chore/enable-flight-encoder:

### Add Flight Compression Support

 - **Configuration Updates**:
   - Added `grpc.flight_compression` option to `config/config.md`, `config/datanode.example.toml`, and `config/frontend.example.toml` to specify compression modes for Arrow IPC service.

 - **Code Enhancements**:
   - Updated `FlightEncoder` in `src/common/grpc/src/flight.rs` to support compression modes.
   - Modified `RegionServer` and `DatanodeBuilder` in `src/datanode/src/datanode.rs` and `src/datanode/src/region_server.rs` to handle `FlightCompression`.
   - Integrated `FlightCompression` in `src/servers/src/grpc.rs` and `src/servers/src/grpc/flight.rs` to manage compression settings.

 - **Testing and Integration**:
   - Updated test utilities and integration tests in `tests-integration/src/grpc/flight.rs` and `tests-integration/src/test_util.rs` to include `FlightCompression`.

Signed-off-by: Lei, HUANG <mrsatangel@gmail.com>

.github/actions/release-cn-artifacts/action.yaml

@@ -64,11 +64,11 @@ inputs:
   upload-max-retry-times:
     description: Max retry times for uploading artifacts to S3
     required: false
-    default: "20"
+    default: "30"
   upload-retry-timeout:
     description: Timeout for uploading artifacts to S3
     required: false
-    default: "30" # minutes
+    default: "120" # minutes
 runs:
   using: composite
   steps:

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

@@ -59,7 +59,7 @@ runs:
         --set base.podTemplate.main.resources.requests.cpu=50m \
         --set base.podTemplate.main.resources.requests.memory=256Mi \
         --set base.podTemplate.main.resources.limits.cpu=2000m \
-        --set base.podTemplate.main.resources.limits.memory=2Gi \
+        --set base.podTemplate.main.resources.limits.memory=3Gi \
         --set frontend.replicas=${{ inputs.frontend-replicas }} \
         --set datanode.replicas=${{ inputs.datanode-replicas }} \
         --set meta.replicas=${{ inputs.meta-replicas }} \

.github/workflows/develop.yml

@@ -250,6 +250,11 @@ jobs:
           name: unstable-fuzz-logs
           path: /tmp/unstable-greptime/
           retention-days: 3
+      - name: Describe pods
+        if: failure()
+        shell: bash
+        run: |
+          kubectl describe pod -n my-greptimedb
 
   build-greptime-ci:
     if: ${{ github.repository == 'GreptimeTeam/greptimedb' }}
@@ -405,6 +410,11 @@ jobs:
         shell: bash
         run: |
           kubectl describe nodes
+      - name: Describe pod
+        if: failure()
+        shell: bash
+        run: |
+          kubectl describe pod -n my-greptimedb
       - name: Export kind logs
         if: failure()
         shell: bash
@@ -554,6 +564,11 @@ jobs:
         shell: bash
         run: |
           kubectl describe nodes
+      - name: Describe pods
+        if: failure()
+        shell: bash
+        run: |
+          kubectl describe pod -n my-greptimedb
       - name: Export kind logs
         if: failure()
         shell: bash

.github/workflows/release.yml

@@ -441,8 +441,8 @@ jobs:
           aws-region: ${{ vars.EC2_RUNNER_REGION }}
           github-token: ${{ secrets.GH_PERSONAL_ACCESS_TOKEN }}
 
-  bump-doc-version:
-    name: Bump doc version
+  bump-downstream-repo-versions:
+    name: Bump downstream repo versions
     if: ${{ github.event_name == 'push' || github.event_name == 'schedule' }}
     needs: [allocate-runners, publish-github-release]
     runs-on: ubuntu-latest
@@ -456,36 +456,16 @@ jobs:
           fetch-depth: 0
           persist-credentials: false
       - uses: ./.github/actions/setup-cyborg
-      - name: Bump doc version
+      - name: Bump downstream repo versions
         working-directory: cyborg
-        run: pnpm tsx bin/bump-doc-version.ts
-        env:
-          VERSION: ${{ needs.allocate-runners.outputs.version }}
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          DOCS_REPO_TOKEN: ${{ secrets.DOCS_REPO_TOKEN }}
-
-  bump-website-version:
-    name: Bump website version
-    if: ${{ github.ref_type == 'tag' && !contains(github.ref_name, 'nightly') && github.event_name != 'schedule' }}
-    needs: [allocate-runners, publish-github-release]
-    runs-on: ubuntu-latest
-    # Permission reference: https://docs.github.com/en/actions/using-jobs/assigning-permissions-to-jobs
-    permissions:
-      issues: write # Allows the action to create issues for cyborg.
-      contents: write # Allows the action to create a release.
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          persist-credentials: false
-      - uses: ./.github/actions/setup-cyborg
-      - name: Bump website version
-        working-directory: cyborg
-        run: pnpm tsx bin/bump-website-version.ts
+        run: pnpm tsx bin/bump-versions.ts
         env:
+          TARGET_REPOS: website,docs,demo
           VERSION: ${{ needs.allocate-runners.outputs.version }}
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           WEBSITE_REPO_TOKEN: ${{ secrets.WEBSITE_REPO_TOKEN }}
+          DOCS_REPO_TOKEN: ${{ secrets.DOCS_REPO_TOKEN }}
+          DEMO_REPO_TOKEN: ${{ secrets.DEMO_REPO_TOKEN }}
 
   bump-helm-charts-version:
     name: Bump helm charts version

Cargo.toml

@@ -30,6 +30,7 @@ members = [
     "src/common/recordbatch",
     "src/common/runtime",
     "src/common/session",
+    "src/common/stat",
     "src/common/substrait",
     "src/common/telemetry",
     "src/common/test-util",
@@ -132,7 +133,7 @@ etcd-client = "0.14"
 fst = "0.4.7"
 futures = "0.3"
 futures-util = "0.3"
-greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "454c52634c3bac27de10bf0d85d5533eed1cf03f" }
+greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "2dca1dc67862d7b410838aef81232274c019b3f6" }
 hex = "0.4"
 http = "1"
 humantime = "2.1"
@@ -148,6 +149,7 @@ meter-core = { git = "https://github.com/GreptimeTeam/greptime-meter.git", rev =
 mockall = "0.13"
 moka = "0.12"
 nalgebra = "0.33"
+nix = { version = "0.30.1", default-features = false, features = ["event", "fs", "process"] }
 notify = "8.0"
 num_cpus = "1.16"
 object_store_opendal = "0.50"
@@ -287,6 +289,7 @@ query = { path = "src/query" }
 servers = { path = "src/servers" }
 session = { path = "src/session" }
 sql = { path = "src/sql" }
+stat = { path = "src/common/stat" }
 store-api = { path = "src/store-api" }
 substrait = { path = "src/common/substrait" }
 table = { path = "src/table" }

config/config.md

@@ -100,7 +100,7 @@
 | `query` | -- | -- | The query engine options. |
 | `query.parallelism` | Integer | `0` | Parallelism of the query engine.<br/>Default to 0, which means the number of CPU cores. |
 | `storage` | -- | -- | The data storage options. |
-| `storage.data_home` | String | `./greptimedb_data/` | 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. |
@@ -195,13 +195,13 @@
 | `slow_query.record_type` | String | Unset | The record type of slow queries. It can be `system_table` or `log`. |
 | `slow_query.threshold` | String | Unset | The threshold of slow query. |
 | `slow_query.sample_ratio` | Float | Unset | The sampling ratio of slow query log. The value should be in the range of (0, 1]. |
-| `export_metrics` | -- | -- | The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
+| `export_metrics` | -- | -- | The standalone can export its metrics and send to Prometheus compatible service (e.g. `greptimedb`) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
 | `export_metrics.enable` | Bool | `false` | whether enable export metrics. |
 | `export_metrics.write_interval` | String | `30s` | The interval of export metrics. |
 | `export_metrics.self_import` | -- | -- | For `standalone` mode, `self_import` is recommended to collect metrics generated by itself<br/>You must create the database before enabling it. |
 | `export_metrics.self_import.db` | String | Unset | -- |
 | `export_metrics.remote_write` | -- | -- | -- |
-| `export_metrics.remote_write.url` | String | `""` | The url the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`. |
+| `export_metrics.remote_write.url` | String | `""` | The prometheus remote write endpoint that the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`. |
 | `export_metrics.remote_write.headers` | InlineTable | -- | HTTP headers of Prometheus remote-write carry. |
 | `tracing` | -- | -- | The tracing options. Only effect when compiled with `tokio-console` feature. |
 | `tracing.tokio_console_addr` | String | Unset | The tokio console address. |
@@ -232,6 +232,7 @@
 | `grpc.bind_addr` | String | `127.0.0.1:4001` | The address to bind the gRPC server. |
 | `grpc.server_addr` | String | `127.0.0.1:4001` | The address advertised to the metasrv, and used for connections from outside the host.<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 `grpc.bind_addr`. |
 | `grpc.runtime_size` | Integer | `8` | The number of server worker threads. |
+| `grpc.flight_compression` | String | `arrow_ipc` | Compression mode for frontend side Arrow IPC service. Available options:<br/>- `none`: disable all compression<br/>- `transport`: only enable gRPC transport compression (zstd)<br/>- `arrow_ipc`: only enable Arrow IPC compression (lz4)<br/>- `all`: enable all compression. |
 | `grpc.tls` | -- | -- | gRPC server TLS options, see `mysql.tls` section. |
 | `grpc.tls.mode` | String | `disable` | TLS mode. |
 | `grpc.tls.cert_path` | String | Unset | Certificate file path. |
@@ -298,13 +299,11 @@
 | `slow_query.threshold` | String | `30s` | The threshold of slow query. It can be human readable time string, for example: `10s`, `100ms`, `1s`. |
 | `slow_query.sample_ratio` | Float | `1.0` | The sampling ratio of slow query log. The value should be in the range of (0, 1]. For example, `0.1` means 10% of the slow queries will be logged and `1.0` means all slow queries will be logged. |
 | `slow_query.ttl` | String | `30d` | The TTL of the `slow_queries` system table. Default is `30d` when `record_type` is `system_table`. |
-| `export_metrics` | -- | -- | The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
+| `export_metrics` | -- | -- | The frontend can export its metrics and send to Prometheus compatible service (e.g. `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
 | `export_metrics.enable` | Bool | `false` | whether enable export metrics. |
 | `export_metrics.write_interval` | String | `30s` | The interval of export metrics. |
-| `export_metrics.self_import` | -- | -- | For `standalone` mode, `self_import` is recommend to collect metrics generated by itself<br/>You must create the database before enabling it. |
-| `export_metrics.self_import.db` | String | Unset | -- |
 | `export_metrics.remote_write` | -- | -- | -- |
-| `export_metrics.remote_write.url` | String | `""` | The url the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`. |
+| `export_metrics.remote_write.url` | String | `""` | The prometheus remote write endpoint that the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`. |
 | `export_metrics.remote_write.headers` | InlineTable | -- | HTTP headers of Prometheus remote-write carry. |
 | `tracing` | -- | -- | The tracing options. Only effect when compiled with `tokio-console` feature. |
 | `tracing.tokio_console_addr` | String | Unset | The tokio console address. |
@@ -314,12 +313,10 @@
 
 | Key | Type | Default | Descriptions |
 | --- | -----| ------- | ----------- |
-| `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`. |
+| `data_home` | String | `./greptimedb_data` | The working home directory. |
 | `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" |
 | `store_key_prefix` | String | `""` | If it's not empty, the metasrv will store all data with this key prefix. |
-| `backend` | String | `etcd_store` | The datastore for meta server.<br/>Available values:<br/>- `etcd_store` (default value)<br/>- `memory_store`<br/>- `postgres_store` |
+| `backend` | String | `etcd_store` | The datastore for meta server.<br/>Available values:<br/>- `etcd_store` (default value)<br/>- `memory_store`<br/>- `postgres_store`<br/>- `mysql_store` |
 | `meta_table_name` | String | `greptime_metakv` | Table name in RDS to store metadata. Effect when using a RDS kvbackend.<br/>**Only used when backend is `postgres_store`.** |
 | `meta_election_lock_id` | Integer | `1` | Advisory lock id in PostgreSQL for election. Effect when using PostgreSQL as kvbackend<br/>Only used when backend is `postgres_store`. |
 | `selector` | String | `round_robin` | Datanode selector type.<br/>- `round_robin` (default value)<br/>- `lease_based`<br/>- `load_based`<br/>For details, please see "https://docs.greptime.com/developer-guide/metasrv/selector". |
@@ -331,6 +328,12 @@
 | `runtime` | -- | -- | The runtime options. |
 | `runtime.global_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
 | `runtime.compact_rt_size` | Integer | `4` | The number of threads to execute the runtime for global write operations. |
+| `grpc` | -- | -- | The gRPC server options. |
+| `grpc.bind_addr` | String | `127.0.0.1:3002` | The address to bind the gRPC server. |
+| `grpc.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`. |
+| `grpc.runtime_size` | Integer | `8` | The number of server worker threads. |
+| `grpc.max_recv_message_size` | String | `512MB` | The maximum receive message size for gRPC server. |
+| `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 | `0s` | HTTP request timeout. Set to 0 to disable timeout. |
@@ -372,13 +375,11 @@
 | `logging.max_log_files` | Integer | `720` | The maximum amount of log files. |
 | `logging.tracing_sample_ratio` | -- | -- | The percentage of tracing will be sampled and exported.<br/>Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.<br/>ratio > 1 are treated as 1. Fractions < 0 are treated as 0 |
 | `logging.tracing_sample_ratio.default_ratio` | Float | `1.0` | -- |
-| `export_metrics` | -- | -- | The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
+| `export_metrics` | -- | -- | The metasrv can export its metrics and send to Prometheus compatible service (e.g. `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
 | `export_metrics.enable` | Bool | `false` | whether enable export metrics. |
 | `export_metrics.write_interval` | String | `30s` | The interval of export metrics. |
-| `export_metrics.self_import` | -- | -- | For `standalone` mode, `self_import` is recommend to collect metrics generated by itself<br/>You must create the database before enabling it. |
-| `export_metrics.self_import.db` | String | Unset | -- |
 | `export_metrics.remote_write` | -- | -- | -- |
-| `export_metrics.remote_write.url` | String | `""` | The url the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`. |
+| `export_metrics.remote_write.url` | String | `""` | The prometheus remote write endpoint that the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`. |
 | `export_metrics.remote_write.headers` | InlineTable | -- | HTTP headers of Prometheus remote-write carry. |
 | `tracing` | -- | -- | The tracing options. Only effect when compiled with `tokio-console` feature. |
 | `tracing.tokio_console_addr` | String | Unset | The tokio console address. |
@@ -404,6 +405,7 @@
 | `grpc.runtime_size` | Integer | `8` | The number of server worker threads. |
 | `grpc.max_recv_message_size` | String | `512MB` | The maximum receive message size for gRPC server. |
 | `grpc.max_send_message_size` | String | `512MB` | The maximum send message size for gRPC server. |
+| `grpc.flight_compression` | String | `arrow_ipc` | Compression mode for datanode side Arrow IPC service. Available options:<br/>- `none`: disable all compression<br/>- `transport`: only enable gRPC transport compression (zstd)<br/>- `arrow_ipc`: only enable Arrow IPC compression (lz4)<br/>- `all`: enable all compression. |
 | `grpc.tls` | -- | -- | gRPC server TLS options, see `mysql.tls` section. |
 | `grpc.tls.mode` | String | `disable` | TLS mode. |
 | `grpc.tls.cert_path` | String | Unset | Certificate file path. |
@@ -446,7 +448,7 @@
 | `query` | -- | -- | The query engine options. |
 | `query.parallelism` | Integer | `0` | Parallelism of the query engine.<br/>Default to 0, which means the number of CPU cores. |
 | `storage` | -- | -- | The data storage options. |
-| `storage.data_home` | String | `./greptimedb_data/` | 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. |
@@ -536,13 +538,11 @@
 | `logging.max_log_files` | Integer | `720` | The maximum amount of log files. |
 | `logging.tracing_sample_ratio` | -- | -- | The percentage of tracing will be sampled and exported.<br/>Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.<br/>ratio > 1 are treated as 1. Fractions < 0 are treated as 0 |
 | `logging.tracing_sample_ratio.default_ratio` | Float | `1.0` | -- |
-| `export_metrics` | -- | -- | The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
+| `export_metrics` | -- | -- | The datanode can export its metrics and send to Prometheus compatible service (e.g. `greptimedb` itself) from remote-write API.<br/>This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape. |
 | `export_metrics.enable` | Bool | `false` | whether enable export metrics. |
 | `export_metrics.write_interval` | String | `30s` | The interval of export metrics. |
-| `export_metrics.self_import` | -- | -- | For `standalone` mode, `self_import` is recommend to collect metrics generated by itself<br/>You must create the database before enabling it. |
-| `export_metrics.self_import.db` | String | Unset | -- |
 | `export_metrics.remote_write` | -- | -- | -- |
-| `export_metrics.remote_write.url` | String | `""` | The url the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`. |
+| `export_metrics.remote_write.url` | String | `""` | The prometheus remote write endpoint that the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`. |
 | `export_metrics.remote_write.headers` | InlineTable | -- | HTTP headers of Prometheus remote-write carry. |
 | `tracing` | -- | -- | The tracing options. Only effect when compiled with `tokio-console` feature. |
 | `tracing.tokio_console_addr` | String | Unset | The tokio console address. |

config/datanode.example.toml

@@ -44,6 +44,12 @@ runtime_size = 8
 max_recv_message_size = "512MB"
 ## The maximum send message size for gRPC server.
 max_send_message_size = "512MB"
+## Compression mode for datanode side Arrow IPC service. Available options:
+## - `none`: disable all compression
+## - `transport`: only enable gRPC transport compression (zstd)
+## - `arrow_ipc`: only enable Arrow IPC compression (lz4)
+## - `all`: enable all compression.
+flight_compression = "arrow_ipc"
 
 ## gRPC server TLS options, see `mysql.tls` section.
 [grpc.tls]
@@ -252,7 +258,7 @@ parallelism = 0
 ## The data storage options.
 [storage]
 ## The working home directory.
-data_home = "./greptimedb_data/"
+data_home = "./greptimedb_data"
 
 ## The storage type used to store the data.
 ## - `File`: the data is stored in the local file system.
@@ -635,24 +641,16 @@ max_log_files = 720
 [logging.tracing_sample_ratio]
 default_ratio = 1.0
 
-## The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.
+## The datanode can export its metrics and send to Prometheus compatible service (e.g. `greptimedb` itself) from remote-write API.
 ## This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape.
 [export_metrics]
-
 ## whether enable export metrics.
 enable = false
-
 ## The interval of export metrics.
 write_interval = "30s"
 
-## For `standalone` mode, `self_import` is recommend to collect metrics generated by itself
-## You must create the database before enabling it.
-[export_metrics.self_import]
-## @toml2docs:none-default
-db = "greptime_metrics"
-
 [export_metrics.remote_write]
-## The url the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`.
+## The prometheus remote write endpoint that the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`.
 url = ""
 
 ## HTTP headers of Prometheus remote-write carry.

config/frontend.example.toml

@@ -54,6 +54,12 @@ bind_addr = "127.0.0.1:4001"
 server_addr = "127.0.0.1:4001"
 ## The number of server worker threads.
 runtime_size = 8
+## Compression mode for frontend side Arrow IPC service. Available options:
+## - `none`: disable all compression
+## - `transport`: only enable gRPC transport compression (zstd)
+## - `arrow_ipc`: only enable Arrow IPC compression (lz4)
+## - `all`: enable all compression.
+flight_compression = "arrow_ipc"
 
 ## gRPC server TLS options, see `mysql.tls` section.
 [grpc.tls]
@@ -247,24 +253,16 @@ sample_ratio = 1.0
 ## The TTL of the `slow_queries` system table. Default is `30d` when `record_type` is `system_table`.
 ttl = "30d"
 
-## The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.
+## The frontend can export its metrics and send to Prometheus compatible service (e.g. `greptimedb` itself) from remote-write API.
 ## This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape.
 [export_metrics]
-
 ## whether enable export metrics.
 enable = false
-
 ## The interval of export metrics.
 write_interval = "30s"
 
-## For `standalone` mode, `self_import` is recommend to collect metrics generated by itself
-## You must create the database before enabling it.
-[export_metrics.self_import]
-## @toml2docs:none-default
-db = "greptime_metrics"
-
 [export_metrics.remote_write]
-## The url the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`.
+## The prometheus remote write endpoint that the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`.
 url = ""
 
 ## HTTP headers of Prometheus remote-write carry.

config/metasrv.example.toml

@@ -1,13 +1,5 @@
 ## The working home directory.
-data_home = "./greptimedb_data/metasrv/"
-
-## The bind address of metasrv.
-bind_addr = "127.0.0.1:3002"
-
-## The communication server address for the frontend and datanode to connect to metasrv.
-## If left empty or unset, the server will automatically use the IP address of the first network interface
-## on the host, with the same port number as the one specified in `bind_addr`.
-server_addr = "127.0.0.1:3002"
+data_home = "./greptimedb_data"
 
 ## Store server address default to etcd store.
 ## For postgres store, the format is:
@@ -24,6 +16,7 @@ store_key_prefix = ""
 ## - `etcd_store` (default value)
 ## - `memory_store`
 ## - `postgres_store`
+## - `mysql_store`
 backend = "etcd_store"
 
 ## Table name in RDS to store metadata. Effect when using a RDS kvbackend.
@@ -67,6 +60,21 @@ node_max_idle_time = "24hours"
 ## The number of threads to execute the runtime for global write operations.
 #+ compact_rt_size = 4
 
+## The gRPC server options.
+[grpc]
+## The address to bind the gRPC server.
+bind_addr = "127.0.0.1:3002"
+## The communication server address for the frontend and datanode to connect to metasrv.
+## If left empty or unset, the server will automatically use the IP address of the first network interface
+## on the host, with the same port number as the one specified in `bind_addr`.
+server_addr = "127.0.0.1:3002"
+## The number of server worker threads.
+runtime_size = 8
+## The maximum receive message size for gRPC server.
+max_recv_message_size = "512MB"
+## The maximum send message size for gRPC server.
+max_send_message_size = "512MB"
+
 ## The HTTP server options.
 [http]
 ## The address to bind the HTTP server.
@@ -229,24 +237,16 @@ max_log_files = 720
 [logging.tracing_sample_ratio]
 default_ratio = 1.0
 
-## The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.
+## The metasrv can export its metrics and send to Prometheus compatible service (e.g. `greptimedb` itself) from remote-write API.
 ## This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape.
 [export_metrics]
-
 ## whether enable export metrics.
 enable = false
-
 ## The interval of export metrics.
 write_interval = "30s"
 
-## For `standalone` mode, `self_import` is recommend to collect metrics generated by itself
-## You must create the database before enabling it.
-[export_metrics.self_import]
-## @toml2docs:none-default
-db = "greptime_metrics"
-
 [export_metrics.remote_write]
-## The url the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`.
+## The prometheus remote write endpoint that the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`.
 url = ""
 
 ## HTTP headers of Prometheus remote-write carry.

config/standalone.example.toml

@@ -350,7 +350,7 @@ parallelism = 0
 ## The data storage options.
 [storage]
 ## The working home directory.
-data_home = "./greptimedb_data/"
+data_home = "./greptimedb_data"
 
 ## The storage type used to store the data.
 ## - `File`: the data is stored in the local file system.
@@ -750,13 +750,11 @@ default_ratio = 1.0
 ## @toml2docs:none-default
 #+ sample_ratio = 1.0
 
-## The datanode can export its metrics and send to Prometheus compatible service (e.g. send to `greptimedb` itself) from remote-write API.
+## The standalone can export its metrics and send to Prometheus compatible service (e.g. `greptimedb`) from remote-write API.
 ## This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape.
 [export_metrics]
-
 ## whether enable export metrics.
 enable = false
-
 ## The interval of export metrics.
 write_interval = "30s"
 
@@ -767,7 +765,7 @@ write_interval = "30s"
 db = "greptime_metrics"
 
 [export_metrics.remote_write]
-## The url the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`.
+## The prometheus remote write endpoint that the metrics send to. The url example can be: `http://127.0.0.1:4000/v1/prometheus/write?db=greptime_metrics`.
 url = ""
 
 ## HTTP headers of Prometheus remote-write carry.

cyborg/bin/bump-doc-version.ts

@@ -1,75 +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.
- */
-
-import * as core from "@actions/core";
-import {obtainClient} from "@/common";
-
-async function triggerWorkflow(workflowId: string, version: string) {
-  const docsClient = obtainClient("DOCS_REPO_TOKEN")
-  try {
-    await docsClient.rest.actions.createWorkflowDispatch({
-      owner: "GreptimeTeam",
-      repo: "docs",
-      workflow_id: workflowId,
-      ref: "main",
-      inputs: {
-        version,
-      },
-    });
-    console.log(`Successfully triggered ${workflowId} workflow with version ${version}`);
-  } catch (error) {
-    core.setFailed(`Failed to trigger workflow: ${error.message}`);
-  }
-}
-
-function determineWorkflow(version: string): [string, string] {
-  // Check if it's a nightly version
-  if (version.includes('nightly')) {
-    return ['bump-nightly-version.yml', version];
-  }
-
-  const parts = version.split('.');
-
-  if (parts.length !== 3) {
-    throw new Error('Invalid version format');
-  }
-
-  // If patch version (last number) is 0, it's a major version
-  // Return only major.minor version
-  if (parts[2] === '0') {
-    return ['bump-version.yml', `${parts[0]}.${parts[1]}`];
-  }
-
-  // Otherwise it's a patch version, use full version
-  return ['bump-patch-version.yml', version];
-}
-
-const version = process.env.VERSION;
-if (!version) {
-  core.setFailed("VERSION environment variable is required");
-  process.exit(1);
-}
-
-// Remove 'v' prefix if exists
-const cleanVersion = version.startsWith('v') ? version.slice(1) : version;
-
-try {
-  const [workflowId, apiVersion] = determineWorkflow(cleanVersion);
-  triggerWorkflow(workflowId, apiVersion);
-} catch (error) {
-  core.setFailed(`Error processing version: ${error.message}`);
-  process.exit(1);
-}

cyborg/bin/bump-versions.ts

@@ -0,0 +1,156 @@
+/*
+ * Copyright 2023 Greptime Team
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+import * as core from "@actions/core";
+import {obtainClient} from "@/common";
+
+interface RepoConfig {
+  tokenEnv: string;
+  repo: string;
+  workflowLogic: (version: string) => [string, string] | null;
+}
+
+const REPO_CONFIGS: Record<string, RepoConfig> = {
+  website: {
+    tokenEnv: "WEBSITE_REPO_TOKEN",
+    repo: "website",
+    workflowLogic: (version: string) => {
+      // Skip nightly versions for website
+      if (version.includes('nightly')) {
+        console.log('Nightly version detected for website, skipping workflow trigger.');
+        return null;
+      }
+      return ['bump-patch-version.yml', version];
+    }
+  },
+  demo: {
+    tokenEnv: "DEMO_REPO_TOKEN",
+    repo: "demo-scene",
+    workflowLogic: (version: string) => {
+      // Skip nightly versions for demo
+      if (version.includes('nightly')) {
+        console.log('Nightly version detected for demo, skipping workflow trigger.');
+        return null;
+      }
+      return ['bump-patch-version.yml', version];
+    }
+  },
+  docs: {
+    tokenEnv: "DOCS_REPO_TOKEN",
+    repo: "docs",
+    workflowLogic: (version: string) => {
+      // Check if it's a nightly version
+      if (version.includes('nightly')) {
+        return ['bump-nightly-version.yml', version];
+      }
+
+      const parts = version.split('.');
+      if (parts.length !== 3) {
+        throw new Error('Invalid version format');
+      }
+
+      // If patch version (last number) is 0, it's a major version
+      // Return only major.minor version
+      if (parts[2] === '0') {
+        return ['bump-version.yml', `${parts[0]}.${parts[1]}`];
+      }
+
+      // Otherwise it's a patch version, use full version
+      return ['bump-patch-version.yml', version];
+    }
+  }
+};
+
+async function triggerWorkflow(repoConfig: RepoConfig, workflowId: string, version: string) {
+  const client = obtainClient(repoConfig.tokenEnv);
+  try {
+    await client.rest.actions.createWorkflowDispatch({
+      owner: "GreptimeTeam",
+      repo: repoConfig.repo,
+      workflow_id: workflowId,
+      ref: "main",
+      inputs: {
+        version,
+      },
+    });
+    console.log(`Successfully triggered ${workflowId} workflow for ${repoConfig.repo} with version ${version}`);
+  } catch (error) {
+    core.setFailed(`Failed to trigger workflow for ${repoConfig.repo}: ${error.message}`);
+    throw error;
+  }
+}
+
+async function processRepo(repoName: string, version: string) {
+  const repoConfig = REPO_CONFIGS[repoName];
+  if (!repoConfig) {
+    throw new Error(`Unknown repository: ${repoName}`);
+  }
+
+  try {
+    const workflowResult = repoConfig.workflowLogic(version);
+    if (workflowResult === null) {
+      // Skip this repo (e.g., nightly version for website)
+      return;
+    }
+
+    const [workflowId, apiVersion] = workflowResult;
+    await triggerWorkflow(repoConfig, workflowId, apiVersion);
+  } catch (error) {
+    core.setFailed(`Error processing ${repoName} with version ${version}: ${error.message}`);
+    throw error;
+  }
+}
+
+async function main() {
+  const version = process.env.VERSION;
+  if (!version) {
+    core.setFailed("VERSION environment variable is required");
+    process.exit(1);
+  }
+
+  // Remove 'v' prefix if exists
+  const cleanVersion = version.startsWith('v') ? version.slice(1) : version;
+
+  // Get target repositories from environment variable
+  // Default to both if not specified
+  const targetRepos = process.env.TARGET_REPOS?.split(',').map(repo => repo.trim()) || ['website', 'docs'];
+
+  console.log(`Processing version ${cleanVersion} for repositories: ${targetRepos.join(', ')}`);
+
+  const errors: string[] = [];
+
+  // Process each repository
+  for (const repo of targetRepos) {
+    try {
+      await processRepo(repo, cleanVersion);
+    } catch (error) {
+      errors.push(`${repo}: ${error.message}`);
+    }
+  }
+
+  if (errors.length > 0) {
+    core.setFailed(`Failed to process some repositories: ${errors.join('; ')}`);
+    process.exit(1);
+  }
+
+  console.log('All repositories processed successfully');
+}
+
+// Execute main function
+main().catch((error) => {
+  core.setFailed(`Unexpected error: ${error.message}`);
+  process.exit(1);
+});

cyborg/bin/bump-website-version.ts

@@ -1,57 +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.
- */
-
-import * as core from "@actions/core";
-import {obtainClient} from "@/common";
-
-async function triggerWorkflow(workflowId: string, version: string) {
-  const websiteClient = obtainClient("WEBSITE_REPO_TOKEN")
-  try {
-    await websiteClient.rest.actions.createWorkflowDispatch({
-      owner: "GreptimeTeam",
-      repo: "website",
-      workflow_id: workflowId,
-      ref: "main",
-      inputs: {
-        version,
-      },
-    });
-    console.log(`Successfully triggered ${workflowId} workflow with version ${version}`);
-  } catch (error) {
-    core.setFailed(`Failed to trigger workflow: ${error.message}`);
-  }
-}
-
-const version = process.env.VERSION;
-if (!version) {
-  core.setFailed("VERSION environment variable is required");
-  process.exit(1);
-}
-
-// Remove 'v' prefix if exists
-const cleanVersion = version.startsWith('v') ? version.slice(1) : version;
-
-if (cleanVersion.includes('nightly')) {
-  console.log('Nightly version detected, skipping workflow trigger.');
-  process.exit(0);
-}
-
-try {
-  triggerWorkflow('bump-patch-version.yml', cleanVersion);
-} catch (error) {
-  core.setFailed(`Error processing version: ${error.message}`);
-  process.exit(1);
-}

grafana/dashboards/metrics/cluster/dashboard.md

@@ -60,7 +60,7 @@
 | Read Stage P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_read_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))` | `timeseries` | Read Stage P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]` |
 | Write Stage P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_write_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))` | `timeseries` | Write Stage P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]` |
 | Compaction OPS per Instance | `sum by(instance, pod) (rate(greptime_mito_compaction_total_elapsed_count{instance=~"$datanode"}[$__rate_interval]))` | `timeseries` | Compaction OPS per Instance. | `prometheus` | `ops` | `[{{ instance }}]-[{{pod}}]` |
-| Compaction Elapsed Time per Instance by Stage | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))`<br/>`sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_sum{instance=~"$datanode"}[$__rate_interval]))/sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_count{instance=~"$datanode"}[$__rate_interval]))` | `timeseries` | Compaction latency by stage | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-p99` |
+| Compaction Elapsed Time per Instance by Stage | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))`<br/>`sum by(instance, pod, stage) (rate(greptime_mito_compaction_stage_elapsed_sum{instance=~"$datanode"}[$__rate_interval]))/sum by(instance, pod, stage) (rate(greptime_mito_compaction_stage_elapsed_count{instance=~"$datanode"}[$__rate_interval]))` | `timeseries` | Compaction latency by stage | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-p99` |
 | Compaction P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le,stage) (rate(greptime_mito_compaction_total_elapsed_bucket{instance=~"$datanode"}[$__rate_interval])))` | `timeseries` | Compaction P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-compaction` |
 | WAL write size | `histogram_quantile(0.95, sum by(le,instance, pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))`<br/>`histogram_quantile(0.99, sum by(le,instance,pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))`<br/>`sum by (instance, pod)(rate(raft_engine_write_size_sum[$__rate_interval]))` | `timeseries` | Write-ahead logs write size as bytes. This chart includes stats of p95 and p99 size by instance, total WAL write rate. | `prometheus` | `bytes` | `[{{instance}}]-[{{pod}}]-req-size-p95` |
 | Cached Bytes per Instance | `greptime_mito_cache_bytes{instance=~"$datanode"}` | `timeseries` | Cached Bytes per Instance. | `prometheus` | `decbytes` | `[{{instance}}]-[{{pod}}]-[{{type}}]` |
@@ -69,7 +69,7 @@
 | Log Store op duration seconds | `histogram_quantile(0.99, sum by(le,logstore,optype,instance, pod) (rate(greptime_logstore_op_elapsed_bucket[$__rate_interval])))` | `timeseries` | Write-ahead log operations latency at p99 | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{logstore}}]-[{{optype}}]-p99` |
 | Inflight Flush | `greptime_mito_inflight_flush_count` | `timeseries` | Ongoing flush task count | `prometheus` | `none` | `[{{instance}}]-[{{pod}}]` |
 | Compaction Input/Output Bytes | `sum by(instance, pod) (greptime_mito_compaction_input_bytes)`<br/>`sum by(instance, pod) (greptime_mito_compaction_output_bytes)` | `timeseries` | Compaction oinput output bytes | `prometheus` | `bytes` | `[{{instance}}]-[{{pod}}]-input` |
-| Region Worker Handle Bulk Insert Requests | `histogram_quantile(0.95, sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_bucket[$__rate_interval])))`<br/>`sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_sum[$__rate_interval]))/sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_count[$__rate_interval]))` | `timeseries` | Per-stage elapsed time for region worker to handle bulk insert region requests. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-P95` |
+| Region Worker Handle Bulk Insert Requests | `histogram_quantile(0.95, sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_bucket[$__rate_interval])))`<br/>`sum by(instance, stage, pod) (rate(greptime_region_worker_handle_write_sum[$__rate_interval]))/sum by(instance, stage, pod) (rate(greptime_region_worker_handle_write_count[$__rate_interval]))` | `timeseries` | Per-stage elapsed time for region worker to handle bulk insert region requests. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-P95` |
 | Region Worker Convert Requests | `histogram_quantile(0.95, sum by(le, instance, stage, pod) (rate(greptime_datanode_convert_region_request_bucket[$__rate_interval])))`<br/>`sum by(le,instance, stage, pod) (rate(greptime_datanode_convert_region_request_sum[$__rate_interval]))/sum by(le,instance, stage, pod) (rate(greptime_datanode_convert_region_request_count[$__rate_interval]))` | `timeseries` | Per-stage elapsed time for region worker to decode requests. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-P95` |
 # OpenDAL
 | Title | Query | Type | Description | Datasource | Unit | Legend Format |
@@ -88,9 +88,19 @@
 # Metasrv
 | Title | Query | Type | Description | Datasource | Unit | Legend Format |
 | --- | --- | --- | --- | --- | --- | --- |
-| Region migration datanode | `greptime_meta_region_migration_stat{datanode_type="src"}`<br/>`greptime_meta_region_migration_stat{datanode_type="desc"}` | `state-timeline` | Counter of region migration by source and destination | `prometheus` | `none` | `from-datanode-{{datanode_id}}` |
-| Region migration error | `greptime_meta_region_migration_error` | `timeseries` | Counter of region migration error | `prometheus` | `none` | `__auto` |
-| Datanode load | `greptime_datanode_load` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `none` | `__auto` |
+| Region migration datanode | `greptime_meta_region_migration_stat{datanode_type="src"}`<br/>`greptime_meta_region_migration_stat{datanode_type="desc"}` | `status-history` | Counter of region migration by source and destination | `prometheus` | -- | `from-datanode-{{datanode_id}}` |
+| Region migration error | `greptime_meta_region_migration_error` | `timeseries` | Counter of region migration error | `prometheus` | `none` | `{{pod}}-{{state}}-{{error_type}}` |
+| Datanode load | `greptime_datanode_load` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `binBps` | `Datanode-{{datanode_id}}-writeload` |
+| Rate of SQL Executions (RDS) | `rate(greptime_meta_rds_pg_sql_execute_elapsed_ms_count[$__rate_interval])` | `timeseries` | Displays the rate of SQL executions processed by the Meta service using the RDS backend. | `prometheus` | `none` | `{{pod}} {{op}} {{type}} {{result}} ` |
+| SQL Execution Latency (RDS) | `histogram_quantile(0.90, sum by(pod, op, type, result, le) (rate(greptime_meta_rds_pg_sql_execute_elapsed_ms_bucket[$__rate_interval])))` | `timeseries` | Measures the response time of SQL executions via the RDS backend.  | `prometheus` | `ms` | `{{pod}} {{op}} {{type}} {{result}} p90` |
+| Handler Execution Latency | `histogram_quantile(0.90, sum by(pod, le, name) (
+  rate(greptime_meta_handler_execute_bucket[$__rate_interval])
+))` | `timeseries` | Shows latency of Meta handlers by pod and handler name, useful for monitoring handler performance and detecting latency spikes.<br/> | `prometheus` | `s` | `{{pod}} {{name}} p90` |
+| Heartbeat Packet Size | `histogram_quantile(0.9, sum by(pod, le) (greptime_meta_heartbeat_stat_memory_size_bucket))` | `timeseries` | Shows p90 heartbeat message sizes, helping track network usage and identify anomalies in heartbeat payload.<br/> | `prometheus` | `bytes` | `{{pod}}` |
+| Meta Heartbeat Receive Rate | `rate(greptime_meta_heartbeat_rate[$__rate_interval])` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `s` | `{{pod}}` |
+| Meta KV Ops Latency | `histogram_quantile(0.99, sum by(pod, le, op, target) (greptime_meta_kv_request_elapsed_bucket))` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `s` | `{{pod}}-{{op}} p99` |
+| Rate of meta KV Ops | `rate(greptime_meta_kv_request_elapsed_count[$__rate_interval])` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `none` | `{{pod}}-{{op}} p99` |
+| DDL Latency | `histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_tables_bucket))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_table))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_view))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_flow))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_drop_table))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_alter_table))` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `s` | `CreateLogicalTables-{{step}} p90` |
 # Flownode
 | Title | Query | Type | Description | Datasource | Unit | Legend Format |
 | --- | --- | --- | --- | --- | --- | --- |

grafana/dashboards/metrics/cluster/dashboard.yaml

@@ -497,7 +497,7 @@ groups:
                 type: prometheus
                 uid: ${metrics}
               legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]-p99'
-            - expr: sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_sum{instance=~"$datanode"}[$__rate_interval]))/sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_count{instance=~"$datanode"}[$__rate_interval]))
+            - expr: sum by(instance, pod, stage) (rate(greptime_mito_compaction_stage_elapsed_sum{instance=~"$datanode"}[$__rate_interval]))/sum by(instance, pod, stage) (rate(greptime_mito_compaction_stage_elapsed_count{instance=~"$datanode"}[$__rate_interval]))
               datasource:
                 type: prometheus
                 uid: ${metrics}
@@ -607,7 +607,7 @@ groups:
                 type: prometheus
                 uid: ${metrics}
               legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]-P95'
-            - expr: sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_sum[$__rate_interval]))/sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_count[$__rate_interval]))
+            - expr: sum by(instance, stage, pod) (rate(greptime_region_worker_handle_write_sum[$__rate_interval]))/sum by(instance, stage, pod) (rate(greptime_region_worker_handle_write_count[$__rate_interval]))
               datasource:
                 type: prometheus
                 uid: ${metrics}
@@ -741,9 +741,8 @@ groups:
     - title: Metasrv
       panels:
         - title: Region migration datanode
-          type: state-timeline
+          type: status-history
           description: Counter of region migration by source and destination
-          unit: none
           queries:
             - expr: greptime_meta_region_migration_stat{datanode_type="src"}
               datasource:
@@ -764,17 +763,127 @@ groups:
               datasource:
                 type: prometheus
                 uid: ${metrics}
-              legendFormat: __auto
+              legendFormat: '{{pod}}-{{state}}-{{error_type}}'
         - title: Datanode load
           type: timeseries
           description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
-          unit: none
+          unit: binBps
           queries:
             - expr: greptime_datanode_load
               datasource:
                 type: prometheus
                 uid: ${metrics}
-              legendFormat: __auto
+              legendFormat: Datanode-{{datanode_id}}-writeload
+        - title: Rate of SQL Executions (RDS)
+          type: timeseries
+          description: Displays the rate of SQL executions processed by the Meta service using the RDS backend.
+          unit: none
+          queries:
+            - expr: rate(greptime_meta_rds_pg_sql_execute_elapsed_ms_count[$__rate_interval])
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}} {{op}} {{type}} {{result}} '
+        - title: SQL Execution Latency (RDS)
+          type: timeseries
+          description: 'Measures the response time of SQL executions via the RDS backend. '
+          unit: ms
+          queries:
+            - expr: histogram_quantile(0.90, sum by(pod, op, type, result, le) (rate(greptime_meta_rds_pg_sql_execute_elapsed_ms_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}} {{op}} {{type}} {{result}} p90'
+        - title: Handler Execution Latency
+          type: timeseries
+          description: |
+            Shows latency of Meta handlers by pod and handler name, useful for monitoring handler performance and detecting latency spikes.
+          unit: s
+          queries:
+            - expr: |-
+                histogram_quantile(0.90, sum by(pod, le, name) (
+                  rate(greptime_meta_handler_execute_bucket[$__rate_interval])
+                ))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}} {{name}} p90'
+        - title: Heartbeat Packet Size
+          type: timeseries
+          description: |
+            Shows p90 heartbeat message sizes, helping track network usage and identify anomalies in heartbeat payload.
+          unit: bytes
+          queries:
+            - expr: histogram_quantile(0.9, sum by(pod, le) (greptime_meta_heartbeat_stat_memory_size_bucket))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}}'
+        - title: Meta Heartbeat Receive Rate
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: s
+          queries:
+            - expr: rate(greptime_meta_heartbeat_rate[$__rate_interval])
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}}'
+        - title: Meta KV Ops Latency
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(pod, le, op, target) (greptime_meta_kv_request_elapsed_bucket))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}}-{{op}} p99'
+        - title: Rate of meta KV Ops
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: none
+          queries:
+            - expr: rate(greptime_meta_kv_request_elapsed_count[$__rate_interval])
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}}-{{op}} p99'
+        - title: DDL Latency
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_tables_bucket))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: CreateLogicalTables-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_table))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: CreateTable-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_view))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: CreateView-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_flow))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: CreateFlow-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_drop_table))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: DropTable-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_alter_table))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: AlterTable-{{step}} p90
     - title: Flownode
       panels:
         - title: Flow Ingest / Output Rate

grafana/dashboards/metrics/standalone/dashboard.md

@@ -60,7 +60,7 @@
 | Read Stage P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_read_stage_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | Read Stage P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]` |
 | Write Stage P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_write_stage_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | Write Stage P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]` |
 | Compaction OPS per Instance | `sum by(instance, pod) (rate(greptime_mito_compaction_total_elapsed_count{}[$__rate_interval]))` | `timeseries` | Compaction OPS per Instance. | `prometheus` | `ops` | `[{{ instance }}]-[{{pod}}]` |
-| Compaction Elapsed Time per Instance by Stage | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{}[$__rate_interval])))`<br/>`sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_sum{}[$__rate_interval]))/sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_count{}[$__rate_interval]))` | `timeseries` | Compaction latency by stage | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-p99` |
+| Compaction Elapsed Time per Instance by Stage | `histogram_quantile(0.99, sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_bucket{}[$__rate_interval])))`<br/>`sum by(instance, pod, stage) (rate(greptime_mito_compaction_stage_elapsed_sum{}[$__rate_interval]))/sum by(instance, pod, stage) (rate(greptime_mito_compaction_stage_elapsed_count{}[$__rate_interval]))` | `timeseries` | Compaction latency by stage | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-p99` |
 | Compaction P99 per Instance | `histogram_quantile(0.99, sum by(instance, pod, le,stage) (rate(greptime_mito_compaction_total_elapsed_bucket{}[$__rate_interval])))` | `timeseries` | Compaction P99 per Instance. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-compaction` |
 | WAL write size | `histogram_quantile(0.95, sum by(le,instance, pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))`<br/>`histogram_quantile(0.99, sum by(le,instance,pod) (rate(raft_engine_write_size_bucket[$__rate_interval])))`<br/>`sum by (instance, pod)(rate(raft_engine_write_size_sum[$__rate_interval]))` | `timeseries` | Write-ahead logs write size as bytes. This chart includes stats of p95 and p99 size by instance, total WAL write rate. | `prometheus` | `bytes` | `[{{instance}}]-[{{pod}}]-req-size-p95` |
 | Cached Bytes per Instance | `greptime_mito_cache_bytes{}` | `timeseries` | Cached Bytes per Instance. | `prometheus` | `decbytes` | `[{{instance}}]-[{{pod}}]-[{{type}}]` |
@@ -69,7 +69,7 @@
 | Log Store op duration seconds | `histogram_quantile(0.99, sum by(le,logstore,optype,instance, pod) (rate(greptime_logstore_op_elapsed_bucket[$__rate_interval])))` | `timeseries` | Write-ahead log operations latency at p99 | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{logstore}}]-[{{optype}}]-p99` |
 | Inflight Flush | `greptime_mito_inflight_flush_count` | `timeseries` | Ongoing flush task count | `prometheus` | `none` | `[{{instance}}]-[{{pod}}]` |
 | Compaction Input/Output Bytes | `sum by(instance, pod) (greptime_mito_compaction_input_bytes)`<br/>`sum by(instance, pod) (greptime_mito_compaction_output_bytes)` | `timeseries` | Compaction oinput output bytes | `prometheus` | `bytes` | `[{{instance}}]-[{{pod}}]-input` |
-| Region Worker Handle Bulk Insert Requests | `histogram_quantile(0.95, sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_bucket[$__rate_interval])))`<br/>`sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_sum[$__rate_interval]))/sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_count[$__rate_interval]))` | `timeseries` | Per-stage elapsed time for region worker to handle bulk insert region requests. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-P95` |
+| Region Worker Handle Bulk Insert Requests | `histogram_quantile(0.95, sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_bucket[$__rate_interval])))`<br/>`sum by(instance, stage, pod) (rate(greptime_region_worker_handle_write_sum[$__rate_interval]))/sum by(instance, stage, pod) (rate(greptime_region_worker_handle_write_count[$__rate_interval]))` | `timeseries` | Per-stage elapsed time for region worker to handle bulk insert region requests. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-P95` |
 | Region Worker Convert Requests | `histogram_quantile(0.95, sum by(le, instance, stage, pod) (rate(greptime_datanode_convert_region_request_bucket[$__rate_interval])))`<br/>`sum by(le,instance, stage, pod) (rate(greptime_datanode_convert_region_request_sum[$__rate_interval]))/sum by(le,instance, stage, pod) (rate(greptime_datanode_convert_region_request_count[$__rate_interval]))` | `timeseries` | Per-stage elapsed time for region worker to decode requests. | `prometheus` | `s` | `[{{instance}}]-[{{pod}}]-[{{stage}}]-P95` |
 # OpenDAL
 | Title | Query | Type | Description | Datasource | Unit | Legend Format |
@@ -88,9 +88,19 @@
 # Metasrv
 | Title | Query | Type | Description | Datasource | Unit | Legend Format |
 | --- | --- | --- | --- | --- | --- | --- |
-| Region migration datanode | `greptime_meta_region_migration_stat{datanode_type="src"}`<br/>`greptime_meta_region_migration_stat{datanode_type="desc"}` | `state-timeline` | Counter of region migration by source and destination | `prometheus` | `none` | `from-datanode-{{datanode_id}}` |
-| Region migration error | `greptime_meta_region_migration_error` | `timeseries` | Counter of region migration error | `prometheus` | `none` | `__auto` |
-| Datanode load | `greptime_datanode_load` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `none` | `__auto` |
+| Region migration datanode | `greptime_meta_region_migration_stat{datanode_type="src"}`<br/>`greptime_meta_region_migration_stat{datanode_type="desc"}` | `status-history` | Counter of region migration by source and destination | `prometheus` | -- | `from-datanode-{{datanode_id}}` |
+| Region migration error | `greptime_meta_region_migration_error` | `timeseries` | Counter of region migration error | `prometheus` | `none` | `{{pod}}-{{state}}-{{error_type}}` |
+| Datanode load | `greptime_datanode_load` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `binBps` | `Datanode-{{datanode_id}}-writeload` |
+| Rate of SQL Executions (RDS) | `rate(greptime_meta_rds_pg_sql_execute_elapsed_ms_count[$__rate_interval])` | `timeseries` | Displays the rate of SQL executions processed by the Meta service using the RDS backend. | `prometheus` | `none` | `{{pod}} {{op}} {{type}} {{result}} ` |
+| SQL Execution Latency (RDS) | `histogram_quantile(0.90, sum by(pod, op, type, result, le) (rate(greptime_meta_rds_pg_sql_execute_elapsed_ms_bucket[$__rate_interval])))` | `timeseries` | Measures the response time of SQL executions via the RDS backend.  | `prometheus` | `ms` | `{{pod}} {{op}} {{type}} {{result}} p90` |
+| Handler Execution Latency | `histogram_quantile(0.90, sum by(pod, le, name) (
+  rate(greptime_meta_handler_execute_bucket[$__rate_interval])
+))` | `timeseries` | Shows latency of Meta handlers by pod and handler name, useful for monitoring handler performance and detecting latency spikes.<br/> | `prometheus` | `s` | `{{pod}} {{name}} p90` |
+| Heartbeat Packet Size | `histogram_quantile(0.9, sum by(pod, le) (greptime_meta_heartbeat_stat_memory_size_bucket))` | `timeseries` | Shows p90 heartbeat message sizes, helping track network usage and identify anomalies in heartbeat payload.<br/> | `prometheus` | `bytes` | `{{pod}}` |
+| Meta Heartbeat Receive Rate | `rate(greptime_meta_heartbeat_rate[$__rate_interval])` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `s` | `{{pod}}` |
+| Meta KV Ops Latency | `histogram_quantile(0.99, sum by(pod, le, op, target) (greptime_meta_kv_request_elapsed_bucket))` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `s` | `{{pod}}-{{op}} p99` |
+| Rate of meta KV Ops | `rate(greptime_meta_kv_request_elapsed_count[$__rate_interval])` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `none` | `{{pod}}-{{op}} p99` |
+| DDL Latency | `histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_tables_bucket))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_table))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_view))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_flow))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_drop_table))`<br/>`histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_alter_table))` | `timeseries` | Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads. | `prometheus` | `s` | `CreateLogicalTables-{{step}} p90` |
 # Flownode
 | Title | Query | Type | Description | Datasource | Unit | Legend Format |
 | --- | --- | --- | --- | --- | --- | --- |

grafana/dashboards/metrics/standalone/dashboard.yaml

@@ -497,7 +497,7 @@ groups:
                 type: prometheus
                 uid: ${metrics}
               legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]-p99'
-            - expr: sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_sum{}[$__rate_interval]))/sum by(instance, pod, le, stage) (rate(greptime_mito_compaction_stage_elapsed_count{}[$__rate_interval]))
+            - expr: sum by(instance, pod, stage) (rate(greptime_mito_compaction_stage_elapsed_sum{}[$__rate_interval]))/sum by(instance, pod, stage) (rate(greptime_mito_compaction_stage_elapsed_count{}[$__rate_interval]))
               datasource:
                 type: prometheus
                 uid: ${metrics}
@@ -607,7 +607,7 @@ groups:
                 type: prometheus
                 uid: ${metrics}
               legendFormat: '[{{instance}}]-[{{pod}}]-[{{stage}}]-P95'
-            - expr: sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_sum[$__rate_interval]))/sum by(le,instance, stage, pod) (rate(greptime_region_worker_handle_write_count[$__rate_interval]))
+            - expr: sum by(instance, stage, pod) (rate(greptime_region_worker_handle_write_sum[$__rate_interval]))/sum by(instance, stage, pod) (rate(greptime_region_worker_handle_write_count[$__rate_interval]))
               datasource:
                 type: prometheus
                 uid: ${metrics}
@@ -741,9 +741,8 @@ groups:
     - title: Metasrv
       panels:
         - title: Region migration datanode
-          type: state-timeline
+          type: status-history
           description: Counter of region migration by source and destination
-          unit: none
           queries:
             - expr: greptime_meta_region_migration_stat{datanode_type="src"}
               datasource:
@@ -764,17 +763,127 @@ groups:
               datasource:
                 type: prometheus
                 uid: ${metrics}
-              legendFormat: __auto
+              legendFormat: '{{pod}}-{{state}}-{{error_type}}'
         - title: Datanode load
           type: timeseries
           description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
-          unit: none
+          unit: binBps
           queries:
             - expr: greptime_datanode_load
               datasource:
                 type: prometheus
                 uid: ${metrics}
-              legendFormat: __auto
+              legendFormat: Datanode-{{datanode_id}}-writeload
+        - title: Rate of SQL Executions (RDS)
+          type: timeseries
+          description: Displays the rate of SQL executions processed by the Meta service using the RDS backend.
+          unit: none
+          queries:
+            - expr: rate(greptime_meta_rds_pg_sql_execute_elapsed_ms_count[$__rate_interval])
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}} {{op}} {{type}} {{result}} '
+        - title: SQL Execution Latency (RDS)
+          type: timeseries
+          description: 'Measures the response time of SQL executions via the RDS backend. '
+          unit: ms
+          queries:
+            - expr: histogram_quantile(0.90, sum by(pod, op, type, result, le) (rate(greptime_meta_rds_pg_sql_execute_elapsed_ms_bucket[$__rate_interval])))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}} {{op}} {{type}} {{result}} p90'
+        - title: Handler Execution Latency
+          type: timeseries
+          description: |
+            Shows latency of Meta handlers by pod and handler name, useful for monitoring handler performance and detecting latency spikes.
+          unit: s
+          queries:
+            - expr: |-
+                histogram_quantile(0.90, sum by(pod, le, name) (
+                  rate(greptime_meta_handler_execute_bucket[$__rate_interval])
+                ))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}} {{name}} p90'
+        - title: Heartbeat Packet Size
+          type: timeseries
+          description: |
+            Shows p90 heartbeat message sizes, helping track network usage and identify anomalies in heartbeat payload.
+          unit: bytes
+          queries:
+            - expr: histogram_quantile(0.9, sum by(pod, le) (greptime_meta_heartbeat_stat_memory_size_bucket))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}}'
+        - title: Meta Heartbeat Receive Rate
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: s
+          queries:
+            - expr: rate(greptime_meta_heartbeat_rate[$__rate_interval])
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}}'
+        - title: Meta KV Ops Latency
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.99, sum by(pod, le, op, target) (greptime_meta_kv_request_elapsed_bucket))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}}-{{op}} p99'
+        - title: Rate of meta KV Ops
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: none
+          queries:
+            - expr: rate(greptime_meta_kv_request_elapsed_count[$__rate_interval])
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: '{{pod}}-{{op}} p99'
+        - title: DDL Latency
+          type: timeseries
+          description: Gauge of load information of each datanode, collected via heartbeat between datanode and metasrv. This information is for metasrv to schedule workloads.
+          unit: s
+          queries:
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_tables_bucket))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: CreateLogicalTables-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_table))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: CreateTable-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_view))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: CreateView-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_create_flow))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: CreateFlow-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_drop_table))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: DropTable-{{step}} p90
+            - expr: histogram_quantile(0.9, sum by(le, pod, step) (greptime_meta_procedure_alter_table))
+              datasource:
+                type: prometheus
+                uid: ${metrics}
+              legendFormat: AlterTable-{{step}} p90
     - title: Flownode
       panels:
         - title: Flow Ingest / Output Rate

src/cli/src/export.rs

@@ -19,9 +19,11 @@ use std::time::Duration;
 
 use async_trait::async_trait;
 use clap::{Parser, ValueEnum};
+use common_base::secrets::{ExposeSecret, SecretString};
 use common_error::ext::BoxedError;
 use common_telemetry::{debug, error, info};
 use object_store::layers::LoggingLayer;
+use object_store::services::Oss;
 use object_store::{services, ObjectStore};
 use serde_json::Value;
 use snafu::{OptionExt, ResultExt};
@@ -110,15 +112,15 @@ pub struct ExportCommand {
     #[clap(long)]
     s3: bool,
 
-    /// if both `s3_ddl_local_dir` and `s3` are set, `s3_ddl_local_dir` will be only used for
-    /// exported SQL files, and the data will be exported to s3.
+    /// if both `ddl_local_dir` and remote storage (s3/oss) are set, `ddl_local_dir` will be only used for
+    /// exported SQL files, and the data will be exported to remote storage.
     ///
-    /// Note that `s3_ddl_local_dir` export sql files to **LOCAL** file system, this is useful if export client don't have
-    /// direct access to s3.
+    /// Note that `ddl_local_dir` export sql files to **LOCAL** file system, this is useful if export client don't have
+    /// direct access to remote storage.
     ///
-    /// if `s3` is set but `s3_ddl_local_dir` is not set, both SQL&data will be exported to s3.
+    /// if remote storage is set but `ddl_local_dir` is not set, both SQL&data will be exported to remote storage.
     #[clap(long)]
-    s3_ddl_local_dir: Option<String>,
+    ddl_local_dir: Option<String>,
 
     /// The s3 bucket name
     /// if s3 is set, this is required
@@ -149,6 +151,30 @@ pub struct ExportCommand {
     /// if s3 is set, this is required
     #[clap(long)]
     s3_region: Option<String>,
+
+    /// if export data to oss
+    #[clap(long)]
+    oss: bool,
+
+    /// The oss bucket name
+    /// if oss is set, this is required
+    #[clap(long)]
+    oss_bucket: Option<String>,
+
+    /// The oss endpoint
+    /// if oss is set, this is required
+    #[clap(long)]
+    oss_endpoint: Option<String>,
+
+    /// The oss access key id
+    /// if oss is set, this is required
+    #[clap(long)]
+    oss_access_key_id: Option<String>,
+
+    /// The oss access key secret
+    /// if oss is set, this is required
+    #[clap(long)]
+    oss_access_key_secret: Option<String>,
 }
 
 impl ExportCommand {
@@ -162,7 +188,7 @@ impl ExportCommand {
         {
             return Err(BoxedError::new(S3ConfigNotSetSnafu {}.build()));
         }
-        if !self.s3 && self.output_dir.is_none() {
+        if !self.s3 && !self.oss && self.output_dir.is_none() {
             return Err(BoxedError::new(OutputDirNotSetSnafu {}.build()));
         }
         let (catalog, schema) =
@@ -187,13 +213,32 @@ impl ExportCommand {
             start_time: self.start_time.clone(),
             end_time: self.end_time.clone(),
             s3: self.s3,
-            s3_ddl_local_dir: self.s3_ddl_local_dir.clone(),
+            ddl_local_dir: self.ddl_local_dir.clone(),
             s3_bucket: self.s3_bucket.clone(),
             s3_root: self.s3_root.clone(),
             s3_endpoint: self.s3_endpoint.clone(),
-            s3_access_key: self.s3_access_key.clone(),
-            s3_secret_key: self.s3_secret_key.clone(),
+            // Wrap sensitive values in SecretString
+            s3_access_key: self
+                .s3_access_key
+                .as_ref()
+                .map(|k| SecretString::from(k.clone())),
+            s3_secret_key: self
+                .s3_secret_key
+                .as_ref()
+                .map(|k| SecretString::from(k.clone())),
             s3_region: self.s3_region.clone(),
+            oss: self.oss,
+            oss_bucket: self.oss_bucket.clone(),
+            oss_endpoint: self.oss_endpoint.clone(),
+            // Wrap sensitive values in SecretString
+            oss_access_key_id: self
+                .oss_access_key_id
+                .as_ref()
+                .map(|k| SecretString::from(k.clone())),
+            oss_access_key_secret: self
+                .oss_access_key_secret
+                .as_ref()
+                .map(|k| SecretString::from(k.clone())),
         }))
     }
 }
@@ -209,23 +254,30 @@ pub struct Export {
     start_time: Option<String>,
     end_time: Option<String>,
     s3: bool,
-    s3_ddl_local_dir: Option<String>,
+    ddl_local_dir: Option<String>,
     s3_bucket: Option<String>,
     s3_root: Option<String>,
     s3_endpoint: Option<String>,
-    s3_access_key: Option<String>,
-    s3_secret_key: Option<String>,
+    // Changed to SecretString for sensitive data
+    s3_access_key: Option<SecretString>,
+    s3_secret_key: Option<SecretString>,
     s3_region: Option<String>,
+    oss: bool,
+    oss_bucket: Option<String>,
+    oss_endpoint: Option<String>,
+    // Changed to SecretString for sensitive data
+    oss_access_key_id: Option<SecretString>,
+    oss_access_key_secret: Option<SecretString>,
 }
 
 impl Export {
     fn catalog_path(&self) -> PathBuf {
-        if self.s3 {
+        if self.s3 || self.oss {
             PathBuf::from(&self.catalog)
         } else if let Some(dir) = &self.output_dir {
             PathBuf::from(dir).join(&self.catalog)
         } else {
-            unreachable!("catalog_path: output_dir must be set when not using s3")
+            unreachable!("catalog_path: output_dir must be set when not using remote storage")
         }
     }
 
@@ -427,7 +479,7 @@ impl Export {
                     .await?;
 
                 // Create directory if needed for file system storage
-                if !export_self.s3 {
+                if !export_self.s3 && !export_self.oss {
                     let db_dir = format!("{}/{}/", export_self.catalog, schema);
                     operator.create_dir(&db_dir).await.context(OpenDalSnafu)?;
                 }
@@ -473,6 +525,8 @@ impl Export {
     async fn build_operator(&self) -> Result<ObjectStore> {
         if self.s3 {
             self.build_s3_operator().await
+        } else if self.oss {
+            self.build_oss_operator().await
         } else {
             self.build_fs_operator().await
         }
@@ -480,9 +534,8 @@ impl Export {
 
     /// build operator with preference for file system
     async fn build_prefer_fs_operator(&self) -> Result<ObjectStore> {
-        // is under s3 mode and s3_ddl_dir is set, use it as root
-        if self.s3 && self.s3_ddl_local_dir.is_some() {
-            let root = self.s3_ddl_local_dir.as_ref().unwrap().clone();
+        if (self.s3 || self.oss) && self.ddl_local_dir.is_some() {
+            let root = self.ddl_local_dir.as_ref().unwrap().clone();
             let op = ObjectStore::new(services::Fs::default().root(&root))
                 .context(OpenDalSnafu)?
                 .layer(LoggingLayer::default())
@@ -490,6 +543,8 @@ impl Export {
             Ok(op)
         } else if self.s3 {
             self.build_s3_operator().await
+        } else if self.oss {
+            self.build_oss_operator().await
         } else {
             self.build_fs_operator().await
         }
@@ -515,11 +570,35 @@ impl Export {
         }
 
         if let Some(key_id) = self.s3_access_key.as_ref() {
-            builder = builder.access_key_id(key_id);
+            builder = builder.access_key_id(key_id.expose_secret());
         }
 
         if let Some(secret_key) = self.s3_secret_key.as_ref() {
-            builder = builder.secret_access_key(secret_key);
+            builder = builder.secret_access_key(secret_key.expose_secret());
+        }
+
+        let op = ObjectStore::new(builder)
+            .context(OpenDalSnafu)?
+            .layer(LoggingLayer::default())
+            .finish();
+        Ok(op)
+    }
+
+    async fn build_oss_operator(&self) -> Result<ObjectStore> {
+        let mut builder = Oss::default()
+            .bucket(self.oss_bucket.as_ref().expect("oss_bucket must be set"))
+            .endpoint(
+                self.oss_endpoint
+                    .as_ref()
+                    .expect("oss_endpoint must be set"),
+            );
+
+        // Use expose_secret() to access the actual secret value
+        if let Some(key_id) = self.oss_access_key_id.as_ref() {
+            builder = builder.access_key_id(key_id.expose_secret());
+        }
+        if let Some(secret_key) = self.oss_access_key_secret.as_ref() {
+            builder = builder.access_key_secret(secret_key.expose_secret());
         }
 
         let op = ObjectStore::new(builder)
@@ -562,8 +641,8 @@ impl Export {
             tasks.push(async move {
                 let _permit = semaphore_moved.acquire().await.unwrap();
 
-                // Create directory if not using S3
-                if !export_self.s3 {
+                // Create directory if not using remote storage
+                if !export_self.s3 && !export_self.oss {
                     let db_dir = format!("{}/{}/", export_self.catalog, schema);
                     operator.create_dir(&db_dir).await.context(OpenDalSnafu)?;
                 }
@@ -575,7 +654,11 @@ impl Export {
                     r#"COPY DATABASE "{}"."{}" TO '{}' WITH ({}){};"#,
                     export_self.catalog, schema, path, with_options_clone, connection_part
                 );
-                info!("Executing sql: {sql}");
+
+                // Log SQL command but mask sensitive information
+                let safe_sql = export_self.mask_sensitive_sql(&sql);
+                info!("Executing sql: {}", safe_sql);
+
                 export_self.database_client.sql_in_public(&sql).await?;
                 info!(
                     "Finished exporting {}.{} data to {}",
@@ -615,6 +698,29 @@ impl Export {
         Ok(())
     }
 
+    /// Mask sensitive information in SQL commands for safe logging
+    fn mask_sensitive_sql(&self, sql: &str) -> String {
+        let mut masked_sql = sql.to_string();
+
+        // Mask S3 credentials
+        if let Some(access_key) = &self.s3_access_key {
+            masked_sql = masked_sql.replace(access_key.expose_secret(), "[REDACTED]");
+        }
+        if let Some(secret_key) = &self.s3_secret_key {
+            masked_sql = masked_sql.replace(secret_key.expose_secret(), "[REDACTED]");
+        }
+
+        // Mask OSS credentials
+        if let Some(access_key_id) = &self.oss_access_key_id {
+            masked_sql = masked_sql.replace(access_key_id.expose_secret(), "[REDACTED]");
+        }
+        if let Some(access_key_secret) = &self.oss_access_key_secret {
+            masked_sql = masked_sql.replace(access_key_secret.expose_secret(), "[REDACTED]");
+        }
+
+        masked_sql
+    }
+
     fn get_file_path(&self, schema: &str, file_name: &str) -> String {
         format!("{}/{}/{}", self.catalog, schema, file_name)
     }
@@ -631,6 +737,13 @@ impl Export {
                 },
                 file_path
             )
+        } else if self.oss {
+            format!(
+                "oss://{}/{}/{}",
+                self.oss_bucket.as_ref().unwrap_or(&String::new()),
+                self.catalog,
+                file_path
+            )
         } else {
             format!(
                 "{}/{}",
@@ -675,15 +788,36 @@ impl Export {
             };
 
             // Safety: All s3 options are required
+            // Use expose_secret() to access the actual secret values
             let connection_options = format!(
                 "ACCESS_KEY_ID='{}', SECRET_ACCESS_KEY='{}', REGION='{}'{}",
-                self.s3_access_key.as_ref().unwrap(),
-                self.s3_secret_key.as_ref().unwrap(),
+                self.s3_access_key.as_ref().unwrap().expose_secret(),
+                self.s3_secret_key.as_ref().unwrap().expose_secret(),
                 self.s3_region.as_ref().unwrap(),
                 endpoint_option
             );
 
             (s3_path, format!(" CONNECTION ({})", connection_options))
+        } else if self.oss {
+            let oss_path = format!(
+                "oss://{}/{}/{}/",
+                self.oss_bucket.as_ref().unwrap(),
+                self.catalog,
+                schema
+            );
+            let endpoint_option = if let Some(endpoint) = self.oss_endpoint.as_ref() {
+                format!(", ENDPOINT='{}'", endpoint)
+            } else {
+                String::new()
+            };
+
+            let connection_options = format!(
+                "ACCESS_KEY_ID='{}', ACCESS_KEY_SECRET='{}'{}",
+                self.oss_access_key_id.as_ref().unwrap().expose_secret(),
+                self.oss_access_key_secret.as_ref().unwrap().expose_secret(),
+                endpoint_option
+            );
+            (oss_path, format!(" CONNECTION ({})", connection_options))
         } else {
             (
                 self.catalog_path()

src/client/src/client.rs

@@ -167,9 +167,7 @@ impl Client {
 
         let client = FlightServiceClient::new(channel)
             .max_decoding_message_size(self.max_grpc_recv_message_size())
-            .max_encoding_message_size(self.max_grpc_send_message_size())
-            .accept_compressed(CompressionEncoding::Zstd)
-            .send_compressed(CompressionEncoding::Zstd);
+            .max_encoding_message_size(self.max_grpc_send_message_size());
 
         Ok(FlightClient { addr, client })
     }
@@ -178,9 +176,7 @@ impl Client {
         let (addr, channel) = self.find_channel()?;
         let client = PbRegionClient::new(channel)
             .max_decoding_message_size(self.max_grpc_recv_message_size())
-            .max_encoding_message_size(self.max_grpc_send_message_size())
-            .accept_compressed(CompressionEncoding::Zstd)
-            .send_compressed(CompressionEncoding::Zstd);
+            .max_encoding_message_size(self.max_grpc_send_message_size());
         Ok((addr, client))
     }
 

src/cmd/Cargo.toml

@@ -80,6 +80,7 @@ servers.workspace = true
 session.workspace = true
 similar-asserts.workspace = true
 snafu.workspace = true
+stat.workspace = true
 store-api.workspace = true
 substrait.workspace = true
 table.workspace = true

src/cmd/src/datanode.rs

@@ -14,12 +14,13 @@
 
 pub mod builder;
 
+use std::path::Path;
 use std::time::Duration;
 
 use async_trait::async_trait;
 use clap::Parser;
 use common_config::Configurable;
-use common_telemetry::logging::TracingOptions;
+use common_telemetry::logging::{TracingOptions, DEFAULT_LOGGING_DIR};
 use common_telemetry::{info, warn};
 use common_wal::config::DatanodeWalConfig;
 use datanode::datanode::Datanode;
@@ -248,6 +249,14 @@ impl StartCommand {
             raft_engine_config.dir.replace(wal_dir.clone());
         }
 
+        // If the logging dir is not set, use the default logs dir in the data home.
+        if opts.logging.dir.is_empty() {
+            opts.logging.dir = Path::new(&opts.storage.data_home)
+                .join(DEFAULT_LOGGING_DIR)
+                .to_string_lossy()
+                .to_string();
+        }
+
         if let Some(http_addr) = &self.http_addr {
             opts.http.addr.clone_from(http_addr);
         }

src/cmd/src/datanode/builder.rs

@@ -28,7 +28,7 @@ use tracing_appender::non_blocking::WorkerGuard;
 
 use crate::datanode::{DatanodeOptions, Instance, APP_NAME};
 use crate::error::{MetaClientInitSnafu, MissingConfigSnafu, Result, StartDatanodeSnafu};
-use crate::log_versions;
+use crate::{create_resource_limit_metrics, log_versions};
 
 /// Builder for Datanode instance.
 pub struct InstanceBuilder {
@@ -68,6 +68,7 @@ impl InstanceBuilder {
         );
 
         log_versions(version(), short_version(), APP_NAME);
+        create_resource_limit_metrics(APP_NAME);
 
         plugins::setup_datanode_plugins(plugins, &opts.plugins, dn_opts)
             .await

src/cmd/src/flownode.rs

@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::path::Path;
 use std::sync::Arc;
 use std::time::Duration;
 
@@ -21,7 +22,7 @@ use catalog::kvbackend::{CachedKvBackendBuilder, KvBackendCatalogManager, MetaKv
 use clap::Parser;
 use client::client_manager::NodeClients;
 use common_base::Plugins;
-use common_config::Configurable;
+use common_config::{Configurable, DEFAULT_DATA_HOME};
 use common_grpc::channel_manager::ChannelConfig;
 use common_meta::cache::{CacheRegistryBuilder, LayeredCacheRegistryBuilder};
 use common_meta::heartbeat::handler::invalidate_table_cache::InvalidateCacheHandler;
@@ -30,7 +31,7 @@ use common_meta::heartbeat::handler::HandlerGroupExecutor;
 use common_meta::key::flow::FlowMetadataManager;
 use common_meta::key::TableMetadataManager;
 use common_telemetry::info;
-use common_telemetry::logging::TracingOptions;
+use common_telemetry::logging::{TracingOptions, DEFAULT_LOGGING_DIR};
 use common_version::{short_version, version};
 use flow::{
     get_flow_auth_options, FlownodeBuilder, FlownodeInstance, FlownodeServiceBuilder,
@@ -45,7 +46,7 @@ use crate::error::{
     MissingConfigSnafu, Result, ShutdownFlownodeSnafu, StartFlownodeSnafu,
 };
 use crate::options::{GlobalOptions, GreptimeOptions};
-use crate::{log_versions, App};
+use crate::{create_resource_limit_metrics, log_versions, App};
 
 pub const APP_NAME: &str = "greptime-flownode";
 
@@ -186,6 +187,14 @@ impl StartCommand {
             opts.logging.dir.clone_from(dir);
         }
 
+        // If the logging dir is not set, use the default logs dir in the data home.
+        if opts.logging.dir.is_empty() {
+            opts.logging.dir = Path::new(DEFAULT_DATA_HOME)
+                .join(DEFAULT_LOGGING_DIR)
+                .to_string_lossy()
+                .to_string();
+        }
+
         if global_options.log_level.is_some() {
             opts.logging.level.clone_from(&global_options.log_level);
         }
@@ -246,7 +255,9 @@ impl StartCommand {
             opts.component.node_id.map(|x| x.to_string()),
             None,
         );
+
         log_versions(version(), short_version(), APP_NAME);
+        create_resource_limit_metrics(APP_NAME);
 
         info!("Flownode start command: {:#?}", self);
         info!("Flownode options: {:#?}", opts);

src/cmd/src/frontend.rs

@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::path::Path;
 use std::sync::Arc;
 use std::time::Duration;
 
@@ -22,14 +23,14 @@ use catalog::kvbackend::{CachedKvBackendBuilder, KvBackendCatalogManager, MetaKv
 use clap::Parser;
 use client::client_manager::NodeClients;
 use common_base::Plugins;
-use common_config::Configurable;
+use common_config::{Configurable, DEFAULT_DATA_HOME};
 use common_grpc::channel_manager::ChannelConfig;
 use common_meta::cache::{CacheRegistryBuilder, LayeredCacheRegistryBuilder};
 use common_meta::heartbeat::handler::invalidate_table_cache::InvalidateCacheHandler;
 use common_meta::heartbeat::handler::parse_mailbox_message::ParseMailboxMessageHandler;
 use common_meta::heartbeat::handler::HandlerGroupExecutor;
 use common_telemetry::info;
-use common_telemetry::logging::TracingOptions;
+use common_telemetry::logging::{TracingOptions, DEFAULT_LOGGING_DIR};
 use common_time::timezone::set_default_timezone;
 use common_version::{short_version, version};
 use frontend::frontend::Frontend;
@@ -44,7 +45,7 @@ use tracing_appender::non_blocking::WorkerGuard;
 
 use crate::error::{self, Result};
 use crate::options::{GlobalOptions, GreptimeOptions};
-use crate::{log_versions, App};
+use crate::{create_resource_limit_metrics, log_versions, App};
 
 type FrontendOptions = GreptimeOptions<frontend::frontend::FrontendOptions>;
 
@@ -194,6 +195,14 @@ impl StartCommand {
             opts.logging.dir.clone_from(dir);
         }
 
+        // If the logging dir is not set, use the default logs dir in the data home.
+        if opts.logging.dir.is_empty() {
+            opts.logging.dir = Path::new(DEFAULT_DATA_HOME)
+                .join(DEFAULT_LOGGING_DIR)
+                .to_string_lossy()
+                .to_string();
+        }
+
         if global_options.log_level.is_some() {
             opts.logging.level.clone_from(&global_options.log_level);
         }
@@ -270,7 +279,9 @@ impl StartCommand {
             opts.component.node_id.clone(),
             opts.component.slow_query.as_ref(),
         );
+
         log_versions(version(), short_version(), APP_NAME);
+        create_resource_limit_metrics(APP_NAME);
 
         info!("Frontend start command: {:#?}", self);
         info!("Frontend options: {:#?}", opts);

src/cmd/src/lib.rs

@@ -16,6 +16,7 @@
 
 use async_trait::async_trait;
 use common_telemetry::{error, info};
+use stat::{get_cpu_limit, get_memory_limit};
 
 use crate::error::Result;
 
@@ -31,6 +32,12 @@ pub mod standalone;
 lazy_static::lazy_static! {
     static ref APP_VERSION: prometheus::IntGaugeVec =
         prometheus::register_int_gauge_vec!("greptime_app_version", "app version", &["version", "short_version", "app"]).unwrap();
+
+    static ref CPU_LIMIT: prometheus::IntGaugeVec =
+        prometheus::register_int_gauge_vec!("greptime_cpu_limit_in_millicores", "cpu limit in millicores", &["app"]).unwrap();
+
+    static ref MEMORY_LIMIT: prometheus::IntGaugeVec =
+        prometheus::register_int_gauge_vec!("greptime_memory_limit_in_bytes", "memory limit in bytes", &["app"]).unwrap();
 }
 
 /// wait for the close signal, for unix platform it's SIGINT or SIGTERM
@@ -114,6 +121,24 @@ pub fn log_versions(version: &str, short_version: &str, app: &str) {
     log_env_flags();
 }
 
+pub fn create_resource_limit_metrics(app: &str) {
+    if let Some(cpu_limit) = get_cpu_limit() {
+        info!(
+            "GreptimeDB start with cpu limit in millicores: {}",
+            cpu_limit
+        );
+        CPU_LIMIT.with_label_values(&[app]).set(cpu_limit);
+    }
+
+    if let Some(memory_limit) = get_memory_limit() {
+        info!(
+            "GreptimeDB start with memory limit in bytes: {}",
+            memory_limit
+        );
+        MEMORY_LIMIT.with_label_values(&[app]).set(memory_limit);
+    }
+}
+
 fn log_env_flags() {
     info!("command line arguments");
     for argument in std::env::args() {

src/cmd/src/metasrv.rs

@@ -13,6 +13,7 @@
 // limitations under the License.
 
 use std::fmt;
+use std::path::Path;
 use std::time::Duration;
 
 use async_trait::async_trait;
@@ -20,7 +21,7 @@ use clap::Parser;
 use common_base::Plugins;
 use common_config::Configurable;
 use common_telemetry::info;
-use common_telemetry::logging::TracingOptions;
+use common_telemetry::logging::{TracingOptions, DEFAULT_LOGGING_DIR};
 use common_version::{short_version, version};
 use meta_srv::bootstrap::MetasrvInstance;
 use meta_srv::metasrv::BackendImpl;
@@ -29,7 +30,7 @@ use tracing_appender::non_blocking::WorkerGuard;
 
 use crate::error::{self, LoadLayeredConfigSnafu, Result, StartMetaServerSnafu};
 use crate::options::{GlobalOptions, GreptimeOptions};
-use crate::{log_versions, App};
+use crate::{create_resource_limit_metrics, log_versions, App};
 
 type MetasrvOptions = GreptimeOptions<meta_srv::metasrv::MetasrvOptions>;
 
@@ -236,12 +237,20 @@ impl StartCommand {
             tokio_console_addr: global_options.tokio_console_addr.clone(),
         };
 
+        #[allow(deprecated)]
         if let Some(addr) = &self.rpc_bind_addr {
             opts.bind_addr.clone_from(addr);
+            opts.grpc.bind_addr.clone_from(addr);
+        } else if !opts.bind_addr.is_empty() {
+            opts.grpc.bind_addr.clone_from(&opts.bind_addr);
         }
 
+        #[allow(deprecated)]
         if let Some(addr) = &self.rpc_server_addr {
             opts.server_addr.clone_from(addr);
+            opts.grpc.server_addr.clone_from(addr);
+        } else if !opts.server_addr.is_empty() {
+            opts.grpc.server_addr.clone_from(&opts.server_addr);
         }
 
         if let Some(addrs) = &self.store_addrs {
@@ -274,6 +283,14 @@ impl StartCommand {
             opts.data_home.clone_from(data_home);
         }
 
+        // If the logging dir is not set, use the default logs dir in the data home.
+        if opts.logging.dir.is_empty() {
+            opts.logging.dir = Path::new(&opts.data_home)
+                .join(DEFAULT_LOGGING_DIR)
+                .to_string_lossy()
+                .to_string();
+        }
+
         if !self.store_key_prefix.is_empty() {
             opts.store_key_prefix.clone_from(&self.store_key_prefix)
         }
@@ -302,13 +319,15 @@ impl StartCommand {
             None,
             None,
         );
+
         log_versions(version(), short_version(), APP_NAME);
+        create_resource_limit_metrics(APP_NAME);
 
         info!("Metasrv start command: {:#?}", self);
 
         let plugin_opts = opts.plugins;
         let mut opts = opts.component;
-        opts.detect_server_addr();
+        opts.grpc.detect_server_addr();
 
         info!("Metasrv options: {:#?}", opts);
 
@@ -352,7 +371,7 @@ mod tests {
         };
 
         let options = cmd.load_options(&Default::default()).unwrap().component;
-        assert_eq!("127.0.0.1:3002".to_string(), options.bind_addr);
+        assert_eq!("127.0.0.1:3002".to_string(), options.grpc.bind_addr);
         assert_eq!(vec!["127.0.0.1:2380".to_string()], options.store_addrs);
         assert_eq!(SelectorType::LoadBased, options.selector);
     }
@@ -385,8 +404,8 @@ mod tests {
         };
 
         let options = cmd.load_options(&Default::default()).unwrap().component;
-        assert_eq!("127.0.0.1:3002".to_string(), options.bind_addr);
-        assert_eq!("127.0.0.1:3002".to_string(), options.server_addr);
+        assert_eq!("127.0.0.1:3002".to_string(), options.grpc.bind_addr);
+        assert_eq!("127.0.0.1:3002".to_string(), options.grpc.server_addr);
         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());
@@ -498,10 +517,10 @@ mod tests {
                 let opts = command.load_options(&Default::default()).unwrap().component;
 
                 // Should be read from env, env > default values.
-                assert_eq!(opts.bind_addr, "127.0.0.1:14002");
+                assert_eq!(opts.grpc.bind_addr, "127.0.0.1:14002");
 
                 // Should be read from config file, config file > env > default values.
-                assert_eq!(opts.server_addr, "127.0.0.1:3002");
+                assert_eq!(opts.grpc.server_addr, "127.0.0.1:3002");
 
                 // Should be read from cli, cli > config file > env > default values.
                 assert_eq!(opts.http.addr, "127.0.0.1:14000");

src/cmd/src/standalone.rs

@@ -13,6 +13,7 @@
 // limitations under the License.
 
 use std::net::SocketAddr;
+use std::path::Path;
 use std::sync::Arc;
 use std::{fs, path};
 
@@ -49,7 +50,9 @@ use common_meta::sequence::SequenceBuilder;
 use common_meta::wal_options_allocator::{build_wal_options_allocator, WalOptionsAllocatorRef};
 use common_procedure::{ProcedureInfo, ProcedureManagerRef};
 use common_telemetry::info;
-use common_telemetry::logging::{LoggingOptions, SlowQueryOptions, TracingOptions};
+use common_telemetry::logging::{
+    LoggingOptions, SlowQueryOptions, TracingOptions, DEFAULT_LOGGING_DIR,
+};
 use common_time::timezone::set_default_timezone;
 use common_version::{short_version, version};
 use common_wal::config::DatanodeWalConfig;
@@ -83,7 +86,7 @@ use tracing_appender::non_blocking::WorkerGuard;
 
 use crate::error::{Result, StartFlownodeSnafu};
 use crate::options::{GlobalOptions, GreptimeOptions};
-use crate::{error, log_versions, App};
+use crate::{create_resource_limit_metrics, error, log_versions, App};
 
 pub const APP_NAME: &str = "greptime-standalone";
 
@@ -407,6 +410,14 @@ impl StartCommand {
             opts.storage.data_home.clone_from(data_home);
         }
 
+        // If the logging dir is not set, use the default logs dir in the data home.
+        if opts.logging.dir.is_empty() {
+            opts.logging.dir = Path::new(&opts.storage.data_home)
+                .join(DEFAULT_LOGGING_DIR)
+                .to_string_lossy()
+                .to_string();
+        }
+
         if let Some(addr) = &self.rpc_bind_addr {
             // frontend grpc addr conflict with datanode default grpc addr
             let datanode_grpc_addr = DatanodeOptions::default().grpc.bind_addr;
@@ -457,7 +468,9 @@ impl StartCommand {
             None,
             opts.component.slow_query.as_ref(),
         );
+
         log_versions(version(), short_version(), APP_NAME);
+        create_resource_limit_metrics(APP_NAME);
 
         info!("Standalone start command: {:#?}", self);
         info!("Standalone options: {opts:#?}");

src/cmd/tests/load_config_test.rs

@@ -12,13 +12,14 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::path::Path;
 use std::time::Duration;
 
 use cmd::options::GreptimeOptions;
 use cmd::standalone::StandaloneOptions;
-use common_config::Configurable;
+use common_config::{Configurable, DEFAULT_DATA_HOME};
 use common_options::datanode::{ClientOptions, DatanodeClientOptions};
-use common_telemetry::logging::{LoggingOptions, DEFAULT_OTLP_ENDPOINT};
+use common_telemetry::logging::{LoggingOptions, DEFAULT_LOGGING_DIR, DEFAULT_OTLP_ENDPOINT};
 use common_wal::config::raft_engine::RaftEngineConfig;
 use common_wal::config::DatanodeWalConfig;
 use datanode::config::{DatanodeOptions, RegionEngineConfig, StorageConfig};
@@ -32,6 +33,7 @@ use mito2::config::MitoConfig;
 use servers::export_metrics::ExportMetricsOption;
 use servers::grpc::GrpcOptions;
 use servers::http::HttpOptions;
+use store_api::path_utils::WAL_DIR;
 
 #[allow(deprecated)]
 #[test]
@@ -56,13 +58,18 @@ fn test_load_datanode_example_config() {
                 metadata_cache_tti: Duration::from_secs(300),
             }),
             wal: DatanodeWalConfig::RaftEngine(RaftEngineConfig {
-                dir: Some("./greptimedb_data/wal".to_string()),
+                dir: Some(
+                    Path::new(DEFAULT_DATA_HOME)
+                        .join(WAL_DIR)
+                        .to_string_lossy()
+                        .to_string(),
+                ),
                 sync_period: Some(Duration::from_secs(10)),
                 recovery_parallelism: 2,
                 ..Default::default()
             }),
             storage: StorageConfig {
-                data_home: "./greptimedb_data/".to_string(),
+                data_home: DEFAULT_DATA_HOME.to_string(),
                 ..Default::default()
             },
             region_engine: vec![
@@ -79,12 +86,16 @@ fn test_load_datanode_example_config() {
             ],
             logging: LoggingOptions {
                 level: Some("info".to_string()),
+                dir: Path::new(DEFAULT_DATA_HOME)
+                    .join(DEFAULT_LOGGING_DIR)
+                    .to_string_lossy()
+                    .to_string(),
                 otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                 tracing_sample_ratio: Some(Default::default()),
                 ..Default::default()
             },
             export_metrics: ExportMetricsOption {
-                self_import: Some(Default::default()),
+                self_import: None,
                 remote_write: Some(Default::default()),
                 ..Default::default()
             },
@@ -121,6 +132,10 @@ fn test_load_frontend_example_config() {
             }),
             logging: LoggingOptions {
                 level: Some("info".to_string()),
+                dir: Path::new(DEFAULT_DATA_HOME)
+                    .join(DEFAULT_LOGGING_DIR)
+                    .to_string_lossy()
+                    .to_string(),
                 otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                 tracing_sample_ratio: Some(Default::default()),
                 ..Default::default()
@@ -133,7 +148,7 @@ fn test_load_frontend_example_config() {
                 },
             },
             export_metrics: ExportMetricsOption {
-                self_import: Some(Default::default()),
+                self_import: None,
                 remote_write: Some(Default::default()),
                 ..Default::default()
             },
@@ -160,10 +175,17 @@ fn test_load_metasrv_example_config() {
     let expected = GreptimeOptions::<MetasrvOptions> {
         component: MetasrvOptions {
             selector: SelectorType::default(),
-            data_home: "./greptimedb_data/metasrv/".to_string(),
-            server_addr: "127.0.0.1:3002".to_string(),
+            data_home: DEFAULT_DATA_HOME.to_string(),
+            grpc: GrpcOptions {
+                bind_addr: "127.0.0.1:3002".to_string(),
+                server_addr: "127.0.0.1:3002".to_string(),
+                ..Default::default()
+            },
             logging: LoggingOptions {
-                dir: "./greptimedb_data/logs".to_string(),
+                dir: Path::new(DEFAULT_DATA_HOME)
+                    .join(DEFAULT_LOGGING_DIR)
+                    .to_string_lossy()
+                    .to_string(),
                 level: Some("info".to_string()),
                 otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                 tracing_sample_ratio: Some(Default::default()),
@@ -177,7 +199,7 @@ fn test_load_metasrv_example_config() {
                 },
             },
             export_metrics: ExportMetricsOption {
-                self_import: Some(Default::default()),
+                self_import: None,
                 remote_write: Some(Default::default()),
                 ..Default::default()
             },
@@ -198,7 +220,12 @@ fn test_load_standalone_example_config() {
         component: StandaloneOptions {
             default_timezone: Some("UTC".to_string()),
             wal: DatanodeWalConfig::RaftEngine(RaftEngineConfig {
-                dir: Some("./greptimedb_data/wal".to_string()),
+                dir: Some(
+                    Path::new(DEFAULT_DATA_HOME)
+                        .join(WAL_DIR)
+                        .to_string_lossy()
+                        .to_string(),
+                ),
                 sync_period: Some(Duration::from_secs(10)),
                 recovery_parallelism: 2,
                 ..Default::default()
@@ -216,11 +243,15 @@ fn test_load_standalone_example_config() {
                 }),
             ],
             storage: StorageConfig {
-                data_home: "./greptimedb_data/".to_string(),
+                data_home: DEFAULT_DATA_HOME.to_string(),
                 ..Default::default()
             },
             logging: LoggingOptions {
                 level: Some("info".to_string()),
+                dir: Path::new(DEFAULT_DATA_HOME)
+                    .join(DEFAULT_LOGGING_DIR)
+                    .to_string_lossy()
+                    .to_string(),
                 otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                 tracing_sample_ratio: Some(Default::default()),
                 ..Default::default()

src/common/config/src/lib.rs

@@ -26,6 +26,9 @@ pub fn metadata_store_dir(store_dir: &str) -> String {
     format!("{store_dir}/metadata")
 }
 
+/// The default data home directory.
+pub const DEFAULT_DATA_HOME: &str = "./greptimedb_data";
+
 #[derive(Debug, Clone, Copy, Serialize, Deserialize, PartialEq, Eq)]
 #[serde(default)]
 pub struct KvBackendConfig {

src/common/datasource/src/object_store.rs

@@ -13,7 +13,9 @@
 // limitations under the License.
 
 pub mod fs;
+pub mod oss;
 pub mod s3;
+
 use std::collections::HashMap;
 
 use lazy_static::lazy_static;
@@ -25,10 +27,12 @@ use url::{ParseError, Url};
 use self::fs::build_fs_backend;
 use self::s3::build_s3_backend;
 use crate::error::{self, Result};
+use crate::object_store::oss::build_oss_backend;
 use crate::util::find_dir_and_filename;
 
 pub const FS_SCHEMA: &str = "FS";
 pub const S3_SCHEMA: &str = "S3";
+pub const OSS_SCHEMA: &str = "OSS";
 
 /// Returns `(schema, Option<host>, path)`
 pub fn parse_url(url: &str) -> Result<(String, Option<String>, String)> {
@@ -64,6 +68,12 @@ pub fn build_backend(url: &str, connection: &HashMap<String, String>) -> Result<
             })?;
             Ok(build_s3_backend(&host, &root, connection)?)
         }
+        OSS_SCHEMA => {
+            let host = host.context(error::EmptyHostPathSnafu {
+                url: url.to_string(),
+            })?;
+            Ok(build_oss_backend(&host, &root, connection)?)
+        }
         FS_SCHEMA => Ok(build_fs_backend(&root)?),
 
         _ => error::UnsupportedBackendProtocolSnafu {

src/common/datasource/src/object_store/oss.rs

@@ -0,0 +1,118 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::collections::HashMap;
+
+use object_store::services::Oss;
+use object_store::ObjectStore;
+use snafu::ResultExt;
+
+use crate::error::{self, Result};
+
+const BUCKET: &str = "bucket";
+const ENDPOINT: &str = "endpoint";
+const ACCESS_KEY_ID: &str = "access_key_id";
+const ACCESS_KEY_SECRET: &str = "access_key_secret";
+const ROOT: &str = "root";
+const ALLOW_ANONYMOUS: &str = "allow_anonymous";
+
+/// Check if the key is supported in OSS configuration.
+pub fn is_supported_in_oss(key: &str) -> bool {
+    [
+        ROOT,
+        ALLOW_ANONYMOUS,
+        BUCKET,
+        ENDPOINT,
+        ACCESS_KEY_ID,
+        ACCESS_KEY_SECRET,
+    ]
+    .contains(&key)
+}
+
+/// Build an OSS backend using the provided bucket, root, and connection parameters.
+pub fn build_oss_backend(
+    bucket: &str,
+    root: &str,
+    connection: &HashMap<String, String>,
+) -> Result<ObjectStore> {
+    let mut builder = Oss::default().bucket(bucket).root(root);
+
+    if let Some(endpoint) = connection.get(ENDPOINT) {
+        builder = builder.endpoint(endpoint);
+    }
+
+    if let Some(access_key_id) = connection.get(ACCESS_KEY_ID) {
+        builder = builder.access_key_id(access_key_id);
+    }
+
+    if let Some(access_key_secret) = connection.get(ACCESS_KEY_SECRET) {
+        builder = builder.access_key_secret(access_key_secret);
+    }
+
+    if let Some(allow_anonymous) = connection.get(ALLOW_ANONYMOUS) {
+        let allow = allow_anonymous.as_str().parse::<bool>().map_err(|e| {
+            error::InvalidConnectionSnafu {
+                msg: format!(
+                    "failed to parse the option {}={}, {}",
+                    ALLOW_ANONYMOUS, allow_anonymous, e
+                ),
+            }
+            .build()
+        })?;
+        if allow {
+            builder = builder.allow_anonymous();
+        }
+    }
+
+    let op = ObjectStore::new(builder)
+        .context(error::BuildBackendSnafu)?
+        .layer(object_store::layers::LoggingLayer::default())
+        .layer(object_store::layers::TracingLayer)
+        .layer(object_store::layers::build_prometheus_metrics_layer(true))
+        .finish();
+
+    Ok(op)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_is_supported_in_oss() {
+        assert!(is_supported_in_oss(ROOT));
+        assert!(is_supported_in_oss(ALLOW_ANONYMOUS));
+        assert!(is_supported_in_oss(BUCKET));
+        assert!(is_supported_in_oss(ENDPOINT));
+        assert!(is_supported_in_oss(ACCESS_KEY_ID));
+        assert!(is_supported_in_oss(ACCESS_KEY_SECRET));
+        assert!(!is_supported_in_oss("foo"));
+        assert!(!is_supported_in_oss("BAR"));
+    }
+
+    #[test]
+    fn test_build_oss_backend_all_fields_valid() {
+        let mut connection = HashMap::new();
+        connection.insert(
+            ENDPOINT.to_string(),
+            "http://oss-ap-southeast-1.aliyuncs.com".to_string(),
+        );
+        connection.insert(ACCESS_KEY_ID.to_string(), "key_id".to_string());
+        connection.insert(ACCESS_KEY_SECRET.to_string(), "key_secret".to_string());
+        connection.insert(ALLOW_ANONYMOUS.to_string(), "true".to_string());
+
+        let result = build_oss_backend("my-bucket", "my-root", &connection);
+        assert!(result.is_ok());
+    }
+}

src/common/function/src/adjust_flow.rs

@@ -0,0 +1,90 @@
+// 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_query::error::{
+    InvalidFuncArgsSnafu, MissingFlowServiceHandlerSnafu, Result, UnsupportedInputDataTypeSnafu,
+};
+use common_query::prelude::Signature;
+use datafusion::logical_expr::Volatility;
+use datatypes::value::{Value, ValueRef};
+use session::context::QueryContextRef;
+use snafu::ensure;
+use store_api::storage::ConcreteDataType;
+
+use crate::handlers::FlowServiceHandlerRef;
+use crate::helper::parse_catalog_flow;
+
+fn adjust_signature() -> Signature {
+    Signature::exact(
+        vec![
+            ConcreteDataType::string_datatype(), // flow name
+            ConcreteDataType::uint64_datatype(), // min_run_interval in seconds
+            ConcreteDataType::uint64_datatype(), // max filter number per query
+        ],
+        Volatility::Immutable,
+    )
+}
+
+#[admin_fn(
+    name = AdjustFlowFunction,
+    display_name = adjust_flow,
+    sig_fn = adjust_signature,
+    ret = uint64
+)]
+pub(crate) async fn adjust_flow(
+    flow_service_handler: &FlowServiceHandlerRef,
+    query_ctx: &QueryContextRef,
+    params: &[ValueRef<'_>],
+) -> Result<Value> {
+    ensure!(
+        params.len() == 3,
+        InvalidFuncArgsSnafu {
+            err_msg: format!(
+                "The length of the args is not correct, expect 3, have: {}",
+                params.len()
+            ),
+        }
+    );
+
+    let (flow_name, min_run_interval, max_filter_num) = match (params[0], params[1], params[2]) {
+        (
+            ValueRef::String(flow_name),
+            ValueRef::UInt64(min_run_interval),
+            ValueRef::UInt64(max_filter_num),
+        ) => (flow_name, min_run_interval, max_filter_num),
+        _ => {
+            return UnsupportedInputDataTypeSnafu {
+                function: "adjust_flow",
+                datatypes: params.iter().map(|v| v.data_type()).collect::<Vec<_>>(),
+            }
+            .fail();
+        }
+    };
+
+    let (catalog_name, flow_name) = parse_catalog_flow(flow_name, query_ctx)?;
+
+    let res = flow_service_handler
+        .adjust(
+            &catalog_name,
+            &flow_name,
+            min_run_interval,
+            max_filter_num as usize,
+            query_ctx.clone(),
+        )
+        .await?;
+    let affected_rows = res.affected_rows;
+
+    Ok(Value::from(affected_rows))
+}

src/common/function/src/admin.rs

@@ -26,6 +26,7 @@ use flush_compact_table::{CompactTableFunction, FlushTableFunction};
 use migrate_region::MigrateRegionFunction;
 use remove_region_follower::RemoveRegionFollowerFunction;
 
+use crate::adjust_flow::AdjustFlowFunction;
 use crate::flush_flow::FlushFlowFunction;
 use crate::function_registry::FunctionRegistry;
 
@@ -43,5 +44,6 @@ impl AdminFunction {
         registry.register_async(Arc::new(FlushTableFunction));
         registry.register_async(Arc::new(CompactTableFunction));
         registry.register_async(Arc::new(FlushFlowFunction));
+        registry.register_async(Arc::new(AdjustFlowFunction));
     }
 }

src/common/function/src/aggrs.rs

@@ -0,0 +1,18 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+pub mod approximate;
+#[cfg(feature = "geo")]
+pub mod geo;
+pub mod vector;

src/common/function/src/aggrs/approximate.rs

@@ -0,0 +1,32 @@
+// 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 crate::function_registry::FunctionRegistry;
+
+pub(crate) mod hll;
+mod uddsketch;
+
+pub(crate) struct ApproximateFunction;
+
+impl ApproximateFunction {
+    pub fn register(registry: &FunctionRegistry) {
+        // uddsketch
+        registry.register_aggr(uddsketch::UddSketchState::state_udf_impl());
+        registry.register_aggr(uddsketch::UddSketchState::merge_udf_impl());
+
+        // hll
+        registry.register_aggr(hll::HllState::state_udf_impl());
+        registry.register_aggr(hll::HllState::merge_udf_impl());
+    }
+}

src/common/function/src/aggrs/geo.rs

@@ -0,0 +1,27 @@
+// 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 crate::function_registry::FunctionRegistry;
+
+mod encoding;
+mod geo_path;
+
+pub(crate) struct GeoFunction;
+
+impl GeoFunction {
+    pub fn register(registry: &FunctionRegistry) {
+        registry.register_aggr(geo_path::GeoPathAccumulator::uadf_impl());
+        registry.register_aggr(encoding::JsonPathAccumulator::uadf_impl());
+    }
+}

src/common/function/src/aggrs/geo/encoding.rs

@@ -19,9 +19,12 @@ use common_error::status_code::StatusCode;
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
 use common_query::error::{self, InvalidInputStateSnafu, Result};
 use common_query::logical_plan::accumulator::AggrFuncTypeStore;
-use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
+use common_query::logical_plan::{
+    create_aggregate_function, Accumulator, AggregateFunctionCreator,
+};
 use common_query::prelude::AccumulatorCreatorFunction;
 use common_time::Timestamp;
+use datafusion_expr::AggregateUDF;
 use datatypes::prelude::ConcreteDataType;
 use datatypes::value::{ListValue, Value};
 use datatypes::vectors::VectorRef;
@@ -47,6 +50,16 @@ impl JsonPathAccumulator {
             timestamp_type,
         }
     }
+
+    /// Create a new `AggregateUDF` for the `json_encode_path` aggregate function.
+    pub fn uadf_impl() -> AggregateUDF {
+        create_aggregate_function(
+            "json_encode_path".to_string(),
+            3,
+            Arc::new(JsonPathEncodeFunctionCreator::default()),
+        )
+        .into()
+    }
 }
 
 impl Accumulator for JsonPathAccumulator {

src/common/function/src/aggrs/geo/geo_path.rs

@@ -47,7 +47,7 @@ impl GeoPathAccumulator {
         Self::default()
     }
 
-    pub fn udf_impl() -> AggregateUDF {
+    pub fn uadf_impl() -> AggregateUDF {
         create_udaf(
             GEO_PATH_NAME,
             // Input types: lat, lng, timestamp

src/common/function/src/aggrs/vector.rs

@@ -0,0 +1,29 @@
+// 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 crate::aggrs::vector::product::VectorProduct;
+use crate::aggrs::vector::sum::VectorSum;
+use crate::function_registry::FunctionRegistry;
+
+mod product;
+mod sum;
+
+pub(crate) struct VectorFunction;
+
+impl VectorFunction {
+    pub fn register(registry: &FunctionRegistry) {
+        registry.register_aggr(VectorSum::uadf_impl());
+        registry.register_aggr(VectorProduct::uadf_impl());
+    }
+}

src/common/function/src/aggrs/vector/product.rs

@@ -16,8 +16,11 @@ use std::sync::Arc;
 
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
 use common_query::error::{CreateAccumulatorSnafu, Error, InvalidFuncArgsSnafu};
-use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
+use common_query::logical_plan::{
+    create_aggregate_function, Accumulator, AggregateFunctionCreator,
+};
 use common_query::prelude::AccumulatorCreatorFunction;
+use datafusion_expr::AggregateUDF;
 use datatypes::prelude::{ConcreteDataType, Value, *};
 use datatypes::vectors::VectorRef;
 use nalgebra::{Const, DVectorView, Dyn, OVector};
@@ -75,6 +78,16 @@ impl AggregateFunctionCreator for VectorProductCreator {
 }
 
 impl VectorProduct {
+    /// Create a new `AggregateUDF` for the `vec_product` aggregate function.
+    pub fn uadf_impl() -> AggregateUDF {
+        create_aggregate_function(
+            "vec_product".to_string(),
+            1,
+            Arc::new(VectorProductCreator::default()),
+        )
+        .into()
+    }
+
     fn inner(&mut self, len: usize) -> &mut OVector<f32, Dyn> {
         self.product.get_or_insert_with(|| {
             OVector::from_iterator_generic(Dyn(len), Const::<1>, (0..len).map(|_| 1.0))

src/common/function/src/aggrs/vector/sum.rs

@@ -16,8 +16,11 @@ use std::sync::Arc;
 
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
 use common_query::error::{CreateAccumulatorSnafu, Error, InvalidFuncArgsSnafu};
-use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
+use common_query::logical_plan::{
+    create_aggregate_function, Accumulator, AggregateFunctionCreator,
+};
 use common_query::prelude::AccumulatorCreatorFunction;
+use datafusion_expr::AggregateUDF;
 use datatypes::prelude::{ConcreteDataType, Value, *};
 use datatypes::vectors::VectorRef;
 use nalgebra::{Const, DVectorView, Dyn, OVector};
@@ -25,6 +28,7 @@ use snafu::ensure;
 
 use crate::scalars::vector::impl_conv::{as_veclit, as_veclit_if_const, veclit_to_binlit};
 
+/// The accumulator for the `vec_sum` aggregate function.
 #[derive(Debug, Default)]
 pub struct VectorSum {
     sum: Option<OVector<f32, Dyn>>,
@@ -74,6 +78,16 @@ impl AggregateFunctionCreator for VectorSumCreator {
 }
 
 impl VectorSum {
+    /// Create a new `AggregateUDF` for the `vec_sum` aggregate function.
+    pub fn uadf_impl() -> AggregateUDF {
+        create_aggregate_function(
+            "vec_sum".to_string(),
+            1,
+            Arc::new(VectorSumCreator::default()),
+        )
+        .into()
+    }
+
     fn inner(&mut self, len: usize) -> &mut OVector<f32, Dyn> {
         self.sum
             .get_or_insert_with(|| OVector::zeros_generic(Dyn(len), Const::<1>))

src/common/function/src/flush_flow.rs

@@ -12,21 +12,19 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use common_error::ext::BoxedError;
 use common_macro::admin_fn;
 use common_query::error::{
-    ExecuteSnafu, InvalidFuncArgsSnafu, MissingFlowServiceHandlerSnafu, Result,
-    UnsupportedInputDataTypeSnafu,
+    InvalidFuncArgsSnafu, MissingFlowServiceHandlerSnafu, Result, UnsupportedInputDataTypeSnafu,
 };
 use common_query::prelude::Signature;
 use datafusion::logical_expr::Volatility;
 use datatypes::value::{Value, ValueRef};
 use session::context::QueryContextRef;
-use snafu::{ensure, ResultExt};
-use sql::parser::ParserContext;
+use snafu::ensure;
 use store_api::storage::ConcreteDataType;
 
 use crate::handlers::FlowServiceHandlerRef;
+use crate::helper::parse_catalog_flow;
 
 fn flush_signature() -> Signature {
     Signature::uniform(
@@ -47,20 +45,6 @@ pub(crate) async fn flush_flow(
     query_ctx: &QueryContextRef,
     params: &[ValueRef<'_>],
 ) -> Result<Value> {
-    let (catalog_name, flow_name) = parse_flush_flow(params, query_ctx)?;
-
-    let res = flow_service_handler
-        .flush(&catalog_name, &flow_name, query_ctx.clone())
-        .await?;
-    let affected_rows = res.affected_rows;
-
-    Ok(Value::from(affected_rows))
-}
-
-fn parse_flush_flow(
-    params: &[ValueRef<'_>],
-    query_ctx: &QueryContextRef,
-) -> Result<(String, String)> {
     ensure!(
         params.len() == 1,
         InvalidFuncArgsSnafu {
@@ -70,7 +54,6 @@ fn parse_flush_flow(
             ),
         }
     );
-
     let ValueRef::String(flow_name) = params[0] else {
         return UnsupportedInputDataTypeSnafu {
             function: "flush_flow",
@@ -78,27 +61,14 @@ fn parse_flush_flow(
         }
         .fail();
     };
-    let obj_name = ParserContext::parse_table_name(flow_name, query_ctx.sql_dialect())
-        .map_err(BoxedError::new)
-        .context(ExecuteSnafu)?;
+    let (catalog_name, flow_name) = parse_catalog_flow(flow_name, query_ctx)?;
 
-    let (catalog_name, flow_name) = match &obj_name.0[..] {
-        [flow_name] => (
-            query_ctx.current_catalog().to_string(),
-            flow_name.value.clone(),
-        ),
-        [catalog, flow_name] => (catalog.value.clone(), flow_name.value.clone()),
-        _ => {
-            return InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "expect flow name to be <catalog>.<flow-name> or <flow-name>, actual: {}",
-                    obj_name
-                ),
-            }
-            .fail()
-        }
-    };
-    Ok((catalog_name, flow_name))
+    let res = flow_service_handler
+        .flush(&catalog_name, &flow_name, query_ctx.clone())
+        .await?;
+    let affected_rows = res.affected_rows;
+
+    Ok(Value::from(affected_rows))
 }
 
 #[cfg(test)]
@@ -154,10 +124,7 @@ mod test {
             ("catalog.flow_name", ("catalog", "flow_name")),
         ];
         for (input, expected) in testcases.iter() {
-            let args = vec![*input];
-            let args = args.into_iter().map(ValueRef::String).collect::<Vec<_>>();
-
-            let result = parse_flush_flow(&args, &QueryContext::arc()).unwrap();
+            let result = parse_catalog_flow(input, &QueryContext::arc()).unwrap();
             assert_eq!(*expected, (result.0.as_str(), result.1.as_str()));
         }
     }

src/common/function/src/function_factory.rs

@@ -0,0 +1,63 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::sync::Arc;
+
+use datafusion_expr::ScalarUDF;
+
+use crate::function::{FunctionContext, FunctionRef};
+use crate::scalars::udf::create_udf;
+
+/// A factory for creating `ScalarUDF` that require a function context.
+#[derive(Clone)]
+pub struct ScalarFunctionFactory {
+    name: String,
+    factory: Arc<dyn Fn(FunctionContext) -> ScalarUDF + Send + Sync>,
+}
+
+impl ScalarFunctionFactory {
+    /// Returns the name of the function.
+    pub fn name(&self) -> &str {
+        &self.name
+    }
+
+    /// Returns a `ScalarUDF` when given a function context.
+    pub fn provide(&self, ctx: FunctionContext) -> ScalarUDF {
+        (self.factory)(ctx)
+    }
+}
+
+impl From<ScalarUDF> for ScalarFunctionFactory {
+    fn from(df_udf: ScalarUDF) -> Self {
+        let name = df_udf.name().to_string();
+        let func = Arc::new(move |_ctx| df_udf.clone());
+        Self {
+            name,
+            factory: func,
+        }
+    }
+}
+
+impl From<FunctionRef> for ScalarFunctionFactory {
+    fn from(func: FunctionRef) -> Self {
+        let name = func.name().to_string();
+        let func = Arc::new(move |ctx: FunctionContext| {
+            create_udf(func.clone(), ctx.query_ctx, ctx.state)
+        });
+        Self {
+            name,
+            factory: func,
+        }
+    }
+}

src/common/function/src/function_registry.rs

@@ -16,11 +16,14 @@
 use std::collections::HashMap;
 use std::sync::{Arc, RwLock};
 
+use datafusion_expr::AggregateUDF;
 use once_cell::sync::Lazy;
 
 use crate::admin::AdminFunction;
-use crate::function::{AsyncFunctionRef, FunctionRef};
-use crate::scalars::aggregate::{AggregateFunctionMetaRef, AggregateFunctions};
+use crate::aggrs::approximate::ApproximateFunction;
+use crate::aggrs::vector::VectorFunction as VectorAggrFunction;
+use crate::function::{AsyncFunctionRef, Function, FunctionRef};
+use crate::function_factory::ScalarFunctionFactory;
 use crate::scalars::date::DateFunction;
 use crate::scalars::expression::ExpressionFunction;
 use crate::scalars::hll_count::HllCalcFunction;
@@ -31,18 +34,19 @@ use crate::scalars::matches_term::MatchesTermFunction;
 use crate::scalars::math::MathFunction;
 use crate::scalars::timestamp::TimestampFunction;
 use crate::scalars::uddsketch_calc::UddSketchCalcFunction;
-use crate::scalars::vector::VectorFunction;
+use crate::scalars::vector::VectorFunction as VectorScalarFunction;
 use crate::system::SystemFunction;
 
 #[derive(Default)]
 pub struct FunctionRegistry {
-    functions: RwLock<HashMap<String, FunctionRef>>,
+    functions: RwLock<HashMap<String, ScalarFunctionFactory>>,
     async_functions: RwLock<HashMap<String, AsyncFunctionRef>>,
-    aggregate_functions: RwLock<HashMap<String, AggregateFunctionMetaRef>>,
+    aggregate_functions: RwLock<HashMap<String, AggregateUDF>>,
 }
 
 impl FunctionRegistry {
-    pub fn register(&self, func: FunctionRef) {
+    pub fn register(&self, func: impl Into<ScalarFunctionFactory>) {
+        let func = func.into();
         let _ = self
             .functions
             .write()
@@ -50,6 +54,10 @@ impl FunctionRegistry {
             .insert(func.name().to_string(), func);
     }
 
+    pub fn register_scalar(&self, func: impl Function + 'static) {
+        self.register(Arc::new(func) as FunctionRef);
+    }
+
     pub fn register_async(&self, func: AsyncFunctionRef) {
         let _ = self
             .async_functions
@@ -58,6 +66,14 @@ impl FunctionRegistry {
             .insert(func.name().to_string(), func);
     }
 
+    pub fn register_aggr(&self, func: AggregateUDF) {
+        let _ = self
+            .aggregate_functions
+            .write()
+            .unwrap()
+            .insert(func.name().to_string(), func);
+    }
+
     pub fn get_async_function(&self, name: &str) -> Option<AsyncFunctionRef> {
         self.async_functions.read().unwrap().get(name).cloned()
     }
@@ -71,27 +87,16 @@ impl FunctionRegistry {
             .collect()
     }
 
-    pub fn register_aggregate_function(&self, func: AggregateFunctionMetaRef) {
-        let _ = self
-            .aggregate_functions
-            .write()
-            .unwrap()
-            .insert(func.name(), func);
-    }
-
-    pub fn get_aggr_function(&self, name: &str) -> Option<AggregateFunctionMetaRef> {
-        self.aggregate_functions.read().unwrap().get(name).cloned()
-    }
-
-    pub fn get_function(&self, name: &str) -> Option<FunctionRef> {
+    #[cfg(test)]
+    pub fn get_function(&self, name: &str) -> Option<ScalarFunctionFactory> {
         self.functions.read().unwrap().get(name).cloned()
     }
 
-    pub fn functions(&self) -> Vec<FunctionRef> {
+    pub fn scalar_functions(&self) -> Vec<ScalarFunctionFactory> {
         self.functions.read().unwrap().values().cloned().collect()
     }
 
-    pub fn aggregate_functions(&self) -> Vec<AggregateFunctionMetaRef> {
+    pub fn aggregate_functions(&self) -> Vec<AggregateUDF> {
         self.aggregate_functions
             .read()
             .unwrap()
@@ -112,9 +117,6 @@ pub static FUNCTION_REGISTRY: Lazy<Arc<FunctionRegistry>> = Lazy::new(|| {
     UddSketchCalcFunction::register(&function_registry);
     HllCalcFunction::register(&function_registry);
 
-    // Aggregate functions
-    AggregateFunctions::register(&function_registry);
-
     // Full text search function
     MatchesFunction::register(&function_registry);
     MatchesTermFunction::register(&function_registry);
@@ -127,15 +129,21 @@ pub static FUNCTION_REGISTRY: Lazy<Arc<FunctionRegistry>> = Lazy::new(|| {
     JsonFunction::register(&function_registry);
 
     // Vector related functions
-    VectorFunction::register(&function_registry);
+    VectorScalarFunction::register(&function_registry);
+    VectorAggrFunction::register(&function_registry);
 
     // Geo functions
     #[cfg(feature = "geo")]
     crate::scalars::geo::GeoFunctions::register(&function_registry);
+    #[cfg(feature = "geo")]
+    crate::aggrs::geo::GeoFunction::register(&function_registry);
 
     // Ip functions
     IpFunctions::register(&function_registry);
 
+    // Approximate functions
+    ApproximateFunction::register(&function_registry);
+
     Arc::new(function_registry)
 });
 
@@ -147,12 +155,11 @@ mod tests {
     #[test]
     fn test_function_registry() {
         let registry = FunctionRegistry::default();
-        let func = Arc::new(TestAndFunction);
 
         assert!(registry.get_function("test_and").is_none());
-        assert!(registry.functions().is_empty());
-        registry.register(func);
+        assert!(registry.scalar_functions().is_empty());
+        registry.register_scalar(TestAndFunction);
         let _ = registry.get_function("test_and").unwrap();
-        assert_eq!(1, registry.functions().len());
+        assert_eq!(1, registry.scalar_functions().len());
     }
 }

src/common/function/src/handlers.rs

@@ -87,6 +87,15 @@ pub trait FlowServiceHandler: Send + Sync {
         flow: &str,
         ctx: QueryContextRef,
     ) -> Result<api::v1::flow::FlowResponse>;
+
+    async fn adjust(
+        &self,
+        catalog: &str,
+        flow: &str,
+        min_run_interval_secs: u64,
+        max_filter_num_per_query: usize,
+        ctx: QueryContextRef,
+    ) -> Result<api::v1::flow::FlowResponse>;
 }
 
 pub type TableMutationHandlerRef = Arc<dyn TableMutationHandler>;

src/common/function/src/helper.rs

@@ -12,12 +12,15 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use common_query::error::{InvalidInputTypeSnafu, Result};
+use common_error::ext::BoxedError;
+use common_query::error::{ExecuteSnafu, InvalidFuncArgsSnafu, InvalidInputTypeSnafu, Result};
 use common_query::prelude::{Signature, TypeSignature, Volatility};
 use datatypes::prelude::ConcreteDataType;
 use datatypes::types::cast::cast;
 use datatypes::value::ValueRef;
+use session::context::QueryContextRef;
 use snafu::ResultExt;
+use sql::parser::ParserContext;
 
 /// Create a function signature with oneof signatures of interleaving two arguments.
 pub fn one_of_sigs2(args1: Vec<ConcreteDataType>, args2: Vec<ConcreteDataType>) -> Signature {
@@ -43,3 +46,30 @@ pub fn cast_u64(value: &ValueRef) -> Result<Option<u64>> {
         })
         .map(|v| v.as_u64())
 }
+
+pub fn parse_catalog_flow(
+    flow_name: &str,
+    query_ctx: &QueryContextRef,
+) -> Result<(String, String)> {
+    let obj_name = ParserContext::parse_table_name(flow_name, query_ctx.sql_dialect())
+        .map_err(BoxedError::new)
+        .context(ExecuteSnafu)?;
+
+    let (catalog_name, flow_name) = match &obj_name.0[..] {
+        [flow_name] => (
+            query_ctx.current_catalog().to_string(),
+            flow_name.value.clone(),
+        ),
+        [catalog, flow_name] => (catalog.value.clone(), flow_name.value.clone()),
+        _ => {
+            return InvalidFuncArgsSnafu {
+                err_msg: format!(
+                    "expect flow name to be <catalog>.<flow-name> or <flow-name>, actual: {}",
+                    obj_name
+                ),
+            }
+            .fail()
+        }
+    };
+    Ok((catalog_name, flow_name))
+}

src/common/function/src/lib.rs

@@ -15,16 +15,18 @@
 #![feature(let_chains)]
 #![feature(try_blocks)]
 
+mod adjust_flow;
 mod admin;
 mod flush_flow;
 mod macros;
-pub mod scalars;
 mod system;
 
-pub mod aggr;
+pub mod aggrs;
 pub mod function;
+pub mod function_factory;
 pub mod function_registry;
 pub mod handlers;
 pub mod helper;
+pub mod scalars;
 pub mod state;
 pub mod utils;

src/common/function/src/scalars.rs

@@ -12,7 +12,6 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-pub mod aggregate;
 pub(crate) mod date;
 pub mod expression;
 #[cfg(feature = "geo")]

src/common/function/src/scalars/aggregate.rs

@@ -1,89 +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.
-
-//! # Deprecate Warning:
-//!
-//! This module is deprecated and will be removed in the future.
-//! All UDAF implementation here are not maintained and should
-//! not be used before they are refactored into the `src/aggr`
-//! version.
-
-use std::sync::Arc;
-
-use common_query::logical_plan::AggregateFunctionCreatorRef;
-
-use crate::function_registry::FunctionRegistry;
-use crate::scalars::vector::product::VectorProductCreator;
-use crate::scalars::vector::sum::VectorSumCreator;
-
-/// A function creates `AggregateFunctionCreator`.
-/// "Aggregator" *is* AggregatorFunction. Since the later one is long, we named an short alias for it.
-/// The two names might be used interchangeably.
-type AggregatorCreatorFunction = Arc<dyn Fn() -> AggregateFunctionCreatorRef + Send + Sync>;
-
-/// `AggregateFunctionMeta` dynamically creates AggregateFunctionCreator.
-#[derive(Clone)]
-pub struct AggregateFunctionMeta {
-    name: String,
-    args_count: u8,
-    creator: AggregatorCreatorFunction,
-}
-
-pub type AggregateFunctionMetaRef = Arc<AggregateFunctionMeta>;
-
-impl AggregateFunctionMeta {
-    pub fn new(name: &str, args_count: u8, creator: AggregatorCreatorFunction) -> Self {
-        Self {
-            name: name.to_string(),
-            args_count,
-            creator,
-        }
-    }
-
-    pub fn name(&self) -> String {
-        self.name.to_string()
-    }
-
-    pub fn args_count(&self) -> u8 {
-        self.args_count
-    }
-
-    pub fn create(&self) -> AggregateFunctionCreatorRef {
-        (self.creator)()
-    }
-}
-
-pub(crate) struct AggregateFunctions;
-
-impl AggregateFunctions {
-    pub fn register(registry: &FunctionRegistry) {
-        registry.register_aggregate_function(Arc::new(AggregateFunctionMeta::new(
-            "vec_sum",
-            1,
-            Arc::new(|| Arc::new(VectorSumCreator::default())),
-        )));
-        registry.register_aggregate_function(Arc::new(AggregateFunctionMeta::new(
-            "vec_product",
-            1,
-            Arc::new(|| Arc::new(VectorProductCreator::default())),
-        )));
-
-        #[cfg(feature = "geo")]
-        registry.register_aggregate_function(Arc::new(AggregateFunctionMeta::new(
-            "json_encode_path",
-            3,
-            Arc::new(|| Arc::new(super::geo::encoding::JsonPathEncodeFunctionCreator::default())),
-        )));
-    }
-}

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

@@ -12,7 +12,6 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::sync::Arc;
 mod date_add;
 mod date_format;
 mod date_sub;
@@ -27,8 +26,8 @@ pub(crate) struct DateFunction;
 
 impl DateFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(DateAddFunction));
-        registry.register(Arc::new(DateSubFunction));
-        registry.register(Arc::new(DateFormatFunction));
+        registry.register_scalar(DateAddFunction);
+        registry.register_scalar(DateSubFunction);
+        registry.register_scalar(DateFormatFunction);
     }
 }

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

@@ -17,8 +17,6 @@ mod ctx;
 mod is_null;
 mod unary;
 
-use std::sync::Arc;
-
 pub use binary::scalar_binary_op;
 pub use ctx::EvalContext;
 pub use unary::scalar_unary_op;
@@ -30,6 +28,6 @@ pub(crate) struct ExpressionFunction;
 
 impl ExpressionFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(IsNullFunction));
+        registry.register_scalar(IsNullFunction);
     }
 }

src/common/function/src/scalars/geo.rs

@@ -12,11 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::sync::Arc;
-pub(crate) mod encoding;
 mod geohash;
 mod h3;
-mod helpers;
+pub(crate) mod helpers;
 mod measure;
 mod relation;
 mod s2;
@@ -29,57 +27,57 @@ pub(crate) struct GeoFunctions;
 impl GeoFunctions {
     pub fn register(registry: &FunctionRegistry) {
         // geohash
-        registry.register(Arc::new(geohash::GeohashFunction));
-        registry.register(Arc::new(geohash::GeohashNeighboursFunction));
+        registry.register_scalar(geohash::GeohashFunction);
+        registry.register_scalar(geohash::GeohashNeighboursFunction);
 
         // h3 index
-        registry.register(Arc::new(h3::H3LatLngToCell));
-        registry.register(Arc::new(h3::H3LatLngToCellString));
+        registry.register_scalar(h3::H3LatLngToCell);
+        registry.register_scalar(h3::H3LatLngToCellString);
 
         // h3 index inspection
-        registry.register(Arc::new(h3::H3CellBase));
-        registry.register(Arc::new(h3::H3CellIsPentagon));
-        registry.register(Arc::new(h3::H3StringToCell));
-        registry.register(Arc::new(h3::H3CellToString));
-        registry.register(Arc::new(h3::H3CellCenterLatLng));
-        registry.register(Arc::new(h3::H3CellResolution));
+        registry.register_scalar(h3::H3CellBase);
+        registry.register_scalar(h3::H3CellIsPentagon);
+        registry.register_scalar(h3::H3StringToCell);
+        registry.register_scalar(h3::H3CellToString);
+        registry.register_scalar(h3::H3CellCenterLatLng);
+        registry.register_scalar(h3::H3CellResolution);
 
         // h3 hierarchical grid
-        registry.register(Arc::new(h3::H3CellCenterChild));
-        registry.register(Arc::new(h3::H3CellParent));
-        registry.register(Arc::new(h3::H3CellToChildren));
-        registry.register(Arc::new(h3::H3CellToChildrenSize));
-        registry.register(Arc::new(h3::H3CellToChildPos));
-        registry.register(Arc::new(h3::H3ChildPosToCell));
-        registry.register(Arc::new(h3::H3CellContains));
+        registry.register_scalar(h3::H3CellCenterChild);
+        registry.register_scalar(h3::H3CellParent);
+        registry.register_scalar(h3::H3CellToChildren);
+        registry.register_scalar(h3::H3CellToChildrenSize);
+        registry.register_scalar(h3::H3CellToChildPos);
+        registry.register_scalar(h3::H3ChildPosToCell);
+        registry.register_scalar(h3::H3CellContains);
 
         // h3 grid traversal
-        registry.register(Arc::new(h3::H3GridDisk));
-        registry.register(Arc::new(h3::H3GridDiskDistances));
-        registry.register(Arc::new(h3::H3GridDistance));
-        registry.register(Arc::new(h3::H3GridPathCells));
+        registry.register_scalar(h3::H3GridDisk);
+        registry.register_scalar(h3::H3GridDiskDistances);
+        registry.register_scalar(h3::H3GridDistance);
+        registry.register_scalar(h3::H3GridPathCells);
 
         // h3 measurement
-        registry.register(Arc::new(h3::H3CellDistanceSphereKm));
-        registry.register(Arc::new(h3::H3CellDistanceEuclideanDegree));
+        registry.register_scalar(h3::H3CellDistanceSphereKm);
+        registry.register_scalar(h3::H3CellDistanceEuclideanDegree);
 
         // s2
-        registry.register(Arc::new(s2::S2LatLngToCell));
-        registry.register(Arc::new(s2::S2CellLevel));
-        registry.register(Arc::new(s2::S2CellToToken));
-        registry.register(Arc::new(s2::S2CellParent));
+        registry.register_scalar(s2::S2LatLngToCell);
+        registry.register_scalar(s2::S2CellLevel);
+        registry.register_scalar(s2::S2CellToToken);
+        registry.register_scalar(s2::S2CellParent);
 
         // spatial data type
-        registry.register(Arc::new(wkt::LatLngToPointWkt));
+        registry.register_scalar(wkt::LatLngToPointWkt);
 
         // spatial relation
-        registry.register(Arc::new(relation::STContains));
-        registry.register(Arc::new(relation::STWithin));
-        registry.register(Arc::new(relation::STIntersects));
+        registry.register_scalar(relation::STContains);
+        registry.register_scalar(relation::STWithin);
+        registry.register_scalar(relation::STIntersects);
 
         // spatial measure
-        registry.register(Arc::new(measure::STDistance));
-        registry.register(Arc::new(measure::STDistanceSphere));
-        registry.register(Arc::new(measure::STArea));
+        registry.register_scalar(measure::STDistance);
+        registry.register_scalar(measure::STDistanceSphere);
+        registry.register_scalar(measure::STArea);
     }
 }

src/common/function/src/scalars/geo/helpers.rs

@@ -37,7 +37,7 @@ macro_rules! ensure_columns_len {
     };
 }
 
-pub(super) use ensure_columns_len;
+pub(crate) use ensure_columns_len;
 
 macro_rules! ensure_columns_n {
     ($columns:ident, $n:literal) => {
@@ -58,7 +58,7 @@ macro_rules! ensure_columns_n {
     };
 }
 
-pub(super) use ensure_columns_n;
+pub(crate) use ensure_columns_n;
 
 macro_rules! ensure_and_coerce {
     ($compare:expr, $coerce:expr) => {{
@@ -72,4 +72,4 @@ macro_rules! ensure_and_coerce {
     }};
 }
 
-pub(super) use ensure_and_coerce;
+pub(crate) use ensure_and_coerce;

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

@@ -16,7 +16,6 @@
 
 use std::fmt;
 use std::fmt::Display;
-use std::sync::Arc;
 
 use common_query::error::{DowncastVectorSnafu, InvalidFuncArgsSnafu, Result};
 use common_query::prelude::{Signature, Volatility};
@@ -27,7 +26,7 @@ use datatypes::vectors::{BinaryVector, MutableVector, UInt64VectorBuilder, Vecto
 use hyperloglogplus::HyperLogLog;
 use snafu::OptionExt;
 
-use crate::aggr::HllStateType;
+use crate::aggrs::approximate::hll::HllStateType;
 use crate::function::{Function, FunctionContext};
 use crate::function_registry::FunctionRegistry;
 
@@ -44,7 +43,7 @@ pub struct HllCalcFunction;
 
 impl HllCalcFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(HllCalcFunction));
+        registry.register_scalar(HllCalcFunction);
     }
 }
 
@@ -117,6 +116,8 @@ impl Function for HllCalcFunction {
 
 #[cfg(test)]
 mod tests {
+    use std::sync::Arc;
+
     use datatypes::vectors::BinaryVector;
 
     use super::*;

src/common/function/src/scalars/ip.rs

@@ -17,8 +17,6 @@ mod ipv4;
 mod ipv6;
 mod range;
 
-use std::sync::Arc;
-
 use cidr::{Ipv4ToCidr, Ipv6ToCidr};
 use ipv4::{Ipv4NumToString, Ipv4StringToNum};
 use ipv6::{Ipv6NumToString, Ipv6StringToNum};
@@ -31,15 +29,15 @@ pub(crate) struct IpFunctions;
 impl IpFunctions {
     pub fn register(registry: &FunctionRegistry) {
         // Register IPv4 functions
-        registry.register(Arc::new(Ipv4NumToString));
-        registry.register(Arc::new(Ipv4StringToNum));
-        registry.register(Arc::new(Ipv4ToCidr));
-        registry.register(Arc::new(Ipv4InRange));
+        registry.register_scalar(Ipv4NumToString);
+        registry.register_scalar(Ipv4StringToNum);
+        registry.register_scalar(Ipv4ToCidr);
+        registry.register_scalar(Ipv4InRange);
 
         // Register IPv6 functions
-        registry.register(Arc::new(Ipv6NumToString));
-        registry.register(Arc::new(Ipv6StringToNum));
-        registry.register(Arc::new(Ipv6ToCidr));
-        registry.register(Arc::new(Ipv6InRange));
+        registry.register_scalar(Ipv6NumToString);
+        registry.register_scalar(Ipv6StringToNum);
+        registry.register_scalar(Ipv6ToCidr);
+        registry.register_scalar(Ipv6InRange);
     }
 }

src/common/function/src/scalars/json.rs

@@ -12,7 +12,6 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::sync::Arc;
 pub mod json_get;
 mod json_is;
 mod json_path_exists;
@@ -33,23 +32,23 @@ pub(crate) struct JsonFunction;
 
 impl JsonFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(JsonToStringFunction));
-        registry.register(Arc::new(ParseJsonFunction));
+        registry.register_scalar(JsonToStringFunction);
+        registry.register_scalar(ParseJsonFunction);
 
-        registry.register(Arc::new(JsonGetInt));
-        registry.register(Arc::new(JsonGetFloat));
-        registry.register(Arc::new(JsonGetString));
-        registry.register(Arc::new(JsonGetBool));
+        registry.register_scalar(JsonGetInt);
+        registry.register_scalar(JsonGetFloat);
+        registry.register_scalar(JsonGetString);
+        registry.register_scalar(JsonGetBool);
 
-        registry.register(Arc::new(JsonIsNull));
-        registry.register(Arc::new(JsonIsInt));
-        registry.register(Arc::new(JsonIsFloat));
-        registry.register(Arc::new(JsonIsString));
-        registry.register(Arc::new(JsonIsBool));
-        registry.register(Arc::new(JsonIsArray));
-        registry.register(Arc::new(JsonIsObject));
+        registry.register_scalar(JsonIsNull);
+        registry.register_scalar(JsonIsInt);
+        registry.register_scalar(JsonIsFloat);
+        registry.register_scalar(JsonIsString);
+        registry.register_scalar(JsonIsBool);
+        registry.register_scalar(JsonIsArray);
+        registry.register_scalar(JsonIsObject);
 
-        registry.register(Arc::new(json_path_exists::JsonPathExistsFunction));
-        registry.register(Arc::new(json_path_match::JsonPathMatchFunction));
+        registry.register_scalar(json_path_exists::JsonPathExistsFunction);
+        registry.register_scalar(json_path_match::JsonPathMatchFunction);
     }
 }

src/common/function/src/scalars/matches.rs

@@ -38,11 +38,11 @@ use crate::function_registry::FunctionRegistry;
 ///
 /// Usage: matches(`<col>`, `<pattern>`) -> boolean
 #[derive(Clone, Debug, Default)]
-pub(crate) struct MatchesFunction;
+pub struct MatchesFunction;
 
 impl MatchesFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(MatchesFunction));
+        registry.register_scalar(MatchesFunction);
     }
 }
 

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

@@ -77,7 +77,7 @@ pub struct MatchesTermFunction;
 
 impl MatchesTermFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(MatchesTermFunction));
+        registry.register_scalar(MatchesTermFunction);
     }
 }
 

src/common/function/src/scalars/math.rs

@@ -18,7 +18,6 @@ mod pow;
 mod rate;
 
 use std::fmt;
-use std::sync::Arc;
 
 pub use clamp::{ClampFunction, ClampMaxFunction, ClampMinFunction};
 use common_query::error::{GeneralDataFusionSnafu, Result};
@@ -39,13 +38,13 @@ pub(crate) struct MathFunction;
 
 impl MathFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(ModuloFunction));
-        registry.register(Arc::new(PowFunction));
-        registry.register(Arc::new(RateFunction));
-        registry.register(Arc::new(RangeFunction));
-        registry.register(Arc::new(ClampFunction));
-        registry.register(Arc::new(ClampMinFunction));
-        registry.register(Arc::new(ClampMaxFunction));
+        registry.register_scalar(ModuloFunction);
+        registry.register_scalar(PowFunction);
+        registry.register_scalar(RateFunction);
+        registry.register_scalar(RangeFunction);
+        registry.register_scalar(ClampFunction);
+        registry.register_scalar(ClampMinFunction);
+        registry.register_scalar(ClampMaxFunction);
     }
 }
 

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

@@ -12,7 +12,6 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::sync::Arc;
 mod to_unixtime;
 
 use to_unixtime::ToUnixtimeFunction;
@@ -23,6 +22,6 @@ pub(crate) struct TimestampFunction;
 
 impl TimestampFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(ToUnixtimeFunction));
+        registry.register_scalar(ToUnixtimeFunction);
     }
 }

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

@@ -16,7 +16,6 @@
 
 use std::fmt;
 use std::fmt::Display;
-use std::sync::Arc;
 
 use common_query::error::{DowncastVectorSnafu, InvalidFuncArgsSnafu, Result};
 use common_query::prelude::{Signature, Volatility};
@@ -44,7 +43,7 @@ pub struct UddSketchCalcFunction;
 
 impl UddSketchCalcFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(UddSketchCalcFunction));
+        registry.register_scalar(UddSketchCalcFunction);
     }
 }
 

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

@@ -17,10 +17,8 @@ mod distance;
 mod elem_product;
 mod elem_sum;
 pub mod impl_conv;
-pub(crate) mod product;
 mod scalar_add;
 mod scalar_mul;
-pub(crate) mod sum;
 mod vector_add;
 mod vector_dim;
 mod vector_div;
@@ -30,37 +28,34 @@ mod vector_norm;
 mod vector_sub;
 mod vector_subvector;
 
-use std::sync::Arc;
-
 use crate::function_registry::FunctionRegistry;
-
 pub(crate) struct VectorFunction;
 
 impl VectorFunction {
     pub fn register(registry: &FunctionRegistry) {
         // conversion
-        registry.register(Arc::new(convert::ParseVectorFunction));
-        registry.register(Arc::new(convert::VectorToStringFunction));
+        registry.register_scalar(convert::ParseVectorFunction);
+        registry.register_scalar(convert::VectorToStringFunction);
 
         // distance
-        registry.register(Arc::new(distance::CosDistanceFunction));
-        registry.register(Arc::new(distance::DotProductFunction));
-        registry.register(Arc::new(distance::L2SqDistanceFunction));
+        registry.register_scalar(distance::CosDistanceFunction);
+        registry.register_scalar(distance::DotProductFunction);
+        registry.register_scalar(distance::L2SqDistanceFunction);
 
         // scalar calculation
-        registry.register(Arc::new(scalar_add::ScalarAddFunction));
-        registry.register(Arc::new(scalar_mul::ScalarMulFunction));
+        registry.register_scalar(scalar_add::ScalarAddFunction);
+        registry.register_scalar(scalar_mul::ScalarMulFunction);
 
         // vector calculation
-        registry.register(Arc::new(vector_add::VectorAddFunction));
-        registry.register(Arc::new(vector_sub::VectorSubFunction));
-        registry.register(Arc::new(vector_mul::VectorMulFunction));
-        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));
+        registry.register_scalar(vector_add::VectorAddFunction);
+        registry.register_scalar(vector_sub::VectorSubFunction);
+        registry.register_scalar(vector_mul::VectorMulFunction);
+        registry.register_scalar(vector_div::VectorDivFunction);
+        registry.register_scalar(vector_norm::VectorNormFunction);
+        registry.register_scalar(vector_dim::VectorDimFunction);
+        registry.register_scalar(vector_kth_elem::VectorKthElemFunction);
+        registry.register_scalar(vector_subvector::VectorSubvectorFunction);
+        registry.register_scalar(elem_sum::ElemSumFunction);
+        registry.register_scalar(elem_product::ElemProductFunction);
     }
 }

src/common/function/src/state.rs

@@ -148,6 +148,17 @@ impl FunctionState {
             ) -> Result<api::v1::flow::FlowResponse> {
                 todo!()
             }
+
+            async fn adjust(
+                &self,
+                _catalog: &str,
+                _flow: &str,
+                _min_run_interval_secs: u64,
+                _max_filter_num_per_query: usize,
+                _ctx: QueryContextRef,
+            ) -> Result<api::v1::flow::FlowResponse> {
+                todo!()
+            }
         }
 
         Self {

src/common/function/src/system.rs

@@ -36,13 +36,13 @@ pub(crate) struct SystemFunction;
 
 impl SystemFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(BuildFunction));
-        registry.register(Arc::new(VersionFunction));
-        registry.register(Arc::new(CurrentSchemaFunction));
-        registry.register(Arc::new(DatabaseFunction));
-        registry.register(Arc::new(SessionUserFunction));
-        registry.register(Arc::new(ReadPreferenceFunction));
-        registry.register(Arc::new(TimezoneFunction));
+        registry.register_scalar(BuildFunction);
+        registry.register_scalar(VersionFunction);
+        registry.register_scalar(CurrentSchemaFunction);
+        registry.register_scalar(DatabaseFunction);
+        registry.register_scalar(SessionUserFunction);
+        registry.register_scalar(ReadPreferenceFunction);
+        registry.register_scalar(TimezoneFunction);
         registry.register_async(Arc::new(ProcedureStateFunction));
         PGCatalogFunction::register(registry);
     }

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

@@ -16,8 +16,6 @@ mod pg_get_userbyid;
 mod table_is_visible;
 mod version;
 
-use std::sync::Arc;
-
 use pg_get_userbyid::PGGetUserByIdFunction;
 use table_is_visible::PGTableIsVisibleFunction;
 use version::PGVersionFunction;
@@ -35,8 +33,8 @@ pub(super) struct PGCatalogFunction;
 
 impl PGCatalogFunction {
     pub fn register(registry: &FunctionRegistry) {
-        registry.register(Arc::new(PGTableIsVisibleFunction));
-        registry.register(Arc::new(PGGetUserByIdFunction));
-        registry.register(Arc::new(PGVersionFunction));
+        registry.register_scalar(PGTableIsVisibleFunction);
+        registry.register_scalar(PGGetUserByIdFunction);
+        registry.register_scalar(PGVersionFunction);
     }
 }

src/common/grpc/src/flight.rs

@@ -64,6 +64,19 @@ impl Default for FlightEncoder {
 }
 
 impl FlightEncoder {
+    /// Creates new [FlightEncoder] with compression disabled.
+    pub fn with_compression_disabled() -> Self {
+        let write_options = writer::IpcWriteOptions::default()
+            .try_with_compression(None)
+            .unwrap();
+
+        Self {
+            write_options,
+            data_gen: writer::IpcDataGenerator::default(),
+            dictionary_tracker: writer::DictionaryTracker::new(false),
+        }
+    }
+
     pub fn encode(&mut self, flight_message: FlightMessage) -> FlightData {
         match flight_message {
             FlightMessage::Schema(schema) => SchemaAsIpc::new(&schema, &self.write_options).into(),

src/common/meta/src/distributed_time_constants.rs

@@ -35,6 +35,9 @@ pub const FLOWNODE_LEASE_SECS: u64 = DATANODE_LEASE_SECS;
 /// The lease seconds of metasrv leader.
 pub const META_LEASE_SECS: u64 = 5;
 
+/// The keep-alive interval of the Postgres connection.
+pub const POSTGRES_KEEP_ALIVE_SECS: u64 = 30;
+
 /// In a lease, there are two opportunities for renewal.
 pub const META_KEEP_ALIVE_INTERVAL_SECS: u64 = META_LEASE_SECS / 2;
 

src/common/meta/src/key.rs

@@ -188,7 +188,71 @@ pub const CACHE_KEY_PREFIXES: [&str; 5] = [
     NODE_ADDRESS_PREFIX,
 ];
 
-pub type RegionDistribution = BTreeMap<DatanodeId, Vec<RegionNumber>>;
+/// A set of regions with the same role.
+#[derive(Debug, Clone, PartialEq, Eq, Default, Serialize)]
+pub struct RegionRoleSet {
+    /// Leader regions.
+    pub leader_regions: Vec<RegionNumber>,
+    /// Follower regions.
+    pub follower_regions: Vec<RegionNumber>,
+}
+
+impl<'de> Deserialize<'de> for RegionRoleSet {
+    fn deserialize<D>(deserializer: D) -> std::result::Result<Self, D::Error>
+    where
+        D: serde::Deserializer<'de>,
+    {
+        #[derive(Deserialize)]
+        #[serde(untagged)]
+        enum RegionRoleSetOrLeaderOnly {
+            Full {
+                leader_regions: Vec<RegionNumber>,
+                follower_regions: Vec<RegionNumber>,
+            },
+            LeaderOnly(Vec<RegionNumber>),
+        }
+        match RegionRoleSetOrLeaderOnly::deserialize(deserializer)? {
+            RegionRoleSetOrLeaderOnly::Full {
+                leader_regions,
+                follower_regions,
+            } => Ok(RegionRoleSet::new(leader_regions, follower_regions)),
+            RegionRoleSetOrLeaderOnly::LeaderOnly(leader_regions) => {
+                Ok(RegionRoleSet::new(leader_regions, vec![]))
+            }
+        }
+    }
+}
+
+impl RegionRoleSet {
+    /// Create a new region role set.
+    pub fn new(leader_regions: Vec<RegionNumber>, follower_regions: Vec<RegionNumber>) -> Self {
+        Self {
+            leader_regions,
+            follower_regions,
+        }
+    }
+
+    /// Add a leader region to the set.
+    pub fn add_leader_region(&mut self, region_number: RegionNumber) {
+        self.leader_regions.push(region_number);
+    }
+
+    /// Add a follower region to the set.
+    pub fn add_follower_region(&mut self, region_number: RegionNumber) {
+        self.follower_regions.push(region_number);
+    }
+
+    /// Sort the regions.
+    pub fn sort(&mut self) {
+        self.follower_regions.sort();
+        self.leader_regions.sort();
+    }
+}
+
+/// The distribution of regions.
+///
+/// The key is the datanode id, the value is the region role set.
+pub type RegionDistribution = BTreeMap<DatanodeId, RegionRoleSet>;
 
 /// The id of flow.
 pub type FlowId = u32;
@@ -1368,7 +1432,8 @@ mod tests {
     use crate::key::table_name::TableNameKey;
     use crate::key::table_route::TableRouteValue;
     use crate::key::{
-        DeserializedValueWithBytes, TableMetadataManager, ViewInfoValue, TOPIC_REGION_PREFIX,
+        DeserializedValueWithBytes, RegionDistribution, RegionRoleSet, TableMetadataManager,
+        ViewInfoValue, TOPIC_REGION_PREFIX,
     };
     use crate::kv_backend::memory::MemoryKvBackend;
     use crate::kv_backend::KvBackend;
@@ -1995,7 +2060,8 @@ mod tests {
                 .unwrap()
                 .unwrap();
 
-            assert_eq!(got.regions, regions)
+            assert_eq!(got.regions, regions.leader_regions);
+            assert_eq!(got.follower_regions, regions.follower_regions);
         }
     }
 
@@ -2412,4 +2478,28 @@ mod tests {
         assert_eq!(current_view_info.columns, new_columns);
         assert_eq!(current_view_info.plan_columns, new_plan_columns);
     }
+
+    #[test]
+    fn test_region_role_set_deserialize() {
+        let s = r#"{"leader_regions": [1, 2, 3], "follower_regions": [4, 5, 6]}"#;
+        let region_role_set: RegionRoleSet = serde_json::from_str(s).unwrap();
+        assert_eq!(region_role_set.leader_regions, vec![1, 2, 3]);
+        assert_eq!(region_role_set.follower_regions, vec![4, 5, 6]);
+
+        let s = r#"[1, 2, 3]"#;
+        let region_role_set: RegionRoleSet = serde_json::from_str(s).unwrap();
+        assert_eq!(region_role_set.leader_regions, vec![1, 2, 3]);
+        assert!(region_role_set.follower_regions.is_empty());
+    }
+
+    #[test]
+    fn test_region_distribution_deserialize() {
+        let s = r#"{"1": [1,2,3], "2": {"leader_regions": [7, 8, 9], "follower_regions": [10, 11, 12]}}"#;
+        let region_distribution: RegionDistribution = serde_json::from_str(s).unwrap();
+        assert_eq!(region_distribution.len(), 2);
+        assert_eq!(region_distribution[&1].leader_regions, vec![1, 2, 3]);
+        assert!(region_distribution[&1].follower_regions.is_empty());
+        assert_eq!(region_distribution[&2].leader_regions, vec![7, 8, 9]);
+        assert_eq!(region_distribution[&2].follower_regions, vec![10, 11, 12]);
+    }
 }

src/common/meta/src/key/datanode_table.rs

@@ -24,7 +24,7 @@ use table::metadata::TableId;
 use crate::error::{DatanodeTableInfoNotFoundSnafu, InvalidMetadataSnafu, Result};
 use crate::key::table_route::PhysicalTableRouteValue;
 use crate::key::{
-    MetadataKey, MetadataValue, RegionDistribution, DATANODE_TABLE_KEY_PATTERN,
+    MetadataKey, MetadataValue, RegionDistribution, RegionRoleSet, DATANODE_TABLE_KEY_PATTERN,
     DATANODE_TABLE_KEY_PREFIX,
 };
 use crate::kv_backend::txn::{Txn, TxnOp};
@@ -118,23 +118,31 @@ impl Display for DatanodeTableKey {
 pub struct DatanodeTableValue {
     pub table_id: TableId,
     pub regions: Vec<RegionNumber>,
+    #[serde(default)]
+    pub follower_regions: Vec<RegionNumber>,
     #[serde(flatten)]
     pub region_info: RegionInfo,
     version: u64,
 }
 
 impl DatanodeTableValue {
-    pub fn new(table_id: TableId, regions: Vec<RegionNumber>, region_info: RegionInfo) -> Self {
+    pub fn new(table_id: TableId, region_role_set: RegionRoleSet, region_info: RegionInfo) -> Self {
+        let RegionRoleSet {
+            leader_regions,
+            follower_regions,
+        } = region_role_set;
+
         Self {
             table_id,
-            regions,
+            regions: leader_regions,
+            follower_regions,
             region_info,
             version: 0,
         }
     }
 }
 
-/// Decodes `KeyValue` to ((),`DatanodeTableValue`)
+/// Decodes [`KeyValue`] to [`DatanodeTableValue`].
 pub fn datanode_table_value_decoder(kv: KeyValue) -> Result<DatanodeTableValue> {
     DatanodeTableValue::try_from_raw_value(&kv.value)
 }
@@ -373,10 +381,11 @@ mod tests {
         let value = DatanodeTableValue {
             table_id: 42,
             regions: vec![1, 2, 3],
+            follower_regions: vec![],
             region_info: RegionInfo::default(),
             version: 1,
         };
-        let literal = br#"{"table_id":42,"regions":[1,2,3],"engine":"","region_storage_path":"","region_options":{},"region_wal_options":{},"version":1}"#;
+        let literal = br#"{"table_id":42,"regions":[1,2,3],"follower_regions":[],"engine":"","region_storage_path":"","region_options":{},"region_wal_options":{},"version":1}"#;
 
         let raw_value = value.try_as_raw_value().unwrap();
         assert_eq!(raw_value, literal);
@@ -467,6 +476,7 @@ mod tests {
         let table_value = DatanodeTableValue {
             table_id: 1,
             regions: vec![],
+            follower_regions: vec![],
             region_info,
             version: 1,
         };

src/common/meta/src/rpc/router.rs

@@ -40,17 +40,23 @@ pub fn region_distribution(region_routes: &[RegionRoute]) -> RegionDistribution
     let mut regions_id_map = RegionDistribution::new();
     for route in region_routes.iter() {
         if let Some(peer) = route.leader_peer.as_ref() {
-            let region_id = route.region.id.region_number();
-            regions_id_map.entry(peer.id).or_default().push(region_id);
+            let region_number = route.region.id.region_number();
+            regions_id_map
+                .entry(peer.id)
+                .or_default()
+                .add_leader_region(region_number);
         }
         for peer in route.follower_peers.iter() {
-            let region_id = route.region.id.region_number();
-            regions_id_map.entry(peer.id).or_default().push(region_id);
+            let region_number = route.region.id.region_number();
+            regions_id_map
+                .entry(peer.id)
+                .or_default()
+                .add_follower_region(region_number);
         }
     }
-    for (_, regions) in regions_id_map.iter_mut() {
-        // id asc
-        regions.sort()
+    for (_, region_role_set) in regions_id_map.iter_mut() {
+        // Sort the regions in ascending order.
+        region_role_set.sort()
     }
     regions_id_map
 }
@@ -455,6 +461,7 @@ impl From<PbPartition> for Partition {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use crate::key::RegionRoleSet;
 
     #[test]
     fn test_leader_is_downgraded() {
@@ -611,8 +618,8 @@ mod tests {
 
         let distribution = region_distribution(&region_routes);
         assert_eq!(distribution.len(), 3);
-        assert_eq!(distribution[&1], vec![1, 2]);
-        assert_eq!(distribution[&2], vec![1, 2]);
-        assert_eq!(distribution[&3], vec![1, 2]);
+        assert_eq!(distribution[&1], RegionRoleSet::new(vec![1], vec![2]));
+        assert_eq!(distribution[&2], RegionRoleSet::new(vec![2], vec![1]));
+        assert_eq!(distribution[&3], RegionRoleSet::new(vec![], vec![1, 2]));
     }
 }

src/common/stat/Cargo.toml

@@ -0,0 +1,11 @@
+[package]
+name = "stat"
+version.workspace = true
+edition.workspace = true
+license.workspace = true
+
+[dependencies]
+nix.workspace = true
+
+[lints]
+workspace = true

src/common/stat/src/cgroups.rs

@@ -0,0 +1,183 @@
+// 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.
+
+#![allow(dead_code)]
+
+use std::fs::read_to_string;
+use std::path::Path;
+
+#[cfg(target_os = "linux")]
+use nix::sys::{statfs, statfs::statfs};
+
+/// `MAX_VALUE` is used to indicate that the resource is unlimited.
+pub const MAX_VALUE: i64 = -1;
+
+const CGROUP_UNIFIED_MOUNTPOINT: &str = "/sys/fs/cgroup";
+
+const MEMORY_MAX_FILE_CGROUP_V2: &str = "memory.max";
+const MEMORY_MAX_FILE_CGROUP_V1: &str = "memory.limit_in_bytes";
+const CPU_MAX_FILE_CGROUP_V2: &str = "cpu.max";
+const CPU_QUOTA_FILE_CGROUP_V1: &str = "cpu.cfs_quota_us";
+const CPU_PERIOD_FILE_CGROUP_V1: &str = "cpu.cfs_period_us";
+
+// `MAX_VALUE_CGROUP_V2` string in `/sys/fs/cgroup/cpu.max` and `/sys/fs/cgroup/memory.max` to indicate that the resource is unlimited.
+const MAX_VALUE_CGROUP_V2: &str = "max";
+
+// For cgroup v1, if the memory is unlimited, it will return a very large value(different from platform) that close to 2^63.
+// For easier comparison, if the memory limit is larger than 1PB we consider it as unlimited.
+const MAX_MEMORY_IN_BYTES: i64 = 1125899906842624; // 1PB
+
+/// Get the limit of memory in bytes.
+///
+/// - If the memory is unlimited, return `-1`.
+/// - Return `None` if it fails to read the memory limit or not on linux.
+pub fn get_memory_limit() -> Option<i64> {
+    #[cfg(target_os = "linux")]
+    {
+        let memory_max_file = if is_cgroup_v2()? {
+            // Read `/sys/fs/cgroup/memory.max` to get the memory limit.
+            MEMORY_MAX_FILE_CGROUP_V2
+        } else {
+            // Read `/sys/fs/cgroup/memory.limit_in_bytes` to get the memory limit.
+            MEMORY_MAX_FILE_CGROUP_V1
+        };
+
+        // For cgroup v1, it will return a very large value(different from platform) if the memory is unlimited.
+        let memory_limit =
+            read_value_from_file(Path::new(CGROUP_UNIFIED_MOUNTPOINT).join(memory_max_file))?;
+
+        // If memory limit exceeds 1PB(cgroup v1), consider it as unlimited.
+        if memory_limit > MAX_MEMORY_IN_BYTES {
+            return Some(MAX_VALUE);
+        }
+        Some(memory_limit)
+    }
+
+    #[cfg(not(target_os = "linux"))]
+    None
+}
+
+/// Get the limit of cpu in millicores.
+///
+/// - If the cpu is unlimited, return `-1`.
+/// - Return `None` if it fails to read the cpu limit or not on linux.
+pub fn get_cpu_limit() -> Option<i64> {
+    #[cfg(target_os = "linux")]
+    if is_cgroup_v2()? {
+        // Read `/sys/fs/cgroup/cpu.max` to get the cpu limit.
+        get_cgroup_v2_cpu_limit(Path::new(CGROUP_UNIFIED_MOUNTPOINT).join(CPU_MAX_FILE_CGROUP_V2))
+    } else {
+        // Read `/sys/fs/cgroup/cpu.cfs_quota_us` and `/sys/fs/cgroup/cpu.cfs_period_us` to get the cpu limit.
+        let quota = read_value_from_file(
+            Path::new(CGROUP_UNIFIED_MOUNTPOINT).join(CPU_QUOTA_FILE_CGROUP_V1),
+        )?;
+
+        if quota == MAX_VALUE {
+            return Some(MAX_VALUE);
+        }
+
+        let period = read_value_from_file(
+            Path::new(CGROUP_UNIFIED_MOUNTPOINT).join(CPU_PERIOD_FILE_CGROUP_V1),
+        )?;
+
+        // Return the cpu limit in millicores.
+        Some(quota * 1000 / period)
+    }
+
+    #[cfg(not(target_os = "linux"))]
+    None
+}
+
+// Check whether the cgroup is v2.
+// - Return `true` if the cgroup is v2, otherwise return `false`.
+// - Return `None` if the detection fails or not on linux.
+fn is_cgroup_v2() -> Option<bool> {
+    #[cfg(target_os = "linux")]
+    {
+        let path = Path::new(CGROUP_UNIFIED_MOUNTPOINT);
+        let fs_stat = statfs(path).ok()?;
+        Some(fs_stat.filesystem_type() == statfs::CGROUP2_SUPER_MAGIC)
+    }
+
+    #[cfg(not(target_os = "linux"))]
+    None
+}
+
+fn read_value_from_file<P: AsRef<Path>>(path: P) -> Option<i64> {
+    let content = read_to_string(&path).ok()?;
+
+    // If the content starts with "max", return `MAX_VALUE`.
+    if content.starts_with(MAX_VALUE_CGROUP_V2) {
+        return Some(MAX_VALUE);
+    }
+
+    content.trim().parse::<i64>().ok()
+}
+
+fn get_cgroup_v2_cpu_limit<P: AsRef<Path>>(path: P) -> Option<i64> {
+    let content = read_to_string(&path).ok()?;
+
+    let fields = content.trim().split(' ').collect::<Vec<&str>>();
+    if fields.len() != 2 {
+        return None;
+    }
+
+    // If the cpu is unlimited, it will be `-1`.
+    let quota = fields[0].trim();
+    if quota == MAX_VALUE_CGROUP_V2 {
+        return Some(MAX_VALUE);
+    }
+
+    let quota = quota.parse::<i64>().ok()?;
+
+    let period = fields[1].trim();
+    let period = period.parse::<i64>().ok()?;
+
+    // Return the cpu limit in millicores.
+    Some(quota * 1000 / period)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_read_value_from_file() {
+        assert_eq!(
+            read_value_from_file(Path::new("testdata").join("memory.max")).unwrap(),
+            100000
+        );
+        assert_eq!(
+            read_value_from_file(Path::new("testdata").join("memory.max.unlimited")).unwrap(),
+            MAX_VALUE
+        );
+        assert_eq!(read_value_from_file(Path::new("non_existent_file")), None);
+    }
+
+    #[test]
+    fn test_get_cgroup_v2_cpu_limit() {
+        assert_eq!(
+            get_cgroup_v2_cpu_limit(Path::new("testdata").join("cpu.max")).unwrap(),
+            1500
+        );
+        assert_eq!(
+            get_cgroup_v2_cpu_limit(Path::new("testdata").join("cpu.max.unlimited")).unwrap(),
+            MAX_VALUE
+        );
+        assert_eq!(
+            get_cgroup_v2_cpu_limit(Path::new("non_existent_file")),
+            None
+        );
+    }
+}

src/common/stat/src/lib.rs

@@ -12,11 +12,6 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-mod geo_path;
-mod hll;
-mod uddsketch_state;
+mod cgroups;
 
-pub use geo_path::{GeoPathAccumulator, GEO_PATH_NAME};
-pub(crate) use hll::HllStateType;
-pub use hll::{HllState, HLL_MERGE_NAME, HLL_NAME};
-pub use uddsketch_state::{UddSketchState, UDDSKETCH_MERGE_NAME, UDDSKETCH_STATE_NAME};
+pub use cgroups::*;

src/common/stat/testdata/cpu.max

@@ -0,0 +1 @@
+150000 100000

src/common/stat/testdata/cpu.max.unlimited

@@ -0,0 +1 @@
+max 100000

src/common/stat/testdata/memory.max

@@ -0,0 +1 @@
+100000

src/common/stat/testdata/memory.max.unlimited

@@ -0,0 +1 @@
+max

src/common/telemetry/src/logging.rs

@@ -37,6 +37,9 @@ use crate::tracing_sampler::{create_sampler, TracingSampleOptions};
 
 pub const DEFAULT_OTLP_ENDPOINT: &str = "http://localhost:4317";
 
+/// The default logs directory.
+pub const DEFAULT_LOGGING_DIR: &str = "logs";
+
 // Handle for reloading log level
 pub static RELOAD_HANDLE: OnceCell<tracing_subscriber::reload::Handle<Targets, Registry>> =
     OnceCell::new();
@@ -133,7 +136,8 @@ impl Eq for LoggingOptions {}
 impl Default for LoggingOptions {
     fn default() -> Self {
         Self {
-            dir: "./greptimedb_data/logs".to_string(),
+            // The directory path will be configured at application startup, typically using the data home directory as a base.
+            dir: "".to_string(),
             level: None,
             log_format: LogFormat::Text,
             enable_otlp_tracing: false,

src/datanode/Cargo.toml

@@ -6,6 +6,7 @@ license.workspace = true
 
 [features]
 testing = []
+enterprise = []
 
 [lints]
 workspace = true

src/datanode/src/config.rs

@@ -18,7 +18,7 @@ use core::time::Duration;
 
 use common_base::readable_size::ReadableSize;
 use common_base::secrets::{ExposeSecret, SecretString};
-use common_config::Configurable;
+use common_config::{Configurable, DEFAULT_DATA_HOME};
 pub use common_procedure::options::ProcedureConfig;
 use common_telemetry::logging::{LoggingOptions, TracingOptions};
 use common_wal::config::DatanodeWalConfig;
@@ -36,9 +36,6 @@ use servers::http::HttpOptions;
 
 pub const DEFAULT_OBJECT_STORE_CACHE_SIZE: ReadableSize = ReadableSize::gb(5);
 
-/// Default data home in file storage
-const DEFAULT_DATA_HOME: &str = "./greptimedb_data";
-
 /// Object storage config
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 #[serde(tag = "type")]

src/datanode/src/datanode.rs

@@ -372,6 +372,7 @@ impl DatanodeBuilder {
             opts.max_concurrent_queries,
             //TODO: revaluate the hardcoded timeout on the next version of datanode concurrency limiter.
             Duration::from_millis(100),
+            opts.grpc.flight_compression,
         );
 
         let object_store_manager = Self::build_object_store_manager(&opts.storage).await?;
@@ -559,6 +560,8 @@ async fn open_all_regions(
     init_regions_parallelism: usize,
 ) -> Result<()> {
     let mut regions = vec![];
+    #[cfg(feature = "enterprise")]
+    let mut follower_regions = vec![];
     for table_value in table_values {
         for region_number in table_value.regions {
             // Augments region options with wal options if a wal options is provided.
@@ -576,6 +579,24 @@ async fn open_all_regions(
                 region_options,
             ));
         }
+
+        #[cfg(feature = "enterprise")]
+        for region_number in table_value.follower_regions {
+            // Augments region options with wal options if a wal options is provided.
+            let mut region_options = table_value.region_info.region_options.clone();
+            prepare_wal_options(
+                &mut region_options,
+                RegionId::new(table_value.table_id, region_number),
+                &table_value.region_info.region_wal_options,
+            );
+
+            follower_regions.push((
+                RegionId::new(table_value.table_id, region_number),
+                table_value.region_info.engine.clone(),
+                table_value.region_info.region_storage_path.clone(),
+                region_options,
+            ));
+        }
     }
     let num_regions = regions.len();
     info!("going to open {} region(s)", num_regions);
@@ -617,6 +638,43 @@ async fn open_all_regions(
             }
         }
     }
+
+    #[cfg(feature = "enterprise")]
+    if !follower_regions.is_empty() {
+        info!(
+            "going to open {} follower region(s)",
+            follower_regions.len()
+        );
+        let mut region_requests = Vec::with_capacity(follower_regions.len());
+        for (region_id, engine, store_path, options) in follower_regions {
+            let region_dir = region_dir(&store_path, region_id);
+            region_requests.push((
+                region_id,
+                RegionOpenRequest {
+                    engine,
+                    region_dir,
+                    options,
+                    skip_wal_replay: true,
+                },
+            ));
+        }
+
+        let open_regions = region_server
+            .handle_batch_open_requests(init_regions_parallelism, region_requests)
+            .await?;
+
+        ensure!(
+            open_regions.len() == num_regions,
+            error::UnexpectedSnafu {
+                violated: format!(
+                    "Expected to open {} of follower regions, only {} of regions has opened",
+                    num_regions,
+                    open_regions.len()
+                )
+            }
+        );
+    }
+
     info!("all regions are opened");
 
     Ok(())
@@ -632,6 +690,7 @@ mod tests {
     use common_base::Plugins;
     use common_meta::cache::LayeredCacheRegistryBuilder;
     use common_meta::key::datanode_table::DatanodeTableManager;
+    use common_meta::key::RegionRoleSet;
     use common_meta::kv_backend::memory::MemoryKvBackend;
     use common_meta::kv_backend::KvBackendRef;
     use mito2::engine::MITO_ENGINE_NAME;
@@ -651,7 +710,7 @@ mod tests {
                 "foo/bar/weny",
                 HashMap::from([("foo".to_string(), "bar".to_string())]),
                 HashMap::default(),
-                BTreeMap::from([(0, vec![0, 1, 2])]),
+                BTreeMap::from([(0, RegionRoleSet::new(vec![0, 1, 2], vec![]))]),
             )
             .unwrap();
 

src/datanode/src/region_server.rs

@@ -50,6 +50,7 @@ use query::QueryEngineRef;
 use servers::error::{self as servers_error, ExecuteGrpcRequestSnafu, Result as ServerResult};
 use servers::grpc::flight::{FlightCraft, FlightRecordBatchStream, TonicStream};
 use servers::grpc::region_server::RegionServerHandler;
+use servers::grpc::FlightCompression;
 use session::context::{QueryContextBuilder, QueryContextRef};
 use snafu::{ensure, OptionExt, ResultExt};
 use store_api::metric_engine_consts::{
@@ -80,6 +81,7 @@ use crate::event_listener::RegionServerEventListenerRef;
 #[derive(Clone)]
 pub struct RegionServer {
     inner: Arc<RegionServerInner>,
+    flight_compression: FlightCompression,
 }
 
 pub struct RegionStat {
@@ -93,6 +95,7 @@ impl RegionServer {
         query_engine: QueryEngineRef,
         runtime: Runtime,
         event_listener: RegionServerEventListenerRef,
+        flight_compression: FlightCompression,
     ) -> Self {
         Self::with_table_provider(
             query_engine,
@@ -101,6 +104,7 @@ impl RegionServer {
             Arc::new(DummyTableProviderFactory),
             0,
             Duration::from_millis(0),
+            flight_compression,
         )
     }
 
@@ -111,6 +115,7 @@ impl RegionServer {
         table_provider_factory: TableProviderFactoryRef,
         max_concurrent_queries: usize,
         concurrent_query_limiter_timeout: Duration,
+        flight_compression: FlightCompression,
     ) -> Self {
         Self {
             inner: Arc::new(RegionServerInner::new(
@@ -123,6 +128,7 @@ impl RegionServer {
                     concurrent_query_limiter_timeout,
                 ),
             )),
+            flight_compression,
         }
     }
 
@@ -536,7 +542,11 @@ impl FlightCraft for RegionServer {
             .trace(tracing_context.attach(info_span!("RegionServer::handle_read")))
             .await?;
 
-        let stream = Box::pin(FlightRecordBatchStream::new(result, tracing_context));
+        let stream = Box::pin(FlightRecordBatchStream::new(
+            result,
+            tracing_context,
+            self.flight_compression,
+        ));
         Ok(Response::new(stream))
     }
 }

src/datanode/src/tests.rs

@@ -19,16 +19,16 @@ use std::time::Duration;
 use api::region::RegionResponse;
 use async_trait::async_trait;
 use common_error::ext::BoxedError;
-use common_function::function::FunctionRef;
-use common_function::scalars::aggregate::AggregateFunctionMetaRef;
+use common_function::function_factory::ScalarFunctionFactory;
 use common_query::Output;
 use common_runtime::runtime::{BuilderBuild, RuntimeTrait};
 use common_runtime::Runtime;
-use datafusion_expr::LogicalPlan;
+use datafusion_expr::{AggregateUDF, LogicalPlan};
 use query::dataframe::DataFrame;
 use query::planner::LogicalPlanner;
 use query::query_engine::{DescribeResult, QueryEngineState};
 use query::{QueryEngine, QueryEngineContext};
+use servers::grpc::FlightCompression;
 use session::context::QueryContextRef;
 use store_api::metadata::RegionMetadataRef;
 use store_api::region_engine::{
@@ -76,9 +76,9 @@ impl QueryEngine for MockQueryEngine {
         unimplemented!()
     }
 
-    fn register_aggregate_function(&self, _func: AggregateFunctionMetaRef) {}
+    fn register_aggregate_function(&self, _func: AggregateUDF) {}
 
-    fn register_function(&self, _func: FunctionRef) {}
+    fn register_scalar_function(&self, _func: ScalarFunctionFactory) {}
 
     fn read_table(&self, _table: TableRef) -> query::error::Result<DataFrame> {
         unimplemented!()
@@ -98,6 +98,7 @@ pub fn mock_region_server() -> RegionServer {
         Arc::new(MockQueryEngine),
         Runtime::builder().build().unwrap(),
         Box::new(NoopRegionServerEventListener),
+        FlightCompression::default(),
     )
 }
 

src/flow/Cargo.toml

@@ -61,6 +61,7 @@ prost.workspace = true
 query.workspace = true
 rand.workspace = true
 serde.workspace = true
+serde_json.workspace = true
 servers.workspace = true
 session.workspace = true
 smallvec.workspace = true

src/flow/src/adapter/flownode_impl.rs

@@ -18,7 +18,7 @@ use std::sync::atomic::AtomicBool;
 use std::sync::Arc;
 
 use api::v1::flow::{
-    flow_request, CreateRequest, DropRequest, FlowRequest, FlowResponse, FlushFlow,
+    flow_request, AdjustFlow, CreateRequest, DropRequest, FlowRequest, FlowResponse, FlushFlow,
 };
 use api::v1::region::InsertRequests;
 use catalog::CatalogManager;
@@ -32,6 +32,7 @@ use common_telemetry::{error, info, trace, warn};
 use datatypes::value::Value;
 use futures::TryStreamExt;
 use itertools::Itertools;
+use serde::{Deserialize, Serialize};
 use session::context::QueryContextBuilder;
 use snafu::{ensure, IntoError, OptionExt, ResultExt};
 use store_api::storage::{RegionId, TableId};
@@ -809,6 +810,25 @@ impl common_meta::node_manager::Flownode for FlowDualEngine {
                     ..Default::default()
                 })
             }
+            Some(flow_request::Body::Adjust(AdjustFlow { flow_id, options })) => {
+                #[derive(Debug, Serialize, Deserialize)]
+                struct Options {
+                    min_run_interval_secs: u64,
+                    max_filter_num_per_query: usize,
+                }
+                let options: Options = serde_json::from_str(&options).with_context(|_| {
+                    common_meta::error::DeserializeFromJsonSnafu { input: options }
+                })?;
+                self.batching_engine
+                    .adjust_flow(
+                        flow_id.unwrap().id as u64,
+                        options.min_run_interval_secs,
+                        options.max_filter_num_per_query,
+                    )
+                    .await
+                    .map_err(to_meta_err(snafu::location!()))?;
+                Ok(Default::default())
+            }
             other => common_meta::error::InvalidFlowRequestBodySnafu { body: other }.fail(),
         }
     }
@@ -841,93 +861,6 @@ fn to_meta_err(
     }
 }
 
-#[async_trait::async_trait]
-impl common_meta::node_manager::Flownode for StreamingEngine {
-    async fn handle(&self, request: FlowRequest) -> MetaResult<FlowResponse> {
-        let query_ctx = request
-            .header
-            .and_then(|h| h.query_context)
-            .map(|ctx| ctx.into());
-        match request.body {
-            Some(flow_request::Body::Create(CreateRequest {
-                flow_id: Some(task_id),
-                source_table_ids,
-                sink_table_name: Some(sink_table_name),
-                create_if_not_exists,
-                expire_after,
-                comment,
-                sql,
-                flow_options,
-                or_replace,
-            })) => {
-                let source_table_ids = source_table_ids.into_iter().map(|id| id.id).collect_vec();
-                let sink_table_name = [
-                    sink_table_name.catalog_name,
-                    sink_table_name.schema_name,
-                    sink_table_name.table_name,
-                ];
-                let expire_after = expire_after.map(|e| e.value);
-                let args = CreateFlowArgs {
-                    flow_id: task_id.id as u64,
-                    sink_table_name,
-                    source_table_ids,
-                    create_if_not_exists,
-                    or_replace,
-                    expire_after,
-                    comment: Some(comment),
-                    sql: sql.clone(),
-                    flow_options,
-                    query_ctx,
-                };
-                let ret = self
-                    .create_flow(args)
-                    .await
-                    .map_err(BoxedError::new)
-                    .with_context(|_| CreateFlowSnafu { sql: sql.clone() })
-                    .map_err(to_meta_err(snafu::location!()))?;
-                METRIC_FLOW_TASK_COUNT.inc();
-                Ok(FlowResponse {
-                    affected_flows: ret
-                        .map(|id| greptime_proto::v1::FlowId { id: id as u32 })
-                        .into_iter()
-                        .collect_vec(),
-                    ..Default::default()
-                })
-            }
-            Some(flow_request::Body::Drop(DropRequest {
-                flow_id: Some(flow_id),
-            })) => {
-                self.remove_flow(flow_id.id as u64)
-                    .await
-                    .map_err(to_meta_err(snafu::location!()))?;
-                METRIC_FLOW_TASK_COUNT.dec();
-                Ok(Default::default())
-            }
-            Some(flow_request::Body::Flush(FlushFlow {
-                flow_id: Some(flow_id),
-            })) => {
-                let row = self
-                    .flush_flow_inner(flow_id.id as u64)
-                    .await
-                    .map_err(to_meta_err(snafu::location!()))?;
-                Ok(FlowResponse {
-                    affected_flows: vec![flow_id],
-                    affected_rows: row as u64,
-                    ..Default::default()
-                })
-            }
-            other => common_meta::error::InvalidFlowRequestBodySnafu { body: other }.fail(),
-        }
-    }
-
-    async fn handle_inserts(&self, request: InsertRequests) -> MetaResult<FlowResponse> {
-        self.handle_inserts_inner(request)
-            .await
-            .map(|_| Default::default())
-            .map_err(to_meta_err(snafu::location!()))
-    }
-}
-
 impl FlowEngine for StreamingEngine {
     async fn create_flow(&self, args: CreateFlowArgs) -> Result<Option<FlowId>, Error> {
         self.create_flow_inner(args).await

src/flow/src/batching_mode/engine.rs

@@ -388,6 +388,20 @@ impl BatchingEngine {
     pub async fn flow_exist_inner(&self, flow_id: FlowId) -> bool {
         self.tasks.read().await.contains_key(&flow_id)
     }
+
+    pub async fn adjust_flow(
+        &self,
+        flow_id: FlowId,
+        min_run_interval_secs: u64,
+        max_filter_num_per_query: usize,
+    ) -> Result<(), Error> {
+        let task = self.tasks.read().await.get(&flow_id).cloned();
+        let task = task.with_context(|| FlowNotFoundSnafu { id: flow_id })?;
+        debug!("Adjusting flow {flow_id} with min_run_interval_secs={} and max_filter_num_per_query={}", min_run_interval_secs, max_filter_num_per_query);
+        task.adjust(min_run_interval_secs, max_filter_num_per_query);
+
+        Ok(())
+    }
 }
 
 impl FlowEngine for BatchingEngine {

src/flow/src/batching_mode/frontend_client.rs

@@ -286,7 +286,7 @@ impl FrontendClient {
 
     /// Get the frontend with recent enough(less than 1 minute from now) `last_activity_ts`
     /// and is able to process query
-    async fn get_random_active_frontend(
+    pub(crate) async fn get_random_active_frontend(
         &self,
         catalog: &str,
         schema: &str,
@@ -382,7 +382,7 @@ impl FrontendClient {
             }),
             catalog,
             schema,
-            &mut None,
+            None,
             task,
         )
         .await
@@ -394,16 +394,28 @@ impl FrontendClient {
         req: api::v1::greptime_request::Request,
         catalog: &str,
         schema: &str,
-        peer_desc: &mut Option<PeerDesc>,
+        use_peer: Option<Peer>,
         task: Option<&BatchingTask>,
     ) -> Result<u32, Error> {
         match self {
-            FrontendClient::Distributed { fe_stats, .. } => {
-                let db = self.get_random_active_frontend(catalog, schema).await?;
-
-                *peer_desc = Some(PeerDesc::Dist {
-                    peer: db.peer.clone(),
-                });
+            FrontendClient::Distributed {
+                fe_stats, chnl_mgr, ..
+            } => {
+                let db = if let Some(peer) = use_peer {
+                    DatabaseWithPeer::new(
+                        Database::new(
+                            catalog,
+                            schema,
+                            Client::with_manager_and_urls(
+                                chnl_mgr.clone(),
+                                vec![peer.addr.clone()],
+                            ),
+                        ),
+                        peer,
+                    )
+                } else {
+                    self.get_random_active_frontend(catalog, schema).await?
+                };
 
                 let flow_id = task.map(|t| t.config.flow_id).unwrap_or_default();
                 let _guard = fe_stats.observe(&db.peer.addr, flow_id);

src/flow/src/batching_mode/state.rs

@@ -32,6 +32,7 @@ use crate::batching_mode::MIN_REFRESH_DURATION;
 use crate::error::{DatatypesSnafu, InternalSnafu, TimeSnafu, UnexpectedSnafu};
 use crate::metrics::{
     METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME_RANGE, METRIC_FLOW_BATCHING_ENGINE_QUERY_WINDOW_CNT,
+    METRIC_FLOW_BATCHING_ENGINE_STALLED_QUERY_WINDOW_CNT,
 };
 use crate::{Error, FlowId};
 
@@ -52,6 +53,13 @@ pub struct TaskState {
     pub(crate) shutdown_rx: oneshot::Receiver<()>,
     /// Task handle
     pub(crate) task_handle: Option<tokio::task::JoinHandle<()>>,
+    /// Slow Query metrics update task handle
+    pub(crate) slow_query_metric_task: Option<tokio::task::JoinHandle<()>>,
+
+    /// min run interval in seconds
+    pub(crate) min_run_interval: Option<u64>,
+    /// max filter number per query
+    pub(crate) max_filter_num: Option<usize>,
 }
 impl TaskState {
     pub fn new(query_ctx: QueryContextRef, shutdown_rx: oneshot::Receiver<()>) -> Self {
@@ -63,6 +71,9 @@ impl TaskState {
             exec_state: ExecState::Idle,
             shutdown_rx,
             task_handle: None,
+            slow_query_metric_task: None,
+            min_run_interval: None,
+            max_filter_num: None,
         }
     }
 
@@ -87,20 +98,17 @@ impl TaskState {
     pub fn get_next_start_query_time(
         &self,
         flow_id: FlowId,
-        time_window_size: &Option<Duration>,
+        _time_window_size: &Option<Duration>,
         max_timeout: Option<Duration>,
     ) -> Instant {
-        let last_duration = max_timeout
+        let next_duration = max_timeout
             .unwrap_or(self.last_query_duration)
             .min(self.last_query_duration)
-            .max(MIN_REFRESH_DURATION);
-
-        let next_duration = time_window_size
-            .map(|t| {
-                let half = t / 2;
-                half.max(last_duration)
-            })
-            .unwrap_or(last_duration);
+            .max(
+                self.min_run_interval
+                    .map(Duration::from_secs)
+                    .unwrap_or(MIN_REFRESH_DURATION),
+            );
 
         // if have dirty time window, execute immediately to clean dirty time window
         if self.dirty_time_windows.windows.is_empty() {
@@ -206,47 +214,69 @@ impl DirtyTimeWindows {
 
         // get the first `window_cnt` time windows
         let max_time_range = window_size * window_cnt as i32;
-        let nth = {
-            let mut cur_time_range = chrono::Duration::zero();
-            let mut nth_key = None;
-            for (idx, (start, end)) in self.windows.iter().enumerate() {
-                // if time range is too long, stop
-                if cur_time_range > max_time_range {
-                    nth_key = Some(*start);
-                    break;
-                }
 
-                // if we have enough time windows, stop
-                if idx >= window_cnt {
-                    nth_key = Some(*start);
-                    break;
-                }
+        let mut to_be_query = BTreeMap::new();
+        let mut new_windows = self.windows.clone();
+        let mut cur_time_range = chrono::Duration::zero();
+        for (idx, (start, end)) in self.windows.iter().enumerate() {
+            let first_end = start
+                .add_duration(window_size.to_std().unwrap())
+                .context(TimeSnafu)?;
+            let end = end.unwrap_or(first_end);
 
-                if let Some(end) = end {
-                    if let Some(x) = end.sub(start) {
-                        cur_time_range += x;
-                    }
-                }
+            // if time range is too long, stop
+            if cur_time_range >= max_time_range {
+                break;
             }
 
-            nth_key
-        };
-        let first_nth = {
-            if let Some(nth) = nth {
-                let mut after = self.windows.split_off(&nth);
-                std::mem::swap(&mut self.windows, &mut after);
-
-                after
-            } else {
-                std::mem::take(&mut self.windows)
+            // if we have enough time windows, stop
+            if idx >= window_cnt {
+                break;
             }
-        };
+
+            if let Some(x) = end.sub(start) {
+                if cur_time_range + x <= max_time_range {
+                    to_be_query.insert(*start, Some(end));
+                    new_windows.remove(start);
+                    cur_time_range += x;
+                } else {
+                    // too large a window, split it
+                    // split at window_size * times
+                    let surplus = max_time_range - cur_time_range;
+                    let times = surplus.num_seconds() / window_size.num_seconds();
+
+                    let split_offset = window_size * times as i32;
+                    let split_at = start
+                        .add_duration(split_offset.to_std().unwrap())
+                        .context(TimeSnafu)?;
+                    to_be_query.insert(*start, Some(split_at));
+
+                    // remove the original window
+                    new_windows.remove(start);
+                    new_windows.insert(split_at, Some(end));
+                    cur_time_range += split_offset;
+                    break;
+                }
+            }
+        }
+
+        self.windows = new_windows;
 
         METRIC_FLOW_BATCHING_ENGINE_QUERY_WINDOW_CNT
-            .with_label_values(&[flow_id.to_string().as_str()])
-            .observe(first_nth.len() as f64);
+            .with_label_values(&[
+                flow_id.to_string().as_str(),
+                format!("{}", window_size).as_str(),
+            ])
+            .observe(to_be_query.len() as f64);
 
-        let full_time_range = first_nth
+        METRIC_FLOW_BATCHING_ENGINE_STALLED_QUERY_WINDOW_CNT
+            .with_label_values(&[
+                flow_id.to_string().as_str(),
+                format!("{}", window_size).as_str(),
+            ])
+            .observe(self.windows.len() as f64);
+
+        let full_time_range = to_be_query
             .iter()
             .fold(chrono::Duration::zero(), |acc, (start, end)| {
                 if let Some(end) = end {
@@ -257,11 +287,14 @@ impl DirtyTimeWindows {
             })
             .num_seconds() as f64;
         METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME_RANGE
-            .with_label_values(&[flow_id.to_string().as_str()])
+            .with_label_values(&[
+                flow_id.to_string().as_str(),
+                format!("{}", window_size).as_str(),
+            ])
             .observe(full_time_range);
 
         let mut expr_lst = vec![];
-        for (start, end) in first_nth.into_iter() {
+        for (start, end) in to_be_query.into_iter() {
             // align using time window exprs
             let (start, end) = if let Some(ctx) = task_ctx {
                 let Some(time_window_expr) = &ctx.config.time_window_expr else {
@@ -495,6 +528,64 @@ mod test {
                     "((ts >= CAST('1970-01-01 00:00:00' AS TIMESTAMP)) AND (ts < CAST('1970-01-01 00:00:21' AS TIMESTAMP)))",
                 )
             ),
+            // split range
+            (
+                Vec::from_iter((0..20).map(|i|Timestamp::new_second(i*3)).chain(std::iter::once(
+                    Timestamp::new_second(60 + 3 * (DirtyTimeWindows::MERGE_DIST as i64 + 1)),
+                ))),
+                (chrono::Duration::seconds(3), None),
+                BTreeMap::from([
+                (
+                    Timestamp::new_second(0),
+                    Some(Timestamp::new_second(
+                        60
+                    )),
+                ),
+                (
+                    Timestamp::new_second(60 + 3 * (DirtyTimeWindows::MERGE_DIST as i64 + 1)),
+                    Some(Timestamp::new_second(
+                        60 + 3 * (DirtyTimeWindows::MERGE_DIST as i64 + 1) + 3
+                    )),
+                )]),
+                Some(
+                    "((ts >= CAST('1970-01-01 00:00:00' AS TIMESTAMP)) AND (ts < CAST('1970-01-01 00:01:00' AS TIMESTAMP)))",
+                )
+            ),
+            // split 2 min into 1 min
+            (
+                Vec::from_iter((0..40).map(|i|Timestamp::new_second(i*3))),
+                (chrono::Duration::seconds(3), None),
+                BTreeMap::from([
+                (
+                    Timestamp::new_second(0),
+                    Some(Timestamp::new_second(
+                        40 * 3
+                    )),
+                )]),
+                Some(
+                    "((ts >= CAST('1970-01-01 00:00:00' AS TIMESTAMP)) AND (ts < CAST('1970-01-01 00:01:00' AS TIMESTAMP)))",
+                )
+            ),
+            // split 3s + 1min into 3s + 57s
+            (
+                Vec::from_iter(std::iter::once(Timestamp::new_second(0)).chain((0..40).map(|i|Timestamp::new_second(20+i*3)))),
+                (chrono::Duration::seconds(3), None),
+                BTreeMap::from([
+                (
+                    Timestamp::new_second(0),
+                    Some(Timestamp::new_second(
+                        3
+                    )),
+                ),(
+                    Timestamp::new_second(20),
+                    Some(Timestamp::new_second(
+                        140
+                    )),
+                )]),
+                Some(
+                    "(((ts >= CAST('1970-01-01 00:00:00' AS TIMESTAMP)) AND (ts < CAST('1970-01-01 00:00:03' AS TIMESTAMP))) OR ((ts >= CAST('1970-01-01 00:00:20' AS TIMESTAMP)) AND (ts < CAST('1970-01-01 00:01:17' AS TIMESTAMP))))",
+                )
+            ),
             // expired
             (
                 vec![
@@ -511,6 +602,8 @@ mod test {
                 None
             ),
         ];
+        // let len = testcases.len();
+        // let testcases = testcases[(len - 2)..(len - 1)].to_vec();
         for (lower_bounds, (window_size, expire_lower_bound), expected, expected_filter_expr) in
             testcases
         {

src/flow/src/batching_mode/task.rs

@@ -61,7 +61,8 @@ use crate::error::{
     SubstraitEncodeLogicalPlanSnafu, UnexpectedSnafu,
 };
 use crate::metrics::{
-    METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME, METRIC_FLOW_BATCHING_ENGINE_SLOW_QUERY,
+    METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME, METRIC_FLOW_BATCHING_ENGINE_REAL_TIME_SLOW_QUERY_CNT,
+    METRIC_FLOW_BATCHING_ENGINE_SLOW_QUERY,
 };
 use crate::{Error, FlowId};
 
@@ -81,6 +82,14 @@ pub struct TaskConfig {
     query_type: QueryType,
 }
 
+impl TaskConfig {
+    pub fn time_window_size(&self) -> Option<Duration> {
+        self.time_window_expr
+            .as_ref()
+            .and_then(|expr| *expr.time_window_size())
+    }
+}
+
 fn determine_query_type(query: &str, query_ctx: &QueryContextRef) -> Result<QueryType, Error> {
     let stmts =
         ParserContext::create_with_dialect(query, query_ctx.sql_dialect(), ParseOptions::default())
@@ -144,6 +153,12 @@ impl BatchingTask {
         })
     }
 
+    pub fn adjust(&self, min_run_interval_secs: u64, max_filter_num_per_query: usize) {
+        let mut state = self.state.write().unwrap();
+        state.min_run_interval = Some(min_run_interval_secs);
+        state.max_filter_num = Some(max_filter_num_per_query);
+    }
+
     /// mark time window range (now - expire_after, now) as dirty (or (0, now) if expire_after not set)
     ///
     /// useful for flush_flow to flush dirty time windows range
@@ -328,11 +343,53 @@ impl BatchingTask {
             })?;
 
         let plan = expanded_plan;
-        let mut peer_desc = None;
+
+        let db = frontend_client
+            .get_random_active_frontend(catalog, schema)
+            .await?;
+        let peer_desc = db.peer.clone();
+
+        let (tx, mut rx) = oneshot::channel();
+        let peer_inner = peer_desc.clone();
+        let window_size_pretty = format!(
+            "{}s",
+            self.config.time_window_size().unwrap_or_default().as_secs()
+        );
+        let inner_window_size_pretty = window_size_pretty.clone();
+        let flow_id = self.config.flow_id;
+        let slow_query_metric_task = tokio::task::spawn(async move {
+            tokio::time::sleep(SLOW_QUERY_THRESHOLD).await;
+            METRIC_FLOW_BATCHING_ENGINE_REAL_TIME_SLOW_QUERY_CNT
+                .with_label_values(&[
+                    flow_id.to_string().as_str(),
+                    &peer_inner.to_string(),
+                    inner_window_size_pretty.as_str(),
+                ])
+                .add(1.0);
+            while rx.try_recv() == Err(TryRecvError::Empty) {
+                // sleep for a while before next update
+                tokio::time::sleep(MIN_REFRESH_DURATION).await;
+            }
+            METRIC_FLOW_BATCHING_ENGINE_REAL_TIME_SLOW_QUERY_CNT
+                .with_label_values(&[
+                    flow_id.to_string().as_str(),
+                    &peer_inner.to_string(),
+                    inner_window_size_pretty.as_str(),
+                ])
+                .sub(1.0);
+        });
+        self.state.write().unwrap().slow_query_metric_task = Some(slow_query_metric_task);
 
         let res = {
             let _timer = METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME
-                .with_label_values(&[flow_id.to_string().as_str()])
+                .with_label_values(&[
+                    flow_id.to_string().as_str(),
+                    format!(
+                        "{}s",
+                        self.config.time_window_size().unwrap_or_default().as_secs()
+                    )
+                    .as_str(),
+                ])
                 .start_timer();
 
             // hack and special handling the insert logical plan
@@ -361,10 +418,12 @@ impl BatchingTask {
             };
 
             frontend_client
-                .handle(req, catalog, schema, &mut peer_desc, Some(self))
+                .handle(req, catalog, schema, Some(db.peer), Some(self))
                 .await
         };
 
+        // signaling the slow query metric task to stop
+        let _ = tx.send(());
         let elapsed = instant.elapsed();
         if let Ok(affected_rows) = &res {
             debug!(
@@ -387,7 +446,12 @@ impl BatchingTask {
             METRIC_FLOW_BATCHING_ENGINE_SLOW_QUERY
                 .with_label_values(&[
                     flow_id.to_string().as_str(),
-                    &peer_desc.unwrap_or_default().to_string(),
+                    &peer_desc.to_string(),
+                    format!(
+                        "{}s",
+                        self.config.time_window_size().unwrap_or_default().as_secs()
+                    )
+                    .as_str(),
                 ])
                 .observe(elapsed.as_secs_f64());
         }
@@ -580,19 +644,20 @@ impl BatchingTask {
                 ),
             })?;
 
-        let expr = self
-            .state
-            .write()
-            .unwrap()
-            .dirty_time_windows
-            .gen_filter_exprs(
+        let expr = {
+            let mut state = self.state.write().unwrap();
+            let max_window_cnt = state
+                .max_filter_num
+                .unwrap_or(DirtyTimeWindows::MAX_FILTER_NUM);
+            state.dirty_time_windows.gen_filter_exprs(
                 &col_name,
                 Some(l),
                 window_size,
-                DirtyTimeWindows::MAX_FILTER_NUM,
+                max_window_cnt,
                 self.config.flow_id,
                 Some(self),
-            )?;
+            )?
+        };
 
         debug!(
             "Flow id={:?}, Generated filter expr: {:?}",

src/flow/src/metrics.rs

@@ -31,22 +31,37 @@ lazy_static! {
     pub static ref METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME: HistogramVec = register_histogram_vec!(
         "greptime_flow_batching_engine_query_time_secs",
         "flow batching engine query time(seconds)",
-        &["flow_id"],
+        &["flow_id", "time_window_granularity"],
         vec![0.0, 5., 10., 20., 40., 80., 160., 320., 640.,]
     )
     .unwrap();
     pub static ref METRIC_FLOW_BATCHING_ENGINE_SLOW_QUERY: HistogramVec = register_histogram_vec!(
         "greptime_flow_batching_engine_slow_query_secs",
-        "flow batching engine slow query(seconds)",
-        &["flow_id", "peer"],
+        "flow batching engine slow query(seconds), updated after query finished",
+        &["flow_id", "peer", "time_window_granularity"],
         vec![60., 2. * 60., 3. * 60., 5. * 60., 10. * 60.]
     )
     .unwrap();
+    pub static ref METRIC_FLOW_BATCHING_ENGINE_REAL_TIME_SLOW_QUERY_CNT: GaugeVec =
+        register_gauge_vec!(
+            "greptime_flow_batching_engine_real_time_slow_query_number",
+            "flow batching engine real time slow query number, updated in real time",
+            &["flow_id", "peer", "time_window_granularity"],
+        )
+        .unwrap();
+    pub static ref METRIC_FLOW_BATCHING_ENGINE_STALLED_QUERY_WINDOW_CNT: HistogramVec =
+        register_histogram_vec!(
+            "greptime_flow_batching_engine_stalled_query_window_cnt",
+            "flow batching engine stalled query time window count",
+            &["flow_id", "time_window_granularity"],
+            vec![0.0, 5., 10., 20., 40.]
+        )
+        .unwrap();
     pub static ref METRIC_FLOW_BATCHING_ENGINE_QUERY_WINDOW_CNT: HistogramVec =
         register_histogram_vec!(
             "greptime_flow_batching_engine_query_window_cnt",
             "flow batching engine query time window count",
-            &["flow_id"],
+            &["flow_id", "time_window_granularity"],
             vec![0.0, 5., 10., 20., 40.]
         )
         .unwrap();
@@ -54,7 +69,7 @@ lazy_static! {
         register_histogram_vec!(
             "greptime_flow_batching_engine_query_time_range_secs",
             "flow batching engine query time range(seconds)",
-            &["flow_id"],
+            &["flow_id", "time_window_granularity"],
             vec![60., 4. * 60., 16. * 60., 64. * 60., 256. * 60.]
         )
         .unwrap();

src/flow/src/transform.rs

@@ -17,7 +17,7 @@ use std::collections::BTreeMap;
 use std::sync::Arc;
 
 use common_error::ext::BoxedError;
-use common_function::function::FunctionContext;
+use common_function::function::{FunctionContext, FunctionRef};
 use datafusion_substrait::extensions::Extensions;
 use datatypes::data_type::ConcreteDataType as CDT;
 use query::QueryEngine;
@@ -108,9 +108,13 @@ impl FunctionExtensions {
 
 /// register flow-specific functions to the query engine
 pub fn register_function_to_query_engine(engine: &Arc<dyn QueryEngine>) {
-    engine.register_function(Arc::new(TumbleFunction::new("tumble")));
-    engine.register_function(Arc::new(TumbleFunction::new(TUMBLE_START)));
-    engine.register_function(Arc::new(TumbleFunction::new(TUMBLE_END)));
+    let tumble_fn = Arc::new(TumbleFunction::new("tumble")) as FunctionRef;
+    let tumble_start_fn = Arc::new(TumbleFunction::new(TUMBLE_START)) as FunctionRef;
+    let tumble_end_fn = Arc::new(TumbleFunction::new(TUMBLE_END)) as FunctionRef;
+
+    engine.register_scalar_function(tumble_fn.into());
+    engine.register_scalar_function(tumble_start_fn.into());
+    engine.register_scalar_function(tumble_end_fn.into());
 }
 
 #[derive(Debug)]

src/frontend/src/instance/jaeger.rs

@@ -130,7 +130,13 @@ impl JaegerQueryHandler for Instance {
         .await?)
     }
 
-    async fn get_trace(&self, ctx: QueryContextRef, trace_id: &str) -> ServerResult<Output> {
+    async fn get_trace(
+        &self,
+        ctx: QueryContextRef,
+        trace_id: &str,
+        start_time: Option<i64>,
+        end_time: Option<i64>,
+    ) -> ServerResult<Output> {
         // It's equivalent to
         //
         // ```
@@ -139,13 +145,25 @@ impl JaegerQueryHandler for Instance {
         // FROM
         //   {db}.{trace_table}
         // WHERE
-        //   trace_id = '{trace_id}'
+        //   trace_id = '{trace_id}' AND
+        //   timestamp >= {start_time} AND
+        //   timestamp <= {end_time}
         // ORDER BY
         //   timestamp DESC
         // ```.
         let selects = vec![wildcard()];
 
-        let filters = vec![col(TRACE_ID_COLUMN).eq(lit(trace_id))];
+        let mut filters = vec![col(TRACE_ID_COLUMN).eq(lit(trace_id))];
+
+        if let Some(start_time) = start_time {
+            // Microseconds to nanoseconds.
+            filters.push(col(TIMESTAMP_COLUMN).gt_eq(lit_timestamp_nano(start_time * 1_000)));
+        }
+
+        if let Some(end_time) = end_time {
+            // Microseconds to nanoseconds.
+            filters.push(col(TIMESTAMP_COLUMN).lt_eq(lit_timestamp_nano(end_time * 1_000)));
+        }
 
         Ok(query_trace_table(
             ctx,

src/frontend/src/server.rs

@@ -154,6 +154,7 @@ where
             ServerGrpcQueryHandlerAdapter::arc(self.instance.clone()),
             user_provider.clone(),
             runtime,
+            opts.grpc.flight_compression,
         );
 
         let grpc_server = builder

src/meta-srv/src/bootstrap.rs

@@ -31,8 +31,6 @@ use common_meta::kv_backend::rds::MySqlStore;
 #[cfg(feature = "pg_kvbackend")]
 use common_meta::kv_backend::rds::PgStore;
 use common_meta::kv_backend::{KvBackendRef, ResettableKvBackendRef};
-#[cfg(feature = "pg_kvbackend")]
-use common_telemetry::error;
 use common_telemetry::info;
 #[cfg(feature = "pg_kvbackend")]
 use deadpool_postgres::{Config, Runtime};
@@ -144,7 +142,8 @@ impl MetasrvInstance {
         let (serve_state_tx, serve_state_rx) = oneshot::channel();
 
         let socket_addr =
-            bootstrap_metasrv_with_router(&self.opts.bind_addr, router, serve_state_tx, rx).await?;
+            bootstrap_metasrv_with_router(&self.opts.grpc.bind_addr, router, serve_state_tx, rx)
+                .await?;
         self.bind_addr = Some(socket_addr);
 
         let addr = self.opts.http.addr.parse().context(error::ParseAddrSnafu {
@@ -260,7 +259,7 @@ pub async fn metasrv_builder(
             let etcd_client = create_etcd_client(&opts.store_addrs).await?;
             let kv_backend = EtcdStore::with_etcd_client(etcd_client.clone(), opts.max_txn_ops);
             let election = EtcdElection::with_etcd_client(
-                &opts.server_addr,
+                &opts.grpc.server_addr,
                 etcd_client,
                 opts.store_key_prefix.clone(),
             )
@@ -270,22 +269,41 @@ pub async fn metasrv_builder(
         }
         #[cfg(feature = "pg_kvbackend")]
         (None, BackendImpl::PostgresStore) => {
-            let pool = create_postgres_pool(&opts.store_addrs).await?;
-            let kv_backend = PgStore::with_pg_pool(pool, &opts.meta_table_name, opts.max_txn_ops)
-                .await
-                .context(error::KvBackendSnafu)?;
-            // Client for election should be created separately since we need a different session keep-alive idle time.
-            let election_client = create_postgres_client(opts).await?;
+            use std::time::Duration;
+
+            use common_meta::distributed_time_constants::POSTGRES_KEEP_ALIVE_SECS;
+
+            use crate::election::rds::postgres::ElectionPgClient;
+
+            let candidate_lease_ttl = Duration::from_secs(CANDIDATE_LEASE_SECS);
+            let execution_timeout = Duration::from_secs(META_LEASE_SECS);
+            let statement_timeout = Duration::from_secs(META_LEASE_SECS);
+            let meta_lease_ttl = Duration::from_secs(META_LEASE_SECS);
+
+            let mut cfg = Config::new();
+            cfg.keepalives = Some(true);
+            cfg.keepalives_idle = Some(Duration::from_secs(POSTGRES_KEEP_ALIVE_SECS));
+            // We use a separate pool for election since we need a different session keep-alive idle time.
+            let pool = create_postgres_pool_with(&opts.store_addrs, cfg).await?;
+
+            let election_client =
+                ElectionPgClient::new(pool, execution_timeout, meta_lease_ttl, statement_timeout)?;
             let election = PgElection::with_pg_client(
-                opts.server_addr.clone(),
+                opts.grpc.server_addr.clone(),
                 election_client,
                 opts.store_key_prefix.clone(),
-                CANDIDATE_LEASE_SECS,
-                META_LEASE_SECS,
+                candidate_lease_ttl,
+                meta_lease_ttl,
                 &opts.meta_table_name,
                 opts.meta_election_lock_id,
             )
             .await?;
+
+            let pool = create_postgres_pool(&opts.store_addrs).await?;
+            let kv_backend = PgStore::with_pg_pool(pool, &opts.meta_table_name, opts.max_txn_ops)
+                .await
+                .context(error::KvBackendSnafu)?;
+
             (kv_backend, Some(election))
         }
         #[cfg(feature = "mysql_kvbackend")]
@@ -299,7 +317,7 @@ pub async fn metasrv_builder(
             let election_table_name = opts.meta_table_name.clone() + "_election";
             let election_client = create_mysql_client(opts).await?;
             let election = MySqlElection::with_mysql_client(
-                opts.server_addr.clone(),
+                opts.grpc.server_addr.clone(),
                 election_client,
                 opts.store_key_prefix.clone(),
                 CANDIDATE_LEASE_SECS,
@@ -372,31 +390,24 @@ pub async fn create_etcd_client(store_addrs: &[String]) -> Result<Client> {
 }
 
 #[cfg(feature = "pg_kvbackend")]
-async fn create_postgres_client(opts: &MetasrvOptions) -> Result<tokio_postgres::Client> {
-    let postgres_url = opts
-        .store_addrs
-        .first()
-        .context(error::InvalidArgumentsSnafu {
-            err_msg: "empty store addrs",
-        })?;
-    let (client, connection) = tokio_postgres::connect(postgres_url, NoTls)
-        .await
-        .context(error::ConnectPostgresSnafu)?;
-
-    tokio::spawn(async move {
-        if let Err(e) = connection.await {
-            error!(e; "connection error");
-        }
-    });
-    Ok(client)
+/// Creates a pool for the Postgres backend.
+///
+/// It only use first store addr to create a pool.
+pub async fn create_postgres_pool(store_addrs: &[String]) -> Result<deadpool_postgres::Pool> {
+    create_postgres_pool_with(store_addrs, Config::new()).await
 }
 
 #[cfg(feature = "pg_kvbackend")]
-pub async fn create_postgres_pool(store_addrs: &[String]) -> Result<deadpool_postgres::Pool> {
+/// Creates a pool for the Postgres backend.
+///
+/// It only use first store addr to create a pool, and use the given config to create a pool.
+pub async fn create_postgres_pool_with(
+    store_addrs: &[String],
+    mut cfg: Config,
+) -> Result<deadpool_postgres::Pool> {
     let postgres_url = store_addrs.first().context(error::InvalidArgumentsSnafu {
         err_msg: "empty store addrs",
     })?;
-    let mut cfg = Config::new();
     cfg.url = Some(postgres_url.to_string());
     let pool = cfg
         .create_pool(Some(Runtime::Tokio1), NoTls)

src/meta-srv/src/election.rs

@@ -157,6 +157,11 @@ pub trait Election: Send + Sync {
     /// but only one can be the leader at a time.
     async fn campaign(&self) -> Result<()>;
 
+    /// Resets the campaign.
+    ///
+    /// Reset the client and the leader flag if needed.
+    async fn reset_campaign(&self) {}
+
     /// Returns the leader value for the current election.
     async fn leader(&self) -> Result<Self::Leader>;