::new("number", CDT::uint32_datatype(), false),
-        datatypes::schema::ColumnSchema::new("ts", CDT::timestamp_millisecond_datatype(), false),
+        datatypes::schema::ColumnSchema::new("ts", CDT::timestamp_millisecond_datatype(), false)
+            .with_time_index(true),
     ];
     let mut columns = vec![];
     let numbers = (1..=10).collect_vec();
@@ -114,6 +115,37 @@ pub fn create_test_query_engine() -> Arc<dyn QueryEngine> {
     };
     catalog_list.register_table_sync(req_with_ts).unwrap();
 
+    let schema = vec![
+        datatypes::schema::ColumnSchema::new("NUMBER", CDT::uint32_datatype(), false),
+        datatypes::schema::ColumnSchema::new("ts", CDT::timestamp_millisecond_datatype(), false)
+            .with_time_index(true),
+    ];
+    let mut columns = vec![];
+    let numbers = (1..=10).collect_vec();
+    let column: VectorRef = Arc::new(<u32 as Scalar>::VectorType::from_vec(numbers));
+    columns.push(column);
+
+    let ts = (1..=10).collect_vec();
+    let mut builder = TimestampMillisecondVectorBuilder::with_capacity(10);
+    ts.into_iter()
+        .map(|v| builder.push(Some(TimestampMillisecond::new(v))))
+        .count();
+    let column: VectorRef = builder.to_vector_cloned();
+    columns.push(column);
+
+    let schema = Arc::new(Schema::new(schema));
+    let recordbatch = common_recordbatch::RecordBatch::new(schema, columns).unwrap();
+    let table = MemTable::table("UPPERCASE_NUMBERS_WITH_TS", recordbatch);
+
+    let req_with_ts = RegisterTableRequest {
+        catalog: DEFAULT_CATALOG_NAME.to_string(),
+        schema: DEFAULT_SCHEMA_NAME.to_string(),
+        table_name: "UPPERCASE_NUMBERS_WITH_TS".to_string(),
+        table_id: 1025,
+        table,
+    };
+    catalog_list.register_table_sync(req_with_ts).unwrap();
+
     let factory = query::QueryEngineFactory::new(catalog_list, None, None, None, None, false);
 
     let engine = factory.query_engine();

src/frontend/src/frontend.rs

@@ -12,17 +12,26 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::sync::Arc;
+
 use common_base::readable_size::ReadableSize;
 use common_config::config::Configurable;
 use common_options::datanode::DatanodeClientOptions;
 use common_telemetry::logging::{LoggingOptions, TracingOptions};
 use meta_client::MetaClientOptions;
 use serde::{Deserialize, Serialize};
-use servers::export_metrics::ExportMetricsOption;
+use servers::export_metrics::{ExportMetricsOption, ExportMetricsTask};
 use servers::grpc::GrpcOptions;
 use servers::heartbeat_options::HeartbeatOptions;
 use servers::http::HttpOptions;
+use servers::server::ServerHandlers;
+use snafu::ResultExt;
 
+use crate::error;
+use crate::error::Result;
+use crate::heartbeat::HeartbeatTask;
+use crate::instance::prom_store::ExportMetricHandler;
+use crate::instance::Instance;
 use crate::service_config::{
     InfluxdbOptions, JaegerOptions, MysqlOptions, OpentsdbOptions, OtlpOptions, PostgresOptions,
     PromStoreOptions,
@@ -84,6 +93,50 @@ impl Configurable for FrontendOptions {
     }
 }
 
+/// The [`Frontend`] struct is the main entry point for the frontend service
+/// which contains server handlers, frontend instance and some background tasks.
+pub struct Frontend {
+    pub instance: Arc<Instance>,
+    pub servers: ServerHandlers,
+    pub heartbeat_task: Option<HeartbeatTask>,
+    pub export_metrics_task: Option<ExportMetricsTask>,
+}
+
+impl Frontend {
+    pub async fn start(&self) -> Result<()> {
+        if let Some(t) = &self.heartbeat_task {
+            t.start().await?;
+        }
+
+        if let Some(t) = self.export_metrics_task.as_ref() {
+            if t.send_by_handler {
+                let inserter = self.instance.inserter().clone();
+                let statement_executor = self.instance.statement_executor().clone();
+                let handler = ExportMetricHandler::new_handler(inserter, statement_executor);
+                t.start(Some(handler)).context(error::StartServerSnafu)?
+            } else {
+                t.start(None).context(error::StartServerSnafu)?;
+            }
+        }
+
+        self.servers
+            .start_all()
+            .await
+            .context(error::StartServerSnafu)
+    }
+
+    pub async fn shutdown(&self) -> Result<()> {
+        self.servers
+            .shutdown_all()
+            .await
+            .context(error::ShutdownServerSnafu)
+    }
+
+    pub fn server_handlers(&self) -> &ServerHandlers {
+        &self.servers
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

src/frontend/src/instance.rs

@@ -20,7 +20,7 @@ mod log_handler;
 mod logs;
 mod opentsdb;
 mod otlp;
-mod prom_store;
+pub mod prom_store;
 mod promql;
 mod region_query;
 pub mod standalone;
@@ -47,7 +47,7 @@ use datafusion_expr::LogicalPlan;
 use log_store::raft_engine::RaftEngineBackend;
 use operator::delete::DeleterRef;
 use operator::insert::InserterRef;
-use operator::statement::StatementExecutor;
+use operator::statement::{StatementExecutor, StatementExecutorRef};
 use pipeline::pipeline_operator::PipelineOperator;
 use prometheus::HistogramTimer;
 use promql_parser::label::Matcher;
@@ -59,18 +59,11 @@ use query::stats::StatementStatistics;
 use query::QueryEngineRef;
 use servers::error as server_error;
 use servers::error::{AuthSnafu, ExecuteQuerySnafu, ParsePromQLSnafu};
-use servers::export_metrics::ExportMetricsTask;
 use servers::interceptor::{
     PromQueryInterceptor, PromQueryInterceptorRef, SqlQueryInterceptor, SqlQueryInterceptorRef,
 };
 use servers::prometheus_handler::PrometheusHandler;
-use servers::query_handler::grpc::GrpcQueryHandler;
 use servers::query_handler::sql::SqlQueryHandler;
-use servers::query_handler::{
-    InfluxdbLineProtocolHandler, JaegerQueryHandler, LogQueryHandler, OpenTelemetryProtocolHandler,
-    OpentsdbProtocolHandler, PipelineHandler, PromStoreProtocolHandler,
-};
-use servers::server::ServerHandlers;
 use session::context::QueryContextRef;
 use session::table_name::table_idents_to_full_name;
 use snafu::prelude::*;
@@ -81,50 +74,25 @@ use sql::statements::statement::Statement;
 use sqlparser::ast::ObjectName;
 pub use standalone::StandaloneDatanodeManager;
 
-use self::prom_store::ExportMetricHandler;
 use crate::error::{
     self, Error, ExecLogicalPlanSnafu, ExecutePromqlSnafu, ExternalSnafu, InvalidSqlSnafu,
     ParseSqlSnafu, PermissionSnafu, PlanStatementSnafu, Result, SqlExecInterceptedSnafu,
-    StartServerSnafu, TableOperationSnafu,
+    TableOperationSnafu,
 };
-use crate::frontend::FrontendOptions;
-use crate::heartbeat::HeartbeatTask;
 use crate::limiter::LimiterRef;
 
-#[async_trait]
-pub trait FrontendInstance:
-    GrpcQueryHandler<Error = Error>
-    + SqlQueryHandler<Error = Error>
-    + OpentsdbProtocolHandler
-    + InfluxdbLineProtocolHandler
-    + PromStoreProtocolHandler
-    + OpenTelemetryProtocolHandler
-    + PrometheusHandler
-    + PipelineHandler
-    + LogQueryHandler
-    + JaegerQueryHandler
-    + Send
-    + Sync
-    + 'static
-{
-    async fn start(&self) -> Result<()>;
-}
-
-pub type FrontendInstanceRef = Arc<dyn FrontendInstance>;
-
+/// The frontend instance contains necessary components, and implements many
+/// traits, like [`servers::query_handler::grpc::GrpcQueryHandler`],
+/// [`servers::query_handler::sql::SqlQueryHandler`], etc.
 #[derive(Clone)]
 pub struct Instance {
-    options: FrontendOptions,
     catalog_manager: CatalogManagerRef,
     pipeline_operator: Arc<PipelineOperator>,
     statement_executor: Arc<StatementExecutor>,
     query_engine: QueryEngineRef,
     plugins: Plugins,
-    servers: ServerHandlers,
-    heartbeat_task: Option<HeartbeatTask>,
     inserter: InserterRef,
     deleter: DeleterRef,
-    export_metrics_task: Option<ExportMetricsTask>,
     table_metadata_manager: TableMetadataManagerRef,
     stats: StatementStatistics,
     limiter: Option<LimiterRef>,
@@ -157,15 +125,6 @@ impl Instance {
         Ok((kv_backend, procedure_manager))
     }
 
-    pub fn build_servers(&mut self, servers: ServerHandlers) -> Result<()> {
-        self.export_metrics_task =
-            ExportMetricsTask::try_new(&self.options.export_metrics, Some(&self.plugins))
-                .context(StartServerSnafu)?;
-
-        self.servers = servers;
-        Ok(())
-    }
-
     pub fn catalog_manager(&self) -> &CatalogManagerRef {
         &self.catalog_manager
     }
@@ -174,50 +133,20 @@ impl Instance {
         &self.query_engine
     }
 
-    pub fn plugins(&self) -> Plugins {
-        self.plugins.clone()
+    pub fn plugins(&self) -> &Plugins {
+        &self.plugins
     }
 
-    pub async fn shutdown(&self) -> Result<()> {
-        self.servers
-            .shutdown_all()
-            .await
-            .context(error::ShutdownServerSnafu)
-    }
-
-    pub fn server_handlers(&self) -> &ServerHandlers {
-        &self.servers
-    }
-
-    pub fn statement_executor(&self) -> Arc<StatementExecutor> {
-        self.statement_executor.clone()
+    pub fn statement_executor(&self) -> &StatementExecutorRef {
+        &self.statement_executor
     }
 
     pub fn table_metadata_manager(&self) -> &TableMetadataManagerRef {
         &self.table_metadata_manager
     }
-}
 
-#[async_trait]
-impl FrontendInstance for Instance {
-    async fn start(&self) -> Result<()> {
-        if let Some(heartbeat_task) = &self.heartbeat_task {
-            heartbeat_task.start().await?;
-        }
-
-        if let Some(t) = self.export_metrics_task.as_ref() {
-            if t.send_by_handler {
-                let handler = ExportMetricHandler::new_handler(
-                    self.inserter.clone(),
-                    self.statement_executor.clone(),
-                );
-                t.start(Some(handler)).context(StartServerSnafu)?
-            } else {
-                t.start(None).context(StartServerSnafu)?;
-            }
-        }
-
-        self.servers.start_all().await.context(StartServerSnafu)
+    pub fn inserter(&self) -> &InserterRef {
+        &self.inserter
     }
 }
 
@@ -595,6 +524,9 @@ pub fn check_permission(
         Statement::ShowIndex(stmt) => {
             validate_db_permission!(stmt, query_ctx);
         }
+        Statement::ShowRegion(stmt) => {
+            validate_db_permission!(stmt, query_ctx);
+        }
         Statement::ShowViews(stmt) => {
             validate_db_permission!(stmt, query_ctx);
         }

src/frontend/src/instance/builder.rs

@@ -35,12 +35,10 @@ use partition::manager::PartitionRuleManager;
 use pipeline::pipeline_operator::PipelineOperator;
 use query::stats::StatementStatistics;
 use query::QueryEngineFactory;
-use servers::server::ServerHandlers;
 use snafu::OptionExt;
 
 use crate::error::{self, Result};
 use crate::frontend::FrontendOptions;
-use crate::heartbeat::HeartbeatTask;
 use crate::instance::region_query::FrontendRegionQueryHandler;
 use crate::instance::Instance;
 use crate::limiter::Limiter;
@@ -55,7 +53,6 @@ pub struct FrontendBuilder {
     node_manager: NodeManagerRef,
     plugins: Option<Plugins>,
     procedure_executor: ProcedureExecutorRef,
-    heartbeat_task: Option<HeartbeatTask>,
     stats: StatementStatistics,
 }
 
@@ -78,7 +75,6 @@ impl FrontendBuilder {
             node_manager,
             plugins: None,
             procedure_executor,
-            heartbeat_task: None,
             stats,
         }
     }
@@ -97,13 +93,6 @@ impl FrontendBuilder {
         }
     }
 
-    pub fn with_heartbeat_task(self, heartbeat_task: HeartbeatTask) -> Self {
-        Self {
-            heartbeat_task: Some(heartbeat_task),
-            ..self
-        }
-    }
-
     pub async fn try_build(self) -> Result<Instance> {
         let kv_backend = self.kv_backend;
         let node_manager = self.node_manager;
@@ -202,17 +191,13 @@ impl FrontendBuilder {
             });
 
         Ok(Instance {
-            options: self.options,
             catalog_manager: self.catalog_manager,
             pipeline_operator,
             statement_executor,
             query_engine,
             plugins,
-            servers: ServerHandlers::default(),
-            heartbeat_task: self.heartbeat_task,
             inserter,
             deleter,
-            export_metrics_task: None,
             table_metadata_manager: Arc::new(TableMetadataManager::new(kv_backend)),
             stats: self.stats,
             limiter,

src/frontend/src/instance/jaeger.rs

@@ -28,14 +28,14 @@ use common_recordbatch::adapter::RecordBatchStreamAdapter;
 use datafusion::dataframe::DataFrame;
 use datafusion::execution::context::SessionContext;
 use datafusion::execution::SessionStateBuilder;
-use datafusion_expr::{col, lit, lit_timestamp_nano, Expr};
+use datafusion_expr::{col, lit, lit_timestamp_nano, wildcard, Expr};
 use query::QueryEngineRef;
 use serde_json::Value as JsonValue;
 use servers::error::{
     CatalogSnafu, CollectRecordbatchSnafu, DataFusionSnafu, Result as ServerResult,
     TableNotFoundSnafu,
 };
-use servers::http::jaeger::{QueryTraceParams, FIND_TRACES_COLS};
+use servers::http::jaeger::{QueryTraceParams, JAEGER_QUERY_TABLE_NAME_KEY};
 use servers::otlp::trace::{
     DURATION_NANO_COLUMN, SERVICE_NAME_COLUMN, SPAN_ATTRIBUTES_COLUMN, SPAN_KIND_COLUMN,
     SPAN_KIND_PREFIX, SPAN_NAME_COLUMN, TIMESTAMP_COLUMN, TRACE_ID_COLUMN, TRACE_TABLE_NAME,
@@ -43,6 +43,7 @@ use servers::otlp::trace::{
 use servers::query_handler::JaegerQueryHandler;
 use session::context::QueryContextRef;
 use snafu::{OptionExt, ResultExt};
+use table::requests::{TABLE_DATA_MODEL, TABLE_DATA_MODEL_TRACE_V1};
 use table::table::adapter::DfTableProviderAdapter;
 
 use super::Instance;
@@ -82,7 +83,19 @@ impl JaegerQueryHandler for Instance {
             ))));
         }
 
-        // It's equivalent to `SELECT span_name, span_kind FROM {db}.{trace_table} WHERE service_name = '{service_name}'`.
+        // It's equivalent to
+        //
+        // ```
+        // SELECT
+        //   span_name,
+        //   span_kind
+        // FROM
+        //   {db}.{trace_table}
+        // WHERE
+        //   service_name = '{service_name}'
+        // ORDER BY
+        //   timestamp
+        // ```.
         Ok(query_trace_table(
             ctx,
             self.catalog_manager(),
@@ -101,9 +114,19 @@ impl JaegerQueryHandler for Instance {
     }
 
     async fn get_trace(&self, ctx: QueryContextRef, trace_id: &str) -> ServerResult<Output> {
-        // It's equivalent to `SELECT trace_id, timestamp, duration_nano, service_name, span_name, span_id, span_attributes, resource_attributes, parent_span_id
-        // FROM {db}.{trace_table} WHERE trace_id = '{trace_id}'`.
-        let selects: Vec<Expr> = FIND_TRACES_COLS.clone();
+        // It's equivalent to
+        //
+        // ```
+        // SELECT
+        //   *
+        // FROM
+        //   {db}.{trace_table}
+        // WHERE
+        //   trace_id = '{trace_id}'
+        // ORDER BY
+        //   timestamp
+        // ```.
+        let selects = vec![wildcard()];
 
         let filters = vec![col(TRACE_ID_COLUMN).eq(lit(trace_id))];
 
@@ -125,7 +148,7 @@ impl JaegerQueryHandler for Instance {
         ctx: QueryContextRef,
         query_params: QueryTraceParams,
     ) -> ServerResult<Output> {
-        let selects: Vec<Expr> = FIND_TRACES_COLS.clone();
+        let selects = vec![wildcard()];
 
         let mut filters = vec![];
 
@@ -174,17 +197,34 @@ async fn query_trace_table(
     tags: Option<HashMap<String, JsonValue>>,
     distinct: bool,
 ) -> ServerResult<Output> {
-    let db = ctx.get_db_string();
+    let table_name = ctx
+        .extension(JAEGER_QUERY_TABLE_NAME_KEY)
+        .unwrap_or(TRACE_TABLE_NAME);
+
     let table = catalog_manager
-        .table(ctx.current_catalog(), &db, TRACE_TABLE_NAME, Some(&ctx))
+        .table(
+            ctx.current_catalog(),
+            &ctx.current_schema(),
+            table_name,
+            Some(&ctx),
+        )
         .await
         .context(CatalogSnafu)?
         .with_context(|| TableNotFoundSnafu {
-            table: TRACE_TABLE_NAME,
+            table: table_name,
             catalog: ctx.current_catalog(),
-            schema: db,
+            schema: ctx.current_schema(),
         })?;
 
+    let is_data_model_v1 = table
+        .table_info()
+        .meta
+        .options
+        .extra_options
+        .get(TABLE_DATA_MODEL)
+        .map(|s| s.as_str())
+        == Some(TABLE_DATA_MODEL_TRACE_V1);
+
     let df_context = create_df_context(query_engine, ctx.clone())?;
 
     let dataframe = df_context
@@ -196,7 +236,9 @@ async fn query_trace_table(
     // Apply all filters.
     let dataframe = filters
         .into_iter()
-        .chain(tags.map_or(Ok(vec![]), |t| tags_filters(&dataframe, t))?)
+        .chain(tags.map_or(Ok(vec![]), |t| {
+            tags_filters(&dataframe, t, is_data_model_v1)
+        })?)
         .try_fold(dataframe, |df, expr| {
             df.filter(expr).context(DataFusionSnafu)
         })?;
@@ -205,7 +247,10 @@ async fn query_trace_table(
     let dataframe = if distinct {
         dataframe.distinct().context(DataFusionSnafu)?
     } else {
+        // for non distinct query, sort by timestamp to make results stable
         dataframe
+            .sort_by(vec![col(TIMESTAMP_COLUMN)])
+            .context(DataFusionSnafu)?
     };
 
     // Apply the limit if needed.
@@ -237,7 +282,7 @@ fn create_df_context(
         SessionStateBuilder::new_from_existing(query_engine.engine_state().session_state()).build(),
     );
 
-    // The following JSON UDFs will be used for tags filters.
+    // The following JSON UDFs will be used for tags filters on v0 data model.
     let udfs: Vec<FunctionRef> = vec![
         Arc::new(JsonGetInt),
         Arc::new(JsonGetFloat),
@@ -256,7 +301,7 @@ fn create_df_context(
     Ok(df_context)
 }
 
-fn tags_filters(
+fn json_tag_filters(
     dataframe: &DataFrame,
     tags: HashMap<String, JsonValue>,
 ) -> ServerResult<Vec<Expr>> {
@@ -322,3 +367,41 @@ fn tags_filters(
 
     Ok(filters)
 }
+
+fn flatten_tag_filters(tags: HashMap<String, JsonValue>) -> ServerResult<Vec<Expr>> {
+    let filters = tags
+        .into_iter()
+        .filter_map(|(key, value)| {
+            let key = format!("\"span_attributes.{}\"", key);
+            match value {
+                JsonValue::String(value) => Some(col(key).eq(lit(value))),
+                JsonValue::Number(value) => {
+                    if value.is_f64() {
+                        // safe to unwrap as checked previously
+                        Some(col(key).eq(lit(value.as_f64().unwrap())))
+                    } else {
+                        Some(col(key).eq(lit(value.as_i64().unwrap())))
+                    }
+                }
+                JsonValue::Bool(value) => Some(col(key).eq(lit(value))),
+                JsonValue::Null => Some(col(key).is_null()),
+                // not supported at the moment
+                JsonValue::Array(_value) => None,
+                JsonValue::Object(_value) => None,
+            }
+        })
+        .collect();
+    Ok(filters)
+}
+
+fn tags_filters(
+    dataframe: &DataFrame,
+    tags: HashMap<String, JsonValue>,
+    is_data_model_v1: bool,
+) -> ServerResult<Vec<Expr>> {
+    if is_data_model_v1 {
+        flatten_tag_filters(tags)
+    } else {
+        json_tag_filters(dataframe, tags)
+    }
+}

src/frontend/src/instance/log_handler.rs

@@ -20,7 +20,7 @@ use auth::{PermissionChecker, PermissionCheckerRef, PermissionReq};
 use client::Output;
 use common_error::ext::BoxedError;
 use pipeline::pipeline_operator::PipelineOperator;
-use pipeline::{GreptimeTransformer, Pipeline, PipelineInfo, PipelineVersion};
+use pipeline::{Pipeline, PipelineInfo, PipelineVersion};
 use servers::error::{
     AuthSnafu, Error as ServerError, ExecuteGrpcRequestSnafu, InFlightWriteBytesExceededSnafu,
     PipelineSnafu, Result as ServerResult,
@@ -56,7 +56,7 @@ impl PipelineHandler for Instance {
         name: &str,
         version: PipelineVersion,
         query_ctx: QueryContextRef,
-    ) -> ServerResult<Arc<Pipeline<GreptimeTransformer>>> {
+    ) -> ServerResult<Arc<Pipeline>> {
         self.pipeline_operator
             .get_pipeline(query_ctx, name, version)
             .await
@@ -100,7 +100,7 @@ impl PipelineHandler for Instance {
             .await
     }
 
-    fn build_pipeline(&self, pipeline: &str) -> ServerResult<Pipeline<GreptimeTransformer>> {
+    fn build_pipeline(&self, pipeline: &str) -> ServerResult<Pipeline> {
         PipelineOperator::build_pipeline(pipeline).context(PipelineSnafu)
     }
 }

src/frontend/src/instance/logs.rs

@@ -12,6 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::ops::Deref;
+
 use auth::{PermissionChecker, PermissionCheckerRef, PermissionReq};
 use client::Output;
 use common_error::ext::BoxedError;
@@ -20,7 +22,7 @@ use server_error::Result as ServerResult;
 use servers::error::{self as server_error, AuthSnafu, ExecuteQuerySnafu};
 use servers::interceptor::{LogQueryInterceptor, LogQueryInterceptorRef};
 use servers::query_handler::LogQueryHandler;
-use session::context::QueryContextRef;
+use session::context::{QueryContext, QueryContextRef};
 use snafu::ResultExt;
 use tonic::async_trait;
 
@@ -64,4 +66,8 @@ impl LogQueryHandler for Instance {
 
         Ok(interceptor.as_ref().post_query(output, ctx.clone())?)
     }
+
+    fn catalog_manager(&self, _ctx: &QueryContext) -> ServerResult<&dyn catalog::CatalogManager> {
+        Ok(self.catalog_manager.deref())
+    }
 }

src/frontend/src/instance/otlp.rs

@@ -90,6 +90,8 @@ impl OpenTelemetryProtocolHandler for Instance {
             .get::<OpenTelemetryProtocolInterceptorRef<servers::error::Error>>();
         interceptor_ref.pre_execute(ctx.clone())?;
 
+        let is_trace_v1_model = matches!(pipeline, PipelineWay::OtlpTraceDirectV1);
+
         let (requests, rows) = otlp::trace::to_grpc_insert_requests(
             request,
             pipeline,
@@ -101,10 +103,17 @@ impl OpenTelemetryProtocolHandler for Instance {
 
         OTLP_TRACES_ROWS.inc_by(rows as u64);
 
-        self.handle_trace_inserts(requests, ctx)
-            .await
-            .map_err(BoxedError::new)
-            .context(error::ExecuteGrpcQuerySnafu)
+        if is_trace_v1_model {
+            self.handle_trace_inserts(requests, ctx)
+                .await
+                .map_err(BoxedError::new)
+                .context(error::ExecuteGrpcQuerySnafu)
+        } else {
+            self.handle_log_inserts(requests, ctx)
+                .await
+                .map_err(BoxedError::new)
+                .context(error::ExecuteGrpcQuerySnafu)
+        }
     }
 
     #[tracing::instrument(skip_all)]

src/frontend/src/server.rs

@@ -36,26 +36,24 @@ use snafu::ResultExt;
 
 use crate::error::{self, Result, StartServerSnafu, TomlFormatSnafu};
 use crate::frontend::FrontendOptions;
-use crate::instance::FrontendInstance;
+use crate::instance::Instance;
 
-pub struct Services<T, U>
+pub struct Services<T>
 where
     T: Into<FrontendOptions> + Configurable + Clone,
-    U: FrontendInstance,
 {
     opts: T,
-    instance: Arc<U>,
+    instance: Arc<Instance>,
     grpc_server_builder: Option<GrpcServerBuilder>,
     http_server_builder: Option<HttpServerBuilder>,
     plugins: Plugins,
 }
 
-impl<T, U> Services<T, U>
+impl<T> Services<T>
 where
     T: Into<FrontendOptions> + Configurable + Clone,
-    U: FrontendInstance,
 {
-    pub fn new(opts: T, instance: Arc<U>, plugins: Plugins) -> Self {
+    pub fn new(opts: T, instance: Arc<Instance>, plugins: Plugins) -> Self {
         Self {
             opts,
             instance,

src/index/src/bloom_filter/applier.rs

@@ -67,7 +67,7 @@ impl BloomFilterApplier {
         for ((_, mut group), bloom) in locs
             .iter()
             .zip(start_seg..end_seg)
-            .group_by(|(x, _)| **x)
+            .chunk_by(|(x, _)| **x)
             .into_iter()
             .zip(bfs.iter())
         {

src/index/src/fulltext_index/create/bloom_filter.rs

@@ -91,7 +91,8 @@ impl FulltextIndexCreator for BloomFilterFulltextIndexCreator {
 
         let (index_finish, puffin_add_blob) = futures::join!(
             creator.finish(tx.compat_write()),
-            puffin_writer.put_blob(blob_key, rx.compat(), put_options)
+            // TODO(zhongzc): add fulltext config properties
+            puffin_writer.put_blob(blob_key, rx.compat(), put_options, Default::default())
         );
 
         match (

src/index/src/fulltext_index/create/tantivy.rs

@@ -164,6 +164,8 @@ impl FulltextIndexCreator for TantivyFulltextIndexCreator {
 
 #[cfg(test)]
 mod tests {
+    use std::collections::HashMap;
+
     use common_test_util::temp_dir::create_temp_dir;
     use futures::AsyncRead;
     use tantivy::collector::DocSetCollector;
@@ -182,6 +184,7 @@ mod tests {
             _key: &str,
             _raw_data: R,
             _options: PutOptions,
+            _properties: HashMap<String, String>,
         ) -> puffin::error::Result<u64>
         where
             R: AsyncRead + Send,

src/index/src/inverted_index/create/sort/external_sort.rs

@@ -437,9 +437,9 @@ mod tests {
     }
 
     fn random_option_bytes(size: usize) -> Option<Vec<u8>> {
-        let mut rng = rand::thread_rng();
+        let mut rng = rand::rng();
 
-        if rng.gen() {
+        if rng.random() {
             let mut buffer = vec![0u8; size];
             rng.fill(&mut buffer[..]);
             Some(buffer)
@@ -469,11 +469,11 @@ mod tests {
         segment_row_count: usize,
     ) -> (DictionaryValues, ValueSegIds) {
         let mut n = row_count;
-        let mut rng = rand::thread_rng();
+        let mut rng = rand::rng();
         let mut dic_values = Vec::new();
 
         while n > 0 {
-            let size = rng.gen_range(1..=n);
+            let size = rng.random_range(1..=n);
             let value = random_option_bytes(100);
             dic_values.push((value, size));
             n -= size;

src/log-store/src/kafka/log_store.rs

@@ -535,7 +535,7 @@ mod tests {
             .flatten()
             .cloned()
             .collect::<Vec<_>>();
-        all_entries.shuffle(&mut rand::thread_rng());
+        all_entries.shuffle(&mut rand::rng());
 
         let response = logstore.append_batch(all_entries.clone()).await.unwrap();
         // 5 region
@@ -575,7 +575,7 @@ mod tests {
             warn!("The endpoints is empty, skipping the test 'test_append_batch_basic_large'");
             return;
         };
-        let data_size_kb = rand::thread_rng().gen_range(9..31usize);
+        let data_size_kb = rand::rng().random_range(9..31usize);
         info!("Entry size: {}Ki", data_size_kb);
         let broker_endpoints = broker_endpoints
             .split(',')
@@ -608,7 +608,7 @@ mod tests {
             .cloned()
             .collect::<Vec<_>>();
         assert_matches!(all_entries[0], Entry::MultiplePart(_));
-        all_entries.shuffle(&mut rand::thread_rng());
+        all_entries.shuffle(&mut rand::rng());
 
         let response = logstore.append_batch(all_entries.clone()).await.unwrap();
         // 5 region

src/log-store/src/kafka/producer.rs

@@ -24,6 +24,10 @@ use tokio::sync::mpsc::{self, Receiver, Sender};
 use crate::error::{self, Result};
 use crate::kafka::index::IndexCollector;
 use crate::kafka::worker::{BackgroundProducerWorker, ProduceResultHandle, WorkerRequest};
+use crate::metrics::{
+    METRIC_KAFKA_CLIENT_BYTES_TOTAL, METRIC_KAFKA_CLIENT_PRODUCE_ELAPSED,
+    METRIC_KAFKA_CLIENT_TRAFFIC_TOTAL,
+};
 
 pub type OrderedBatchProducerRef = Arc<OrderedBatchProducer>;
 
@@ -106,6 +110,18 @@ impl ProducerClient for PartitionClient {
         records: Vec<Record>,
         compression: Compression,
     ) -> rskafka::client::error::Result<Vec<i64>> {
+        let total_size = records.iter().map(|r| r.approximate_size()).sum::<usize>();
+        let partition = self.partition().to_string();
+        METRIC_KAFKA_CLIENT_BYTES_TOTAL
+            .with_label_values(&[self.topic(), &partition])
+            .inc_by(total_size as u64);
+        METRIC_KAFKA_CLIENT_TRAFFIC_TOTAL
+            .with_label_values(&[self.topic(), &partition])
+            .inc();
+        let _timer = METRIC_KAFKA_CLIENT_PRODUCE_ELAPSED
+            .with_label_values(&[self.topic(), &partition])
+            .start_timer();
+
         self.produce(records, compression).await
     }
 

src/log-store/src/metrics.rs

@@ -19,6 +19,10 @@ use prometheus::*;
 pub const LOGSTORE_LABEL: &str = "logstore";
 /// Operation type label.
 pub const OPTYPE_LABEL: &str = "optype";
+/// Kafka topic label.
+pub const TOPIC_LABEL: &str = "topic";
+/// Kafka partition label.
+pub const PARTITION_LABEL: &str = "partition";
 
 lazy_static! {
     /// Counters of bytes of each operation on a logstore.
@@ -62,4 +66,23 @@ lazy_static! {
     /// Timer of the append_batch operation on the raft-engine logstore.
     /// This timer only measures the duration of the read operation, not measures the total duration of replay.
     pub static ref METRIC_RAFT_ENGINE_READ_ELAPSED: Histogram = METRIC_LOGSTORE_OP_ELAPSED.with_label_values(&["raft-engine", "read"]);
+
+    pub static ref METRIC_KAFKA_CLIENT_BYTES_TOTAL: IntCounterVec = register_int_counter_vec!(
+        "greptime_logstore_kafka_client_bytes_total",
+        "kafka logstore's bytes traffic total",
+        &[LOGSTORE_LABEL, PARTITION_LABEL],
+    )
+    .unwrap();
+    pub static ref METRIC_KAFKA_CLIENT_TRAFFIC_TOTAL: IntCounterVec = register_int_counter_vec!(
+        "greptime_logstore_kafka_client_traffic_total",
+        "kafka logstore's request count traffic total",
+        &[LOGSTORE_LABEL, PARTITION_LABEL],
+    )
+    .unwrap();
+    pub static ref METRIC_KAFKA_CLIENT_PRODUCE_ELAPSED: HistogramVec = register_histogram_vec!(
+        "greptime_logstore_kafka_client_produce_elapsed",
+        "kafka logstore produce operation elapsed",
+        &[LOGSTORE_LABEL, PARTITION_LABEL],
+    )
+    .unwrap();
 }

src/meta-client/Cargo.toml

@@ -10,6 +10,7 @@ workspace = true
 [dependencies]
 api.workspace = true
 async-trait.workspace = true
+common-base.workspace = true
 common-error.workspace = true
 common-grpc.workspace = true
 common-macro.workspace = true

src/meta-client/src/client.rs

@@ -21,9 +21,11 @@ mod cluster;
 mod store;
 mod util;
 
+use std::fmt::Debug;
 use std::sync::Arc;
 
 use api::v1::meta::{ProcedureDetailResponse, Role};
+pub use ask_leader::AskLeader;
 use cluster::Client as ClusterClient;
 pub use cluster::ClusterKvBackend;
 use common_error::ext::BoxedError;
@@ -33,13 +35,16 @@ use common_meta::cluster::{
 };
 use common_meta::datanode::{DatanodeStatKey, DatanodeStatValue, RegionStat};
 use common_meta::ddl::{ExecutorContext, ProcedureExecutor};
-use common_meta::error::{self as meta_error, ExternalSnafu, Result as MetaResult};
+use common_meta::error::{
+    self as meta_error, ExternalSnafu, Result as MetaResult, UnsupportedSnafu,
+};
 use common_meta::key::flow::flow_state::{FlowStat, FlowStateManager};
 use common_meta::kv_backend::KvBackendRef;
 use common_meta::range_stream::PaginationStream;
 use common_meta::rpc::ddl::{SubmitDdlTaskRequest, SubmitDdlTaskResponse};
 use common_meta::rpc::procedure::{
-    MigrateRegionRequest, MigrateRegionResponse, ProcedureStateResponse,
+    AddRegionFollowerRequest, MigrateRegionRequest, MigrateRegionResponse, ProcedureStateResponse,
+    RemoveRegionFollowerRequest,
 };
 use common_meta::rpc::store::{
     BatchDeleteRequest, BatchDeleteResponse, BatchGetRequest, BatchGetResponse, BatchPutRequest,
@@ -74,6 +79,7 @@ pub struct MetaClientBuilder {
     enable_store: bool,
     enable_procedure: bool,
     enable_access_cluster_info: bool,
+    region_follower: Option<RegionFollowerClientRef>,
     channel_manager: Option<ChannelManager>,
     ddl_channel_manager: Option<ChannelManager>,
     heartbeat_channel_manager: Option<ChannelManager>,
@@ -162,6 +168,13 @@ impl MetaClientBuilder {
         }
     }
 
+    pub fn with_region_follower(self, region_follower: RegionFollowerClientRef) -> Self {
+        Self {
+            region_follower: Some(region_follower),
+            ..self
+        }
+    }
+
     pub fn build(self) -> MetaClient {
         let mut client = if let Some(mgr) = self.channel_manager {
             MetaClient::with_channel_manager(self.id, mgr)
@@ -204,6 +217,10 @@ impl MetaClientBuilder {
             ))
         }
 
+        if let Some(region_follower) = self.region_follower {
+            client.region_follower = Some(region_follower);
+        }
+
         client
     }
 }
@@ -216,6 +233,19 @@ pub struct MetaClient {
     store: Option<StoreClient>,
     procedure: Option<ProcedureClient>,
     cluster: Option<ClusterClient>,
+    region_follower: Option<RegionFollowerClientRef>,
+}
+
+pub type RegionFollowerClientRef = Arc<dyn RegionFollowerClient>;
+
+/// A trait for clients that can manage region followers.
+#[async_trait::async_trait]
+pub trait RegionFollowerClient: Sync + Send + Debug {
+    async fn add_region_follower(&self, request: AddRegionFollowerRequest) -> Result<()>;
+
+    async fn remove_region_follower(&self, request: RemoveRegionFollowerRequest) -> Result<()>;
+
+    async fn start(&self, urls: &[&str]) -> Result<()>;
 }
 
 #[async_trait::async_trait]
@@ -242,6 +272,44 @@ impl ProcedureExecutor for MetaClient {
             .context(meta_error::ExternalSnafu)
     }
 
+    async fn add_region_follower(
+        &self,
+        _ctx: &ExecutorContext,
+        request: AddRegionFollowerRequest,
+    ) -> MetaResult<()> {
+        if let Some(region_follower) = &self.region_follower {
+            region_follower
+                .add_region_follower(request)
+                .await
+                .map_err(BoxedError::new)
+                .context(meta_error::ExternalSnafu)
+        } else {
+            UnsupportedSnafu {
+                operation: "add_region_follower",
+            }
+            .fail()
+        }
+    }
+
+    async fn remove_region_follower(
+        &self,
+        _ctx: &ExecutorContext,
+        request: RemoveRegionFollowerRequest,
+    ) -> MetaResult<()> {
+        if let Some(region_follower) = &self.region_follower {
+            region_follower
+                .remove_region_follower(request)
+                .await
+                .map_err(BoxedError::new)
+                .context(meta_error::ExternalSnafu)
+        } else {
+            UnsupportedSnafu {
+                operation: "remove_region_follower",
+            }
+            .fail()
+        }
+    }
+
     async fn query_procedure_state(
         &self,
         _ctx: &ExecutorContext,
@@ -284,7 +352,7 @@ impl ClusterInfo for MetaClient {
             followers
                 .into_iter()
                 .map(|node| NodeInfo {
-                    peer: node.peer.map(|p| p.into()).unwrap_or_default(),
+                    peer: node.peer.unwrap_or_default(),
                     last_activity_ts,
                     status: NodeStatus::Metasrv(MetasrvStatus { is_leader: false }),
                     version: node.version,
@@ -292,7 +360,7 @@ impl ClusterInfo for MetaClient {
                     start_time_ms: node.start_time_ms,
                 })
                 .chain(leader.into_iter().map(|node| NodeInfo {
-                    peer: node.peer.map(|p| p.into()).unwrap_or_default(),
+                    peer: node.peer.unwrap_or_default(),
                     last_activity_ts,
                     status: NodeStatus::Metasrv(MetasrvStatus { is_leader: true }),
                     version: node.version,
@@ -335,6 +403,15 @@ impl ClusterInfo for MetaClient {
 
         Ok(region_stats)
     }
+
+    async fn list_flow_stats(&self) -> Result<Option<FlowStat>> {
+        let cluster_backend = ClusterKvBackend::new(Arc::new(self.cluster_client()?));
+        let cluster_backend = Arc::new(cluster_backend) as KvBackendRef;
+        let flow_state_manager = FlowStateManager::new(cluster_backend);
+        let res = flow_state_manager.get().await.context(GetFlowStatSnafu)?;
+
+        Ok(res.map(|r| r.into()))
+    }
 }
 
 fn decode_stats(kv: KeyValue) -> MetaResult<DatanodeStatValue> {
@@ -344,15 +421,6 @@ fn decode_stats(kv: KeyValue) -> MetaResult<DatanodeStatValue> {
 }
 
 impl MetaClient {
-    pub async fn list_flow_stats(&self) -> Result<Option<FlowStat>> {
-        let cluster_backend = ClusterKvBackend::new(Arc::new(self.cluster_client()?));
-        let cluster_backend = Arc::new(cluster_backend) as KvBackendRef;
-        let flow_state_manager = FlowStateManager::new(cluster_backend);
-        let res = flow_state_manager.get().await.context(GetFlowStatSnafu)?;
-
-        Ok(res.map(|r| r.into()))
-    }
-
     pub fn new(id: Id) -> Self {
         Self {
             id,
@@ -375,6 +443,11 @@ impl MetaClient {
     {
         info!("MetaClient channel config: {:?}", self.channel_config());
 
+        if let Some(client) = &mut self.region_follower {
+            let urls = urls.as_ref().iter().map(|u| u.as_ref()).collect::<Vec<_>>();
+            client.start(&urls).await?;
+            info!("Region follower client started");
+        }
         if let Some(client) = &mut self.heartbeat {
             client.start(urls.clone()).await?;
             info!("Heartbeat client started");
@@ -985,11 +1058,11 @@ mod tests {
         let tx = new_client("test_cluster_client").await;
         let in_memory = tx.in_memory().unwrap();
         let cluster_client = tx.client.cluster_client().unwrap();
-        let mut rng = rand::thread_rng();
+        let mut rng = rand::rng();
 
         // Generates rough 10MB data, which is larger than the default grpc message size limit.
         for i in 0..10 {
-            let data: Vec<u8> = (0..1024 * 1024).map(|_| rng.gen()).collect();
+            let data: Vec<u8> = (0..1024 * 1024).map(|_| rng.random()).collect();
             in_memory
                 .put(
                     PutRequest::new()

src/meta-client/src/client/ask_leader.rs

@@ -75,7 +75,7 @@ impl AskLeader {
             let leadership_group = self.leadership_group.read().unwrap();
             leadership_group.peers.clone()
         };
-        peers.shuffle(&mut rand::thread_rng());
+        peers.shuffle(&mut rand::rng());
 
         let req = AskLeaderRequest {
             header: Some(RequestHeader::new(

src/meta-client/src/client/load_balance.rs

@@ -22,8 +22,8 @@ where
         return None;
     }
 
-    let mut rng = rand::thread_rng();
-    let i = rng.gen_range(0..len);
+    let mut rng = rand::rng();
+    let i = rng.random_range(0..len);
 
     func(i)
 }

src/meta-client/src/lib.rs

@@ -15,8 +15,10 @@
 use std::sync::Arc;
 use std::time::Duration;
 
+use client::RegionFollowerClientRef;
+use common_base::Plugins;
 use common_grpc::channel_manager::{ChannelConfig, ChannelManager};
-use common_telemetry::info;
+use common_telemetry::{debug, info};
 use serde::{Deserialize, Serialize};
 
 use crate::client::MetaClientBuilder;
@@ -73,6 +75,7 @@ pub type MetaClientRef = Arc<client::MetaClient>;
 pub async fn create_meta_client(
     client_type: MetaClientType,
     meta_client_options: &MetaClientOptions,
+    plugins: Option<&Plugins>,
 ) -> error::Result<MetaClientRef> {
     info!(
         "Creating {:?} instance with Metasrv addrs {:?}",
@@ -98,6 +101,13 @@ pub async fn create_meta_client(
     if let MetaClientType::Frontend = client_type {
         let ddl_config = base_config.clone().timeout(meta_client_options.ddl_timeout);
         builder = builder.ddl_channel_manager(ChannelManager::with_config(ddl_config));
+        if let Some(plugins) = plugins {
+            let region_follower = plugins.get::<RegionFollowerClientRef>();
+            if let Some(region_follower) = region_follower {
+                debug!("Region follower client found in plugins");
+                builder = builder.with_region_follower(region_follower);
+            }
+        }
     }
 
     builder = builder