chore: bump version to 0.9.5 (#4853)

* fix: flush metric metadata region

* chore: apply suggestions from CR

.github/workflows/develop.yml

@@ -442,6 +442,13 @@ jobs:
             minio: true
             kafka: true
             values: "with-remote-wal.yaml"
+        include:
+          - target: "fuzz_migrate_mito_regions"
+            mode:
+              name: "Local WAL"
+              minio: true
+              kafka: false
+              values: "with-minio.yaml"
     steps:
       - name: Remove unused software
         run: |
@@ -530,7 +537,7 @@ jobs:
         with:
           image-registry: localhost:5001
           values-filename: ${{ matrix.mode.values }}
-          enable-region-failover: true
+          enable-region-failover: ${{ matrix.mode.kafka }}
       - name: Port forward (mysql)
         run: |
           kubectl port-forward service/my-greptimedb-frontend 4002:4002 -n my-greptimedb&

Cargo.toml

@@ -20,6 +20,7 @@ members = [
     "src/common/mem-prof",
     "src/common/meta",
     "src/common/plugins",
+    "src/common/pprof",
     "src/common/procedure",
     "src/common/procedure-test",
     "src/common/query",
@@ -64,7 +65,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "0.9.3"
+version = "0.9.5"
 edition = "2021"
 license = "Apache-2.0"
 
@@ -120,7 +121,7 @@ etcd-client = { version = "0.13" }
 fst = "0.4.7"
 futures = "0.3"
 futures-util = "0.3"
-greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "36334744c7020734dcb4a6b8d24d52ae7ed53fe1" }
+greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "b4d301184eb0d01fd4d1042fcc7c5dfb54f3c1e3" }
 humantime = "2.1"
 humantime-serde = "1.1"
 itertools = "0.10"
@@ -137,12 +138,13 @@ opentelemetry-proto = { version = "0.5", features = [
     "metrics",
     "trace",
     "with-serde",
+    "logs",
 ] }
 parquet = { version = "51.0.0", default-features = false, features = ["arrow", "async", "object_store"] }
 paste = "1.0"
 pin-project = "1.0"
 prometheus = { version = "0.13.3", features = ["process"] }
-promql-parser = { version = "0.4" }
+promql-parser = { version = "0.4.1" }
 prost = "0.12"
 raft-engine = { version = "0.4.1", default-features = false }
 rand = "0.8"
@@ -208,6 +210,7 @@ common-macro = { path = "src/common/macro" }
 common-mem-prof = { path = "src/common/mem-prof" }
 common-meta = { path = "src/common/meta" }
 common-plugins = { path = "src/common/plugins" }
+common-pprof = { path = "src/common/pprof" }
 common-procedure = { path = "src/common/procedure" }
 common-procedure-test = { path = "src/common/procedure-test" }
 common-query = { path = "src/common/query" }

config/config.md

@@ -83,6 +83,7 @@
 | `wal.backoff_max` | String | `10s` | The maximum backoff delay.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.backoff_base` | Integer | `2` | The exponential backoff rate, i.e. next backoff = base * current backoff.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.backoff_deadline` | String | `5mins` | The deadline of retries.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.overwrite_entry_start_id` | Bool | `false` | Ignore missing entries during read WAL.<br/>**It's only used when the provider is `kafka`**.<br/><br/>This option ensures that when Kafka messages are deleted, the system <br/>can still successfully replay memtable data without throwing an <br/>out-of-range error. <br/>However, enabling this option might lead to unexpected data loss, <br/>as the system will skip over missing entries instead of treating <br/>them as critical errors. |
 | `metadata_store` | -- | -- | Metadata storage options. |
 | `metadata_store.file_size` | String | `256MB` | Kv file size in bytes. |
 | `metadata_store.purge_threshold` | String | `4GB` | Kv purge threshold. |
@@ -161,8 +162,13 @@
 | `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
+| `logging.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` | -- |
+| `logging.slow_query` | -- | -- | The slow query log options. |
+| `logging.slow_query.enable` | Bool | `false` | Whether to enable slow query log. |
+| `logging.slow_query.threshold` | String | Unset | The threshold of slow query. |
+| `logging.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.enable` | Bool | `false` | whether enable export metrics. |
 | `export_metrics.write_interval` | String | `30s` | The interval of export metrics. |
@@ -247,8 +253,13 @@
 | `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
+| `logging.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` | -- |
+| `logging.slow_query` | -- | -- | The slow query log options. |
+| `logging.slow_query.enable` | Bool | `false` | Whether to enable slow query log. |
+| `logging.slow_query.threshold` | String | Unset | The threshold of slow query. |
+| `logging.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.enable` | Bool | `false` | whether enable export metrics. |
 | `export_metrics.write_interval` | String | `30s` | The interval of export metrics. |
@@ -312,8 +323,13 @@
 | `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
+| `logging.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` | -- |
+| `logging.slow_query` | -- | -- | The slow query log options. |
+| `logging.slow_query.enable` | Bool | `false` | Whether to enable slow query log. |
+| `logging.slow_query.threshold` | String | Unset | The threshold of slow query. |
+| `logging.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.enable` | Bool | `false` | whether enable export metrics. |
 | `export_metrics.write_interval` | String | `30s` | The interval of export metrics. |
@@ -394,6 +410,7 @@
 | `wal.backoff_deadline` | String | `5mins` | The deadline of retries.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.create_index` | Bool | `true` | Whether to enable WAL index creation.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.dump_index_interval` | String | `60s` | The interval for dumping WAL indexes.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.overwrite_entry_start_id` | Bool | `false` | Ignore missing entries during read WAL.<br/>**It's only used when the provider is `kafka`**.<br/><br/>This option ensures that when Kafka messages are deleted, the system <br/>can still successfully replay memtable data without throwing an <br/>out-of-range error. <br/>However, enabling this option might lead to unexpected data loss, <br/>as the system will skip over missing entries instead of treating <br/>them as critical errors. |
 | `storage` | -- | -- | The data storage options. |
 | `storage.data_home` | String | `/tmp/greptimedb/` | The working home directory. |
 | `storage.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. |
@@ -464,8 +481,13 @@
 | `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
+| `logging.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` | -- |
+| `logging.slow_query` | -- | -- | The slow query log options. |
+| `logging.slow_query.enable` | Bool | `false` | Whether to enable slow query log. |
+| `logging.slow_query.threshold` | String | Unset | The threshold of slow query. |
+| `logging.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.enable` | Bool | `false` | whether enable export metrics. |
 | `export_metrics.write_interval` | String | `30s` | The interval of export metrics. |
@@ -510,7 +532,12 @@
 | `logging.otlp_endpoint` | String | `http://localhost:4317` | The OTLP tracing endpoint. |
 | `logging.append_stdout` | Bool | `true` | Whether to append logs to stdout. |
 | `logging.log_format` | String | `text` | The log format. Can be `text`/`json`. |
+| `logging.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` | -- |
+| `logging.slow_query` | -- | -- | The slow query log options. |
+| `logging.slow_query.enable` | Bool | `false` | Whether to enable slow query log. |
+| `logging.slow_query.threshold` | String | Unset | The threshold of slow query. |
+| `logging.slow_query.sample_ratio` | Float | Unset | The sampling ratio of slow query log. The value should be in the range of (0, 1]. |
 | `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

@@ -213,6 +213,17 @@ create_index = true
 ## **It's only used when the provider is `kafka`**.
 dump_index_interval = "60s"
 
+## Ignore missing entries during read WAL.
+## **It's only used when the provider is `kafka`**.
+## 
+## This option ensures that when Kafka messages are deleted, the system 
+## can still successfully replay memtable data without throwing an 
+## out-of-range error. 
+## However, enabling this option might lead to unexpected data loss, 
+## as the system will skip over missing entries instead of treating 
+## them as critical errors.
+overwrite_entry_start_id = false
+
 # The Kafka SASL configuration.
 # **It's only used when the provider is `kafka`**.
 # Available SASL mechanisms:
@@ -580,12 +591,28 @@ append_stdout = true
 ## The log format. Can be `text`/`json`.
 log_format = "text"
 
+## The maximum amount of log files.
+max_log_files = 720
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
 [logging.tracing_sample_ratio]
 default_ratio = 1.0
 
+## The slow query log options.
+[logging.slow_query]
+## Whether to enable slow query log.
+enable = false
+
+## The threshold of slow query.
+## @toml2docs:none-default
+threshold = "10s"
+
+## The sampling ratio of slow query log. The value should be in the range of (0, 1].
+## @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.
 ## This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape.
 [export_metrics]

config/flownode.example.toml

@@ -78,12 +78,28 @@ append_stdout = true
 ## The log format. Can be `text`/`json`.
 log_format = "text"
 
+## The maximum amount of log files.
+max_log_files = 720
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
 [logging.tracing_sample_ratio]
 default_ratio = 1.0
 
+## The slow query log options.
+[logging.slow_query]
+## Whether to enable slow query log.
+enable = false
+
+## The threshold of slow query.
+## @toml2docs:none-default
+threshold = "10s"
+
+## The sampling ratio of slow query log. The value should be in the range of (0, 1].
+## @toml2docs:none-default
+sample_ratio = 1.0
+
 ## The tracing options. Only effect when compiled with `tokio-console` feature.
 [tracing]
 ## The tokio console address.

config/frontend.example.toml

@@ -185,12 +185,28 @@ append_stdout = true
 ## The log format. Can be `text`/`json`.
 log_format = "text"
 
+## The maximum amount of log files.
+max_log_files = 720
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
 [logging.tracing_sample_ratio]
 default_ratio = 1.0
 
+## The slow query log options.
+[logging.slow_query]
+## Whether to enable slow query log.
+enable = false
+
+## The threshold of slow query.
+## @toml2docs:none-default
+threshold = "10s"
+
+## The sampling ratio of slow query log. The value should be in the range of (0, 1].
+## @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.
 ## This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape.
 [export_metrics]

config/metasrv.example.toml

@@ -172,12 +172,28 @@ append_stdout = true
 ## The log format. Can be `text`/`json`.
 log_format = "text"
 
+## The maximum amount of log files.
+max_log_files = 720
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
 [logging.tracing_sample_ratio]
 default_ratio = 1.0
 
+## The slow query log options.
+[logging.slow_query]
+## Whether to enable slow query log.
+enable = false
+
+## The threshold of slow query.
+## @toml2docs:none-default
+threshold = "10s"
+
+## The sampling ratio of slow query log. The value should be in the range of (0, 1].
+## @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.
 ## This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape.
 [export_metrics]

config/standalone.example.toml

@@ -237,6 +237,17 @@ backoff_base = 2
 ## **It's only used when the provider is `kafka`**.
 backoff_deadline = "5mins"
 
+## Ignore missing entries during read WAL.
+## **It's only used when the provider is `kafka`**.
+## 
+## This option ensures that when Kafka messages are deleted, the system 
+## can still successfully replay memtable data without throwing an 
+## out-of-range error. 
+## However, enabling this option might lead to unexpected data loss, 
+## as the system will skip over missing entries instead of treating 
+## them as critical errors.
+overwrite_entry_start_id = false
+
 # The Kafka SASL configuration.
 # **It's only used when the provider is `kafka`**.
 # Available SASL mechanisms:
@@ -624,12 +635,28 @@ append_stdout = true
 ## The log format. Can be `text`/`json`.
 log_format = "text"
 
+## The maximum amount of log files.
+max_log_files = 720
+
 ## The percentage of tracing will be sampled and exported.
 ## Valid range `[0, 1]`, 1 means all traces are sampled, 0 means all traces are not sampled, the default value is 1.
 ## ratio > 1 are treated as 1. Fractions < 0 are treated as 0
 [logging.tracing_sample_ratio]
 default_ratio = 1.0
 
+## The slow query log options.
+[logging.slow_query]
+## Whether to enable slow query log.
+enable = false
+
+## The threshold of slow query.
+## @toml2docs:none-default
+threshold = "10s"
+
+## The sampling ratio of slow query log. The value should be in the range of (0, 1].
+## @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.
 ## This is only used for `greptimedb` to export its own metrics internally. It's different from prometheus scrape.
 [export_metrics]

docs/benchmarks/log/README.md

@@ -48,4 +48,4 @@ Please refer to [SQL query](./query.sql) for GreptimeDB and Clickhouse, and [que
 
 ## Addition
 - You can tune GreptimeDB's configuration to get better performance.
-- You can setup GreptimeDB to use S3 as storage, see [here](https://docs.greptime.com/user-guide/operations/configuration/#storage-options).
+- You can setup GreptimeDB to use S3 as storage, see [here](https://docs.greptime.com/user-guide/deployments/configuration#storage-options).

docs/how-to/how-to-profile-cpu.md

@@ -1,15 +1,9 @@
 # Profiling CPU
 
-## Build GreptimeDB with `pprof` feature
-
-```bash
-cargo build --features=pprof
-```
-
 ## HTTP API
 Sample at 99 Hertz, for 5 seconds, output report in [protobuf format](https://github.com/google/pprof/blob/master/proto/profile.proto).
 ```bash
-curl -s '0:4000/v1/prof/cpu' > /tmp/pprof.out
+curl -s '0:4000/debug/prof/cpu' > /tmp/pprof.out
 ```
 
 Then you can use `pprof` command with the protobuf file.
@@ -19,10 +13,10 @@ go tool pprof -top /tmp/pprof.out
 
 Sample at 99 Hertz, for 60 seconds, output report in flamegraph format.
 ```bash
-curl -s '0:4000/v1/prof/cpu?seconds=60&output=flamegraph' > /tmp/pprof.svg
+curl -s '0:4000/debug/prof/cpu?seconds=60&output=flamegraph' > /tmp/pprof.svg
 ```
 
 Sample at 49 Hertz, for 10 seconds, output report in text format.
 ```bash
-curl -s '0:4000/v1/prof/cpu?seconds=10&frequency=49&output=text' > /tmp/pprof.txt
+curl -s '0:4000/debug/prof/cpu?seconds=10&frequency=49&output=text' > /tmp/pprof.txt
 ```

docs/how-to/how-to-profile-memory.md

@@ -12,16 +12,10 @@ brew install jemalloc
 sudo apt install libjemalloc-dev
 ```
 
-### [flamegraph](https://github.com/brendangregg/FlameGraph) 
+### [flamegraph](https://github.com/brendangregg/FlameGraph)
 
 ```bash
-curl https://raw.githubusercontent.com/brendangregg/FlameGraph/master/flamegraph.pl > ./flamegraph.pl 
-```
-
-### Build GreptimeDB with `mem-prof` feature.
-
-```bash
-cargo build --features=mem-prof
+curl https://raw.githubusercontent.com/brendangregg/FlameGraph/master/flamegraph.pl > ./flamegraph.pl
 ```
 
 ## Profiling
@@ -35,7 +29,7 @@ MALLOC_CONF=prof:true,lg_prof_interval:28 ./target/debug/greptime standalone sta
 Dump memory profiling data through HTTP API:
 
 ```bash
-curl localhost:4000/v1/prof/mem > greptime.hprof
+curl localhost:4000/debug/prof/mem > greptime.hprof
 ```
 
 You can periodically dump profiling data and compare them to find the delta memory usage.
@@ -45,6 +39,9 @@ You can periodically dump profiling data and compare them to find the delta memo
 To create flamegraph according to dumped profiling data:
 
 ```bash
-jeprof --svg <path_to_greptimedb_binary> --base=<baseline_prof> <profile_data> > output.svg
-```
+sudo apt install -y libjemalloc-dev
 
+jeprof <path_to_greptime_binary> <profile_data> --collapse | ./flamegraph.pl > mem-prof.svg
+
+jeprof <path_to_greptime_binary> --base <baseline_prof> <profile_data> --collapse | ./flamegraph.pl > output.svg
+```

src/api/src/helper.rs

@@ -17,10 +17,11 @@ use std::sync::Arc;
 use common_base::BitVec;
 use common_decimal::decimal128::{DECIMAL128_DEFAULT_SCALE, DECIMAL128_MAX_PRECISION};
 use common_decimal::Decimal128;
-use common_time::interval::IntervalUnit;
 use common_time::time::Time;
 use common_time::timestamp::TimeUnit;
-use common_time::{Date, DateTime, Interval, Timestamp};
+use common_time::{
+    Date, DateTime, IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth, Timestamp,
+};
 use datatypes::prelude::{ConcreteDataType, ValueRef};
 use datatypes::scalars::ScalarVector;
 use datatypes::types::{
@@ -456,13 +457,11 @@ pub fn push_vals(column: &mut Column, origin_count: usize, vector: VectorRef) {
             TimeUnit::Microsecond => values.time_microsecond_values.push(val.value()),
             TimeUnit::Nanosecond => values.time_nanosecond_values.push(val.value()),
         },
-        Value::Interval(val) => match val.unit() {
-            IntervalUnit::YearMonth => values.interval_year_month_values.push(val.to_i32()),
-            IntervalUnit::DayTime => values.interval_day_time_values.push(val.to_i64()),
-            IntervalUnit::MonthDayNano => values
-                .interval_month_day_nano_values
-                .push(convert_i128_to_interval(val.to_i128())),
-        },
+        Value::IntervalYearMonth(val) => values.interval_year_month_values.push(val.to_i32()),
+        Value::IntervalDayTime(val) => values.interval_day_time_values.push(val.to_i64()),
+        Value::IntervalMonthDayNano(val) => values
+            .interval_month_day_nano_values
+            .push(convert_month_day_nano_to_pb(val)),
         Value::Decimal128(val) => values.decimal128_values.push(convert_to_pb_decimal128(val)),
         Value::List(_) | Value::Duration(_) => unreachable!(),
     });
@@ -507,14 +506,12 @@ fn ddl_request_type(request: &DdlRequest) -> &'static str {
     }
 }
 
-/// Converts an i128 value to google protobuf type [IntervalMonthDayNano].
-pub fn convert_i128_to_interval(v: i128) -> v1::IntervalMonthDayNano {
-    let interval = Interval::from_i128(v);
-    let (months, days, nanoseconds) = interval.to_month_day_nano();
+/// Converts an interval to google protobuf type [IntervalMonthDayNano].
+pub fn convert_month_day_nano_to_pb(v: IntervalMonthDayNano) -> v1::IntervalMonthDayNano {
     v1::IntervalMonthDayNano {
-        months,
-        days,
-        nanoseconds,
+        months: v.months,
+        days: v.days,
+        nanoseconds: v.nanoseconds,
     }
 }
 
@@ -562,11 +559,15 @@ pub fn pb_value_to_value_ref<'a>(
         ValueData::TimeMillisecondValue(t) => ValueRef::Time(Time::new_millisecond(*t)),
         ValueData::TimeMicrosecondValue(t) => ValueRef::Time(Time::new_microsecond(*t)),
         ValueData::TimeNanosecondValue(t) => ValueRef::Time(Time::new_nanosecond(*t)),
-        ValueData::IntervalYearMonthValue(v) => ValueRef::Interval(Interval::from_i32(*v)),
-        ValueData::IntervalDayTimeValue(v) => ValueRef::Interval(Interval::from_i64(*v)),
+        ValueData::IntervalYearMonthValue(v) => {
+            ValueRef::IntervalYearMonth(IntervalYearMonth::from_i32(*v))
+        }
+        ValueData::IntervalDayTimeValue(v) => {
+            ValueRef::IntervalDayTime(IntervalDayTime::from_i64(*v))
+        }
         ValueData::IntervalMonthDayNanoValue(v) => {
-            let interval = Interval::from_month_day_nano(v.months, v.days, v.nanoseconds);
-            ValueRef::Interval(interval)
+            let interval = IntervalMonthDayNano::new(v.months, v.days, v.nanoseconds);
+            ValueRef::IntervalMonthDayNano(interval)
         }
         ValueData::Decimal128Value(v) => {
             // get precision and scale from datatype_extension
@@ -657,7 +658,7 @@ pub fn pb_values_to_vector_ref(data_type: &ConcreteDataType, values: Values) ->
             IntervalType::MonthDayNano(_) => {
                 Arc::new(IntervalMonthDayNanoVector::from_iter_values(
                     values.interval_month_day_nano_values.iter().map(|x| {
-                        Interval::from_month_day_nano(x.months, x.days, x.nanoseconds).to_i128()
+                        IntervalMonthDayNano::new(x.months, x.days, x.nanoseconds).to_i128()
                     }),
                 ))
             }
@@ -802,18 +803,18 @@ pub fn pb_values_to_values(data_type: &ConcreteDataType, values: Values) -> Vec<
         ConcreteDataType::Interval(IntervalType::YearMonth(_)) => values
             .interval_year_month_values
             .into_iter()
-            .map(|v| Value::Interval(Interval::from_i32(v)))
+            .map(|v| Value::IntervalYearMonth(IntervalYearMonth::from_i32(v)))
             .collect(),
         ConcreteDataType::Interval(IntervalType::DayTime(_)) => values
             .interval_day_time_values
             .into_iter()
-            .map(|v| Value::Interval(Interval::from_i64(v)))
+            .map(|v| Value::IntervalDayTime(IntervalDayTime::from_i64(v)))
             .collect(),
         ConcreteDataType::Interval(IntervalType::MonthDayNano(_)) => values
             .interval_month_day_nano_values
             .into_iter()
             .map(|v| {
-                Value::Interval(Interval::from_month_day_nano(
+                Value::IntervalMonthDayNano(IntervalMonthDayNano::new(
                     v.months,
                     v.days,
                     v.nanoseconds,
@@ -941,18 +942,16 @@ pub fn to_proto_value(value: Value) -> Option<v1::Value> {
                 value_data: Some(ValueData::TimeNanosecondValue(v.value())),
             },
         },
-        Value::Interval(v) => match v.unit() {
-            IntervalUnit::YearMonth => v1::Value {
-                value_data: Some(ValueData::IntervalYearMonthValue(v.to_i32())),
-            },
-            IntervalUnit::DayTime => v1::Value {
-                value_data: Some(ValueData::IntervalDayTimeValue(v.to_i64())),
-            },
-            IntervalUnit::MonthDayNano => v1::Value {
-                value_data: Some(ValueData::IntervalMonthDayNanoValue(
-                    convert_i128_to_interval(v.to_i128()),
-                )),
-            },
+        Value::IntervalYearMonth(v) => v1::Value {
+            value_data: Some(ValueData::IntervalYearMonthValue(v.to_i32())),
+        },
+        Value::IntervalDayTime(v) => v1::Value {
+            value_data: Some(ValueData::IntervalDayTimeValue(v.to_i64())),
+        },
+        Value::IntervalMonthDayNano(v) => v1::Value {
+            value_data: Some(ValueData::IntervalMonthDayNanoValue(
+                convert_month_day_nano_to_pb(v),
+            )),
         },
         Value::Decimal128(v) => v1::Value {
             value_data: Some(ValueData::Decimal128Value(convert_to_pb_decimal128(v))),
@@ -1044,13 +1043,11 @@ pub fn value_to_grpc_value(value: Value) -> GrpcValue {
                 TimeUnit::Microsecond => ValueData::TimeMicrosecondValue(v.value()),
                 TimeUnit::Nanosecond => ValueData::TimeNanosecondValue(v.value()),
             }),
-            Value::Interval(v) => Some(match v.unit() {
-                IntervalUnit::YearMonth => ValueData::IntervalYearMonthValue(v.to_i32()),
-                IntervalUnit::DayTime => ValueData::IntervalDayTimeValue(v.to_i64()),
-                IntervalUnit::MonthDayNano => {
-                    ValueData::IntervalMonthDayNanoValue(convert_i128_to_interval(v.to_i128()))
-                }
-            }),
+            Value::IntervalYearMonth(v) => Some(ValueData::IntervalYearMonthValue(v.to_i32())),
+            Value::IntervalDayTime(v) => Some(ValueData::IntervalDayTimeValue(v.to_i64())),
+            Value::IntervalMonthDayNano(v) => Some(ValueData::IntervalMonthDayNanoValue(
+                convert_month_day_nano_to_pb(v),
+            )),
             Value::Decimal128(v) => Some(ValueData::Decimal128Value(convert_to_pb_decimal128(v))),
             Value::List(_) | Value::Duration(_) => unreachable!(),
         },
@@ -1061,6 +1058,7 @@ pub fn value_to_grpc_value(value: Value) -> GrpcValue {
 mod tests {
     use std::sync::Arc;
 
+    use common_time::interval::IntervalUnit;
     use datatypes::types::{
         Int32Type, IntervalDayTimeType, IntervalMonthDayNanoType, IntervalYearMonthType,
         TimeMillisecondType, TimeSecondType, TimestampMillisecondType, TimestampSecondType,
@@ -1506,11 +1504,11 @@ mod tests {
 
     #[test]
     fn test_convert_i128_to_interval() {
-        let i128_val = 3000;
-        let interval = convert_i128_to_interval(i128_val);
+        let i128_val = 3;
+        let interval = convert_month_day_nano_to_pb(IntervalMonthDayNano::from_i128(i128_val));
         assert_eq!(interval.months, 0);
         assert_eq!(interval.days, 0);
-        assert_eq!(interval.nanoseconds, 3000);
+        assert_eq!(interval.nanoseconds, 3);
     }
 
     #[test]
@@ -1590,9 +1588,9 @@ mod tests {
             },
         );
         let expect = vec![
-            Value::Interval(Interval::from_year_month(1_i32)),
-            Value::Interval(Interval::from_year_month(2_i32)),
-            Value::Interval(Interval::from_year_month(3_i32)),
+            Value::IntervalYearMonth(IntervalYearMonth::new(1_i32)),
+            Value::IntervalYearMonth(IntervalYearMonth::new(2_i32)),
+            Value::IntervalYearMonth(IntervalYearMonth::new(3_i32)),
         ];
         assert_eq!(expect, actual);
 
@@ -1605,9 +1603,9 @@ mod tests {
             },
         );
         let expect = vec![
-            Value::Interval(Interval::from_i64(1_i64)),
-            Value::Interval(Interval::from_i64(2_i64)),
-            Value::Interval(Interval::from_i64(3_i64)),
+            Value::IntervalDayTime(IntervalDayTime::from_i64(1_i64)),
+            Value::IntervalDayTime(IntervalDayTime::from_i64(2_i64)),
+            Value::IntervalDayTime(IntervalDayTime::from_i64(3_i64)),
         ];
         assert_eq!(expect, actual);
 
@@ -1636,9 +1634,9 @@ mod tests {
             },
         );
         let expect = vec![
-            Value::Interval(Interval::from_month_day_nano(1, 2, 3)),
-            Value::Interval(Interval::from_month_day_nano(5, 6, 7)),
-            Value::Interval(Interval::from_month_day_nano(9, 10, 11)),
+            Value::IntervalMonthDayNano(IntervalMonthDayNano::new(1, 2, 3)),
+            Value::IntervalMonthDayNano(IntervalMonthDayNano::new(5, 6, 7)),
+            Value::IntervalMonthDayNano(IntervalMonthDayNano::new(9, 10, 11)),
         ];
         assert_eq!(expect, actual);
     }

src/catalog/src/error.rs

@@ -89,9 +89,8 @@ pub enum Error {
         location: Location,
     },
 
-    #[snafu(display("Failed to get procedure client in {mode} mode"))]
-    GetProcedureClient {
-        mode: String,
+    #[snafu(display("Failed to get information extension client"))]
+    GetInformationExtension {
         #[snafu(implicit)]
         location: Location,
     },
@@ -301,7 +300,7 @@ impl ErrorExt for Error {
             | Error::CacheNotFound { .. }
             | Error::CastManager { .. }
             | Error::Json { .. }
-            | Error::GetProcedureClient { .. }
+            | Error::GetInformationExtension { .. }
             | Error::ProcedureIdNotFound { .. } => StatusCode::Unexpected,
 
             Error::ViewPlanColumnsChanged { .. } => StatusCode::InvalidArguments,

src/catalog/src/kvbackend/manager.rs

@@ -21,7 +21,6 @@ use common_catalog::consts::{
     DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME, INFORMATION_SCHEMA_NAME, NUMBERS_TABLE_ID,
     PG_CATALOG_NAME,
 };
-use common_config::Mode;
 use common_error::ext::BoxedError;
 use common_meta::cache::{LayeredCacheRegistryRef, ViewInfoCacheRef};
 use common_meta::key::catalog_name::CatalogNameKey;
@@ -34,7 +33,6 @@ use common_meta::kv_backend::KvBackendRef;
 use common_procedure::ProcedureManagerRef;
 use futures_util::stream::BoxStream;
 use futures_util::{StreamExt, TryStreamExt};
-use meta_client::client::MetaClient;
 use moka::sync::Cache;
 use partition::manager::{PartitionRuleManager, PartitionRuleManagerRef};
 use session::context::{Channel, QueryContext};
@@ -50,7 +48,7 @@ use crate::error::{
     CacheNotFoundSnafu, GetTableCacheSnafu, InvalidTableInfoInCatalogSnafu, ListCatalogsSnafu,
     ListSchemasSnafu, ListTablesSnafu, Result, TableMetadataManagerSnafu,
 };
-use crate::information_schema::InformationSchemaProvider;
+use crate::information_schema::{InformationExtensionRef, InformationSchemaProvider};
 use crate::kvbackend::TableCacheRef;
 use crate::system_schema::pg_catalog::PGCatalogProvider;
 use crate::system_schema::SystemSchemaProvider;
@@ -63,9 +61,8 @@ use crate::CatalogManager;
 /// comes from `SystemCatalog`, which is static and read-only.
 #[derive(Clone)]
 pub struct KvBackendCatalogManager {
-    mode: Mode,
-    /// Only available in `Distributed` mode.
-    meta_client: Option<Arc<MetaClient>>,
+    /// Provides the extension methods for the `information_schema` tables
+    information_extension: InformationExtensionRef,
     /// Manages partition rules.
     partition_manager: PartitionRuleManagerRef,
     /// Manages table metadata.
@@ -82,15 +79,13 @@ const CATALOG_CACHE_MAX_CAPACITY: u64 = 128;
 
 impl KvBackendCatalogManager {
     pub fn new(
-        mode: Mode,
-        meta_client: Option<Arc<MetaClient>>,
+        information_extension: InformationExtensionRef,
         backend: KvBackendRef,
         cache_registry: LayeredCacheRegistryRef,
         procedure_manager: Option<ProcedureManagerRef>,
     ) -> Arc<Self> {
         Arc::new_cyclic(|me| Self {
-            mode,
-            meta_client,
+            information_extension,
             partition_manager: Arc::new(PartitionRuleManager::new(
                 backend.clone(),
                 cache_registry
@@ -118,20 +113,15 @@ impl KvBackendCatalogManager {
         })
     }
 
-    /// Returns the server running mode.
-    pub fn running_mode(&self) -> &Mode {
-        &self.mode
-    }
-
     pub fn view_info_cache(&self) -> Result<ViewInfoCacheRef> {
         self.cache_registry.get().context(CacheNotFoundSnafu {
             name: "view_info_cache",
         })
     }
 
-    /// Returns the `[MetaClient]`.
-    pub fn meta_client(&self) -> Option<Arc<MetaClient>> {
-        self.meta_client.clone()
+    /// Returns the [`InformationExtension`].
+    pub fn information_extension(&self) -> InformationExtensionRef {
+        self.information_extension.clone()
     }
 
     pub fn partition_manager(&self) -> PartitionRuleManagerRef {

src/catalog/src/system_schema/information_schema.rs

@@ -32,7 +32,11 @@ use std::collections::HashMap;
 use std::sync::{Arc, Weak};
 
 use common_catalog::consts::{self, DEFAULT_CATALOG_NAME, INFORMATION_SCHEMA_NAME};
+use common_error::ext::ErrorExt;
+use common_meta::cluster::NodeInfo;
+use common_meta::datanode::RegionStat;
 use common_meta::key::flow::FlowMetadataManager;
+use common_procedure::ProcedureInfo;
 use common_recordbatch::SendableRecordBatchStream;
 use datatypes::schema::SchemaRef;
 use lazy_static::lazy_static;
@@ -45,7 +49,7 @@ use views::InformationSchemaViews;
 
 use self::columns::InformationSchemaColumns;
 use super::{SystemSchemaProviderInner, SystemTable, SystemTableRef};
-use crate::error::Result;
+use crate::error::{Error, Result};
 use crate::system_schema::information_schema::cluster_info::InformationSchemaClusterInfo;
 use crate::system_schema::information_schema::flows::InformationSchemaFlows;
 use crate::system_schema::information_schema::information_memory_table::get_schema_columns;
@@ -318,3 +322,39 @@ where
         InformationTable::to_stream(self, request)
     }
 }
+
+pub type InformationExtensionRef = Arc<dyn InformationExtension<Error = Error> + Send + Sync>;
+
+/// The `InformationExtension` trait provides the extension methods for the `information_schema` tables.
+#[async_trait::async_trait]
+pub trait InformationExtension {
+    type Error: ErrorExt;
+
+    /// Gets the nodes information.
+    async fn nodes(&self) -> std::result::Result<Vec<NodeInfo>, Self::Error>;
+
+    /// Gets the procedures information.
+    async fn procedures(&self) -> std::result::Result<Vec<(String, ProcedureInfo)>, Self::Error>;
+
+    /// Gets the region statistics.
+    async fn region_stats(&self) -> std::result::Result<Vec<RegionStat>, Self::Error>;
+}
+
+pub struct NoopInformationExtension;
+
+#[async_trait::async_trait]
+impl InformationExtension for NoopInformationExtension {
+    type Error = Error;
+
+    async fn nodes(&self) -> std::result::Result<Vec<NodeInfo>, Self::Error> {
+        Ok(vec![])
+    }
+
+    async fn procedures(&self) -> std::result::Result<Vec<(String, ProcedureInfo)>, Self::Error> {
+        Ok(vec![])
+    }
+
+    async fn region_stats(&self) -> std::result::Result<Vec<RegionStat>, Self::Error> {
+        Ok(vec![])
+    }
+}

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

@@ -17,13 +17,10 @@ use std::time::Duration;
 
 use arrow_schema::SchemaRef as ArrowSchemaRef;
 use common_catalog::consts::INFORMATION_SCHEMA_CLUSTER_INFO_TABLE_ID;
-use common_config::Mode;
 use common_error::ext::BoxedError;
-use common_meta::cluster::{ClusterInfo, NodeInfo, NodeStatus};
-use common_meta::peer::Peer;
+use common_meta::cluster::NodeInfo;
 use common_recordbatch::adapter::RecordBatchStreamAdapter;
 use common_recordbatch::{RecordBatch, SendableRecordBatchStream};
-use common_telemetry::warn;
 use common_time::timestamp::Timestamp;
 use datafusion::execution::TaskContext;
 use datafusion::physical_plan::stream::RecordBatchStreamAdapter as DfRecordBatchStreamAdapter;
@@ -40,7 +37,7 @@ use snafu::ResultExt;
 use store_api::storage::{ScanRequest, TableId};
 
 use super::CLUSTER_INFO;
-use crate::error::{CreateRecordBatchSnafu, InternalSnafu, ListNodesSnafu, Result};
+use crate::error::{CreateRecordBatchSnafu, InternalSnafu, Result};
 use crate::system_schema::information_schema::{InformationTable, Predicates};
 use crate::system_schema::utils;
 use crate::CatalogManager;
@@ -70,7 +67,6 @@ const INIT_CAPACITY: usize = 42;
 pub(super) struct InformationSchemaClusterInfo {
     schema: SchemaRef,
     catalog_manager: Weak<dyn CatalogManager>,
-    start_time_ms: u64,
 }
 
 impl InformationSchemaClusterInfo {
@@ -78,7 +74,6 @@ impl InformationSchemaClusterInfo {
         Self {
             schema: Self::schema(),
             catalog_manager,
-            start_time_ms: common_time::util::current_time_millis() as u64,
         }
     }
 
@@ -100,11 +95,7 @@ impl InformationSchemaClusterInfo {
     }
 
     fn builder(&self) -> InformationSchemaClusterInfoBuilder {
-        InformationSchemaClusterInfoBuilder::new(
-            self.schema.clone(),
-            self.catalog_manager.clone(),
-            self.start_time_ms,
-        )
+        InformationSchemaClusterInfoBuilder::new(self.schema.clone(), self.catalog_manager.clone())
     }
 }
 
@@ -144,7 +135,6 @@ impl InformationTable for InformationSchemaClusterInfo {
 
 struct InformationSchemaClusterInfoBuilder {
     schema: SchemaRef,
-    start_time_ms: u64,
     catalog_manager: Weak<dyn CatalogManager>,
 
     peer_ids: Int64VectorBuilder,
@@ -158,11 +148,7 @@ struct InformationSchemaClusterInfoBuilder {
 }
 
 impl InformationSchemaClusterInfoBuilder {
-    fn new(
-        schema: SchemaRef,
-        catalog_manager: Weak<dyn CatalogManager>,
-        start_time_ms: u64,
-    ) -> Self {
+    fn new(schema: SchemaRef, catalog_manager: Weak<dyn CatalogManager>) -> Self {
         Self {
             schema,
             catalog_manager,
@@ -174,56 +160,17 @@ impl InformationSchemaClusterInfoBuilder {
             start_times: TimestampMillisecondVectorBuilder::with_capacity(INIT_CAPACITY),
             uptimes: StringVectorBuilder::with_capacity(INIT_CAPACITY),
             active_times: StringVectorBuilder::with_capacity(INIT_CAPACITY),
-            start_time_ms,
         }
     }
 
     /// Construct the `information_schema.cluster_info` virtual table
     async fn make_cluster_info(&mut self, request: Option<ScanRequest>) -> Result<RecordBatch> {
         let predicates = Predicates::from_scan_request(&request);
-        let mode = utils::running_mode(&self.catalog_manager)?.unwrap_or(Mode::Standalone);
-
-        match mode {
-            Mode::Standalone => {
-                let build_info = common_version::build_info();
-
-                self.add_node_info(
-                    &predicates,
-                    NodeInfo {
-                        // For the standalone:
-                        // - id always 0
-                        // - empty string for peer_addr
-                        peer: Peer {
-                            id: 0,
-                            addr: "".to_string(),
-                        },
-                        last_activity_ts: -1,
-                        status: NodeStatus::Standalone,
-                        version: build_info.version.to_string(),
-                        git_commit: build_info.commit_short.to_string(),
-                        // Use `self.start_time_ms` instead.
-                        // It's not precise but enough.
-                        start_time_ms: self.start_time_ms,
-                    },
-                );
-            }
-            Mode::Distributed => {
-                if let Some(meta_client) = utils::meta_client(&self.catalog_manager)? {
-                    let node_infos = meta_client
-                        .list_nodes(None)
-                        .await
-                        .map_err(BoxedError::new)
-                        .context(ListNodesSnafu)?;
-
-                    for node_info in node_infos {
-                        self.add_node_info(&predicates, node_info);
-                    }
-                } else {
-                    warn!("Could not find meta client in distributed mode.");
-                }
-            }
+        let information_extension = utils::information_extension(&self.catalog_manager)?;
+        let node_infos = information_extension.nodes().await?;
+        for node_info in node_infos {
+            self.add_node_info(&predicates, node_info);
         }
-
         self.finish()
     }
 

src/catalog/src/system_schema/information_schema/procedure_info.rs

@@ -14,14 +14,10 @@
 
 use std::sync::{Arc, Weak};
 
-use api::v1::meta::{ProcedureMeta, ProcedureStatus};
 use arrow_schema::SchemaRef as ArrowSchemaRef;
 use common_catalog::consts::INFORMATION_SCHEMA_PROCEDURE_INFO_TABLE_ID;
-use common_config::Mode;
 use common_error::ext::BoxedError;
-use common_meta::ddl::{ExecutorContext, ProcedureExecutor};
-use common_meta::rpc::procedure;
-use common_procedure::{ProcedureInfo, ProcedureState};
+use common_procedure::ProcedureInfo;
 use common_recordbatch::adapter::RecordBatchStreamAdapter;
 use common_recordbatch::{RecordBatch, SendableRecordBatchStream};
 use common_time::timestamp::Timestamp;
@@ -38,10 +34,7 @@ use snafu::ResultExt;
 use store_api::storage::{ScanRequest, TableId};
 
 use super::PROCEDURE_INFO;
-use crate::error::{
-    ConvertProtoDataSnafu, CreateRecordBatchSnafu, GetProcedureClientSnafu, InternalSnafu,
-    ListProceduresSnafu, ProcedureIdNotFoundSnafu, Result,
-};
+use crate::error::{CreateRecordBatchSnafu, InternalSnafu, Result};
 use crate::system_schema::information_schema::{InformationTable, Predicates};
 use crate::system_schema::utils;
 use crate::CatalogManager;
@@ -167,45 +160,11 @@ impl InformationSchemaProcedureInfoBuilder {
     /// Construct the `information_schema.procedure_info` virtual table
     async fn make_procedure_info(&mut self, request: Option<ScanRequest>) -> Result<RecordBatch> {
         let predicates = Predicates::from_scan_request(&request);
-        let mode = utils::running_mode(&self.catalog_manager)?.unwrap_or(Mode::Standalone);
-        match mode {
-            Mode::Standalone => {
-                if let Some(procedure_manager) = utils::procedure_manager(&self.catalog_manager)? {
-                    let procedures = procedure_manager
-                        .list_procedures()
-                        .await
-                        .map_err(BoxedError::new)
-                        .context(ListProceduresSnafu)?;
-                    for procedure in procedures {
-                        self.add_procedure(
-                            &predicates,
-                            procedure.state.as_str_name().to_string(),
-                            procedure,
-                        );
-                    }
-                } else {
-                    return GetProcedureClientSnafu { mode: "standalone" }.fail();
-                }
-            }
-            Mode::Distributed => {
-                if let Some(meta_client) = utils::meta_client(&self.catalog_manager)? {
-                    let procedures = meta_client
-                        .list_procedures(&ExecutorContext::default())
-                        .await
-                        .map_err(BoxedError::new)
-                        .context(ListProceduresSnafu)?;
-                    for procedure in procedures.procedures {
-                        self.add_procedure_info(&predicates, procedure)?;
-                    }
-                } else {
-                    return GetProcedureClientSnafu {
-                        mode: "distributed",
-                    }
-                    .fail();
-                }
-            }
-        };
-
+        let information_extension = utils::information_extension(&self.catalog_manager)?;
+        let procedures = information_extension.procedures().await?;
+        for (status, procedure_info) in procedures {
+            self.add_procedure(&predicates, status, procedure_info);
+        }
         self.finish()
     }
 
@@ -247,34 +206,6 @@ impl InformationSchemaProcedureInfoBuilder {
         self.lock_keys.push(Some(&lock_keys));
     }
 
-    fn add_procedure_info(
-        &mut self,
-        predicates: &Predicates,
-        procedure: ProcedureMeta,
-    ) -> Result<()> {
-        let pid = match procedure.id {
-            Some(pid) => pid,
-            None => return ProcedureIdNotFoundSnafu {}.fail(),
-        };
-        let pid = procedure::pb_pid_to_pid(&pid)
-            .map_err(BoxedError::new)
-            .context(ConvertProtoDataSnafu)?;
-        let status = ProcedureStatus::try_from(procedure.status)
-            .map(|v| v.as_str_name())
-            .unwrap_or("Unknown")
-            .to_string();
-        let procedure_info = ProcedureInfo {
-            id: pid,
-            type_name: procedure.type_name,
-            start_time_ms: procedure.start_time_ms,
-            end_time_ms: procedure.end_time_ms,
-            state: ProcedureState::Running,
-            lock_keys: procedure.lock_keys,
-        };
-        self.add_procedure(predicates, status, procedure_info);
-        Ok(())
-    }
-
     fn finish(&mut self) -> Result<RecordBatch> {
         let columns: Vec<VectorRef> = vec![
             Arc::new(self.procedure_ids.finish()),

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

@@ -224,8 +224,8 @@ impl InformationSchemaRegionPeersBuilder {
             let region_id = RegionId::new(table_id, route.region.id.region_number()).as_u64();
             let peer_id = route.leader_peer.clone().map(|p| p.id);
             let peer_addr = route.leader_peer.clone().map(|p| p.addr);
-            let status = if let Some(status) = route.leader_status {
-                Some(status.as_ref().to_string())
+            let state = if let Some(state) = route.leader_state {
+                Some(state.as_ref().to_string())
             } else {
                 // Alive by default
                 Some("ALIVE".to_string())
@@ -242,7 +242,7 @@ impl InformationSchemaRegionPeersBuilder {
             self.peer_ids.push(peer_id);
             self.peer_addrs.push(peer_addr.as_deref());
             self.is_leaders.push(Some("Yes"));
-            self.statuses.push(status.as_deref());
+            self.statuses.push(state.as_deref());
             self.down_seconds
                 .push(route.leader_down_millis().map(|m| m / 1000));
         }

src/catalog/src/system_schema/information_schema/region_statistics.rs

@@ -16,13 +16,10 @@ use std::sync::{Arc, Weak};
 
 use arrow_schema::SchemaRef as ArrowSchemaRef;
 use common_catalog::consts::INFORMATION_SCHEMA_REGION_STATISTICS_TABLE_ID;
-use common_config::Mode;
 use common_error::ext::BoxedError;
-use common_meta::cluster::ClusterInfo;
 use common_meta::datanode::RegionStat;
 use common_recordbatch::adapter::RecordBatchStreamAdapter;
 use common_recordbatch::{DfSendableRecordBatchStream, RecordBatch, SendableRecordBatchStream};
-use common_telemetry::tracing::warn;
 use datafusion::execution::TaskContext;
 use datafusion::physical_plan::stream::RecordBatchStreamAdapter as DfRecordBatchStreamAdapter;
 use datafusion::physical_plan::streaming::PartitionStream as DfPartitionStream;
@@ -34,7 +31,7 @@ use snafu::ResultExt;
 use store_api::storage::{ScanRequest, TableId};
 
 use super::{InformationTable, REGION_STATISTICS};
-use crate::error::{CreateRecordBatchSnafu, InternalSnafu, ListRegionStatsSnafu, Result};
+use crate::error::{CreateRecordBatchSnafu, InternalSnafu, Result};
 use crate::information_schema::Predicates;
 use crate::system_schema::utils;
 use crate::CatalogManager;
@@ -167,28 +164,11 @@ impl InformationSchemaRegionStatisticsBuilder {
         request: Option<ScanRequest>,
     ) -> Result<RecordBatch> {
         let predicates = Predicates::from_scan_request(&request);
-        let mode = utils::running_mode(&self.catalog_manager)?.unwrap_or(Mode::Standalone);
-
-        match mode {
-            Mode::Standalone => {
-                // TODO(weny): implement it
-            }
-            Mode::Distributed => {
-                if let Some(meta_client) = utils::meta_client(&self.catalog_manager)? {
-                    let region_stats = meta_client
-                        .list_region_stats()
-                        .await
-                        .map_err(BoxedError::new)
-                        .context(ListRegionStatsSnafu)?;
-                    for region_stat in region_stats {
-                        self.add_region_statistic(&predicates, region_stat);
-                    }
-                } else {
-                    warn!("Meta client is not available");
-                }
-            }
+        let information_extension = utils::information_extension(&self.catalog_manager)?;
+        let region_stats = information_extension.region_stats().await?;
+        for region_stat in region_stats {
+            self.add_region_statistic(&predicates, region_stat);
         }
-
         self.finish()
     }
 

src/catalog/src/system_schema/utils.rs

@@ -12,48 +12,33 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-pub mod tables;
+use std::sync::Weak;
 
-use std::sync::{Arc, Weak};
-
-use common_config::Mode;
 use common_meta::key::TableMetadataManagerRef;
-use common_procedure::ProcedureManagerRef;
-use meta_client::client::MetaClient;
 use snafu::OptionExt;
 
-use crate::error::{Result, UpgradeWeakCatalogManagerRefSnafu};
+use crate::error::{GetInformationExtensionSnafu, Result, UpgradeWeakCatalogManagerRefSnafu};
+use crate::information_schema::InformationExtensionRef;
 use crate::kvbackend::KvBackendCatalogManager;
 use crate::CatalogManager;
 
-/// Try to get the server running mode from `[CatalogManager]` weak reference.
-pub fn running_mode(catalog_manager: &Weak<dyn CatalogManager>) -> Result<Option<Mode>> {
+pub mod tables;
+
+/// Try to get the `[InformationExtension]` from `[CatalogManager]` weak reference.
+pub fn information_extension(
+    catalog_manager: &Weak<dyn CatalogManager>,
+) -> Result<InformationExtensionRef> {
     let catalog_manager = catalog_manager
         .upgrade()
         .context(UpgradeWeakCatalogManagerRefSnafu)?;
 
-    Ok(catalog_manager
+    let information_extension = catalog_manager
         .as_any()
         .downcast_ref::<KvBackendCatalogManager>()
-        .map(|manager| manager.running_mode())
-        .copied())
-}
+        .map(|manager| manager.information_extension())
+        .context(GetInformationExtensionSnafu)?;
 
-/// Try to get the `[MetaClient]` from `[CatalogManager]` weak reference.
-pub fn meta_client(catalog_manager: &Weak<dyn CatalogManager>) -> Result<Option<Arc<MetaClient>>> {
-    let catalog_manager = catalog_manager
-        .upgrade()
-        .context(UpgradeWeakCatalogManagerRefSnafu)?;
-
-    let meta_client = match catalog_manager
-        .as_any()
-        .downcast_ref::<KvBackendCatalogManager>()
-    {
-        None => None,
-        Some(manager) => manager.meta_client(),
-    };
-
-    Ok(meta_client)
+    Ok(information_extension)
 }
 
 /// Try to get the `[TableMetadataManagerRef]` from `[CatalogManager]` weak reference.
@@ -69,17 +54,3 @@ pub fn table_meta_manager(
         .downcast_ref::<KvBackendCatalogManager>()
         .map(|manager| manager.table_metadata_manager_ref().clone()))
 }
-
-/// Try to get the `[ProcedureManagerRef]` from `[CatalogManager]` weak reference.
-pub fn procedure_manager(
-    catalog_manager: &Weak<dyn CatalogManager>,
-) -> Result<Option<ProcedureManagerRef>> {
-    let catalog_manager = catalog_manager
-        .upgrade()
-        .context(UpgradeWeakCatalogManagerRefSnafu)?;
-
-    Ok(catalog_manager
-        .as_any()
-        .downcast_ref::<KvBackendCatalogManager>()
-        .and_then(|manager| manager.procedure_manager()))
-}

src/catalog/src/table_source.rs

@@ -259,7 +259,6 @@ mod tests {
 
     use arrow::datatypes::{DataType, Field, Schema, SchemaRef};
     use cache::{build_fundamental_cache_registry, with_default_composite_cache_registry};
-    use common_config::Mode;
     use common_meta::cache::{CacheRegistryBuilder, LayeredCacheRegistryBuilder};
     use common_meta::key::TableMetadataManager;
     use common_meta::kv_backend::memory::MemoryKvBackend;
@@ -269,6 +268,8 @@ mod tests {
     use datafusion::logical_expr::builder::LogicalTableSource;
     use datafusion::logical_expr::{col, lit, LogicalPlan, LogicalPlanBuilder};
 
+    use crate::information_schema::NoopInformationExtension;
+
     struct MockDecoder;
     impl MockDecoder {
         pub fn arc() -> Arc<Self> {
@@ -323,8 +324,7 @@ mod tests {
         );
 
         let catalog_manager = KvBackendCatalogManager::new(
-            Mode::Standalone,
-            None,
+            Arc::new(NoopInformationExtension),
             backend.clone(),
             layered_cache_registry,
             None,

src/cmd/Cargo.toml

@@ -10,7 +10,7 @@ name = "greptime"
 path = "src/bin/greptime.rs"
 
 [features]
-default = ["python"]
+default = ["python", "servers/pprof", "servers/mem-prof"]
 tokio-console = ["common-telemetry/tokio-console"]
 python = ["frontend/python"]
 

src/cmd/src/cli/bench.rs

@@ -158,7 +158,7 @@ fn create_region_routes(regions: Vec<RegionNumber>) -> Vec<RegionRoute> {
                 addr: String::new(),
             }),
             follower_peers: vec![],
-            leader_status: None,
+            leader_state: None,
             leader_down_since: None,
         });
     }

src/cmd/src/cli/repl.rs

@@ -46,12 +46,12 @@ use substrait::{DFLogicalSubstraitConvertor, SubstraitPlan};
 use crate::cli::cmd::ReplCommand;
 use crate::cli::helper::RustylineHelper;
 use crate::cli::AttachCommand;
-use crate::error;
 use crate::error::{
     CollectRecordBatchesSnafu, ParseSqlSnafu, PlanStatementSnafu, PrettyPrintRecordBatchesSnafu,
     ReadlineSnafu, ReplCreationSnafu, RequestDatabaseSnafu, Result, StartMetaClientSnafu,
     SubstraitEncodeLogicalPlanSnafu,
 };
+use crate::{error, DistributedInformationExtension};
 
 /// Captures the state of the repl, gathers commands and executes them one by one
 pub struct Repl {
@@ -275,9 +275,9 @@ async fn create_query_engine(meta_addr: &str) -> Result<DatafusionQueryEngine> {
         .build(),
     );
 
+    let information_extension = Arc::new(DistributedInformationExtension::new(meta_client.clone()));
     let catalog_manager = KvBackendCatalogManager::new(
-        Mode::Distributed,
-        Some(meta_client.clone()),
+        information_extension,
         cached_meta_backend.clone(),
         layered_cache_registry,
         None,

src/cmd/src/flownode.rs

@@ -41,7 +41,7 @@ use crate::error::{
     MissingConfigSnafu, Result, ShutdownFlownodeSnafu, StartFlownodeSnafu,
 };
 use crate::options::{GlobalOptions, GreptimeOptions};
-use crate::{log_versions, App};
+use crate::{log_versions, App, DistributedInformationExtension};
 
 pub const APP_NAME: &str = "greptime-flownode";
 
@@ -269,9 +269,10 @@ impl StartCommand {
             .build(),
         );
 
+        let information_extension =
+            Arc::new(DistributedInformationExtension::new(meta_client.clone()));
         let catalog_manager = KvBackendCatalogManager::new(
-            opts.mode,
-            Some(meta_client.clone()),
+            information_extension,
             cached_meta_backend.clone(),
             layered_cache_registry.clone(),
             None,

src/cmd/src/frontend.rs

@@ -36,8 +36,8 @@ use frontend::instance::builder::FrontendBuilder;
 use frontend::instance::{FrontendInstance, Instance as FeInstance};
 use frontend::server::Services;
 use meta_client::{MetaClientOptions, MetaClientType};
+use query::stats::StatementStatistics;
 use servers::tls::{TlsMode, TlsOption};
-use servers::Mode;
 use snafu::{OptionExt, ResultExt};
 use tracing_appender::non_blocking::WorkerGuard;
 
@@ -46,7 +46,7 @@ use crate::error::{
     Result, StartFrontendSnafu,
 };
 use crate::options::{GlobalOptions, GreptimeOptions};
-use crate::{log_versions, App};
+use crate::{log_versions, App, DistributedInformationExtension};
 
 type FrontendOptions = GreptimeOptions<frontend::frontend::FrontendOptions>;
 
@@ -315,9 +315,10 @@ impl StartCommand {
             .build(),
         );
 
+        let information_extension =
+            Arc::new(DistributedInformationExtension::new(meta_client.clone()));
         let catalog_manager = KvBackendCatalogManager::new(
-            Mode::Distributed,
-            Some(meta_client.clone()),
+            information_extension,
             cached_meta_backend.clone(),
             layered_cache_registry.clone(),
             None,
@@ -352,6 +353,7 @@ impl StartCommand {
             catalog_manager,
             Arc::new(client),
             meta_client,
+            StatementStatistics::new(opts.logging.slow_query.clone()),
         )
         .with_plugin(plugins.clone())
         .with_local_cache_invalidator(layered_cache_registry)

src/cmd/src/lib.rs

@@ -15,7 +15,17 @@
 #![feature(assert_matches, let_chains)]
 
 use async_trait::async_trait;
+use catalog::information_schema::InformationExtension;
+use client::api::v1::meta::ProcedureStatus;
+use common_error::ext::BoxedError;
+use common_meta::cluster::{ClusterInfo, NodeInfo};
+use common_meta::datanode::RegionStat;
+use common_meta::ddl::{ExecutorContext, ProcedureExecutor};
+use common_meta::rpc::procedure;
+use common_procedure::{ProcedureInfo, ProcedureState};
 use common_telemetry::{error, info};
+use meta_client::MetaClientRef;
+use snafu::ResultExt;
 
 use crate::error::Result;
 
@@ -94,3 +104,69 @@ fn log_env_flags() {
         info!("argument: {}", argument);
     }
 }
+
+pub struct DistributedInformationExtension {
+    meta_client: MetaClientRef,
+}
+
+impl DistributedInformationExtension {
+    pub fn new(meta_client: MetaClientRef) -> Self {
+        Self { meta_client }
+    }
+}
+
+#[async_trait::async_trait]
+impl InformationExtension for DistributedInformationExtension {
+    type Error = catalog::error::Error;
+
+    async fn nodes(&self) -> std::result::Result<Vec<NodeInfo>, Self::Error> {
+        self.meta_client
+            .list_nodes(None)
+            .await
+            .map_err(BoxedError::new)
+            .context(catalog::error::ListNodesSnafu)
+    }
+
+    async fn procedures(&self) -> std::result::Result<Vec<(String, ProcedureInfo)>, Self::Error> {
+        let procedures = self
+            .meta_client
+            .list_procedures(&ExecutorContext::default())
+            .await
+            .map_err(BoxedError::new)
+            .context(catalog::error::ListProceduresSnafu)?
+            .procedures;
+        let mut result = Vec::with_capacity(procedures.len());
+        for procedure in procedures {
+            let pid = match procedure.id {
+                Some(pid) => pid,
+                None => return catalog::error::ProcedureIdNotFoundSnafu {}.fail(),
+            };
+            let pid = procedure::pb_pid_to_pid(&pid)
+                .map_err(BoxedError::new)
+                .context(catalog::error::ConvertProtoDataSnafu)?;
+            let status = ProcedureStatus::try_from(procedure.status)
+                .map(|v| v.as_str_name())
+                .unwrap_or("Unknown")
+                .to_string();
+            let procedure_info = ProcedureInfo {
+                id: pid,
+                type_name: procedure.type_name,
+                start_time_ms: procedure.start_time_ms,
+                end_time_ms: procedure.end_time_ms,
+                state: ProcedureState::Running,
+                lock_keys: procedure.lock_keys,
+            };
+            result.push((status, procedure_info));
+        }
+
+        Ok(result)
+    }
+
+    async fn region_stats(&self) -> std::result::Result<Vec<RegionStat>, Self::Error> {
+        self.meta_client
+            .list_region_stats()
+            .await
+            .map_err(BoxedError::new)
+            .context(catalog::error::ListRegionStatsSnafu)
+    }
+}

src/cmd/src/metasrv.rs

@@ -48,6 +48,10 @@ impl Instance {
             _guard: guard,
         }
     }
+
+    pub fn get_inner(&self) -> &MetasrvInstance {
+        &self.instance
+    }
 }
 
 #[async_trait]
@@ -86,6 +90,14 @@ impl Command {
     pub fn load_options(&self, global_options: &GlobalOptions) -> Result<MetasrvOptions> {
         self.subcmd.load_options(global_options)
     }
+
+    pub fn config_file(&self) -> &Option<String> {
+        self.subcmd.config_file()
+    }
+
+    pub fn env_prefix(&self) -> &String {
+        self.subcmd.env_prefix()
+    }
 }
 
 #[derive(Parser)]
@@ -105,6 +117,18 @@ impl SubCommand {
             SubCommand::Start(cmd) => cmd.load_options(global_options),
         }
     }
+
+    fn config_file(&self) -> &Option<String> {
+        match self {
+            SubCommand::Start(cmd) => &cmd.config_file,
+        }
+    }
+
+    fn env_prefix(&self) -> &String {
+        match self {
+            SubCommand::Start(cmd) => &cmd.env_prefix,
+        }
+    }
 }
 
 #[derive(Debug, Default, Parser)]

src/cmd/src/standalone.rs

@@ -17,14 +17,18 @@ use std::{fs, path};
 
 use async_trait::async_trait;
 use cache::{build_fundamental_cache_registry, with_default_composite_cache_registry};
+use catalog::information_schema::InformationExtension;
 use catalog::kvbackend::KvBackendCatalogManager;
 use clap::Parser;
+use client::api::v1::meta::RegionRole;
 use common_base::Plugins;
 use common_catalog::consts::{MIN_USER_FLOW_ID, MIN_USER_TABLE_ID};
 use common_config::{metadata_store_dir, Configurable, KvBackendConfig};
 use common_error::ext::BoxedError;
 use common_meta::cache::LayeredCacheRegistryBuilder;
 use common_meta::cache_invalidator::CacheInvalidatorRef;
+use common_meta::cluster::{NodeInfo, NodeStatus};
+use common_meta::datanode::RegionStat;
 use common_meta::ddl::flow_meta::{FlowMetadataAllocator, FlowMetadataAllocatorRef};
 use common_meta::ddl::table_meta::{TableMetadataAllocator, TableMetadataAllocatorRef};
 use common_meta::ddl::{DdlContext, NoopRegionFailureDetectorControl, ProcedureExecutorRef};
@@ -33,10 +37,11 @@ use common_meta::key::flow::{FlowMetadataManager, FlowMetadataManagerRef};
 use common_meta::key::{TableMetadataManager, TableMetadataManagerRef};
 use common_meta::kv_backend::KvBackendRef;
 use common_meta::node_manager::NodeManagerRef;
+use common_meta::peer::Peer;
 use common_meta::region_keeper::MemoryRegionKeeper;
 use common_meta::sequence::SequenceBuilder;
 use common_meta::wal_options_allocator::{WalOptionsAllocator, WalOptionsAllocatorRef};
-use common_procedure::ProcedureManagerRef;
+use common_procedure::{ProcedureInfo, ProcedureManagerRef};
 use common_telemetry::info;
 use common_telemetry::logging::{LoggingOptions, TracingOptions};
 use common_time::timezone::set_default_timezone;
@@ -44,6 +49,7 @@ use common_version::{short_version, version};
 use common_wal::config::DatanodeWalConfig;
 use datanode::config::{DatanodeOptions, ProcedureConfig, RegionEngineConfig, StorageConfig};
 use datanode::datanode::{Datanode, DatanodeBuilder};
+use datanode::region_server::RegionServer;
 use file_engine::config::EngineConfig as FileEngineConfig;
 use flow::{FlowWorkerManager, FlownodeBuilder, FrontendInvoker};
 use frontend::frontend::FrontendOptions;
@@ -55,6 +61,7 @@ use frontend::service_config::{
 };
 use meta_srv::metasrv::{FLOW_ID_SEQ, TABLE_ID_SEQ};
 use mito2::config::MitoConfig;
+use query::stats::StatementStatistics;
 use serde::{Deserialize, Serialize};
 use servers::export_metrics::ExportMetricsOption;
 use servers::grpc::GrpcOptions;
@@ -477,9 +484,18 @@ impl StartCommand {
             .build(),
         );
 
+        let datanode = DatanodeBuilder::new(dn_opts, plugins.clone())
+            .with_kv_backend(kv_backend.clone())
+            .build()
+            .await
+            .context(StartDatanodeSnafu)?;
+
+        let information_extension = Arc::new(StandaloneInformationExtension::new(
+            datanode.region_server(),
+            procedure_manager.clone(),
+        ));
         let catalog_manager = KvBackendCatalogManager::new(
-            dn_opts.mode,
-            None,
+            information_extension,
             kv_backend.clone(),
             layered_cache_registry.clone(),
             Some(procedure_manager.clone()),
@@ -488,12 +504,6 @@ impl StartCommand {
         let table_metadata_manager =
             Self::create_table_metadata_manager(kv_backend.clone()).await?;
 
-        let datanode = DatanodeBuilder::new(dn_opts, plugins.clone())
-            .with_kv_backend(kv_backend.clone())
-            .build()
-            .await
-            .context(StartDatanodeSnafu)?;
-
         let flow_metadata_manager = Arc::new(FlowMetadataManager::new(kv_backend.clone()));
         let flow_builder = FlownodeBuilder::new(
             Default::default(),
@@ -557,6 +567,7 @@ impl StartCommand {
             catalog_manager.clone(),
             node_manager.clone(),
             ddl_task_executor.clone(),
+            StatementStatistics::new(opts.logging.slow_query.clone()),
         )
         .with_plugin(plugins.clone())
         .try_build()
@@ -642,6 +653,91 @@ impl StartCommand {
     }
 }
 
+pub struct StandaloneInformationExtension {
+    region_server: RegionServer,
+    procedure_manager: ProcedureManagerRef,
+    start_time_ms: u64,
+}
+
+impl StandaloneInformationExtension {
+    pub fn new(region_server: RegionServer, procedure_manager: ProcedureManagerRef) -> Self {
+        Self {
+            region_server,
+            procedure_manager,
+            start_time_ms: common_time::util::current_time_millis() as u64,
+        }
+    }
+}
+
+#[async_trait::async_trait]
+impl InformationExtension for StandaloneInformationExtension {
+    type Error = catalog::error::Error;
+
+    async fn nodes(&self) -> std::result::Result<Vec<NodeInfo>, Self::Error> {
+        let build_info = common_version::build_info();
+        let node_info = NodeInfo {
+            // For the standalone:
+            // - id always 0
+            // - empty string for peer_addr
+            peer: Peer {
+                id: 0,
+                addr: "".to_string(),
+            },
+            last_activity_ts: -1,
+            status: NodeStatus::Standalone,
+            version: build_info.version.to_string(),
+            git_commit: build_info.commit_short.to_string(),
+            // Use `self.start_time_ms` instead.
+            // It's not precise but enough.
+            start_time_ms: self.start_time_ms,
+        };
+        Ok(vec![node_info])
+    }
+
+    async fn procedures(&self) -> std::result::Result<Vec<(String, ProcedureInfo)>, Self::Error> {
+        self.procedure_manager
+            .list_procedures()
+            .await
+            .map_err(BoxedError::new)
+            .map(|procedures| {
+                procedures
+                    .into_iter()
+                    .map(|procedure| {
+                        let status = procedure.state.as_str_name().to_string();
+                        (status, procedure)
+                    })
+                    .collect::<Vec<_>>()
+            })
+            .context(catalog::error::ListProceduresSnafu)
+    }
+
+    async fn region_stats(&self) -> std::result::Result<Vec<RegionStat>, Self::Error> {
+        let stats = self
+            .region_server
+            .reportable_regions()
+            .into_iter()
+            .map(|stat| {
+                let region_stat = self
+                    .region_server
+                    .region_statistic(stat.region_id)
+                    .unwrap_or_default();
+                RegionStat {
+                    id: stat.region_id,
+                    rcus: 0,
+                    wcus: 0,
+                    approximate_bytes: region_stat.estimated_disk_size() as i64,
+                    engine: stat.engine,
+                    role: RegionRole::from(stat.role).into(),
+                    memtable_size: region_stat.memtable_size,
+                    manifest_size: region_stat.manifest_size,
+                    sst_size: region_stat.sst_size,
+                }
+            })
+            .collect::<Vec<_>>();
+        Ok(stats)
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use std::default::Default;

src/common/base/src/plugins.rs

@@ -38,6 +38,18 @@ impl Plugins {
         self.read().get::<T>().cloned()
     }
 
+    pub fn get_or_insert<T, F>(&self, f: F) -> T
+    where
+        T: 'static + Send + Sync + Clone,
+        F: FnOnce() -> T,
+    {
+        let mut binding = self.write();
+        if !binding.contains::<T>() {
+            binding.insert(f());
+        }
+        binding.get::<T>().cloned().unwrap()
+    }
+
     pub fn map_mut<T: 'static + Send + Sync, F, R>(&self, mapper: F) -> R
     where
         F: FnOnce(Option<&mut T>) -> R,

src/common/function/Cargo.toml

@@ -9,7 +9,7 @@ workspace = true
 
 [features]
 default = ["geo"]
-geo = ["geohash", "h3o"]
+geo = ["geohash", "h3o", "s2"]
 
 [dependencies]
 api.workspace = true
@@ -35,6 +35,7 @@ num = "0.4"
 num-traits = "0.2"
 once_cell.workspace = true
 paste = "1.0"
+s2 = { version = "0.0.12", optional = true }
 serde.workspace = true
 serde_json.workspace = true
 session.workspace = true

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

@@ -91,5 +91,12 @@ impl AggregateFunctions {
         register_aggr_func!("argmin", 1, ArgminAccumulatorCreator);
         register_aggr_func!("scipystatsnormcdf", 2, ScipyStatsNormCdfAccumulatorCreator);
         register_aggr_func!("scipystatsnormpdf", 2, ScipyStatsNormPdfAccumulatorCreator);
+
+        #[cfg(feature = "geo")]
+        register_aggr_func!(
+            "json_encode_path",
+            3,
+            super::geo::encoding::JsonPathEncodeFunctionCreator
+        );
     }
 }

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

@@ -16,7 +16,10 @@ use std::cmp::Ordering;
 use std::sync::Arc;
 
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
-use common_query::error::{BadAccumulatorImplSnafu, CreateAccumulatorSnafu, Result};
+use common_query::error::{
+    BadAccumulatorImplSnafu, CreateAccumulatorSnafu, InvalidInputStateSnafu, Result,
+};
+use common_query::logical_plan::accumulator::AggrFuncTypeStore;
 use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
 use common_query::prelude::*;
 use datatypes::prelude::*;

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

@@ -16,7 +16,10 @@ use std::cmp::Ordering;
 use std::sync::Arc;
 
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
-use common_query::error::{BadAccumulatorImplSnafu, CreateAccumulatorSnafu, Result};
+use common_query::error::{
+    BadAccumulatorImplSnafu, CreateAccumulatorSnafu, InvalidInputStateSnafu, Result,
+};
+use common_query::logical_plan::accumulator::AggrFuncTypeStore;
 use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
 use common_query::prelude::*;
 use datatypes::prelude::*;

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

@@ -17,8 +17,10 @@ use std::sync::Arc;
 
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
 use common_query::error::{
-    CreateAccumulatorSnafu, DowncastVectorSnafu, FromScalarValueSnafu, Result,
+    CreateAccumulatorSnafu, DowncastVectorSnafu, FromScalarValueSnafu, InvalidInputStateSnafu,
+    Result,
 };
+use common_query::logical_plan::accumulator::AggrFuncTypeStore;
 use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
 use common_query::prelude::*;
 use datatypes::prelude::*;

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

@@ -17,8 +17,10 @@ use std::sync::Arc;
 
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
 use common_query::error::{
-    BadAccumulatorImplSnafu, CreateAccumulatorSnafu, DowncastVectorSnafu, Result,
+    BadAccumulatorImplSnafu, CreateAccumulatorSnafu, DowncastVectorSnafu, InvalidInputStateSnafu,
+    Result,
 };
+use common_query::logical_plan::accumulator::AggrFuncTypeStore;
 use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
 use common_query::prelude::*;
 use datatypes::prelude::*;

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

@@ -18,8 +18,9 @@ use std::sync::Arc;
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
 use common_query::error::{
     self, BadAccumulatorImplSnafu, CreateAccumulatorSnafu, DowncastVectorSnafu,
-    FromScalarValueSnafu, InvalidInputColSnafu, Result,
+    FromScalarValueSnafu, InvalidInputColSnafu, InvalidInputStateSnafu, Result,
 };
+use common_query::logical_plan::accumulator::AggrFuncTypeStore;
 use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
 use common_query::prelude::*;
 use datatypes::prelude::*;

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

@@ -17,8 +17,10 @@ use std::sync::Arc;
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
 use common_query::error::{
     self, BadAccumulatorImplSnafu, CreateAccumulatorSnafu, DowncastVectorSnafu,
-    FromScalarValueSnafu, GenerateFunctionSnafu, InvalidInputColSnafu, Result,
+    FromScalarValueSnafu, GenerateFunctionSnafu, InvalidInputColSnafu, InvalidInputStateSnafu,
+    Result,
 };
+use common_query::logical_plan::accumulator::AggrFuncTypeStore;
 use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
 use common_query::prelude::*;
 use datatypes::prelude::*;

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

@@ -17,8 +17,10 @@ use std::sync::Arc;
 use common_macro::{as_aggr_func_creator, AggrFuncTypeStore};
 use common_query::error::{
     self, BadAccumulatorImplSnafu, CreateAccumulatorSnafu, DowncastVectorSnafu,
-    FromScalarValueSnafu, GenerateFunctionSnafu, InvalidInputColSnafu, Result,
+    FromScalarValueSnafu, GenerateFunctionSnafu, InvalidInputColSnafu, InvalidInputStateSnafu,
+    Result,
 };
+use common_query::logical_plan::accumulator::AggrFuncTypeStore;
 use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
 use common_query::prelude::*;
 use datatypes::prelude::*;

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

@@ -14,18 +14,19 @@
 
 use std::fmt;
 
-use common_query::error::{InvalidFuncArgsSnafu, Result, UnsupportedInputDataTypeSnafu};
+use common_query::error::{ArrowComputeSnafu, IntoVectorSnafu, InvalidFuncArgsSnafu, Result};
 use common_query::prelude::Signature;
-use datatypes::data_type::DataType;
+use datatypes::arrow::compute::kernels::numeric;
 use datatypes::prelude::ConcreteDataType;
-use datatypes::value::ValueRef;
-use datatypes::vectors::VectorRef;
-use snafu::ensure;
+use datatypes::vectors::{Helper, VectorRef};
+use snafu::{ensure, ResultExt};
 
 use crate::function::{Function, FunctionContext};
 use crate::helper;
 
-/// A function adds an interval value to Timestamp, Date or DateTime, and return the result.
+/// A function adds an interval value to Timestamp, Date, and return the result.
+/// The implementation of datetime type is based on Date64 which is incorrect so this function
+/// doesn't support the datetime type.
 #[derive(Clone, Debug, Default)]
 pub struct DateAddFunction;
 
@@ -44,7 +45,6 @@ impl Function for DateAddFunction {
         helper::one_of_sigs2(
             vec![
                 ConcreteDataType::date_datatype(),
-                ConcreteDataType::datetime_datatype(),
                 ConcreteDataType::timestamp_second_datatype(),
                 ConcreteDataType::timestamp_millisecond_datatype(),
                 ConcreteDataType::timestamp_microsecond_datatype(),
@@ -69,64 +69,14 @@ impl Function for DateAddFunction {
             }
         );
 
-        let left = &columns[0];
-        let right = &columns[1];
+        let left = columns[0].to_arrow_array();
+        let right = columns[1].to_arrow_array();
 
-        let size = left.len();
-        let left_datatype = columns[0].data_type();
-        match left_datatype {
-            ConcreteDataType::Timestamp(_) => {
-                let mut result = left_datatype.create_mutable_vector(size);
-                for i in 0..size {
-                    let ts = left.get(i).as_timestamp();
-                    let interval = right.get(i).as_interval();
-
-                    let new_ts = match (ts, interval) {
-                        (Some(ts), Some(interval)) => ts.add_interval(interval),
-                        _ => ts,
-                    };
-
-                    result.push_value_ref(ValueRef::from(new_ts));
-                }
-
-                Ok(result.to_vector())
-            }
-            ConcreteDataType::Date(_) => {
-                let mut result = left_datatype.create_mutable_vector(size);
-                for i in 0..size {
-                    let date = left.get(i).as_date();
-                    let interval = right.get(i).as_interval();
-                    let new_date = match (date, interval) {
-                        (Some(date), Some(interval)) => date.add_interval(interval),
-                        _ => date,
-                    };
-
-                    result.push_value_ref(ValueRef::from(new_date));
-                }
-
-                Ok(result.to_vector())
-            }
-            ConcreteDataType::DateTime(_) => {
-                let mut result = left_datatype.create_mutable_vector(size);
-                for i in 0..size {
-                    let datetime = left.get(i).as_datetime();
-                    let interval = right.get(i).as_interval();
-                    let new_datetime = match (datetime, interval) {
-                        (Some(datetime), Some(interval)) => datetime.add_interval(interval),
-                        _ => datetime,
-                    };
-
-                    result.push_value_ref(ValueRef::from(new_datetime));
-                }
-
-                Ok(result.to_vector())
-            }
-            _ => UnsupportedInputDataTypeSnafu {
-                function: NAME,
-                datatypes: columns.iter().map(|c| c.data_type()).collect::<Vec<_>>(),
-            }
-            .fail(),
-        }
+        let result = numeric::add(&left, &right).context(ArrowComputeSnafu)?;
+        let arrow_type = result.data_type().clone();
+        Helper::try_into_vector(result).context(IntoVectorSnafu {
+            data_type: arrow_type,
+        })
     }
 }
 
@@ -144,8 +94,7 @@ mod tests {
     use datatypes::prelude::ConcreteDataType;
     use datatypes::value::Value;
     use datatypes::vectors::{
-        DateTimeVector, DateVector, IntervalDayTimeVector, IntervalYearMonthVector,
-        TimestampSecondVector,
+        DateVector, IntervalDayTimeVector, IntervalYearMonthVector, TimestampSecondVector,
     };
 
     use super::{DateAddFunction, *};
@@ -168,16 +117,15 @@ mod tests {
             ConcreteDataType::date_datatype(),
             f.return_type(&[ConcreteDataType::date_datatype()]).unwrap()
         );
-        assert_eq!(
-            ConcreteDataType::datetime_datatype(),
-            f.return_type(&[ConcreteDataType::datetime_datatype()])
-                .unwrap()
-        );
-        assert!(matches!(f.signature(),
+        assert!(
+            matches!(f.signature(),
                          Signature {
                              type_signature: TypeSignature::OneOf(sigs),
                              volatility: Volatility::Immutable
-                         } if  sigs.len() == 18));
+                         } if  sigs.len() == 15),
+            "{:?}",
+            f.signature()
+        );
     }
 
     #[test]
@@ -243,36 +191,4 @@ mod tests {
             }
         }
     }
-
-    #[test]
-    fn test_datetime_date_add() {
-        let f = DateAddFunction;
-
-        let dates = vec![Some(123), None, Some(42), None];
-        // Intervals in months
-        let intervals = vec![1, 2, 3, 1];
-        let results = [Some(2678400123), None, Some(7776000042), None];
-
-        let date_vector = DateTimeVector::from(dates.clone());
-        let interval_vector = IntervalYearMonthVector::from_vec(intervals);
-        let args: Vec<VectorRef> = vec![Arc::new(date_vector), Arc::new(interval_vector)];
-        let vector = f.eval(FunctionContext::default(), &args).unwrap();
-
-        assert_eq!(4, vector.len());
-        for (i, _t) in dates.iter().enumerate() {
-            let v = vector.get(i);
-            let result = results.get(i).unwrap();
-
-            if result.is_none() {
-                assert_eq!(Value::Null, v);
-                continue;
-            }
-            match v {
-                Value::DateTime(date) => {
-                    assert_eq!(date.val(), result.unwrap());
-                }
-                _ => unreachable!(),
-            }
-        }
-    }
 }

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

@@ -14,18 +14,19 @@
 
 use std::fmt;
 
-use common_query::error::{InvalidFuncArgsSnafu, Result, UnsupportedInputDataTypeSnafu};
+use common_query::error::{ArrowComputeSnafu, IntoVectorSnafu, InvalidFuncArgsSnafu, Result};
 use common_query::prelude::Signature;
-use datatypes::data_type::DataType;
+use datatypes::arrow::compute::kernels::numeric;
 use datatypes::prelude::ConcreteDataType;
-use datatypes::value::ValueRef;
-use datatypes::vectors::VectorRef;
-use snafu::ensure;
+use datatypes::vectors::{Helper, VectorRef};
+use snafu::{ensure, ResultExt};
 
 use crate::function::{Function, FunctionContext};
 use crate::helper;
 
-/// A function subtracts an interval value to Timestamp, Date or DateTime, and return the result.
+/// A function subtracts an interval value to Timestamp, Date, and return the result.
+/// The implementation of datetime type is based on Date64 which is incorrect so this function
+/// doesn't support the datetime type.
 #[derive(Clone, Debug, Default)]
 pub struct DateSubFunction;
 
@@ -44,7 +45,6 @@ impl Function for DateSubFunction {
         helper::one_of_sigs2(
             vec![
                 ConcreteDataType::date_datatype(),
-                ConcreteDataType::datetime_datatype(),
                 ConcreteDataType::timestamp_second_datatype(),
                 ConcreteDataType::timestamp_millisecond_datatype(),
                 ConcreteDataType::timestamp_microsecond_datatype(),
@@ -69,65 +69,14 @@ impl Function for DateSubFunction {
             }
         );
 
-        let left = &columns[0];
-        let right = &columns[1];
+        let left = columns[0].to_arrow_array();
+        let right = columns[1].to_arrow_array();
 
-        let size = left.len();
-        let left_datatype = columns[0].data_type();
-
-        match left_datatype {
-            ConcreteDataType::Timestamp(_) => {
-                let mut result = left_datatype.create_mutable_vector(size);
-                for i in 0..size {
-                    let ts = left.get(i).as_timestamp();
-                    let interval = right.get(i).as_interval();
-
-                    let new_ts = match (ts, interval) {
-                        (Some(ts), Some(interval)) => ts.sub_interval(interval),
-                        _ => ts,
-                    };
-
-                    result.push_value_ref(ValueRef::from(new_ts));
-                }
-
-                Ok(result.to_vector())
-            }
-            ConcreteDataType::Date(_) => {
-                let mut result = left_datatype.create_mutable_vector(size);
-                for i in 0..size {
-                    let date = left.get(i).as_date();
-                    let interval = right.get(i).as_interval();
-                    let new_date = match (date, interval) {
-                        (Some(date), Some(interval)) => date.sub_interval(interval),
-                        _ => date,
-                    };
-
-                    result.push_value_ref(ValueRef::from(new_date));
-                }
-
-                Ok(result.to_vector())
-            }
-            ConcreteDataType::DateTime(_) => {
-                let mut result = left_datatype.create_mutable_vector(size);
-                for i in 0..size {
-                    let datetime = left.get(i).as_datetime();
-                    let interval = right.get(i).as_interval();
-                    let new_datetime = match (datetime, interval) {
-                        (Some(datetime), Some(interval)) => datetime.sub_interval(interval),
-                        _ => datetime,
-                    };
-
-                    result.push_value_ref(ValueRef::from(new_datetime));
-                }
-
-                Ok(result.to_vector())
-            }
-            _ => UnsupportedInputDataTypeSnafu {
-                function: NAME,
-                datatypes: columns.iter().map(|c| c.data_type()).collect::<Vec<_>>(),
-            }
-            .fail(),
-        }
+        let result = numeric::sub(&left, &right).context(ArrowComputeSnafu)?;
+        let arrow_type = result.data_type().clone();
+        Helper::try_into_vector(result).context(IntoVectorSnafu {
+            data_type: arrow_type,
+        })
     }
 }
 
@@ -145,8 +94,7 @@ mod tests {
     use datatypes::prelude::ConcreteDataType;
     use datatypes::value::Value;
     use datatypes::vectors::{
-        DateTimeVector, DateVector, IntervalDayTimeVector, IntervalYearMonthVector,
-        TimestampSecondVector,
+        DateVector, IntervalDayTimeVector, IntervalYearMonthVector, TimestampSecondVector,
     };
 
     use super::{DateSubFunction, *};
@@ -174,11 +122,15 @@ mod tests {
             f.return_type(&[ConcreteDataType::datetime_datatype()])
                 .unwrap()
         );
-        assert!(matches!(f.signature(),
+        assert!(
+            matches!(f.signature(),
                          Signature {
                              type_signature: TypeSignature::OneOf(sigs),
                              volatility: Volatility::Immutable
-                         } if  sigs.len() == 18));
+                         } if  sigs.len() == 15),
+            "{:?}",
+            f.signature()
+        );
     }
 
     #[test]
@@ -250,42 +202,4 @@ mod tests {
             }
         }
     }
-
-    #[test]
-    fn test_datetime_date_sub() {
-        let f = DateSubFunction;
-        let millis_per_month = 3600 * 24 * 30 * 1000;
-
-        let dates = vec![
-            Some(123 * millis_per_month),
-            None,
-            Some(42 * millis_per_month),
-            None,
-        ];
-        // Intervals in months
-        let intervals = vec![1, 2, 3, 1];
-        let results = [Some(316137600000), None, Some(100915200000), None];
-
-        let date_vector = DateTimeVector::from(dates.clone());
-        let interval_vector = IntervalYearMonthVector::from_vec(intervals);
-        let args: Vec<VectorRef> = vec![Arc::new(date_vector), Arc::new(interval_vector)];
-        let vector = f.eval(FunctionContext::default(), &args).unwrap();
-
-        assert_eq!(4, vector.len());
-        for (i, _t) in dates.iter().enumerate() {
-            let v = vector.get(i);
-            let result = results.get(i).unwrap();
-
-            if result.is_none() {
-                assert_eq!(Value::Null, v);
-                continue;
-            }
-            match v {
-                Value::DateTime(date) => {
-                    assert_eq!(date.val(), result.unwrap());
-                }
-                _ => unreachable!(),
-            }
-        }
-    }
 }

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

@@ -13,10 +13,11 @@
 // limitations under the License.
 
 use std::sync::Arc;
+pub(crate) mod encoding;
 mod geohash;
 mod h3;
-
-use geohash::{GeohashFunction, GeohashNeighboursFunction};
+mod helpers;
+mod s2;
 
 use crate::function_registry::FunctionRegistry;
 
@@ -25,20 +26,39 @@ pub(crate) struct GeoFunctions;
 impl GeoFunctions {
     pub fn register(registry: &FunctionRegistry) {
         // geohash
-        registry.register(Arc::new(GeohashFunction));
-        registry.register(Arc::new(GeohashNeighboursFunction));
-        // h3 family
+        registry.register(Arc::new(geohash::GeohashFunction));
+        registry.register(Arc::new(geohash::GeohashNeighboursFunction));
+
+        // h3 index
         registry.register(Arc::new(h3::H3LatLngToCell));
         registry.register(Arc::new(h3::H3LatLngToCellString));
+
+        // h3 index inspection
         registry.register(Arc::new(h3::H3CellBase));
-        registry.register(Arc::new(h3::H3CellCenterChild));
-        registry.register(Arc::new(h3::H3CellCenterLat));
-        registry.register(Arc::new(h3::H3CellCenterLng));
         registry.register(Arc::new(h3::H3CellIsPentagon));
-        registry.register(Arc::new(h3::H3CellParent));
-        registry.register(Arc::new(h3::H3CellResolution));
-        registry.register(Arc::new(h3::H3CellToString));
-        registry.register(Arc::new(h3::H3IsNeighbour));
         registry.register(Arc::new(h3::H3StringToCell));
+        registry.register(Arc::new(h3::H3CellToString));
+        registry.register(Arc::new(h3::H3CellCenterLatLng));
+        registry.register(Arc::new(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));
+
+        // 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));
+
+        // 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));
     }
 }

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

@@ -0,0 +1,223 @@
+// 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 common_error::ext::{BoxedError, PlainError};
+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::prelude::AccumulatorCreatorFunction;
+use common_time::Timestamp;
+use datatypes::prelude::ConcreteDataType;
+use datatypes::value::{ListValue, Value};
+use datatypes::vectors::VectorRef;
+use snafu::{ensure, ResultExt};
+
+use super::helpers::{ensure_columns_len, ensure_columns_n};
+
+/// Accumulator of lat, lng, timestamp tuples
+#[derive(Debug)]
+pub struct JsonPathAccumulator {
+    timestamp_type: ConcreteDataType,
+    lat: Vec<Option<f64>>,
+    lng: Vec<Option<f64>>,
+    timestamp: Vec<Option<Timestamp>>,
+}
+
+impl JsonPathAccumulator {
+    fn new(timestamp_type: ConcreteDataType) -> Self {
+        Self {
+            lat: Vec::default(),
+            lng: Vec::default(),
+            timestamp: Vec::default(),
+            timestamp_type,
+        }
+    }
+}
+
+impl Accumulator for JsonPathAccumulator {
+    fn state(&self) -> Result<Vec<Value>> {
+        Ok(vec![
+            Value::List(ListValue::new(
+                self.lat.iter().map(|i| Value::from(*i)).collect(),
+                ConcreteDataType::float64_datatype(),
+            )),
+            Value::List(ListValue::new(
+                self.lng.iter().map(|i| Value::from(*i)).collect(),
+                ConcreteDataType::float64_datatype(),
+            )),
+            Value::List(ListValue::new(
+                self.timestamp.iter().map(|i| Value::from(*i)).collect(),
+                self.timestamp_type.clone(),
+            )),
+        ])
+    }
+
+    fn update_batch(&mut self, columns: &[VectorRef]) -> Result<()> {
+        // update batch as in datafusion just provides the accumulator original
+        //  input.
+        //
+        // columns is vec of [`lat`, `lng`, `timestamp`]
+        // where
+        // - `lat` is a vector of `Value::Float64` or similar type. Each item in
+        //  the vector is a row in given dataset.
+        // - so on so forth for `lng` and `timestamp`
+        ensure_columns_n!(columns, 3);
+
+        let lat = &columns[0];
+        let lng = &columns[1];
+        let ts = &columns[2];
+
+        let size = lat.len();
+
+        for idx in 0..size {
+            self.lat.push(lat.get(idx).as_f64_lossy());
+            self.lng.push(lng.get(idx).as_f64_lossy());
+            self.timestamp.push(ts.get(idx).as_timestamp());
+        }
+
+        Ok(())
+    }
+
+    fn merge_batch(&mut self, states: &[VectorRef]) -> Result<()> {
+        // merge batch as in datafusion gives state accumulated from the data
+        //  returned from child accumulators' state() call
+        // In our particular implementation, the data structure is like
+        //
+        // states is vec of [`lat`, `lng`, `timestamp`]
+        // where
+        // - `lat` is a vector of `Value::List`. Each item in the list is all
+        //  coordinates from a child accumulator.
+        // - so on so forth for `lng` and `timestamp`
+
+        ensure_columns_n!(states, 3);
+
+        let lat_lists = &states[0];
+        let lng_lists = &states[1];
+        let ts_lists = &states[2];
+
+        let len = lat_lists.len();
+
+        for idx in 0..len {
+            if let Some(lat_list) = lat_lists
+                .get(idx)
+                .as_list()
+                .map_err(BoxedError::new)
+                .context(error::ExecuteSnafu)?
+            {
+                for v in lat_list.items() {
+                    self.lat.push(v.as_f64_lossy());
+                }
+            }
+
+            if let Some(lng_list) = lng_lists
+                .get(idx)
+                .as_list()
+                .map_err(BoxedError::new)
+                .context(error::ExecuteSnafu)?
+            {
+                for v in lng_list.items() {
+                    self.lng.push(v.as_f64_lossy());
+                }
+            }
+
+            if let Some(ts_list) = ts_lists
+                .get(idx)
+                .as_list()
+                .map_err(BoxedError::new)
+                .context(error::ExecuteSnafu)?
+            {
+                for v in ts_list.items() {
+                    self.timestamp.push(v.as_timestamp());
+                }
+            }
+        }
+
+        Ok(())
+    }
+
+    fn evaluate(&self) -> Result<Value> {
+        let mut work_vec: Vec<(&Option<f64>, &Option<f64>, &Option<Timestamp>)> = self
+            .lat
+            .iter()
+            .zip(self.lng.iter())
+            .zip(self.timestamp.iter())
+            .map(|((a, b), c)| (a, b, c))
+            .collect();
+
+        // sort by timestamp, we treat null timestamp as 0
+        work_vec.sort_unstable_by_key(|tuple| tuple.2.unwrap_or_else(|| Timestamp::new_second(0)));
+
+        let result = serde_json::to_string(
+            &work_vec
+                .into_iter()
+                // note that we transform to lng,lat for geojson compatibility
+                .map(|(lat, lng, _)| vec![lng, lat])
+                .collect::<Vec<Vec<&Option<f64>>>>(),
+        )
+        .map_err(|e| {
+            BoxedError::new(PlainError::new(
+                format!("Serialization failure: {}", e),
+                StatusCode::EngineExecuteQuery,
+            ))
+        })
+        .context(error::ExecuteSnafu)?;
+
+        Ok(Value::String(result.into()))
+    }
+}
+
+/// This function accept rows of lat, lng and timestamp, sort with timestamp and
+/// encoding them into a geojson-like path.
+///
+/// Example:
+///
+/// ```sql
+/// SELECT json_encode_path(lat, lon, timestamp) FROM table [group by ...];
+/// ```
+///
+#[as_aggr_func_creator]
+#[derive(Debug, Default, AggrFuncTypeStore)]
+pub struct JsonPathEncodeFunctionCreator {}
+
+impl AggregateFunctionCreator for JsonPathEncodeFunctionCreator {
+    fn creator(&self) -> AccumulatorCreatorFunction {
+        let creator: AccumulatorCreatorFunction = Arc::new(move |types: &[ConcreteDataType]| {
+            let ts_type = types[2].clone();
+            Ok(Box::new(JsonPathAccumulator::new(ts_type)))
+        });
+
+        creator
+    }
+
+    fn output_type(&self) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::string_datatype())
+    }
+
+    fn state_types(&self) -> Result<Vec<ConcreteDataType>> {
+        let input_types = self.input_types()?;
+        ensure!(input_types.len() == 3, InvalidInputStateSnafu);
+
+        let timestamp_type = input_types[2].clone();
+
+        Ok(vec![
+            ConcreteDataType::list_datatype(ConcreteDataType::float64_datatype()),
+            ConcreteDataType::list_datatype(ConcreteDataType::float64_datatype()),
+            ConcreteDataType::list_datatype(timestamp_type),
+        ])
+    }
+}

src/common/function/src/scalars/geo/h3.rs

@@ -16,22 +16,75 @@ use std::str::FromStr;
 
 use common_error::ext::{BoxedError, PlainError};
 use common_error::status_code::StatusCode;
-use common_query::error::{self, InvalidFuncArgsSnafu, Result};
+use common_query::error::{self, Result};
 use common_query::prelude::{Signature, TypeSignature};
 use datafusion::logical_expr::Volatility;
 use datatypes::prelude::ConcreteDataType;
-use datatypes::scalars::ScalarVectorBuilder;
-use datatypes::value::Value;
+use datatypes::scalars::{Scalar, ScalarVectorBuilder};
+use datatypes::value::{ListValue, Value};
 use datatypes::vectors::{
-    BooleanVectorBuilder, Float64VectorBuilder, MutableVector, StringVectorBuilder,
-    UInt64VectorBuilder, UInt8VectorBuilder, VectorRef,
+    BooleanVectorBuilder, Int32VectorBuilder, ListVectorBuilder, MutableVector,
+    StringVectorBuilder, UInt64VectorBuilder, UInt8VectorBuilder, VectorRef,
 };
 use derive_more::Display;
 use h3o::{CellIndex, LatLng, Resolution};
-use snafu::{ensure, ResultExt};
+use once_cell::sync::Lazy;
+use snafu::ResultExt;
 
+use super::helpers::{ensure_and_coerce, ensure_columns_len, ensure_columns_n};
 use crate::function::{Function, FunctionContext};
 
+static CELL_TYPES: Lazy<Vec<ConcreteDataType>> = Lazy::new(|| {
+    vec![
+        ConcreteDataType::int64_datatype(),
+        ConcreteDataType::uint64_datatype(),
+    ]
+});
+
+static COORDINATE_TYPES: Lazy<Vec<ConcreteDataType>> = Lazy::new(|| {
+    vec![
+        ConcreteDataType::float32_datatype(),
+        ConcreteDataType::float64_datatype(),
+    ]
+});
+static RESOLUTION_TYPES: Lazy<Vec<ConcreteDataType>> = Lazy::new(|| {
+    vec![
+        ConcreteDataType::int8_datatype(),
+        ConcreteDataType::int16_datatype(),
+        ConcreteDataType::int32_datatype(),
+        ConcreteDataType::int64_datatype(),
+        ConcreteDataType::uint8_datatype(),
+        ConcreteDataType::uint16_datatype(),
+        ConcreteDataType::uint32_datatype(),
+        ConcreteDataType::uint64_datatype(),
+    ]
+});
+static DISTANCE_TYPES: Lazy<Vec<ConcreteDataType>> = Lazy::new(|| {
+    vec![
+        ConcreteDataType::int8_datatype(),
+        ConcreteDataType::int16_datatype(),
+        ConcreteDataType::int32_datatype(),
+        ConcreteDataType::int64_datatype(),
+        ConcreteDataType::uint8_datatype(),
+        ConcreteDataType::uint16_datatype(),
+        ConcreteDataType::uint32_datatype(),
+        ConcreteDataType::uint64_datatype(),
+    ]
+});
+
+static POSITION_TYPES: Lazy<Vec<ConcreteDataType>> = Lazy::new(|| {
+    vec![
+        ConcreteDataType::int8_datatype(),
+        ConcreteDataType::int16_datatype(),
+        ConcreteDataType::int32_datatype(),
+        ConcreteDataType::int64_datatype(),
+        ConcreteDataType::uint8_datatype(),
+        ConcreteDataType::uint16_datatype(),
+        ConcreteDataType::uint32_datatype(),
+        ConcreteDataType::uint64_datatype(),
+    ]
+});
+
 /// Function that returns [h3] encoding cellid for a given geospatial coordinate.
 ///
 /// [h3]: https://h3geo.org/
@@ -50,20 +103,8 @@ impl Function for H3LatLngToCell {
 
     fn signature(&self) -> Signature {
         let mut signatures = Vec::new();
-        for coord_type in &[
-            ConcreteDataType::float32_datatype(),
-            ConcreteDataType::float64_datatype(),
-        ] {
-            for resolution_type in &[
-                ConcreteDataType::int8_datatype(),
-                ConcreteDataType::int16_datatype(),
-                ConcreteDataType::int32_datatype(),
-                ConcreteDataType::int64_datatype(),
-                ConcreteDataType::uint8_datatype(),
-                ConcreteDataType::uint16_datatype(),
-                ConcreteDataType::uint32_datatype(),
-                ConcreteDataType::uint64_datatype(),
-            ] {
+        for coord_type in COORDINATE_TYPES.as_slice() {
+            for resolution_type in RESOLUTION_TYPES.as_slice() {
                 signatures.push(TypeSignature::Exact(vec![
                     // latitude
                     coord_type.clone(),
@@ -78,15 +119,7 @@ impl Function for H3LatLngToCell {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 3,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 3, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 3);
 
         let lat_vec = &columns[0];
         let lon_vec = &columns[1];
@@ -142,20 +175,8 @@ impl Function for H3LatLngToCellString {
 
     fn signature(&self) -> Signature {
         let mut signatures = Vec::new();
-        for coord_type in &[
-            ConcreteDataType::float32_datatype(),
-            ConcreteDataType::float64_datatype(),
-        ] {
-            for resolution_type in &[
-                ConcreteDataType::int8_datatype(),
-                ConcreteDataType::int16_datatype(),
-                ConcreteDataType::int32_datatype(),
-                ConcreteDataType::int64_datatype(),
-                ConcreteDataType::uint8_datatype(),
-                ConcreteDataType::uint16_datatype(),
-                ConcreteDataType::uint32_datatype(),
-                ConcreteDataType::uint64_datatype(),
-            ] {
+        for coord_type in COORDINATE_TYPES.as_slice() {
+            for resolution_type in RESOLUTION_TYPES.as_slice() {
                 signatures.push(TypeSignature::Exact(vec![
                     // latitude
                     coord_type.clone(),
@@ -170,15 +191,7 @@ impl Function for H3LatLngToCellString {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 3,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 3, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 3);
 
         let lat_vec = &columns[0];
         let lon_vec = &columns[1];
@@ -234,15 +247,7 @@ impl Function for H3CellToString {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 1,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 1, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 1);
 
         let cell_vec = &columns[0];
         let size = cell_vec.len();
@@ -280,15 +285,7 @@ impl Function for H3StringToCell {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 1,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 1, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 1);
 
         let string_vec = &columns[0];
         let size = string_vec.len();
@@ -319,18 +316,20 @@ impl Function for H3StringToCell {
     }
 }
 
-/// Function that returns centroid latitude of given cell id
+/// Function that returns centroid latitude and longitude of given cell id
 #[derive(Clone, Debug, Default, Display)]
 #[display("{}", self.name())]
-pub struct H3CellCenterLat;
+pub struct H3CellCenterLatLng;
 
-impl Function for H3CellCenterLat {
+impl Function for H3CellCenterLatLng {
     fn name(&self) -> &str {
-        "h3_cell_center_lat"
+        "h3_cell_center_latlng"
     }
 
     fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
-        Ok(ConcreteDataType::float64_datatype())
+        Ok(ConcreteDataType::list_datatype(
+            ConcreteDataType::float64_datatype(),
+        ))
     }
 
     fn signature(&self) -> Signature {
@@ -338,69 +337,26 @@ impl Function for H3CellCenterLat {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 1,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 1, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 1);
 
         let cell_vec = &columns[0];
         let size = cell_vec.len();
-        let mut results = Float64VectorBuilder::with_capacity(size);
+        let mut results =
+            ListVectorBuilder::with_type_capacity(ConcreteDataType::float64_datatype(), size);
 
         for i in 0..size {
             let cell = cell_from_value(cell_vec.get(i))?;
-            let lat = cell.map(|cell| LatLng::from(cell).lat());
+            let latlng = cell.map(LatLng::from);
 
-            results.push(lat);
-        }
-
-        Ok(results.to_vector())
-    }
-}
-
-/// Function that returns centroid longitude of given cell id
-#[derive(Clone, Debug, Default, Display)]
-#[display("{}", self.name())]
-pub struct H3CellCenterLng;
-
-impl Function for H3CellCenterLng {
-    fn name(&self) -> &str {
-        "h3_cell_center_lng"
-    }
-
-    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
-        Ok(ConcreteDataType::float64_datatype())
-    }
-
-    fn signature(&self) -> Signature {
-        signature_of_cell()
-    }
-
-    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 1,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 1, provided : {}",
-                    columns.len()
-                ),
+            if let Some(latlng) = latlng {
+                let result = ListValue::new(
+                    vec![latlng.lat().into(), latlng.lng().into()],
+                    ConcreteDataType::float64_datatype(),
+                );
+                results.push(Some(result.as_scalar_ref()));
+            } else {
+                results.push(None);
             }
-        );
-
-        let cell_vec = &columns[0];
-        let size = cell_vec.len();
-        let mut results = Float64VectorBuilder::with_capacity(size);
-
-        for i in 0..size {
-            let cell = cell_from_value(cell_vec.get(i))?;
-            let lat = cell.map(|cell| LatLng::from(cell).lng());
-
-            results.push(lat);
         }
 
         Ok(results.to_vector())
@@ -426,15 +382,7 @@ impl Function for H3CellResolution {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 1,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 1, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 1);
 
         let cell_vec = &columns[0];
         let size = cell_vec.len();
@@ -470,15 +418,7 @@ impl Function for H3CellBase {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 1,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 1, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 1);
 
         let cell_vec = &columns[0];
         let size = cell_vec.len();
@@ -514,15 +454,7 @@ impl Function for H3CellIsPentagon {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 1,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 1, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 1);
 
         let cell_vec = &columns[0];
         let size = cell_vec.len();
@@ -558,15 +490,7 @@ impl Function for H3CellCenterChild {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 2,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 2, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 2);
 
         let cell_vec = &columns[0];
         let res_vec = &columns[1];
@@ -606,15 +530,7 @@ impl Function for H3CellParent {
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 2,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 2, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 2);
 
         let cell_vec = &columns[0];
         let res_vec = &columns[1];
@@ -633,48 +549,323 @@ impl Function for H3CellParent {
     }
 }
 
-/// Function that checks if two cells are neighbour
+/// Function that returns children cell list
 #[derive(Clone, Debug, Default, Display)]
 #[display("{}", self.name())]
-pub struct H3IsNeighbour;
+pub struct H3CellToChildren;
 
-impl Function for H3IsNeighbour {
+impl Function for H3CellToChildren {
     fn name(&self) -> &str {
-        "h3_is_neighbour"
+        "h3_cell_to_children"
     }
 
     fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
-        Ok(ConcreteDataType::boolean_datatype())
+        Ok(ConcreteDataType::list_datatype(
+            ConcreteDataType::uint64_datatype(),
+        ))
     }
 
     fn signature(&self) -> Signature {
-        signature_of_double_cell()
+        signature_of_cell_and_resolution()
     }
 
     fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
-        ensure!(
-            columns.len() == 2,
-            InvalidFuncArgsSnafu {
-                err_msg: format!(
-                    "The length of the args is not correct, expect 2, provided : {}",
-                    columns.len()
-                ),
-            }
-        );
+        ensure_columns_n!(columns, 2);
 
         let cell_vec = &columns[0];
-        let cell2_vec = &columns[1];
+        let res_vec = &columns[1];
         let size = cell_vec.len();
-        let mut results = BooleanVectorBuilder::with_capacity(size);
+        let mut results =
+            ListVectorBuilder::with_type_capacity(ConcreteDataType::uint64_datatype(), size);
+
+        for i in 0..size {
+            let cell = cell_from_value(cell_vec.get(i))?;
+            let res = value_to_resolution(res_vec.get(i))?;
+            let result = cell.map(|cell| {
+                let children: Vec<Value> = cell
+                    .children(res)
+                    .map(|child| Value::from(u64::from(child)))
+                    .collect();
+                ListValue::new(children, ConcreteDataType::uint64_datatype())
+            });
+
+            if let Some(list_value) = result {
+                results.push(Some(list_value.as_scalar_ref()));
+            } else {
+                results.push(None);
+            }
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+/// Function that returns children cell count
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct H3CellToChildrenSize;
+
+impl Function for H3CellToChildrenSize {
+    fn name(&self) -> &str {
+        "h3_cell_to_children_size"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::uint64_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        signature_of_cell_and_resolution()
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 2);
+
+        let cell_vec = &columns[0];
+        let res_vec = &columns[1];
+        let size = cell_vec.len();
+        let mut results = UInt64VectorBuilder::with_capacity(size);
+
+        for i in 0..size {
+            let cell = cell_from_value(cell_vec.get(i))?;
+            let res = value_to_resolution(res_vec.get(i))?;
+            let result = cell.map(|cell| cell.children_count(res));
+            results.push(result);
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+/// Function that returns the cell position if its parent at given resolution
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct H3CellToChildPos;
+
+impl Function for H3CellToChildPos {
+    fn name(&self) -> &str {
+        "h3_cell_to_child_pos"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::uint64_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        signature_of_cell_and_resolution()
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 2);
+
+        let cell_vec = &columns[0];
+        let res_vec = &columns[1];
+        let size = cell_vec.len();
+        let mut results = UInt64VectorBuilder::with_capacity(size);
+
+        for i in 0..size {
+            let cell = cell_from_value(cell_vec.get(i))?;
+            let res = value_to_resolution(res_vec.get(i))?;
+            let result = cell.and_then(|cell| cell.child_position(res));
+            results.push(result);
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+/// Function that returns the cell at given position of the parent at given resolution
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct H3ChildPosToCell;
+
+impl Function for H3ChildPosToCell {
+    fn name(&self) -> &str {
+        "h3_child_pos_to_cell"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::uint64_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        let mut signatures =
+            Vec::with_capacity(POSITION_TYPES.len() * CELL_TYPES.len() * RESOLUTION_TYPES.len());
+        for position_type in POSITION_TYPES.as_slice() {
+            for cell_type in CELL_TYPES.as_slice() {
+                for resolution_type in RESOLUTION_TYPES.as_slice() {
+                    signatures.push(TypeSignature::Exact(vec![
+                        position_type.clone(),
+                        cell_type.clone(),
+                        resolution_type.clone(),
+                    ]));
+                }
+            }
+        }
+        Signature::one_of(signatures, Volatility::Stable)
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 3);
+
+        let pos_vec = &columns[0];
+        let cell_vec = &columns[1];
+        let res_vec = &columns[2];
+        let size = cell_vec.len();
+        let mut results = UInt64VectorBuilder::with_capacity(size);
+
+        for i in 0..size {
+            let cell = cell_from_value(cell_vec.get(i))?;
+            let pos = value_to_position(pos_vec.get(i))?;
+            let res = value_to_resolution(res_vec.get(i))?;
+            let result = cell.and_then(|cell| cell.child_at(pos, res).map(u64::from));
+            results.push(result);
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+/// Function that returns cells with k distances of given cell
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct H3GridDisk;
+
+impl Function for H3GridDisk {
+    fn name(&self) -> &str {
+        "h3_grid_disk"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::list_datatype(
+            ConcreteDataType::uint64_datatype(),
+        ))
+    }
+
+    fn signature(&self) -> Signature {
+        signature_of_cell_and_distance()
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 2);
+
+        let cell_vec = &columns[0];
+        let k_vec = &columns[1];
+        let size = cell_vec.len();
+        let mut results =
+            ListVectorBuilder::with_type_capacity(ConcreteDataType::uint64_datatype(), size);
+
+        for i in 0..size {
+            let cell = cell_from_value(cell_vec.get(i))?;
+            let k = value_to_distance(k_vec.get(i))?;
+
+            let result = cell.map(|cell| {
+                let children: Vec<Value> = cell
+                    .grid_disk::<Vec<_>>(k)
+                    .into_iter()
+                    .map(|child| Value::from(u64::from(child)))
+                    .collect();
+                ListValue::new(children, ConcreteDataType::uint64_datatype())
+            });
+
+            if let Some(list_value) = result {
+                results.push(Some(list_value.as_scalar_ref()));
+            } else {
+                results.push(None);
+            }
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+/// Function that returns all cells within k distances of given cell
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct H3GridDiskDistances;
+
+impl Function for H3GridDiskDistances {
+    fn name(&self) -> &str {
+        "h3_grid_disk_distances"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::list_datatype(
+            ConcreteDataType::uint64_datatype(),
+        ))
+    }
+
+    fn signature(&self) -> Signature {
+        signature_of_cell_and_distance()
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 2);
+
+        let cell_vec = &columns[0];
+        let k_vec = &columns[1];
+        let size = cell_vec.len();
+        let mut results =
+            ListVectorBuilder::with_type_capacity(ConcreteDataType::uint64_datatype(), size);
+
+        for i in 0..size {
+            let cell = cell_from_value(cell_vec.get(i))?;
+            let k = value_to_distance(k_vec.get(i))?;
+
+            let result = cell.map(|cell| {
+                let children: Vec<Value> = cell
+                    .grid_disk_distances::<Vec<_>>(k)
+                    .into_iter()
+                    .map(|(child, _distance)| Value::from(u64::from(child)))
+                    .collect();
+                ListValue::new(children, ConcreteDataType::uint64_datatype())
+            });
+
+            if let Some(list_value) = result {
+                results.push(Some(list_value.as_scalar_ref()));
+            } else {
+                results.push(None);
+            }
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+/// Function that returns distance between two cells
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct H3GridDistance;
+
+impl Function for H3GridDistance {
+    fn name(&self) -> &str {
+        "h3_grid_distance"
+    }
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::int32_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        signature_of_double_cells()
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 2);
+
+        let cell_this_vec = &columns[0];
+        let cell_that_vec = &columns[1];
+        let size = cell_this_vec.len();
+
+        let mut results = Int32VectorBuilder::with_capacity(size);
 
         for i in 0..size {
             let result = match (
-                cell_from_value(cell_vec.get(i))?,
-                cell_from_value(cell2_vec.get(i))?,
+                cell_from_value(cell_this_vec.get(i))?,
+                cell_from_value(cell_that_vec.get(i))?,
             ) {
                 (Some(cell_this), Some(cell_that)) => {
-                    let is_neighbour = cell_this
-                        .is_neighbor_with(cell_that)
+                    let dist = cell_this
+                        .grid_distance(cell_that)
                         .map_err(|e| {
                             BoxedError::new(PlainError::new(
                                 format!("H3 error: {}", e),
@@ -682,7 +873,7 @@ impl Function for H3IsNeighbour {
                             ))
                         })
                         .context(error::ExecuteSnafu)?;
-                    Some(is_neighbour)
+                    Some(dist)
                 }
                 _ => None,
             };
@@ -694,6 +885,73 @@ impl Function for H3IsNeighbour {
     }
 }
 
+/// Function that returns path cells between two cells
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct H3GridPathCells;
+
+impl Function for H3GridPathCells {
+    fn name(&self) -> &str {
+        "h3_grid_path_cells"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::list_datatype(
+            ConcreteDataType::uint64_datatype(),
+        ))
+    }
+
+    fn signature(&self) -> Signature {
+        signature_of_double_cells()
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 2);
+
+        let cell_this_vec = &columns[0];
+        let cell_that_vec = &columns[1];
+        let size = cell_this_vec.len();
+        let mut results =
+            ListVectorBuilder::with_type_capacity(ConcreteDataType::uint64_datatype(), size);
+
+        for i in 0..size {
+            let result = match (
+                cell_from_value(cell_this_vec.get(i))?,
+                cell_from_value(cell_that_vec.get(i))?,
+            ) {
+                (Some(cell_this), Some(cell_that)) => {
+                    let cells = cell_this
+                        .grid_path_cells(cell_that)
+                        .and_then(|x| x.collect::<std::result::Result<Vec<CellIndex>, _>>())
+                        .map_err(|e| {
+                            BoxedError::new(PlainError::new(
+                                format!("H3 error: {}", e),
+                                StatusCode::EngineExecuteQuery,
+                            ))
+                        })
+                        .context(error::ExecuteSnafu)?;
+                    Some(ListValue::new(
+                        cells
+                            .into_iter()
+                            .map(|c| Value::from(u64::from(c)))
+                            .collect(),
+                        ConcreteDataType::uint64_datatype(),
+                    ))
+                }
+                _ => None,
+            };
+
+            if let Some(list_value) = result {
+                results.push(Some(list_value.as_scalar_ref()));
+            } else {
+                results.push(None);
+            }
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
 fn value_to_resolution(v: Value) -> Result<Resolution> {
     let r = match v {
         Value::Int8(v) => v as u8,
@@ -716,26 +974,47 @@ fn value_to_resolution(v: Value) -> Result<Resolution> {
         .context(error::ExecuteSnafu)
 }
 
+fn value_to_position(v: Value) -> Result<u64> {
+    match v {
+        Value::Int8(v) => ensure_and_coerce!(v >= 0, v as u64),
+        Value::Int16(v) => ensure_and_coerce!(v >= 0, v as u64),
+        Value::Int32(v) => ensure_and_coerce!(v >= 0, v as u64),
+        Value::Int64(v) => ensure_and_coerce!(v >= 0, v as u64),
+        Value::UInt8(v) => Ok(v as u64),
+        Value::UInt16(v) => Ok(v as u64),
+        Value::UInt32(v) => Ok(v as u64),
+        Value::UInt64(v) => Ok(v),
+        _ => unreachable!(),
+    }
+}
+
+fn value_to_distance(v: Value) -> Result<u32> {
+    match v {
+        Value::Int8(v) => ensure_and_coerce!(v >= 0, v as u32),
+        Value::Int16(v) => ensure_and_coerce!(v >= 0, v as u32),
+        Value::Int32(v) => ensure_and_coerce!(v >= 0, v as u32),
+        Value::Int64(v) => ensure_and_coerce!(v >= 0, v as u32),
+        Value::UInt8(v) => Ok(v as u32),
+        Value::UInt16(v) => Ok(v as u32),
+        Value::UInt32(v) => Ok(v),
+        Value::UInt64(v) => Ok(v as u32),
+        _ => unreachable!(),
+    }
+}
+
 fn signature_of_cell() -> Signature {
-    let mut signatures = Vec::new();
-    for cell_type in &[
-        ConcreteDataType::uint64_datatype(),
-        ConcreteDataType::int64_datatype(),
-    ] {
+    let mut signatures = Vec::with_capacity(CELL_TYPES.len());
+    for cell_type in CELL_TYPES.as_slice() {
         signatures.push(TypeSignature::Exact(vec![cell_type.clone()]));
     }
 
     Signature::one_of(signatures, Volatility::Stable)
 }
 
-fn signature_of_double_cell() -> Signature {
-    let mut signatures = Vec::new();
-    let cell_types = &[
-        ConcreteDataType::uint64_datatype(),
-        ConcreteDataType::int64_datatype(),
-    ];
-    for cell_type in cell_types {
-        for cell_type2 in cell_types {
+fn signature_of_double_cells() -> Signature {
+    let mut signatures = Vec::with_capacity(CELL_TYPES.len() * CELL_TYPES.len());
+    for cell_type in CELL_TYPES.as_slice() {
+        for cell_type2 in CELL_TYPES.as_slice() {
             signatures.push(TypeSignature::Exact(vec![
                 cell_type.clone(),
                 cell_type2.clone(),
@@ -747,21 +1026,9 @@ fn signature_of_double_cell() -> Signature {
 }
 
 fn signature_of_cell_and_resolution() -> Signature {
-    let mut signatures = Vec::new();
-    for cell_type in &[
-        ConcreteDataType::uint64_datatype(),
-        ConcreteDataType::int64_datatype(),
-    ] {
-        for resolution_type in &[
-            ConcreteDataType::int8_datatype(),
-            ConcreteDataType::int16_datatype(),
-            ConcreteDataType::int32_datatype(),
-            ConcreteDataType::int64_datatype(),
-            ConcreteDataType::uint8_datatype(),
-            ConcreteDataType::uint16_datatype(),
-            ConcreteDataType::uint32_datatype(),
-            ConcreteDataType::uint64_datatype(),
-        ] {
+    let mut signatures = Vec::with_capacity(CELL_TYPES.len() * RESOLUTION_TYPES.len());
+    for cell_type in CELL_TYPES.as_slice() {
+        for resolution_type in RESOLUTION_TYPES.as_slice() {
             signatures.push(TypeSignature::Exact(vec![
                 cell_type.clone(),
                 resolution_type.clone(),
@@ -771,6 +1038,19 @@ fn signature_of_cell_and_resolution() -> Signature {
     Signature::one_of(signatures, Volatility::Stable)
 }
 
+fn signature_of_cell_and_distance() -> Signature {
+    let mut signatures = Vec::with_capacity(CELL_TYPES.len() * DISTANCE_TYPES.len());
+    for cell_type in CELL_TYPES.as_slice() {
+        for distance_type in DISTANCE_TYPES.as_slice() {
+            signatures.push(TypeSignature::Exact(vec![
+                cell_type.clone(),
+                distance_type.clone(),
+            ]));
+        }
+    }
+    Signature::one_of(signatures, Volatility::Stable)
+}
+
 fn cell_from_value(v: Value) -> Result<Option<CellIndex>> {
     let cell = match v {
         Value::Int64(v) => Some(

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

@@ -0,0 +1,75 @@
+// 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.
+
+macro_rules! ensure_columns_len {
+    ($columns:ident) => {
+        snafu::ensure!(
+            $columns.windows(2).all(|c| c[0].len() == c[1].len()),
+            common_query::error::InvalidFuncArgsSnafu {
+                err_msg: "The length of input columns are in different size"
+            }
+        )
+    };
+    ($column_a:ident, $column_b:ident, $($column_n:ident),*) => {
+        snafu::ensure!(
+            {
+                let mut result = $column_a.len() == $column_b.len();
+                $(
+                result = result && ($column_a.len() == $column_n.len());
+                )*
+                result
+            }
+            common_query::error::InvalidFuncArgsSnafu {
+                err_msg: "The length of input columns are in different size"
+            }
+        )
+    };
+}
+
+pub(super) use ensure_columns_len;
+
+macro_rules! ensure_columns_n {
+    ($columns:ident, $n:literal) => {
+        snafu::ensure!(
+            $columns.len() == $n,
+            common_query::error::InvalidFuncArgsSnafu {
+                err_msg: format!(
+                    "The length of arguments is not correct, expect {}, provided : {}",
+                    stringify!($n),
+                    $columns.len()
+                ),
+            }
+        );
+
+        if $n > 1 {
+            ensure_columns_len!($columns);
+        }
+    };
+}
+
+pub(super) use ensure_columns_n;
+
+macro_rules! ensure_and_coerce {
+    ($compare:expr, $coerce:expr) => {{
+        snafu::ensure!(
+            $compare,
+            common_query::error::InvalidFuncArgsSnafu {
+                err_msg: "Argument was outside of acceptable range "
+            }
+        );
+        Ok($coerce)
+    }};
+}
+
+pub(super) use ensure_and_coerce;

src/common/function/src/scalars/geo/s2.rs

@@ -0,0 +1,275 @@
+// 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_query::error::{InvalidFuncArgsSnafu, Result};
+use common_query::prelude::{Signature, TypeSignature};
+use datafusion::logical_expr::Volatility;
+use datatypes::prelude::ConcreteDataType;
+use datatypes::scalars::ScalarVectorBuilder;
+use datatypes::value::Value;
+use datatypes::vectors::{MutableVector, StringVectorBuilder, UInt64VectorBuilder, VectorRef};
+use derive_more::Display;
+use once_cell::sync::Lazy;
+use s2::cellid::{CellID, MAX_LEVEL};
+use s2::latlng::LatLng;
+use snafu::ensure;
+
+use crate::function::{Function, FunctionContext};
+use crate::scalars::geo::helpers::{ensure_and_coerce, ensure_columns_len, ensure_columns_n};
+
+static CELL_TYPES: Lazy<Vec<ConcreteDataType>> = Lazy::new(|| {
+    vec![
+        ConcreteDataType::int64_datatype(),
+        ConcreteDataType::uint64_datatype(),
+    ]
+});
+
+static COORDINATE_TYPES: Lazy<Vec<ConcreteDataType>> = Lazy::new(|| {
+    vec![
+        ConcreteDataType::float32_datatype(),
+        ConcreteDataType::float64_datatype(),
+    ]
+});
+
+static LEVEL_TYPES: Lazy<Vec<ConcreteDataType>> = Lazy::new(|| {
+    vec![
+        ConcreteDataType::int8_datatype(),
+        ConcreteDataType::int16_datatype(),
+        ConcreteDataType::int32_datatype(),
+        ConcreteDataType::int64_datatype(),
+        ConcreteDataType::uint8_datatype(),
+        ConcreteDataType::uint16_datatype(),
+        ConcreteDataType::uint32_datatype(),
+        ConcreteDataType::uint64_datatype(),
+    ]
+});
+
+/// Function that returns [s2] encoding cellid for a given geospatial coordinate.
+///
+/// [s2]: http://s2geometry.io
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct S2LatLngToCell;
+
+impl Function for S2LatLngToCell {
+    fn name(&self) -> &str {
+        "s2_latlng_to_cell"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::uint64_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        let mut signatures = Vec::with_capacity(COORDINATE_TYPES.len());
+        for coord_type in COORDINATE_TYPES.as_slice() {
+            signatures.push(TypeSignature::Exact(vec![
+                // latitude
+                coord_type.clone(),
+                // longitude
+                coord_type.clone(),
+            ]));
+        }
+        Signature::one_of(signatures, Volatility::Stable)
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 2);
+
+        let lat_vec = &columns[0];
+        let lon_vec = &columns[1];
+
+        let size = lat_vec.len();
+        let mut results = UInt64VectorBuilder::with_capacity(size);
+
+        for i in 0..size {
+            let lat = lat_vec.get(i).as_f64_lossy();
+            let lon = lon_vec.get(i).as_f64_lossy();
+
+            let result = match (lat, lon) {
+                (Some(lat), Some(lon)) => {
+                    let coord = LatLng::from_degrees(lat, lon);
+                    ensure!(
+                        coord.is_valid(),
+                        InvalidFuncArgsSnafu {
+                            err_msg: "The input coordinates are invalid",
+                        }
+                    );
+                    let cellid = CellID::from(coord);
+                    let encoded: u64 = cellid.0;
+                    Some(encoded)
+                }
+                _ => None,
+            };
+
+            results.push(result);
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+/// Return the level of current s2 cell
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct S2CellLevel;
+
+impl Function for S2CellLevel {
+    fn name(&self) -> &str {
+        "s2_cell_level"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::uint64_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        signature_of_cell()
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 1);
+
+        let cell_vec = &columns[0];
+        let size = cell_vec.len();
+        let mut results = UInt64VectorBuilder::with_capacity(size);
+
+        for i in 0..size {
+            let cell = cell_from_value(cell_vec.get(i));
+            let res = cell.map(|cell| cell.level());
+
+            results.push(res);
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+/// Return the string presentation of the cell
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct S2CellToToken;
+
+impl Function for S2CellToToken {
+    fn name(&self) -> &str {
+        "s2_cell_to_token"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::string_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        signature_of_cell()
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 1);
+
+        let cell_vec = &columns[0];
+        let size = cell_vec.len();
+        let mut results = StringVectorBuilder::with_capacity(size);
+
+        for i in 0..size {
+            let cell = cell_from_value(cell_vec.get(i));
+            let res = cell.map(|cell| cell.to_token());
+
+            results.push(res.as_deref());
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+/// Return parent at given level of current s2 cell
+#[derive(Clone, Debug, Default, Display)]
+#[display("{}", self.name())]
+pub struct S2CellParent;
+
+impl Function for S2CellParent {
+    fn name(&self) -> &str {
+        "s2_cell_parent"
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::uint64_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        signature_of_cell_and_level()
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure_columns_n!(columns, 2);
+
+        let cell_vec = &columns[0];
+        let level_vec = &columns[1];
+        let size = cell_vec.len();
+        let mut results = UInt64VectorBuilder::with_capacity(size);
+
+        for i in 0..size {
+            let cell = cell_from_value(cell_vec.get(i));
+            let level = value_to_level(level_vec.get(i))?;
+            let result = cell.map(|cell| cell.parent(level).0);
+
+            results.push(result);
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+fn signature_of_cell() -> Signature {
+    let mut signatures = Vec::with_capacity(CELL_TYPES.len());
+    for cell_type in CELL_TYPES.as_slice() {
+        signatures.push(TypeSignature::Exact(vec![cell_type.clone()]));
+    }
+
+    Signature::one_of(signatures, Volatility::Stable)
+}
+
+fn signature_of_cell_and_level() -> Signature {
+    let mut signatures = Vec::with_capacity(CELL_TYPES.len() * LEVEL_TYPES.len());
+    for cell_type in CELL_TYPES.as_slice() {
+        for level_type in LEVEL_TYPES.as_slice() {
+            signatures.push(TypeSignature::Exact(vec![
+                cell_type.clone(),
+                level_type.clone(),
+            ]));
+        }
+    }
+    Signature::one_of(signatures, Volatility::Stable)
+}
+
+fn cell_from_value(v: Value) -> Option<CellID> {
+    match v {
+        Value::Int64(v) => Some(CellID(v as u64)),
+        Value::UInt64(v) => Some(CellID(v)),
+        _ => None,
+    }
+}
+
+fn value_to_level(v: Value) -> Result<u64> {
+    match v {
+        Value::Int8(v) => ensure_and_coerce!(v >= 0 && v <= MAX_LEVEL as i8, v as u64),
+        Value::Int16(v) => ensure_and_coerce!(v >= 0 && v <= MAX_LEVEL as i16, v as u64),
+        Value::Int32(v) => ensure_and_coerce!(v >= 0 && v <= MAX_LEVEL as i32, v as u64),
+        Value::Int64(v) => ensure_and_coerce!(v >= 0 && v <= MAX_LEVEL as i64, v as u64),
+        Value::UInt8(v) => ensure_and_coerce!(v <= MAX_LEVEL as u8, v as u64),
+        Value::UInt16(v) => ensure_and_coerce!(v <= MAX_LEVEL as u16, v as u64),
+        Value::UInt32(v) => ensure_and_coerce!(v <= MAX_LEVEL as u32, v as u64),
+        Value::UInt64(v) => ensure_and_coerce!(v <= MAX_LEVEL, v),
+        _ => unreachable!(),
+    }
+}

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

@@ -15,6 +15,7 @@
 use std::sync::Arc;
 mod json_get;
 mod json_is;
+mod json_path_exists;
 mod json_to_string;
 mod parse_json;
 
@@ -46,5 +47,7 @@ impl JsonFunction {
         registry.register(Arc::new(JsonIsBool));
         registry.register(Arc::new(JsonIsArray));
         registry.register(Arc::new(JsonIsObject));
+
+        registry.register(Arc::new(json_path_exists::JsonPathExistsFunction));
     }
 }

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

@@ -0,0 +1,172 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use std::fmt::{self, Display};
+
+use common_query::error::{InvalidFuncArgsSnafu, Result, UnsupportedInputDataTypeSnafu};
+use common_query::prelude::Signature;
+use datafusion::logical_expr::Volatility;
+use datatypes::data_type::ConcreteDataType;
+use datatypes::prelude::VectorRef;
+use datatypes::scalars::ScalarVectorBuilder;
+use datatypes::vectors::{BooleanVectorBuilder, MutableVector};
+use snafu::ensure;
+
+use crate::function::{Function, FunctionContext};
+
+/// Check if the given JSON data contains the given JSON path.
+#[derive(Clone, Debug, Default)]
+pub struct JsonPathExistsFunction;
+
+const NAME: &str = "json_path_exists";
+
+impl Function for JsonPathExistsFunction {
+    fn name(&self) -> &str {
+        NAME
+    }
+
+    fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
+        Ok(ConcreteDataType::boolean_datatype())
+    }
+
+    fn signature(&self) -> Signature {
+        Signature::exact(
+            vec![
+                ConcreteDataType::json_datatype(),
+                ConcreteDataType::string_datatype(),
+            ],
+            Volatility::Immutable,
+        )
+    }
+
+    fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
+        ensure!(
+            columns.len() == 2,
+            InvalidFuncArgsSnafu {
+                err_msg: format!(
+                    "The length of the args is not correct, expect exactly two, have: {}",
+                    columns.len()
+                ),
+            }
+        );
+        let jsons = &columns[0];
+        let paths = &columns[1];
+
+        let size = jsons.len();
+        let datatype = jsons.data_type();
+        let mut results = BooleanVectorBuilder::with_capacity(size);
+
+        match datatype {
+            // JSON data type uses binary vector
+            ConcreteDataType::Binary(_) => {
+                for i in 0..size {
+                    let json = jsons.get_ref(i);
+                    let path = paths.get_ref(i);
+
+                    let json = json.as_binary();
+                    let path = path.as_string();
+                    let result = match (json, path) {
+                        (Ok(Some(json)), Ok(Some(path))) => {
+                            let json_path = jsonb::jsonpath::parse_json_path(path.as_bytes());
+                            match json_path {
+                                Ok(json_path) => jsonb::path_exists(json, json_path).ok(),
+                                Err(_) => None,
+                            }
+                        }
+                        _ => None,
+                    };
+
+                    results.push(result);
+                }
+            }
+            _ => {
+                return UnsupportedInputDataTypeSnafu {
+                    function: NAME,
+                    datatypes: columns.iter().map(|c| c.data_type()).collect::<Vec<_>>(),
+                }
+                .fail();
+            }
+        }
+
+        Ok(results.to_vector())
+    }
+}
+
+impl Display for JsonPathExistsFunction {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "JSON_PATH_EXISTS")
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::Arc;
+
+    use common_query::prelude::TypeSignature;
+    use datatypes::scalars::ScalarVector;
+    use datatypes::vectors::{BinaryVector, StringVector};
+
+    use super::*;
+
+    #[test]
+    fn test_json_path_exists_function() {
+        let json_path_exists = JsonPathExistsFunction;
+
+        assert_eq!("json_path_exists", json_path_exists.name());
+        assert_eq!(
+            ConcreteDataType::boolean_datatype(),
+            json_path_exists
+                .return_type(&[ConcreteDataType::json_datatype()])
+                .unwrap()
+        );
+
+        assert!(matches!(json_path_exists.signature(),
+                         Signature {
+                             type_signature: TypeSignature::Exact(valid_types),
+                             volatility: Volatility::Immutable
+                         } if  valid_types == vec![ConcreteDataType::json_datatype(), ConcreteDataType::string_datatype()]
+        ));
+
+        let json_strings = [
+            r#"{"a": {"b": 2}, "b": 2, "c": 3}"#,
+            r#"{"a": 4, "b": {"c": 6}, "c": 6}"#,
+            r#"{"a": 7, "b": 8, "c": {"a": 7}}"#,
+            r#"{"a": 7, "b": 8, "c": {"a": 7}}"#,
+        ];
+        let paths = vec!["$.a.b.c", "$.b", "$.c.a", ".d"];
+        let results = [false, true, true, false];
+
+        let jsonbs = json_strings
+            .iter()
+            .map(|s| {
+                let value = jsonb::parse_value(s.as_bytes()).unwrap();
+                value.to_vec()
+            })
+            .collect::<Vec<_>>();
+
+        let json_vector = BinaryVector::from_vec(jsonbs);
+        let path_vector = StringVector::from_vec(paths);
+        let args: Vec<VectorRef> = vec![Arc::new(json_vector), Arc::new(path_vector)];
+        let vector = json_path_exists
+            .eval(FunctionContext::default(), &args)
+            .unwrap();
+
+        assert_eq!(4, vector.len());
+        for (i, gt) in results.iter().enumerate() {
+            let result = vector.get_ref(i);
+            let result = result.as_boolean().unwrap().unwrap();
+            assert_eq!(*gt, result);
+        }
+    }
+}

src/common/grpc/src/select.rs

@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use api::helper::{convert_i128_to_interval, convert_to_pb_decimal128};
+use api::helper::{convert_month_day_nano_to_pb, convert_to_pb_decimal128};
 use api::v1::column::Values;
 use common_base::BitVec;
 use datatypes::types::{IntervalType, TimeType, TimestampType, WrapperType};
@@ -211,7 +211,7 @@ pub fn values(arrays: &[VectorRef]) -> Result<Values> {
             ConcreteDataType::Interval(IntervalType::MonthDayNano(_)),
             IntervalMonthDayNanoVector,
             interval_month_day_nano_values,
-            |x| { convert_i128_to_interval(x.into_native()) }
+            |x| { convert_month_day_nano_to_pb(x) }
         ),
         (
             ConcreteDataType::Decimal128(_),

src/common/macro/src/aggr_func.rs

@@ -21,23 +21,19 @@ use syn::{parse_macro_input, DeriveInput, ItemStruct};
 pub(crate) fn impl_aggr_func_type_store(ast: &DeriveInput) -> TokenStream {
     let name = &ast.ident;
     let gen = quote! {
-        use common_query::logical_plan::accumulator::AggrFuncTypeStore;
-        use common_query::error::{InvalidInputStateSnafu, Error as QueryError};
-        use datatypes::prelude::ConcreteDataType;
-
-        impl AggrFuncTypeStore for #name {
-            fn input_types(&self) -> std::result::Result<Vec<ConcreteDataType>, QueryError> {
+        impl common_query::logical_plan::accumulator::AggrFuncTypeStore for #name {
+            fn input_types(&self) -> std::result::Result<Vec<datatypes::prelude::ConcreteDataType>, common_query::error::Error> {
                 let input_types = self.input_types.load();
-                snafu::ensure!(input_types.is_some(), InvalidInputStateSnafu);
+                snafu::ensure!(input_types.is_some(), common_query::error::InvalidInputStateSnafu);
                 Ok(input_types.as_ref().unwrap().as_ref().clone())
             }
 
-            fn set_input_types(&self, input_types: Vec<ConcreteDataType>) -> std::result::Result<(), QueryError> {
+            fn set_input_types(&self, input_types: Vec<datatypes::prelude::ConcreteDataType>) -> std::result::Result<(), common_query::error::Error> {
                 let old = self.input_types.swap(Some(std::sync::Arc::new(input_types.clone())));
                 if let Some(old) = old {
-                    snafu::ensure!(old.len() == input_types.len(), InvalidInputStateSnafu);
+                    snafu::ensure!(old.len() == input_types.len(), common_query::error::InvalidInputStateSnafu);
                     for (x, y) in old.iter().zip(input_types.iter()) {
-                        snafu::ensure!(x == y, InvalidInputStateSnafu);
+                        snafu::ensure!(x == y, common_query::error::InvalidInputStateSnafu);
                     }
                 }
                 Ok(())
@@ -51,7 +47,7 @@ pub(crate) fn impl_as_aggr_func_creator(_args: TokenStream, input: TokenStream)
     let mut item_struct = parse_macro_input!(input as ItemStruct);
     if let syn::Fields::Named(ref mut fields) = item_struct.fields {
         let result = syn::Field::parse_named.parse2(quote! {
-            input_types: arc_swap::ArcSwapOption<Vec<ConcreteDataType>>
+            input_types: arc_swap::ArcSwapOption<Vec<datatypes::prelude::ConcreteDataType>>
         });
         match result {
             Ok(field) => fields.named.push(field),

src/common/macro/tests/test_derive.rs

@@ -24,5 +24,5 @@ struct Foo {}
 fn test_derive() {
     let _ = Foo::default();
     assert_fields!(Foo: input_types);
-    assert_impl_all!(Foo: std::fmt::Debug, Default, AggrFuncTypeStore);
+    assert_impl_all!(Foo: std::fmt::Debug, Default, common_query::logical_plan::accumulator::AggrFuncTypeStore);
 }

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

@@ -187,7 +187,7 @@ mod tests {
                     region: Region::new_test(region_id),
                     leader_peer: Some(Peer::empty(1)),
                     follower_peers: vec![],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 }]),
                 HashMap::new(),

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

@@ -107,21 +107,21 @@ async fn test_on_submit_alter_request() {
                     region: Region::new_test(RegionId::new(table_id, 1)),
                     leader_peer: Some(Peer::empty(1)),
                     follower_peers: vec![Peer::empty(5)],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 },
                 RegionRoute {
                     region: Region::new_test(RegionId::new(table_id, 2)),
                     leader_peer: Some(Peer::empty(2)),
                     follower_peers: vec![Peer::empty(4)],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 },
                 RegionRoute {
                     region: Region::new_test(RegionId::new(table_id, 3)),
                     leader_peer: Some(Peer::empty(3)),
                     follower_peers: vec![],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 },
             ]),
@@ -193,21 +193,21 @@ async fn test_on_submit_alter_request_with_outdated_request() {
                     region: Region::new_test(RegionId::new(table_id, 1)),
                     leader_peer: Some(Peer::empty(1)),
                     follower_peers: vec![Peer::empty(5)],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 },
                 RegionRoute {
                     region: Region::new_test(RegionId::new(table_id, 2)),
                     leader_peer: Some(Peer::empty(2)),
                     follower_peers: vec![Peer::empty(4)],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 },
                 RegionRoute {
                     region: Region::new_test(RegionId::new(table_id, 3)),
                     leader_peer: Some(Peer::empty(3)),
                     follower_peers: vec![],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 },
             ]),

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

@@ -119,21 +119,21 @@ async fn test_on_datanode_drop_regions() {
                     region: Region::new_test(RegionId::new(table_id, 1)),
                     leader_peer: Some(Peer::empty(1)),
                     follower_peers: vec![Peer::empty(5)],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 },
                 RegionRoute {
                     region: Region::new_test(RegionId::new(table_id, 2)),
                     leader_peer: Some(Peer::empty(2)),
                     follower_peers: vec![Peer::empty(4)],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 },
                 RegionRoute {
                     region: Region::new_test(RegionId::new(table_id, 3)),
                     leader_peer: Some(Peer::empty(3)),
                     follower_peers: vec![],
-                    leader_status: None,
+                    leader_state: None,
                     leader_down_since: None,
                 },
             ]),

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

@@ -18,6 +18,7 @@ use common_procedure::error::Error as ProcedureError;
 use snafu::{ensure, OptionExt, ResultExt};
 use store_api::metric_engine_consts::LOGICAL_TABLE_METADATA_KEY;
 use table::metadata::TableId;
+use table::table_reference::TableReference;
 
 use crate::ddl::DetectingRegion;
 use crate::error::{Error, OperateDatanodeSnafu, Result, TableNotFoundSnafu, UnsupportedSnafu};
@@ -109,8 +110,8 @@ pub async fn check_and_get_physical_table_id(
         .table_name_manager()
         .get(physical_table_name)
         .await?
-        .context(TableNotFoundSnafu {
-            table_name: physical_table_name.to_string(),
+        .with_context(|| TableNotFoundSnafu {
+            table_name: TableReference::from(physical_table_name).to_string(),
         })
         .map(|table| table.table_id())
 }
@@ -123,8 +124,8 @@ pub async fn get_physical_table_id(
         .table_name_manager()
         .get(logical_table_name)
         .await?
-        .context(TableNotFoundSnafu {
-            table_name: logical_table_name.to_string(),
+        .with_context(|| TableNotFoundSnafu {
+            table_name: TableReference::from(logical_table_name).to_string(),
         })
         .map(|table| table.table_id())?;
 

src/common/meta/src/error.rs

@@ -147,6 +147,20 @@ pub enum Error {
         source: common_procedure::Error,
     },
 
+    #[snafu(display("Failed to start procedure manager"))]
+    StartProcedureManager {
+        #[snafu(implicit)]
+        location: Location,
+        source: common_procedure::Error,
+    },
+
+    #[snafu(display("Failed to stop procedure manager"))]
+    StopProcedureManager {
+        #[snafu(implicit)]
+        location: Location,
+        source: common_procedure::Error,
+    },
+
     #[snafu(display(
         "Failed to get procedure output, procedure id: {procedure_id}, error: {err_msg}"
     ))]
@@ -715,7 +729,9 @@ impl ErrorExt for Error {
 
             SubmitProcedure { source, .. }
             | QueryProcedure { source, .. }
-            | WaitProcedure { source, .. } => source.status_code(),
+            | WaitProcedure { source, .. }
+            | StartProcedureManager { source, .. }
+            | StopProcedureManager { source, .. } => source.status_code(),
             RegisterProcedureLoader { source, .. } => source.status_code(),
             External { source, .. } => source.status_code(),
             OperateDatanode { source, .. } => source.status_code(),

src/common/meta/src/instruction.rs

@@ -137,14 +137,16 @@ pub struct DowngradeRegion {
     /// `None` stands for don't flush before downgrading the region.
     #[serde(default)]
     pub flush_timeout: Option<Duration>,
+    /// Rejects all write requests after flushing.
+    pub reject_write: bool,
 }
 
 impl Display for DowngradeRegion {
     fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
         write!(
             f,
-            "DowngradeRegion(region_id={}, flush_timeout={:?})",
-            self.region_id, self.flush_timeout,
+            "DowngradeRegion(region_id={}, flush_timeout={:?}, rejct_write={})",
+            self.region_id, self.flush_timeout, self.reject_write
         )
     }
 }

src/common/meta/src/key.rs

@@ -140,7 +140,7 @@ use crate::key::table_route::TableRouteKey;
 use crate::key::txn_helper::TxnOpGetResponseSet;
 use crate::kv_backend::txn::{Txn, TxnOp};
 use crate::kv_backend::KvBackendRef;
-use crate::rpc::router::{region_distribution, RegionRoute, RegionStatus};
+use crate::rpc::router::{region_distribution, LeaderState, RegionRoute};
 use crate::rpc::store::BatchDeleteRequest;
 use crate::DatanodeId;
 
@@ -1126,14 +1126,14 @@ impl TableMetadataManager {
         next_region_route_status: F,
     ) -> Result<()>
     where
-        F: Fn(&RegionRoute) -> Option<Option<RegionStatus>>,
+        F: Fn(&RegionRoute) -> Option<Option<LeaderState>>,
     {
         let mut new_region_routes = current_table_route_value.region_routes()?.clone();
 
         let mut updated = 0;
         for route in &mut new_region_routes {
-            if let Some(status) = next_region_route_status(route) {
-                if route.set_leader_status(status) {
+            if let Some(state) = next_region_route_status(route) {
+                if route.set_leader_state(state) {
                     updated += 1;
                 }
             }
@@ -1280,7 +1280,7 @@ mod tests {
     use crate::key::{DeserializedValueWithBytes, TableMetadataManager, ViewInfoValue};
     use crate::kv_backend::memory::MemoryKvBackend;
     use crate::peer::Peer;
-    use crate::rpc::router::{region_distribution, Region, RegionRoute, RegionStatus};
+    use crate::rpc::router::{region_distribution, LeaderState, Region, RegionRoute};
 
     #[test]
     fn test_deserialized_value_with_bytes() {
@@ -1324,7 +1324,7 @@ mod tests {
             },
             leader_peer: Some(Peer::new(datanode, "a2")),
             follower_peers: vec![],
-            leader_status: None,
+            leader_state: None,
             leader_down_since: None,
         }
     }
@@ -1715,7 +1715,7 @@ mod tests {
                     attrs: BTreeMap::new(),
                 },
                 leader_peer: Some(Peer::new(datanode, "a2")),
-                leader_status: Some(RegionStatus::Downgraded),
+                leader_state: Some(LeaderState::Downgrading),
                 follower_peers: vec![],
                 leader_down_since: Some(current_time_millis()),
             },
@@ -1727,7 +1727,7 @@ mod tests {
                     attrs: BTreeMap::new(),
                 },
                 leader_peer: Some(Peer::new(datanode, "a1")),
-                leader_status: None,
+                leader_state: None,
                 follower_peers: vec![],
                 leader_down_since: None,
             },
@@ -1750,10 +1750,10 @@ mod tests {
 
         table_metadata_manager
             .update_leader_region_status(table_id, &current_table_route_value, |region_route| {
-                if region_route.leader_status.is_some() {
+                if region_route.leader_state.is_some() {
                     None
                 } else {
-                    Some(Some(RegionStatus::Downgraded))
+                    Some(Some(LeaderState::Downgrading))
                 }
             })
             .await
@@ -1768,8 +1768,8 @@ mod tests {
             .unwrap();
 
         assert_eq!(
-            updated_route_value.region_routes().unwrap()[0].leader_status,
-            Some(RegionStatus::Downgraded)
+            updated_route_value.region_routes().unwrap()[0].leader_state,
+            Some(LeaderState::Downgrading)
         );
 
         assert!(updated_route_value.region_routes().unwrap()[0]
@@ -1777,8 +1777,8 @@ mod tests {
             .is_some());
 
         assert_eq!(
-            updated_route_value.region_routes().unwrap()[1].leader_status,
-            Some(RegionStatus::Downgraded)
+            updated_route_value.region_routes().unwrap()[1].leader_state,
+            Some(LeaderState::Downgrading)
         );
         assert!(updated_route_value.region_routes().unwrap()[1]
             .leader_down_since
@@ -1943,21 +1943,21 @@ mod tests {
                         region: Region::new_test(RegionId::new(table_id, 1)),
                         leader_peer: Some(Peer::empty(1)),
                         follower_peers: vec![Peer::empty(5)],
-                        leader_status: None,
+                        leader_state: None,
                         leader_down_since: None,
                     },
                     RegionRoute {
                         region: Region::new_test(RegionId::new(table_id, 2)),
                         leader_peer: Some(Peer::empty(2)),
                         follower_peers: vec![Peer::empty(4)],
-                        leader_status: None,
+                        leader_state: None,
                         leader_down_since: None,
                     },
                     RegionRoute {
                         region: Region::new_test(RegionId::new(table_id, 3)),
                         leader_peer: Some(Peer::empty(3)),
                         follower_peers: vec![],
-                        leader_status: None,
+                        leader_state: None,
                         leader_down_since: None,
                     },
                 ]),
@@ -1996,21 +1996,21 @@ mod tests {
                         region: Region::new_test(RegionId::new(table_id, 1)),
                         leader_peer: Some(Peer::empty(1)),
                         follower_peers: vec![Peer::empty(5)],
-                        leader_status: None,
+                        leader_state: None,
                         leader_down_since: None,
                     },
                     RegionRoute {
                         region: Region::new_test(RegionId::new(table_id, 2)),
                         leader_peer: Some(Peer::empty(2)),
                         follower_peers: vec![Peer::empty(4)],
-                        leader_status: None,
+                        leader_state: None,
                         leader_down_since: None,
                     },
                     RegionRoute {
                         region: Region::new_test(RegionId::new(table_id, 3)),
                         leader_peer: Some(Peer::empty(3)),
                         follower_peers: vec![],
-                        leader_status: None,
+                        leader_state: None,
                         leader_down_since: None,
                     },
                 ]),

src/common/meta/src/key/table_name.rs

@@ -21,6 +21,7 @@ use serde::{Deserialize, Serialize};
 use snafu::OptionExt;
 use table::metadata::TableId;
 use table::table_name::TableName;
+use table::table_reference::TableReference;
 
 use super::{MetadataKey, MetadataValue, TABLE_NAME_KEY_PATTERN, TABLE_NAME_KEY_PREFIX};
 use crate::error::{Error, InvalidMetadataSnafu, Result};
@@ -122,6 +123,16 @@ impl From<TableNameKey<'_>> for TableName {
     }
 }
 
+impl<'a> From<TableNameKey<'a>> for TableReference<'a> {
+    fn from(value: TableNameKey<'a>) -> Self {
+        Self {
+            catalog: value.catalog,
+            schema: value.schema,
+            table: value.table,
+        }
+    }
+}
+
 impl<'a> TryFrom<&'a str> for TableNameKey<'a> {
     type Error = Error;
 

src/common/meta/src/key/table_route.rs

@@ -744,6 +744,7 @@ mod tests {
     use crate::kv_backend::memory::MemoryKvBackend;
     use crate::kv_backend::{KvBackend, TxnService};
     use crate::peer::Peer;
+    use crate::rpc::router::Region;
     use crate::rpc::store::PutRequest;
 
     #[test]
@@ -751,11 +752,43 @@ mod tests {
         let old_raw_v = r#"{"region_routes":[{"region":{"id":1,"name":"r1","partition":null,"attrs":{}},"leader_peer":{"id":2,"addr":"a2"},"follower_peers":[]},{"region":{"id":1,"name":"r1","partition":null,"attrs":{}},"leader_peer":{"id":2,"addr":"a2"},"follower_peers":[]}],"version":0}"#;
         let v = TableRouteValue::try_from_raw_value(old_raw_v.as_bytes()).unwrap();
 
-        let new_raw_v = format!("{:?}", v);
-        assert_eq!(
-            new_raw_v,
-            r#"Physical(PhysicalTableRouteValue { region_routes: [RegionRoute { region: Region { id: 1(0, 1), name: "r1", partition: None, attrs: {} }, leader_peer: Some(Peer { id: 2, addr: "a2" }), follower_peers: [], leader_status: None, leader_down_since: None }, RegionRoute { region: Region { id: 1(0, 1), name: "r1", partition: None, attrs: {} }, leader_peer: Some(Peer { id: 2, addr: "a2" }), follower_peers: [], leader_status: None, leader_down_since: None }], version: 0 })"#
-        );
+        let expected_table_route = TableRouteValue::Physical(PhysicalTableRouteValue {
+            region_routes: vec![
+                RegionRoute {
+                    region: Region {
+                        id: RegionId::new(0, 1),
+                        name: "r1".to_string(),
+                        partition: None,
+                        attrs: Default::default(),
+                    },
+                    leader_peer: Some(Peer {
+                        id: 2,
+                        addr: "a2".to_string(),
+                    }),
+                    follower_peers: vec![],
+                    leader_state: None,
+                    leader_down_since: None,
+                },
+                RegionRoute {
+                    region: Region {
+                        id: RegionId::new(0, 1),
+                        name: "r1".to_string(),
+                        partition: None,
+                        attrs: Default::default(),
+                    },
+                    leader_peer: Some(Peer {
+                        id: 2,
+                        addr: "a2".to_string(),
+                    }),
+                    follower_peers: vec![],
+                    leader_state: None,
+                    leader_down_since: None,
+                },
+            ],
+            version: 0,
+        });
+
+        assert_eq!(v, expected_table_route);
     }
 
     #[test]

src/common/meta/src/leadership_notifier.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.
+
+use std::sync::{Arc, Mutex};
+
+use async_trait::async_trait;
+use common_telemetry::{error, info};
+
+use crate::error::Result;
+
+pub type LeadershipChangeNotifierCustomizerRef = Arc<dyn LeadershipChangeNotifierCustomizer>;
+
+/// A trait for customizing the leadership change notifier.
+pub trait LeadershipChangeNotifierCustomizer: Send + Sync {
+    fn customize(&self, notifier: &mut LeadershipChangeNotifier);
+
+    fn add_listener(&self, listener: Arc<dyn LeadershipChangeListener>);
+}
+
+/// A trait for handling leadership change events in a distributed system.
+#[async_trait]
+pub trait LeadershipChangeListener: Send + Sync {
+    /// Returns the listener name.
+    fn name(&self) -> &str;
+
+    /// Called when the node transitions to the leader role.
+    async fn on_leader_start(&self) -> Result<()>;
+
+    /// Called when the node transitions to the follower role.
+    async fn on_leader_stop(&self) -> Result<()>;
+}
+
+/// A notifier for leadership change events.
+#[derive(Default)]
+pub struct LeadershipChangeNotifier {
+    listeners: Vec<Arc<dyn LeadershipChangeListener>>,
+}
+
+#[derive(Default)]
+pub struct DefaultLeadershipChangeNotifierCustomizer {
+    listeners: Mutex<Vec<Arc<dyn LeadershipChangeListener>>>,
+}
+
+impl DefaultLeadershipChangeNotifierCustomizer {
+    pub fn new() -> Self {
+        Self {
+            listeners: Mutex::new(Vec::new()),
+        }
+    }
+}
+
+impl LeadershipChangeNotifierCustomizer for DefaultLeadershipChangeNotifierCustomizer {
+    fn customize(&self, notifier: &mut LeadershipChangeNotifier) {
+        info!("Customizing leadership change notifier");
+        let listeners = self.listeners.lock().unwrap().clone();
+        notifier.listeners.extend(listeners);
+    }
+
+    fn add_listener(&self, listener: Arc<dyn LeadershipChangeListener>) {
+        self.listeners.lock().unwrap().push(listener);
+    }
+}
+
+impl LeadershipChangeNotifier {
+    /// Adds a listener to the notifier.
+    pub fn add_listener(&mut self, listener: Arc<dyn LeadershipChangeListener>) {
+        self.listeners.push(listener);
+    }
+
+    /// Notify all listeners that the node has become a leader.
+    pub async fn notify_on_leader_start(&self) {
+        for listener in &self.listeners {
+            if let Err(err) = listener.on_leader_start().await {
+                error!(
+                    err;
+                    "Failed to notify listener: {}, event 'on_leader_start'",
+                    listener.name()
+                );
+            }
+        }
+    }
+
+    /// Notify all listeners that the node has become a follower.
+    pub async fn notify_on_leader_stop(&self) {
+        for listener in &self.listeners {
+            if let Err(err) = listener.on_leader_stop().await {
+                error!(
+                    err;
+                    "Failed to notify listener: {}, event: 'on_follower_start'",
+                    listener.name()
+                );
+            }
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::atomic::{AtomicBool, Ordering};
+    use std::sync::Arc;
+
+    use super::*;
+
+    struct MockListener {
+        name: String,
+        on_leader_start_fn: Option<Box<dyn Fn() -> Result<()> + Send + Sync>>,
+        on_follower_start_fn: Option<Box<dyn Fn() -> Result<()> + Send + Sync>>,
+    }
+
+    #[async_trait::async_trait]
+    impl LeadershipChangeListener for MockListener {
+        fn name(&self) -> &str {
+            &self.name
+        }
+
+        async fn on_leader_start(&self) -> Result<()> {
+            if let Some(f) = &self.on_leader_start_fn {
+                return f();
+            }
+            Ok(())
+        }
+
+        async fn on_leader_stop(&self) -> Result<()> {
+            if let Some(f) = &self.on_follower_start_fn {
+                return f();
+            }
+            Ok(())
+        }
+    }
+
+    #[tokio::test]
+    async fn test_leadership_change_notifier() {
+        let mut notifier = LeadershipChangeNotifier::default();
+        let listener1 = Arc::new(MockListener {
+            name: "listener1".to_string(),
+            on_leader_start_fn: None,
+            on_follower_start_fn: None,
+        });
+        let called_on_leader_start = Arc::new(AtomicBool::new(false));
+        let called_on_follower_start = Arc::new(AtomicBool::new(false));
+        let called_on_leader_start_moved = called_on_leader_start.clone();
+        let called_on_follower_start_moved = called_on_follower_start.clone();
+        let listener2 = Arc::new(MockListener {
+            name: "listener2".to_string(),
+            on_leader_start_fn: Some(Box::new(move || {
+                called_on_leader_start_moved.store(true, Ordering::Relaxed);
+                Ok(())
+            })),
+            on_follower_start_fn: Some(Box::new(move || {
+                called_on_follower_start_moved.store(true, Ordering::Relaxed);
+                Ok(())
+            })),
+        });
+
+        notifier.add_listener(listener1);
+        notifier.add_listener(listener2);
+
+        let listener1 = notifier.listeners.first().unwrap();
+        let listener2 = notifier.listeners.get(1).unwrap();
+
+        assert_eq!(listener1.name(), "listener1");
+        assert_eq!(listener2.name(), "listener2");
+
+        notifier.notify_on_leader_start().await;
+        assert!(!called_on_follower_start.load(Ordering::Relaxed));
+        assert!(called_on_leader_start.load(Ordering::Relaxed));
+
+        notifier.notify_on_leader_stop().await;
+        assert!(called_on_follower_start.load(Ordering::Relaxed));
+        assert!(called_on_leader_start.load(Ordering::Relaxed));
+    }
+}

src/common/meta/src/lib.rs

@@ -32,6 +32,7 @@ pub mod heartbeat;
 pub mod instruction;
 pub mod key;
 pub mod kv_backend;
+pub mod leadership_notifier;
 pub mod lock_key;
 pub mod metrics;
 pub mod node_manager;

src/common/meta/src/region_keeper.rs

@@ -58,7 +58,7 @@ impl MemoryRegionKeeper {
         Default::default()
     }
 
-    /// Returns [OpeningRegionGuard] if Region(`region_id`) on Peer(`datanode_id`) does not exist.
+    /// Returns [OperatingRegionGuard] if Region(`region_id`) on Peer(`datanode_id`) does not exist.
     pub fn register(
         &self,
         datanode_id: DatanodeId,

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

@@ -108,16 +108,16 @@ pub fn convert_to_region_peer_map(
         .collect::<HashMap<_, _>>()
 }
 
-/// Returns the HashMap<[RegionNumber], [RegionStatus]>;
-pub fn convert_to_region_leader_status_map(
+/// Returns the HashMap<[RegionNumber], [LeaderState]>;
+pub fn convert_to_region_leader_state_map(
     region_routes: &[RegionRoute],
-) -> HashMap<RegionNumber, RegionStatus> {
+) -> HashMap<RegionNumber, LeaderState> {
     region_routes
         .iter()
         .filter_map(|x| {
-            x.leader_status
+            x.leader_state
                 .as_ref()
-                .map(|status| (x.region.id.region_number(), *status))
+                .map(|state| (x.region.id.region_number(), *state))
         })
         .collect::<HashMap<_, _>>()
 }
@@ -205,7 +205,7 @@ impl TableRoute {
                 region,
                 leader_peer,
                 follower_peers,
-                leader_status: None,
+                leader_state: None,
                 leader_down_since: None,
             });
         }
@@ -259,9 +259,13 @@ pub struct RegionRoute {
     pub follower_peers: Vec<Peer>,
     /// `None` by default.
     #[builder(setter(into, strip_option), default)]
-    #[serde(default, skip_serializing_if = "Option::is_none")]
-    pub leader_status: Option<RegionStatus>,
-    /// The start time when the leader is in `Downgraded` status.
+    #[serde(
+        default,
+        alias = "leader_status",
+        skip_serializing_if = "Option::is_none"
+    )]
+    pub leader_state: Option<LeaderState>,
+    /// The start time when the leader is in `Downgraded` state.
     #[serde(default)]
     #[builder(default = "self.default_leader_down_since()")]
     pub leader_down_since: Option<i64>,
@@ -269,76 +273,79 @@ pub struct RegionRoute {
 
 impl RegionRouteBuilder {
     fn default_leader_down_since(&self) -> Option<i64> {
-        match self.leader_status {
-            Some(Some(RegionStatus::Downgraded)) => Some(current_time_millis()),
+        match self.leader_state {
+            Some(Some(LeaderState::Downgrading)) => Some(current_time_millis()),
             _ => None,
         }
     }
 }
 
-/// The Status of the [Region].
+/// The State of the [`Region`] Leader.
 /// TODO(dennis): It's better to add more fine-grained statuses such as `PENDING` etc.
 #[derive(Debug, Clone, Copy, Deserialize, Serialize, PartialEq, AsRefStr)]
 #[strum(serialize_all = "UPPERCASE")]
-pub enum RegionStatus {
-    /// The following cases in which the [Region] will be downgraded.
+pub enum LeaderState {
+    /// The following cases in which the [`Region`] will be downgraded.
     ///
-    /// - The [Region] is unavailable(e.g., Crashed, Network disconnected).
-    /// - The [Region] was planned to migrate to another [Peer].
-    Downgraded,
+    /// - The [`Region`] may be unavailable (e.g., Crashed, Network disconnected).
+    /// - The [`Region`] was planned to migrate to another [`Peer`].
+    #[serde(alias = "Downgraded")]
+    Downgrading,
 }
 
 impl RegionRoute {
-    /// Returns true if the Leader [Region] is downgraded.
+    /// Returns true if the Leader [`Region`] is downgraded.
     ///
-    /// The following cases in which the [Region] will be downgraded.
+    /// The following cases in which the [`Region`] will be downgraded.
     ///
-    /// - The [Region] is unavailable(e.g., Crashed, Network disconnected).
-    /// - The [Region] was planned to migrate to another [Peer].
+    /// - The [`Region`] is unavailable(e.g., Crashed, Network disconnected).
+    /// - The [`Region`] was planned to migrate to another [`Peer`].
     ///
-    pub fn is_leader_downgraded(&self) -> bool {
-        matches!(self.leader_status, Some(RegionStatus::Downgraded))
+    pub fn is_leader_downgrading(&self) -> bool {
+        matches!(self.leader_state, Some(LeaderState::Downgrading))
     }
 
-    /// Marks the Leader [Region] as downgraded.
+    /// Marks the Leader [`Region`] as [`RegionState::Downgrading`].
     ///
-    /// We should downgrade a [Region] before deactivating it:
+    /// We should downgrade a [`Region`] before deactivating it:
     ///
-    /// - During the [Region] Failover Procedure.
-    /// - Migrating a [Region].
+    /// - During the [`Region`] Failover Procedure.
+    /// - Migrating a [`Region`].
     ///
-    /// **Notes:** Meta Server will stop renewing the lease for the downgraded [Region].
+    /// **Notes:** Meta Server will renewing a special lease(`Downgrading`) for the downgrading [`Region`].
+    ///
+    /// A downgrading region will reject any write requests, and only allow memetable to be flushed to object storage
     ///
     pub fn downgrade_leader(&mut self) {
         self.leader_down_since = Some(current_time_millis());
-        self.leader_status = Some(RegionStatus::Downgraded)
+        self.leader_state = Some(LeaderState::Downgrading)
     }
 
-    /// Returns how long since the leader is in `Downgraded` status.
+    /// Returns how long since the leader is in `Downgraded` state.
     pub fn leader_down_millis(&self) -> Option<i64> {
         self.leader_down_since
             .map(|start| current_time_millis() - start)
     }
 
-    /// Sets the leader status.
+    /// Sets the leader state.
     ///
     /// Returns true if updated.
-    pub fn set_leader_status(&mut self, status: Option<RegionStatus>) -> bool {
-        let updated = self.leader_status != status;
+    pub fn set_leader_state(&mut self, state: Option<LeaderState>) -> bool {
+        let updated = self.leader_state != state;
 
-        match (status, updated) {
-            (Some(RegionStatus::Downgraded), true) => {
+        match (state, updated) {
+            (Some(LeaderState::Downgrading), true) => {
                 self.leader_down_since = Some(current_time_millis());
             }
-            (Some(RegionStatus::Downgraded), false) => {
-                // Do nothing if leader is still in `Downgraded` status.
+            (Some(LeaderState::Downgrading), false) => {
+                // Do nothing if leader is still in `Downgraded` state.
             }
             _ => {
                 self.leader_down_since = None;
             }
         }
 
-        self.leader_status = status;
+        self.leader_state = state;
         updated
     }
 }
@@ -477,15 +484,15 @@ mod tests {
             },
             leader_peer: Some(Peer::new(1, "a1")),
             follower_peers: vec![Peer::new(2, "a2"), Peer::new(3, "a3")],
-            leader_status: None,
+            leader_state: None,
             leader_down_since: None,
         };
 
-        assert!(!region_route.is_leader_downgraded());
+        assert!(!region_route.is_leader_downgrading());
 
         region_route.downgrade_leader();
 
-        assert!(region_route.is_leader_downgraded());
+        assert!(region_route.is_leader_downgrading());
     }
 
     #[test]
@@ -499,7 +506,7 @@ mod tests {
             },
             leader_peer: Some(Peer::new(1, "a1")),
             follower_peers: vec![Peer::new(2, "a2"), Peer::new(3, "a3")],
-            leader_status: None,
+            leader_state: None,
             leader_down_since: None,
         };
 
@@ -510,6 +517,73 @@ mod tests {
         assert_eq!(decoded, region_route);
     }
 
+    #[test]
+    fn test_region_route_compatibility() {
+        let region_route = RegionRoute {
+            region: Region {
+                id: 2.into(),
+                name: "r2".to_string(),
+                partition: None,
+                attrs: BTreeMap::new(),
+            },
+            leader_peer: Some(Peer::new(1, "a1")),
+            follower_peers: vec![Peer::new(2, "a2"), Peer::new(3, "a3")],
+            leader_state: Some(LeaderState::Downgrading),
+            leader_down_since: None,
+        };
+        let input = r#"{"region":{"id":2,"name":"r2","partition":null,"attrs":{}},"leader_peer":{"id":1,"addr":"a1"},"follower_peers":[{"id":2,"addr":"a2"},{"id":3,"addr":"a3"}],"leader_state":"Downgraded","leader_down_since":null}"#;
+        let decoded: RegionRoute = serde_json::from_str(input).unwrap();
+        assert_eq!(decoded, region_route);
+
+        let region_route = RegionRoute {
+            region: Region {
+                id: 2.into(),
+                name: "r2".to_string(),
+                partition: None,
+                attrs: BTreeMap::new(),
+            },
+            leader_peer: Some(Peer::new(1, "a1")),
+            follower_peers: vec![Peer::new(2, "a2"), Peer::new(3, "a3")],
+            leader_state: Some(LeaderState::Downgrading),
+            leader_down_since: None,
+        };
+        let input = r#"{"region":{"id":2,"name":"r2","partition":null,"attrs":{}},"leader_peer":{"id":1,"addr":"a1"},"follower_peers":[{"id":2,"addr":"a2"},{"id":3,"addr":"a3"}],"leader_status":"Downgraded","leader_down_since":null}"#;
+        let decoded: RegionRoute = serde_json::from_str(input).unwrap();
+        assert_eq!(decoded, region_route);
+
+        let region_route = RegionRoute {
+            region: Region {
+                id: 2.into(),
+                name: "r2".to_string(),
+                partition: None,
+                attrs: BTreeMap::new(),
+            },
+            leader_peer: Some(Peer::new(1, "a1")),
+            follower_peers: vec![Peer::new(2, "a2"), Peer::new(3, "a3")],
+            leader_state: Some(LeaderState::Downgrading),
+            leader_down_since: None,
+        };
+        let input = r#"{"region":{"id":2,"name":"r2","partition":null,"attrs":{}},"leader_peer":{"id":1,"addr":"a1"},"follower_peers":[{"id":2,"addr":"a2"},{"id":3,"addr":"a3"}],"leader_state":"Downgrading","leader_down_since":null}"#;
+        let decoded: RegionRoute = serde_json::from_str(input).unwrap();
+        assert_eq!(decoded, region_route);
+
+        let region_route = RegionRoute {
+            region: Region {
+                id: 2.into(),
+                name: "r2".to_string(),
+                partition: None,
+                attrs: BTreeMap::new(),
+            },
+            leader_peer: Some(Peer::new(1, "a1")),
+            follower_peers: vec![Peer::new(2, "a2"), Peer::new(3, "a3")],
+            leader_state: Some(LeaderState::Downgrading),
+            leader_down_since: None,
+        };
+        let input = r#"{"region":{"id":2,"name":"r2","partition":null,"attrs":{}},"leader_peer":{"id":1,"addr":"a1"},"follower_peers":[{"id":2,"addr":"a2"},{"id":3,"addr":"a3"}],"leader_status":"Downgrading","leader_down_since":null}"#;
+        let decoded: RegionRoute = serde_json::from_str(input).unwrap();
+        assert_eq!(decoded, region_route);
+    }
+
     #[test]
     fn test_de_serialize_partition() {
         let p = Partition {

src/common/meta/src/wal_options_allocator.rs

@@ -17,6 +17,7 @@ pub mod kafka;
 use std::collections::HashMap;
 use std::sync::Arc;
 
+use async_trait::async_trait;
 use common_wal::config::MetasrvWalConfig;
 use common_wal::options::{KafkaWalOptions, WalOptions, WAL_OPTIONS_KEY};
 use snafu::ResultExt;
@@ -24,6 +25,7 @@ use store_api::storage::{RegionId, RegionNumber};
 
 use crate::error::{EncodeWalOptionsSnafu, Result};
 use crate::kv_backend::KvBackendRef;
+use crate::leadership_notifier::LeadershipChangeListener;
 use crate::wal_options_allocator::kafka::topic_manager::TopicManager as KafkaTopicManager;
 
 /// Allocates wal options in region granularity.
@@ -94,6 +96,21 @@ impl WalOptionsAllocator {
     }
 }
 
+#[async_trait]
+impl LeadershipChangeListener for WalOptionsAllocator {
+    fn name(&self) -> &str {
+        "WalOptionsAllocator"
+    }
+
+    async fn on_leader_start(&self) -> Result<()> {
+        self.start().await
+    }
+
+    async fn on_leader_stop(&self) -> Result<()> {
+        Ok(())
+    }
+}
+
 /// Allocates a wal options for each region. The allocated wal options is encoded immediately.
 pub fn allocate_region_wal_options(
     regions: Vec<RegionNumber>,

src/common/pprof/Cargo.toml

@@ -0,0 +1,22 @@
+[package]
+name = "common-pprof"
+version.workspace = true
+edition.workspace = true
+license.workspace = true
+
+[dependencies]
+common-error.workspace = true
+common-macro.workspace = true
+prost.workspace = true
+snafu.workspace = true
+tokio.workspace = true
+
+[target.'cfg(unix)'.dependencies]
+pprof = { version = "0.13", features = [
+    "flamegraph",
+    "prost-codec",
+    "protobuf",
+] }
+
+[lints]
+workspace = true

src/common/pprof/src/lib.rs

@@ -0,0 +1,99 @@
+// 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.
+
+#[cfg(unix)]
+pub mod nix;
+
+pub mod error {
+    use std::any::Any;
+
+    use common_error::ext::ErrorExt;
+    use common_error::status_code::StatusCode;
+    use common_macro::stack_trace_debug;
+    use snafu::{Location, Snafu};
+
+    #[derive(Snafu)]
+    #[stack_trace_debug]
+    #[snafu(visibility(pub(crate)))]
+    pub enum Error {
+        #[cfg(unix)]
+        #[snafu(display("Pprof error"))]
+        Pprof {
+            #[snafu(source)]
+            error: pprof::Error,
+            #[snafu(implicit)]
+            location: Location,
+        },
+
+        #[snafu(display("Pprof is unsupported on this platform"))]
+        Unsupported {
+            #[snafu(implicit)]
+            location: Location,
+        },
+    }
+
+    pub type Result<T> = std::result::Result<T, Error>;
+
+    impl ErrorExt for Error {
+        fn status_code(&self) -> StatusCode {
+            match self {
+                #[cfg(unix)]
+                Error::Pprof { .. } => StatusCode::Unexpected,
+                Error::Unsupported { .. } => StatusCode::Unsupported,
+            }
+        }
+
+        fn as_any(&self) -> &dyn Any {
+            self
+        }
+    }
+}
+
+#[cfg(not(unix))]
+pub mod dummy {
+    use std::time::Duration;
+
+    use crate::error::{Result, UnsupportedSnafu};
+
+    /// Dummpy CPU profiler utility.
+    #[derive(Debug)]
+    pub struct Profiling {}
+
+    impl Profiling {
+        /// Creates a new profiler.
+        pub fn new(_duration: Duration, _frequency: i32) -> Profiling {
+            Profiling {}
+        }
+
+        /// Profiles and returns a generated text.
+        pub async fn dump_text(&self) -> Result<String> {
+            UnsupportedSnafu {}.fail()
+        }
+
+        /// Profiles and returns a generated flamegraph.
+        pub async fn dump_flamegraph(&self) -> Result<Vec<u8>> {
+            UnsupportedSnafu {}.fail()
+        }
+
+        /// Profiles and returns a generated proto.
+        pub async fn dump_proto(&self) -> Result<Vec<u8>> {
+            UnsupportedSnafu {}.fail()
+        }
+    }
+}
+
+#[cfg(not(unix))]
+pub use dummy::Profiling;
+#[cfg(unix)]
+pub use nix::Profiling;

src/common/pprof/src/nix.rs

@@ -0,0 +1,78 @@
+// 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::time::Duration;
+
+use pprof::protos::Message;
+use snafu::ResultExt;
+
+use crate::error::{PprofSnafu, Result};
+
+/// CPU profiler utility.
+// Inspired by https://github.com/datafuselabs/databend/blob/67f445e83cd4eceda98f6c1c114858929d564029/src/common/base/src/base/profiling.rs
+#[derive(Debug)]
+pub struct Profiling {
+    /// Sample duration.
+    duration: Duration,
+    /// Sample frequency.
+    frequency: i32,
+}
+
+impl Profiling {
+    /// Creates a new profiler.
+    pub fn new(duration: Duration, frequency: i32) -> Profiling {
+        Profiling {
+            duration,
+            frequency,
+        }
+    }
+
+    /// Profiles and returns a generated pprof report.
+    pub async fn report(&self) -> Result<pprof::Report> {
+        let guard = pprof::ProfilerGuardBuilder::default()
+            .frequency(self.frequency)
+            .blocklist(&["libc", "libgcc", "pthread", "vdso"])
+            .build()
+            .context(PprofSnafu)?;
+        tokio::time::sleep(self.duration).await;
+        guard.report().build().context(PprofSnafu)
+    }
+
+    /// Profiles and returns a generated text.
+    pub async fn dump_text(&self) -> Result<String> {
+        let report = self.report().await?;
+        let text = format!("{report:?}");
+        Ok(text)
+    }
+
+    /// Profiles and returns a generated flamegraph.
+    pub async fn dump_flamegraph(&self) -> Result<Vec<u8>> {
+        let mut body: Vec<u8> = Vec::new();
+
+        let report = self.report().await?;
+        report.flamegraph(&mut body).context(PprofSnafu)?;
+
+        Ok(body)
+    }
+
+    /// Profiles and returns a generated proto.
+    pub async fn dump_proto(&self) -> Result<Vec<u8>> {
+        let report = self.report().await?;
+        // Generate google’s pprof format report.
+        let profile = report.pprof().context(PprofSnafu)?;
+        let body = profile.encode_to_vec();
+
+        Ok(body)
+    }
+}

src/common/recordbatch/src/adapter.rs

@@ -329,6 +329,7 @@ impl ExecutionPlanVisitor for MetricCollector {
                 level: self.current_level,
                 metrics: vec![],
             });
+            self.current_level += 1;
             return Ok(true);
         };
 
@@ -365,8 +366,7 @@ impl ExecutionPlanVisitor for MetricCollector {
     }
 
     fn post_visit(&mut self, _plan: &dyn ExecutionPlan) -> std::result::Result<bool, Self::Error> {
-        // the last minus will underflow
-        self.current_level = self.current_level.wrapping_sub(1);
+        self.current_level -= 1;
         Ok(true)
     }
 }

src/common/telemetry/Cargo.toml

@@ -17,6 +17,7 @@ backtrace = "0.3"
 common-error.workspace = true
 console-subscriber = { version = "0.1", optional = true }
 greptime-proto.workspace = true
+humantime-serde.workspace = true
 lazy_static.workspace = true
 once_cell.workspace = true
 opentelemetry = { version = "0.21.0", default-features = false, features = [

src/common/telemetry/src/logging.rs

@@ -15,6 +15,7 @@
 //! logging stuffs, inspired by databend
 use std::env;
 use std::sync::{Arc, Mutex, Once};
+use std::time::Duration;
 
 use once_cell::sync::{Lazy, OnceCell};
 use opentelemetry::{global, KeyValue};
@@ -26,7 +27,7 @@ use serde::{Deserialize, Serialize};
 use tracing_appender::non_blocking::WorkerGuard;
 use tracing_appender::rolling::{RollingFileAppender, Rotation};
 use tracing_log::LogTracer;
-use tracing_subscriber::filter::Targets;
+use tracing_subscriber::filter::{FilterFn, Targets};
 use tracing_subscriber::fmt::Layer;
 use tracing_subscriber::layer::SubscriberExt;
 use tracing_subscriber::prelude::*;
@@ -53,6 +54,9 @@ pub struct LoggingOptions {
     /// The log format that can be one of "json" or "text". Default is "text".
     pub log_format: LogFormat,
 
+    /// The maximum number of log files set by default.
+    pub max_log_files: usize,
+
     /// Whether to append logs to stdout. Default is true.
     pub append_stdout: bool,
 
@@ -64,6 +68,24 @@ pub struct LoggingOptions {
 
     /// The tracing sample ratio.
     pub tracing_sample_ratio: Option<TracingSampleOptions>,
+
+    /// The logging options of slow query.
+    pub slow_query: SlowQueryOptions,
+}
+
+/// The options of slow query.
+#[derive(Clone, Debug, Serialize, Deserialize, Default)]
+#[serde(default)]
+pub struct SlowQueryOptions {
+    /// Whether to enable slow query log.
+    pub enable: bool,
+
+    /// The threshold of slow queries.
+    #[serde(with = "humantime_serde")]
+    pub threshold: Option<Duration>,
+
+    /// The sample ratio of slow queries.
+    pub sample_ratio: Option<f64>,
 }
 
 #[derive(Clone, Debug, Copy, PartialEq, Eq, Serialize, Deserialize)]
@@ -96,6 +118,9 @@ impl Default for LoggingOptions {
             otlp_endpoint: None,
             tracing_sample_ratio: None,
             append_stdout: true,
+            slow_query: SlowQueryOptions::default(),
+            // Rotation hourly, 24 files per day, keeps info log files of 30 days
+            max_log_files: 720,
         }
     }
 }
@@ -186,8 +211,17 @@ pub fn init_global_logging(
 
         // Configure the file logging layer with rolling policy.
         let file_logging_layer = if !opts.dir.is_empty() {
-            let rolling_appender =
-                RollingFileAppender::new(Rotation::HOURLY, &opts.dir, "greptimedb");
+            let rolling_appender = RollingFileAppender::builder()
+                .rotation(Rotation::HOURLY)
+                .filename_prefix("greptimedb")
+                .max_log_files(opts.max_log_files)
+                .build(&opts.dir)
+                .unwrap_or_else(|e| {
+                    panic!(
+                        "initializing rolling file appender at {} failed: {}",
+                        &opts.dir, e
+                    )
+                });
             let (writer, guard) = tracing_appender::non_blocking(rolling_appender);
             guards.push(guard);
 
@@ -208,8 +242,17 @@ pub fn init_global_logging(
 
         // Configure the error file logging layer with rolling policy.
         let err_file_logging_layer = if !opts.dir.is_empty() {
-            let rolling_appender =
-                RollingFileAppender::new(Rotation::HOURLY, &opts.dir, "greptimedb-err");
+            let rolling_appender = RollingFileAppender::builder()
+                .rotation(Rotation::HOURLY)
+                .filename_prefix("greptimedb-err")
+                .max_log_files(opts.max_log_files)
+                .build(&opts.dir)
+                .unwrap_or_else(|e| {
+                    panic!(
+                        "initializing rolling file appender at {} failed: {}",
+                        &opts.dir, e
+                    )
+                });
             let (writer, guard) = tracing_appender::non_blocking(rolling_appender);
             guards.push(guard);
 
@@ -235,6 +278,51 @@ pub fn init_global_logging(
             None
         };
 
+        let slow_query_logging_layer = if !opts.dir.is_empty() && opts.slow_query.enable {
+            let rolling_appender = RollingFileAppender::builder()
+                .rotation(Rotation::HOURLY)
+                .filename_prefix("greptimedb-slow-queries")
+                .max_log_files(opts.max_log_files)
+                .build(&opts.dir)
+                .unwrap_or_else(|e| {
+                    panic!(
+                        "initializing rolling file appender at {} failed: {}",
+                        &opts.dir, e
+                    )
+                });
+            let (writer, guard) = tracing_appender::non_blocking(rolling_appender);
+            guards.push(guard);
+
+            // Only logs if the field contains "slow".
+            let slow_query_filter = FilterFn::new(|metadata| {
+                metadata
+                    .fields()
+                    .iter()
+                    .any(|field| field.name().contains("slow"))
+            });
+
+            if opts.log_format == LogFormat::Json {
+                Some(
+                    Layer::new()
+                        .json()
+                        .with_writer(writer)
+                        .with_ansi(false)
+                        .with_filter(slow_query_filter)
+                        .boxed(),
+                )
+            } else {
+                Some(
+                    Layer::new()
+                        .with_writer(writer)
+                        .with_ansi(false)
+                        .with_filter(slow_query_filter)
+                        .boxed(),
+                )
+            }
+        } else {
+            None
+        };
+
         // resolve log level settings from:
         // - options from command line or config files
         // - environment variable: RUST_LOG
@@ -279,6 +367,7 @@ pub fn init_global_logging(
                 .with(stdout_logging_layer)
                 .with(file_logging_layer)
                 .with(err_file_logging_layer)
+                .with(slow_query_logging_layer)
         };
 
         // consume the `tracing_opts` to avoid "unused" warnings.
@@ -289,7 +378,8 @@ pub fn init_global_logging(
             .with(dyn_filter)
             .with(stdout_logging_layer)
             .with(file_logging_layer)
-            .with(err_file_logging_layer);
+            .with(err_file_logging_layer)
+            .with(slow_query_logging_layer);
 
         if opts.enable_otlp_tracing {
             global::set_text_map_propagator(TraceContextPropagator::new());

src/common/telemetry/src/macros.rs

@@ -152,6 +152,17 @@ macro_rules! trace {
     };
 }
 
+#[macro_export]
+macro_rules! slow {
+    (target: $target:expr, $($arg:tt)+) => {
+        $crate::log!(target: $target, slow = true, $crate::tracing::Level::INFO, $($arg)+)
+    };
+
+    ($($arg:tt)+) => {
+        $crate::log!($crate::tracing::Level::INFO, slow = true, $($arg)+)
+    };
+}
+
 #[cfg(test)]
 mod tests {
     use common_error::mock::MockError;

src/common/time/src/date.rs

@@ -14,13 +14,13 @@
 
 use std::fmt::{Display, Formatter, Write};
 
-use chrono::{Datelike, Days, LocalResult, Months, NaiveDate, NaiveTime, TimeZone};
+use chrono::{Datelike, Days, LocalResult, Months, NaiveDate, NaiveTime, TimeDelta, TimeZone};
 use serde::{Deserialize, Serialize};
 use serde_json::Value;
 use snafu::ResultExt;
 
 use crate::error::{InvalidDateStrSnafu, ParseDateStrSnafu, Result};
-use crate::interval::Interval;
+use crate::interval::{IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth};
 use crate::timezone::get_timezone;
 use crate::util::datetime_to_utc;
 use crate::Timezone;
@@ -134,29 +134,64 @@ impl Date {
         (self.0 as i64) * 24 * 3600
     }
 
-    /// Adds given Interval to the current date.
-    /// Returns None if the resulting date would be out of range.
-    pub fn add_interval(&self, interval: Interval) -> Option<Date> {
+    // FIXME(yingwen): remove add/sub intervals later
+    /// Adds given [IntervalYearMonth] to the current date.
+    pub fn add_year_month(&self, interval: IntervalYearMonth) -> Option<Date> {
         let naive_date = self.to_chrono_date()?;
 
-        let (months, days, _) = interval.to_month_day_nano();
-
         naive_date
-            .checked_add_months(Months::new(months as u32))?
-            .checked_add_days(Days::new(days as u64))
+            .checked_add_months(Months::new(interval.months as u32))
             .map(Into::into)
     }
 
-    /// Subtracts given Interval to the current date.
-    /// Returns None if the resulting date would be out of range.
-    pub fn sub_interval(&self, interval: Interval) -> Option<Date> {
+    /// Adds given [IntervalDayTime] to the current date.
+    pub fn add_day_time(&self, interval: IntervalDayTime) -> Option<Date> {
         let naive_date = self.to_chrono_date()?;
 
-        let (months, days, _) = interval.to_month_day_nano();
+        naive_date
+            .checked_add_days(Days::new(interval.days as u64))?
+            .checked_add_signed(TimeDelta::milliseconds(interval.milliseconds as i64))
+            .map(Into::into)
+    }
+
+    /// Adds given [IntervalMonthDayNano] to the current date.
+    pub fn add_month_day_nano(&self, interval: IntervalMonthDayNano) -> Option<Date> {
+        let naive_date = self.to_chrono_date()?;
 
         naive_date
-            .checked_sub_months(Months::new(months as u32))?
-            .checked_sub_days(Days::new(days as u64))
+            .checked_add_months(Months::new(interval.months as u32))?
+            .checked_add_days(Days::new(interval.days as u64))?
+            .checked_add_signed(TimeDelta::nanoseconds(interval.nanoseconds))
+            .map(Into::into)
+    }
+
+    /// Subtracts given [IntervalYearMonth] to the current date.
+    pub fn sub_year_month(&self, interval: IntervalYearMonth) -> Option<Date> {
+        let naive_date = self.to_chrono_date()?;
+
+        naive_date
+            .checked_sub_months(Months::new(interval.months as u32))
+            .map(Into::into)
+    }
+
+    /// Subtracts given [IntervalDayTime] to the current date.
+    pub fn sub_day_time(&self, interval: IntervalDayTime) -> Option<Date> {
+        let naive_date = self.to_chrono_date()?;
+
+        naive_date
+            .checked_sub_days(Days::new(interval.days as u64))?
+            .checked_sub_signed(TimeDelta::milliseconds(interval.milliseconds as i64))
+            .map(Into::into)
+    }
+
+    /// Subtracts given [IntervalMonthDayNano] to the current date.
+    pub fn sub_month_day_nano(&self, interval: IntervalMonthDayNano) -> Option<Date> {
+        let naive_date = self.to_chrono_date()?;
+
+        naive_date
+            .checked_sub_months(Months::new(interval.months as u32))?
+            .checked_sub_days(Days::new(interval.days as u64))?
+            .checked_sub_signed(TimeDelta::nanoseconds(interval.nanoseconds))
             .map(Into::into)
     }
 
@@ -246,12 +281,12 @@ mod tests {
     fn test_add_sub_interval() {
         let date = Date::new(1000);
 
-        let interval = Interval::from_year_month(3);
+        let interval = IntervalYearMonth::new(3);
 
-        let new_date = date.add_interval(interval).unwrap();
+        let new_date = date.add_year_month(interval).unwrap();
         assert_eq!(new_date.val(), 1091);
 
-        assert_eq!(date, new_date.sub_interval(interval).unwrap());
+        assert_eq!(date, new_date.sub_year_month(interval).unwrap());
     }
 
     #[test]

src/common/time/src/datetime.rs

@@ -13,16 +13,18 @@
 // limitations under the License.
 
 use std::fmt::{Display, Formatter, Write};
-use std::time::Duration;
 
-use chrono::{Days, LocalResult, Months, NaiveDateTime, TimeZone as ChronoTimeZone, Utc};
+use chrono::{
+    Days, LocalResult, Months, NaiveDateTime, TimeDelta, TimeZone as ChronoTimeZone, Utc,
+};
 use serde::{Deserialize, Serialize};
 use snafu::ResultExt;
 
 use crate::error::{InvalidDateStrSnafu, Result};
+use crate::interval::{IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth};
 use crate::timezone::{get_timezone, Timezone};
 use crate::util::{datetime_to_utc, format_utc_datetime};
-use crate::{Date, Interval};
+use crate::Date;
 
 const DATETIME_FORMAT: &str = "%F %H:%M:%S%.f";
 const DATETIME_FORMAT_WITH_TZ: &str = "%F %H:%M:%S%.f%z";
@@ -160,32 +162,66 @@ impl DateTime {
             None => Utc.from_utc_datetime(&v).naive_local(),
         })
     }
-    /// Adds given Interval to the current datetime.
-    /// Returns None if the resulting datetime would be out of range.
-    pub fn add_interval(&self, interval: Interval) -> Option<Self> {
+
+    // FIXME(yingwen): remove add/sub intervals later
+    /// Adds given [IntervalYearMonth] to the current datetime.
+    pub fn add_year_month(&self, interval: IntervalYearMonth) -> Option<Self> {
         let naive_datetime = self.to_chrono_datetime()?;
-        let (months, days, nsecs) = interval.to_month_day_nano();
 
-        let naive_datetime = naive_datetime
-            .checked_add_months(Months::new(months as u32))?
-            .checked_add_days(Days::new(days as u64))?
-            + Duration::from_nanos(nsecs as u64);
-
-        Some(naive_datetime.into())
+        naive_datetime
+            .checked_add_months(Months::new(interval.months as u32))
+            .map(Into::into)
     }
 
-    /// Subtracts given Interval to the current datetime.
-    /// Returns None if the resulting datetime would be out of range.
-    pub fn sub_interval(&self, interval: Interval) -> Option<Self> {
+    /// Adds given [IntervalDayTime] to the current datetime.
+    pub fn add_day_time(&self, interval: IntervalDayTime) -> Option<Self> {
         let naive_datetime = self.to_chrono_datetime()?;
-        let (months, days, nsecs) = interval.to_month_day_nano();
 
-        let naive_datetime = naive_datetime
-            .checked_sub_months(Months::new(months as u32))?
-            .checked_sub_days(Days::new(days as u64))?
-            - Duration::from_nanos(nsecs as u64);
+        naive_datetime
+            .checked_add_days(Days::new(interval.days as u64))?
+            .checked_add_signed(TimeDelta::milliseconds(interval.milliseconds as i64))
+            .map(Into::into)
+    }
 
-        Some(naive_datetime.into())
+    /// Adds given [IntervalMonthDayNano] to the current datetime.
+    pub fn add_month_day_nano(&self, interval: IntervalMonthDayNano) -> Option<Self> {
+        let naive_datetime = self.to_chrono_datetime()?;
+
+        naive_datetime
+            .checked_add_months(Months::new(interval.months as u32))?
+            .checked_add_days(Days::new(interval.days as u64))?
+            .checked_add_signed(TimeDelta::nanoseconds(interval.nanoseconds))
+            .map(Into::into)
+    }
+
+    /// Subtracts given [IntervalYearMonth] to the current datetime.
+    pub fn sub_year_month(&self, interval: IntervalYearMonth) -> Option<Self> {
+        let naive_datetime = self.to_chrono_datetime()?;
+
+        naive_datetime
+            .checked_sub_months(Months::new(interval.months as u32))
+            .map(Into::into)
+    }
+
+    /// Subtracts given [IntervalDayTime] to the current datetime.
+    pub fn sub_day_time(&self, interval: IntervalDayTime) -> Option<Self> {
+        let naive_datetime = self.to_chrono_datetime()?;
+
+        naive_datetime
+            .checked_sub_days(Days::new(interval.days as u64))?
+            .checked_sub_signed(TimeDelta::milliseconds(interval.milliseconds as i64))
+            .map(Into::into)
+    }
+
+    /// Subtracts given [IntervalMonthDayNano] to the current datetime.
+    pub fn sub_month_day_nano(&self, interval: IntervalMonthDayNano) -> Option<Self> {
+        let naive_datetime = self.to_chrono_datetime()?;
+
+        naive_datetime
+            .checked_sub_months(Months::new(interval.months as u32))?
+            .checked_sub_days(Days::new(interval.days as u64))?
+            .checked_sub_signed(TimeDelta::nanoseconds(interval.nanoseconds))
+            .map(Into::into)
     }
 
     /// Convert to [common_time::date].
@@ -231,12 +267,12 @@ mod tests {
     fn test_add_sub_interval() {
         let datetime = DateTime::new(1000);
 
-        let interval = Interval::from_day_time(1, 200);
+        let interval = IntervalDayTime::new(1, 200);
 
-        let new_datetime = datetime.add_interval(interval).unwrap();
+        let new_datetime = datetime.add_day_time(interval).unwrap();
         assert_eq!(new_datetime.val(), 1000 + 3600 * 24 * 1000 + 200);
 
-        assert_eq!(datetime, new_datetime.sub_interval(interval).unwrap());
+        assert_eq!(datetime, new_datetime.sub_day_time(interval).unwrap());
     }
 
     #[test]

src/common/time/src/interval.rs

@@ -12,18 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::cmp::Ordering;
-use std::default::Default;
-use std::fmt::{self, Display, Formatter, Write};
-use std::hash::{Hash, Hasher};
+use std::hash::Hash;
 
 use arrow::datatypes::IntervalUnit as ArrowIntervalUnit;
 use serde::{Deserialize, Serialize};
-use serde_json::Value;
-use snafu::ResultExt;
-
-use crate::duration::Duration;
-use crate::error::{Result, TimestampOverflowSnafu};
 
 #[derive(
     Debug, Default, Copy, Clone, PartialEq, Eq, Hash, PartialOrd, Ord, Serialize, Deserialize,
@@ -61,268 +53,269 @@ impl From<ArrowIntervalUnit> for IntervalUnit {
     }
 }
 
-/// Interval Type represents a period of time.
-/// It is composed of months, days and nanoseconds.
-/// 3 kinds of interval are supported: year-month, day-time and
-/// month-day-nano, which will be stored in the following format.
-/// Interval data format:
-/// | months | days   | nsecs      |
-/// | 4bytes | 4bytes | 8bytes     |
-#[derive(Debug, Clone, Default, Copy, Serialize, Deserialize)]
-pub struct Interval {
-    months: i32,
-    days: i32,
-    nsecs: i64,
-    unit: IntervalUnit,
+// The `Value` type requires Serialize, Deserialize.
+#[derive(
+    Debug, Default, Copy, Clone, Eq, PartialEq, Hash, Ord, PartialOrd, Serialize, Deserialize,
+)]
+#[repr(C)]
+pub struct IntervalYearMonth {
+    /// Number of months
+    pub months: i32,
 }
 
-// Nanosecond convert to other time unit
-pub const NANOS_PER_SEC: i64 = 1_000_000_000;
-pub const NANOS_PER_MILLI: i64 = 1_000_000;
-pub const NANOS_PER_MICRO: i64 = 1_000;
-pub const NANOS_PER_HOUR: i64 = 60 * 60 * NANOS_PER_SEC;
-pub const NANOS_PER_DAY: i64 = 24 * NANOS_PER_HOUR;
-pub const NANOS_PER_MONTH: i64 = 30 * NANOS_PER_DAY;
-
-pub const DAYS_PER_MONTH: i64 = 30;
-
-impl Interval {
-    /// Creates a new interval from months, days and nanoseconds.
-    /// Precision is nanosecond.
-    pub fn from_month_day_nano(months: i32, days: i32, nsecs: i64) -> Self {
-        Interval {
-            months,
-            days,
-            nsecs,
-            unit: IntervalUnit::MonthDayNano,
-        }
-    }
-
-    /// Creates a new interval from months.
-    pub fn from_year_month(months: i32) -> Self {
-        Interval {
-            months,
-            days: 0,
-            nsecs: 0,
-            unit: IntervalUnit::YearMonth,
-        }
-    }
-
-    /// Creates a new interval from days and milliseconds.
-    pub fn from_day_time(days: i32, millis: i32) -> Self {
-        Interval {
-            months: 0,
-            days,
-            nsecs: (millis as i64) * NANOS_PER_MILLI,
-            unit: IntervalUnit::DayTime,
-        }
-    }
-
-    pub fn to_duration(&self) -> Result<Duration> {
-        Ok(Duration::new_nanosecond(
-            self.to_nanosecond()
-                .try_into()
-                .context(TimestampOverflowSnafu)?,
-        ))
-    }
-
-    /// Return a tuple(months, days, nanoseconds) from the interval.
-    pub fn to_month_day_nano(&self) -> (i32, i32, i64) {
-        (self.months, self.days, self.nsecs)
-    }
-
-    /// Converts the interval to nanoseconds.
-    pub fn to_nanosecond(&self) -> i128 {
-        let days = (self.days as i64) + DAYS_PER_MONTH * (self.months as i64);
-        (self.nsecs as i128) + (NANOS_PER_DAY as i128) * (days as i128)
-    }
-
-    /// Smallest interval value.
-    pub const MIN: Self = Self {
-        months: i32::MIN,
-        days: i32::MIN,
-        nsecs: i64::MIN,
-        unit: IntervalUnit::MonthDayNano,
-    };
-
-    /// Largest interval value.
-    pub const MAX: Self = Self {
-        months: i32::MAX,
-        days: i32::MAX,
-        nsecs: i64::MAX,
-        unit: IntervalUnit::MonthDayNano,
-    };
-
-    /// Returns the justified interval.
-    /// allows you to adjust the interval of 30-day as one month and the interval of 24-hour as one day
-    pub fn justified_interval(&self) -> Self {
-        let mut result = *self;
-        let extra_months_d = self.days as i64 / DAYS_PER_MONTH;
-        let extra_months_nsecs = self.nsecs / NANOS_PER_MONTH;
-        result.days -= (extra_months_d * DAYS_PER_MONTH) as i32;
-        result.nsecs -= extra_months_nsecs * NANOS_PER_MONTH;
-
-        let extra_days = self.nsecs / NANOS_PER_DAY;
-        result.nsecs -= extra_days * NANOS_PER_DAY;
-
-        result.months += extra_months_d as i32 + extra_months_nsecs as i32;
-        result.days += extra_days as i32;
-
-        result
-    }
-
-    /// Convert Interval to nanoseconds,
-    /// to check whether Interval is positive
-    pub fn is_positive(&self) -> bool {
-        self.to_nanosecond() > 0
-    }
-
-    /// is_zero
-    pub fn is_zero(&self) -> bool {
-        self.months == 0 && self.days == 0 && self.nsecs == 0
-    }
-
-    /// get unit
-    pub fn unit(&self) -> IntervalUnit {
-        self.unit
-    }
-
-    /// Multiple Interval by an integer with overflow check.
-    /// Returns justified Interval, or `None` if overflow occurred.
-    pub fn checked_mul_int<I>(&self, rhs: I) -> Option<Self>
-    where
-        I: TryInto<i32>,
-    {
-        let rhs = rhs.try_into().ok()?;
-        let months = self.months.checked_mul(rhs)?;
-        let days = self.days.checked_mul(rhs)?;
-        let nsecs = self.nsecs.checked_mul(rhs as i64)?;
-
-        Some(
-            Self {
-                months,
-                days,
-                nsecs,
-                unit: self.unit,
-            }
-            .justified_interval(),
-        )
-    }
-
-    /// Convert Interval to ISO 8601 string
-    pub fn to_iso8601_string(self) -> String {
-        IntervalFormat::from(self).to_iso8601_string()
-    }
-
-    /// Convert Interval to postgres verbose string
-    pub fn to_postgres_string(self) -> String {
-        IntervalFormat::from(self).to_postgres_string()
-    }
-
-    /// Convert Interval to sql_standard string
-    pub fn to_sql_standard_string(self) -> String {
-        IntervalFormat::from(self).to_sql_standard_string()
-    }
-
-    /// Interval Type and i128 [IntervalUnit::MonthDayNano] Convert
-    /// v consists of months(i32) | days(i32) | nsecs(i64)
-    pub fn from_i128(v: i128) -> Self {
-        Interval {
-            nsecs: v as i64,
-            days: (v >> 64) as i32,
-            months: (v >> 96) as i32,
-            unit: IntervalUnit::MonthDayNano,
-        }
-    }
-
-    /// `Interval` Type and i64 [IntervalUnit::DayTime] Convert
-    /// v consists of days(i32) | milliseconds(i32)
-    pub fn from_i64(v: i64) -> Self {
-        Interval {
-            nsecs: ((v as i32) as i64) * NANOS_PER_MILLI,
-            days: (v >> 32) as i32,
-            months: 0,
-            unit: IntervalUnit::DayTime,
-        }
-    }
-
-    /// `Interval` Type and i32 [IntervalUnit::YearMonth] Convert
-    /// v consists of months(i32)
-    pub fn from_i32(v: i32) -> Self {
-        Interval {
-            nsecs: 0,
-            days: 0,
-            months: v,
-            unit: IntervalUnit::YearMonth,
-        }
-    }
-
-    pub fn to_i128(&self) -> i128 {
-        // 128            96              64                               0
-        // +-------+-------+-------+-------+-------+-------+-------+-------+
-        // |     months    |      days     |          nanoseconds          |
-        // +-------+-------+-------+-------+-------+-------+-------+-------+
-        let months = (self.months as u128 & u32::MAX as u128) << 96;
-        let days = (self.days as u128 & u32::MAX as u128) << 64;
-        let nsecs = self.nsecs as u128 & u64::MAX as u128;
-        (months | days | nsecs) as i128
-    }
-
-    pub fn to_i64(&self) -> i64 {
-        // 64                              32                              0
-        // +-------+-------+-------+-------+-------+-------+-------+-------+
-        // |             days              |         milliseconds          |
-        // +-------+-------+-------+-------+-------+-------+-------+-------+
-        let days = (self.days as u64 & u32::MAX as u64) << 32;
-        let milliseconds = (self.nsecs / NANOS_PER_MILLI) as u64 & u32::MAX as u64;
-        (days | milliseconds) as i64
+impl IntervalYearMonth {
+    pub fn new(months: i32) -> Self {
+        Self { months }
     }
 
     pub fn to_i32(&self) -> i32 {
         self.months
     }
 
+    pub fn from_i32(months: i32) -> Self {
+        Self { months }
+    }
+
     pub fn negative(&self) -> Self {
-        Self {
-            months: -self.months,
-            days: -self.days,
-            nsecs: -self.nsecs,
-            unit: self.unit,
+        Self::new(-self.months)
+    }
+
+    pub fn to_iso8601_string(&self) -> String {
+        IntervalFormat::from(*self).to_iso8601_string()
+    }
+}
+
+impl From<IntervalYearMonth> for IntervalFormat {
+    fn from(interval: IntervalYearMonth) -> Self {
+        IntervalFormat {
+            years: interval.months / 12,
+            months: interval.months % 12,
+            ..Default::default()
         }
     }
 }
 
-impl From<i128> for Interval {
+impl From<i32> for IntervalYearMonth {
+    fn from(v: i32) -> Self {
+        Self::from_i32(v)
+    }
+}
+
+impl From<IntervalYearMonth> for i32 {
+    fn from(v: IntervalYearMonth) -> Self {
+        v.to_i32()
+    }
+}
+
+impl From<IntervalYearMonth> for serde_json::Value {
+    fn from(v: IntervalYearMonth) -> Self {
+        serde_json::Value::from(v.to_i32())
+    }
+}
+
+#[derive(
+    Debug, Default, Copy, Clone, Eq, PartialEq, Hash, Ord, PartialOrd, Serialize, Deserialize,
+)]
+#[repr(C)]
+pub struct IntervalDayTime {
+    /// Number of days
+    pub days: i32,
+    /// Number of milliseconds
+    pub milliseconds: i32,
+}
+
+impl IntervalDayTime {
+    /// The additive identity i.e. `0`.
+    pub const ZERO: Self = Self::new(0, 0);
+
+    /// The multiplicative inverse, i.e. `-1`.
+    pub const MINUS_ONE: Self = Self::new(-1, -1);
+
+    /// The maximum value that can be represented
+    pub const MAX: Self = Self::new(i32::MAX, i32::MAX);
+
+    /// The minimum value that can be represented
+    pub const MIN: Self = Self::new(i32::MIN, i32::MIN);
+
+    pub const fn new(days: i32, milliseconds: i32) -> Self {
+        Self { days, milliseconds }
+    }
+
+    pub fn to_i64(&self) -> i64 {
+        let d = (self.days as u64 & u32::MAX as u64) << 32;
+        let m = self.milliseconds as u64 & u32::MAX as u64;
+        (d | m) as i64
+    }
+
+    pub fn from_i64(value: i64) -> Self {
+        let days = (value >> 32) as i32;
+        let milliseconds = value as i32;
+        Self { days, milliseconds }
+    }
+
+    pub fn negative(&self) -> Self {
+        Self::new(-self.days, -self.milliseconds)
+    }
+
+    pub fn to_iso8601_string(&self) -> String {
+        IntervalFormat::from(*self).to_iso8601_string()
+    }
+
+    pub fn as_millis(&self) -> i64 {
+        self.days as i64 * MS_PER_DAY + self.milliseconds as i64
+    }
+}
+
+impl From<i64> for IntervalDayTime {
+    fn from(v: i64) -> Self {
+        Self::from_i64(v)
+    }
+}
+
+impl From<IntervalDayTime> for i64 {
+    fn from(v: IntervalDayTime) -> Self {
+        v.to_i64()
+    }
+}
+
+impl From<IntervalDayTime> for serde_json::Value {
+    fn from(v: IntervalDayTime) -> Self {
+        serde_json::Value::from(v.to_i64())
+    }
+}
+
+// Millisecond convert to other time unit
+pub const MS_PER_SEC: i64 = 1_000;
+pub const MS_PER_MINUTE: i64 = 60 * MS_PER_SEC;
+pub const MS_PER_HOUR: i64 = 60 * MS_PER_MINUTE;
+pub const MS_PER_DAY: i64 = 24 * MS_PER_HOUR;
+pub const NANOS_PER_MILLI: i64 = 1_000_000;
+
+impl From<IntervalDayTime> for IntervalFormat {
+    fn from(interval: IntervalDayTime) -> Self {
+        IntervalFormat {
+            days: interval.days,
+            hours: interval.milliseconds as i64 / MS_PER_HOUR,
+            minutes: (interval.milliseconds as i64 % MS_PER_HOUR) / MS_PER_MINUTE,
+            seconds: (interval.milliseconds as i64 % MS_PER_MINUTE) / MS_PER_SEC,
+            microseconds: (interval.milliseconds as i64 % MS_PER_SEC) * MS_PER_SEC,
+            ..Default::default()
+        }
+    }
+}
+
+#[derive(
+    Debug, Default, Copy, Clone, Eq, PartialEq, Hash, Ord, PartialOrd, Serialize, Deserialize,
+)]
+#[repr(C)]
+pub struct IntervalMonthDayNano {
+    /// Number of months
+    pub months: i32,
+    /// Number of days
+    pub days: i32,
+    /// Number of nanoseconds
+    pub nanoseconds: i64,
+}
+
+impl IntervalMonthDayNano {
+    /// The additive identity i.e. `0`.
+    pub const ZERO: Self = Self::new(0, 0, 0);
+
+    /// The multiplicative inverse, i.e. `-1`.
+    pub const MINUS_ONE: Self = Self::new(-1, -1, -1);
+
+    /// The maximum value that can be represented
+    pub const MAX: Self = Self::new(i32::MAX, i32::MAX, i64::MAX);
+
+    /// The minimum value that can be represented
+    pub const MIN: Self = Self::new(i32::MIN, i32::MIN, i64::MIN);
+
+    pub const fn new(months: i32, days: i32, nanoseconds: i64) -> Self {
+        Self {
+            months,
+            days,
+            nanoseconds,
+        }
+    }
+
+    pub fn to_i128(&self) -> i128 {
+        let m = (self.months as u128 & u32::MAX as u128) << 96;
+        let d = (self.days as u128 & u32::MAX as u128) << 64;
+        let n = self.nanoseconds as u128 & u64::MAX as u128;
+        (m | d | n) as i128
+    }
+
+    pub fn from_i128(value: i128) -> Self {
+        let months = (value >> 96) as i32;
+        let days = (value >> 64) as i32;
+        let nanoseconds = value as i64;
+        Self {
+            months,
+            days,
+            nanoseconds,
+        }
+    }
+
+    pub fn negative(&self) -> Self {
+        Self::new(-self.months, -self.days, -self.nanoseconds)
+    }
+
+    pub fn to_iso8601_string(&self) -> String {
+        IntervalFormat::from(*self).to_iso8601_string()
+    }
+}
+
+impl From<i128> for IntervalMonthDayNano {
     fn from(v: i128) -> Self {
         Self::from_i128(v)
     }
 }
 
-impl From<Interval> for i128 {
-    fn from(v: Interval) -> Self {
+impl From<IntervalMonthDayNano> for i128 {
+    fn from(v: IntervalMonthDayNano) -> Self {
         v.to_i128()
     }
 }
 
-impl From<Interval> for serde_json::Value {
-    fn from(v: Interval) -> Self {
-        Value::String(v.to_string())
+impl From<IntervalMonthDayNano> for serde_json::Value {
+    fn from(v: IntervalMonthDayNano) -> Self {
+        serde_json::Value::from(v.to_i128().to_string())
     }
 }
 
-impl Display for Interval {
-    fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result {
-        let mut s = String::new();
-        if self.months != 0 {
-            write!(s, "{} months ", self.months)?;
+// Nanosecond convert to other time unit
+pub const NS_PER_SEC: i64 = 1_000_000_000;
+pub const NS_PER_MINUTE: i64 = 60 * NS_PER_SEC;
+pub const NS_PER_HOUR: i64 = 60 * NS_PER_MINUTE;
+pub const NS_PER_DAY: i64 = 24 * NS_PER_HOUR;
+
+impl From<IntervalMonthDayNano> for IntervalFormat {
+    fn from(interval: IntervalMonthDayNano) -> Self {
+        IntervalFormat {
+            years: interval.months / 12,
+            months: interval.months % 12,
+            days: interval.days,
+            hours: interval.nanoseconds / NS_PER_HOUR,
+            minutes: (interval.nanoseconds % NS_PER_HOUR) / NS_PER_MINUTE,
+            seconds: (interval.nanoseconds % NS_PER_MINUTE) / NS_PER_SEC,
+            microseconds: (interval.nanoseconds % NS_PER_SEC) / 1_000,
         }
-        if self.days != 0 {
-            write!(s, "{} days ", self.days)?;
-        }
-        if self.nsecs != 0 {
-            write!(s, "{} nsecs", self.nsecs)?;
-        }
-        write!(f, "{}", s.trim())
+    }
+}
+
+pub fn interval_year_month_to_month_day_nano(interval: IntervalYearMonth) -> IntervalMonthDayNano {
+    IntervalMonthDayNano {
+        months: interval.months,
+        days: 0,
+        nanoseconds: 0,
+    }
+}
+
+pub fn interval_day_time_to_month_day_nano(interval: IntervalDayTime) -> IntervalMonthDayNano {
+    IntervalMonthDayNano {
+        months: 0,
+        days: interval.days,
+        nanoseconds: interval.milliseconds as i64 * NANOS_PER_MILLI,
     }
 }
 
@@ -339,31 +332,6 @@ pub struct IntervalFormat {
     pub microseconds: i64,
 }
 
-impl From<Interval> for IntervalFormat {
-    fn from(val: Interval) -> IntervalFormat {
-        let months = val.months;
-        let days = val.days;
-        let microseconds = val.nsecs / NANOS_PER_MICRO;
-        let years = (months - (months % 12)) / 12;
-        let months = months - years * 12;
-        let hours = (microseconds - (microseconds % 3_600_000_000)) / 3_600_000_000;
-        let microseconds = microseconds - hours * 3_600_000_000;
-        let minutes = (microseconds - (microseconds % 60_000_000)) / 60_000_000;
-        let microseconds = microseconds - minutes * 60_000_000;
-        let seconds = (microseconds - (microseconds % 1_000_000)) / 1_000_000;
-        let microseconds = microseconds - seconds * 1_000_000;
-        IntervalFormat {
-            years,
-            months,
-            days,
-            hours,
-            minutes,
-            seconds,
-            microseconds,
-        }
-    }
-}
-
 impl IntervalFormat {
     /// All the field in the interval is 0
     pub fn is_zero(&self) -> bool {
@@ -540,117 +508,37 @@ fn get_time_part(
     interval
 }
 
-/// IntervalCompare is used to compare two intervals
-/// It makes interval into nanoseconds style.
-#[derive(PartialEq, Eq, Hash, PartialOrd, Ord)]
-struct IntervalCompare(i128);
-
-impl From<Interval> for IntervalCompare {
-    fn from(interval: Interval) -> Self {
-        Self(interval.to_nanosecond())
-    }
-}
-
-impl Ord for Interval {
-    fn cmp(&self, other: &Self) -> Ordering {
-        IntervalCompare::from(*self).cmp(&IntervalCompare::from(*other))
-    }
-}
-
-impl PartialOrd for Interval {
-    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
-        Some(self.cmp(other))
-    }
-}
-
-impl Eq for Interval {}
-
-impl PartialEq for Interval {
-    fn eq(&self, other: &Self) -> bool {
-        self.cmp(other).is_eq()
-    }
-}
-
-impl Hash for Interval {
-    fn hash<H: Hasher>(&self, state: &mut H) {
-        IntervalCompare::from(*self).hash(state)
-    }
-}
-
 #[cfg(test)]
 mod tests {
-    use std::collections::HashMap;
-
     use super::*;
-    use crate::timestamp::TimeUnit;
 
     #[test]
     fn test_from_year_month() {
-        let interval = Interval::from_year_month(1);
+        let interval = IntervalYearMonth::new(1);
         assert_eq!(interval.months, 1);
     }
 
     #[test]
     fn test_from_date_time() {
-        let interval = Interval::from_day_time(1, 2);
+        let interval = IntervalDayTime::new(1, 2);
         assert_eq!(interval.days, 1);
-        assert_eq!(interval.nsecs, 2_000_000);
+        assert_eq!(interval.milliseconds, 2);
     }
 
     #[test]
-    fn test_to_duration() {
-        let interval = Interval::from_day_time(1, 2);
-
-        let duration = interval.to_duration().unwrap();
-        assert_eq!(86400002000000, duration.value());
-        assert_eq!(TimeUnit::Nanosecond, duration.unit());
-
-        let interval = Interval::from_year_month(12);
-
-        let duration = interval.to_duration().unwrap();
-        assert_eq!(31104000000000000, duration.value());
-        assert_eq!(TimeUnit::Nanosecond, duration.unit());
-    }
-
-    #[test]
-    fn test_interval_is_positive() {
-        let interval = Interval::from_year_month(1);
-        assert!(interval.is_positive());
-        let interval = Interval::from_year_month(-1);
-        assert!(!interval.is_positive());
-
-        let interval = Interval::from_day_time(1, i32::MIN);
-        assert!(!interval.is_positive());
-    }
-
-    #[test]
-    fn test_to_nanosecond() {
-        let interval = Interval::from_year_month(1);
-        assert_eq!(interval.to_nanosecond(), 2592000000000000);
-        let interval = Interval::from_day_time(1, 2);
-        assert_eq!(interval.to_nanosecond(), 86400002000000);
-
-        let max_interval = Interval::from_month_day_nano(i32::MAX, i32::MAX, i64::MAX);
-        assert_eq!(max_interval.to_nanosecond(), 5751829423496836854775807);
-
-        let min_interval = Interval::from_month_day_nano(i32::MIN, i32::MIN, i64::MIN);
-        assert_eq!(min_interval.to_nanosecond(), -5751829426175236854775808);
-    }
-
-    #[test]
-    fn test_interval_is_zero() {
-        let interval = Interval::from_month_day_nano(1, 1, 1);
-        assert!(!interval.is_zero());
-        let interval = Interval::from_month_day_nano(0, 0, 0);
-        assert!(interval.is_zero());
+    fn test_from_month_day_nano() {
+        let interval = IntervalMonthDayNano::new(1, 2, 3);
+        assert_eq!(interval.months, 1);
+        assert_eq!(interval.days, 2);
+        assert_eq!(interval.nanoseconds, 3);
     }
 
     #[test]
     fn test_interval_i128_convert() {
         let test_interval_eq = |month, day, nano| {
-            let interval = Interval::from_month_day_nano(month, day, nano);
+            let interval = IntervalMonthDayNano::new(month, day, nano);
             let interval_i128 = interval.to_i128();
-            let interval2 = Interval::from_i128(interval_i128);
+            let interval2 = IntervalMonthDayNano::from_i128(interval_i128);
             assert_eq!(interval, interval2);
         };
 
@@ -666,11 +554,26 @@ mod tests {
         test_interval_eq(i32::MAX, i32::MIN, i64::MIN);
         test_interval_eq(i32::MIN, i32::MAX, i64::MIN);
         test_interval_eq(i32::MIN, i32::MIN, i64::MIN);
+
+        let interval = IntervalMonthDayNano::from_i128(1);
+        assert_eq!(interval, IntervalMonthDayNano::new(0, 0, 1));
+        assert_eq!(1, IntervalMonthDayNano::new(0, 0, 1).to_i128());
+    }
+
+    #[test]
+    fn test_interval_i64_convert() {
+        let interval = IntervalDayTime::from_i64(1);
+        assert_eq!(interval, IntervalDayTime::new(0, 1));
+        assert_eq!(1, IntervalDayTime::new(0, 1).to_i64());
     }
 
     #[test]
     fn test_convert_interval_format() {
-        let interval = Interval::from_month_day_nano(14, 160, 1000000);
+        let interval = IntervalMonthDayNano {
+            months: 14,
+            days: 160,
+            nanoseconds: 1000000,
+        };
         let interval_format = IntervalFormat::from(interval);
         assert_eq!(interval_format.years, 1);
         assert_eq!(interval_format.months, 2);
@@ -681,94 +584,34 @@ mod tests {
         assert_eq!(interval_format.microseconds, 1000);
     }
 
-    #[test]
-    fn test_interval_hash() {
-        let interval = Interval::from_month_day_nano(1, 31, 1);
-        let interval2 = Interval::from_month_day_nano(2, 1, 1);
-        let mut map = HashMap::new();
-        map.insert(interval, 1);
-        assert_eq!(map.get(&interval2), Some(&1));
-    }
-
-    #[test]
-    fn test_interval_mul_int() {
-        let interval = Interval::from_month_day_nano(1, 1, 1);
-        let interval2 = interval.checked_mul_int(2).unwrap();
-        assert_eq!(interval2.months, 2);
-        assert_eq!(interval2.days, 2);
-        assert_eq!(interval2.nsecs, 2);
-
-        // test justified interval
-        let interval = Interval::from_month_day_nano(1, 31, 1);
-        let interval2 = interval.checked_mul_int(2).unwrap();
-        assert_eq!(interval2.months, 4);
-        assert_eq!(interval2.days, 2);
-        assert_eq!(interval2.nsecs, 2);
-
-        // test overflow situation
-        let interval = Interval::from_month_day_nano(i32::MAX, 1, 1);
-        let interval2 = interval.checked_mul_int(2);
-        assert!(interval2.is_none());
-    }
-
-    #[test]
-    fn test_display() {
-        let interval = Interval::from_month_day_nano(1, 1, 1);
-        assert_eq!(interval.to_string(), "1 months 1 days 1 nsecs");
-
-        let interval = Interval::from_month_day_nano(14, 31, 10000000000);
-        assert_eq!(interval.to_string(), "14 months 31 days 10000000000 nsecs");
-    }
-
-    #[test]
-    fn test_interval_justified() {
-        let interval = Interval::from_month_day_nano(1, 131, 1).justified_interval();
-        let interval2 = Interval::from_month_day_nano(5, 11, 1);
-        assert_eq!(interval, interval2);
-
-        let interval = Interval::from_month_day_nano(1, 1, NANOS_PER_MONTH + 2 * NANOS_PER_DAY)
-            .justified_interval();
-        let interval2 = Interval::from_month_day_nano(2, 3, 0);
-        assert_eq!(interval, interval2);
-    }
-
-    #[test]
-    fn test_serde_json() {
-        let interval = Interval::from_month_day_nano(1, 1, 1);
-        let json = serde_json::to_string(&interval).unwrap();
-        assert_eq!(
-            json,
-            "{\"months\":1,\"days\":1,\"nsecs\":1,\"unit\":\"MonthDayNano\"}"
-        );
-        let interval2: Interval = serde_json::from_str(&json).unwrap();
-        assert_eq!(interval, interval2);
-    }
-
     #[test]
     fn test_to_iso8601_string() {
         // Test interval zero
-        let interval = Interval::from_month_day_nano(0, 0, 0);
+        let interval = IntervalMonthDayNano::new(0, 0, 0);
         assert_eq!(interval.to_iso8601_string(), "PT0S");
 
-        let interval = Interval::from_month_day_nano(1, 1, 1);
+        let interval = IntervalMonthDayNano::new(1, 1, 1);
         assert_eq!(interval.to_iso8601_string(), "P0Y1M1DT0H0M0S");
 
-        let interval = Interval::from_month_day_nano(14, 31, 10000000000);
+        let interval = IntervalMonthDayNano::new(14, 31, 10000000000);
         assert_eq!(interval.to_iso8601_string(), "P1Y2M31DT0H0M10S");
 
-        let interval = Interval::from_month_day_nano(14, 31, 23210200000000);
+        let interval = IntervalMonthDayNano::new(14, 31, 23210200000000);
         assert_eq!(interval.to_iso8601_string(), "P1Y2M31DT6H26M50.2S");
     }
 
     #[test]
     fn test_to_postgres_string() {
         // Test interval zero
-        let interval = Interval::from_month_day_nano(0, 0, 0);
-        assert_eq!(interval.to_postgres_string(), "00:00:00");
-
-        let interval = Interval::from_month_day_nano(23, 100, 23210200000000);
+        let interval = IntervalMonthDayNano::new(0, 0, 0);
         assert_eq!(
-            interval.to_postgres_string(),
+            IntervalFormat::from(interval).to_postgres_string(),
+            "00:00:00"
+        );
+
+        let interval = IntervalMonthDayNano::new(23, 100, 23210200000000);
+        assert_eq!(
+            IntervalFormat::from(interval).to_postgres_string(),
             "1 year 11 mons 100 days 06:26:50.200000"
         );
     }
@@ -776,18 +619,21 @@ mod tests {
     #[test]
     fn test_to_sql_standard_string() {
         // Test zero interval
-        let interval = Interval::from_month_day_nano(0, 0, 0);
-        assert_eq!(interval.to_sql_standard_string(), "0");
+        let interval = IntervalMonthDayNano::new(0, 0, 0);
+        assert_eq!(IntervalFormat::from(interval).to_sql_standard_string(), "0");
 
-        let interval = Interval::from_month_day_nano(23, 100, 23210200000000);
+        let interval = IntervalMonthDayNano::new(23, 100, 23210200000000);
         assert_eq!(
-            interval.to_sql_standard_string(),
+            IntervalFormat::from(interval).to_sql_standard_string(),
             "+1-11 +100 +6:26:50.200000"
         );
 
         // Test interval without year, month, day
-        let interval = Interval::from_month_day_nano(0, 0, 23210200000000);
-        assert_eq!(interval.to_sql_standard_string(), "6:26:50.200000");
+        let interval = IntervalMonthDayNano::new(0, 0, 23210200000000);
+        assert_eq!(
+            IntervalFormat::from(interval).to_sql_standard_string(),
+            "6:26:50.200000"
+        );
     }
 
     #[test]

src/common/time/src/lib.rs

@@ -27,7 +27,7 @@ pub mod util;
 pub use date::Date;
 pub use datetime::DateTime;
 pub use duration::Duration;
-pub use interval::Interval;
+pub use interval::{IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth};
 pub use range::RangeMillis;
 pub use timestamp::Timestamp;
 pub use timestamp_millis::TimestampMillis;

src/common/time/src/timestamp.rs

@@ -20,16 +20,17 @@ use std::time::Duration;
 
 use arrow::datatypes::TimeUnit as ArrowTimeUnit;
 use chrono::{
-    DateTime, Days, LocalResult, Months, NaiveDate, NaiveDateTime, NaiveTime,
+    DateTime, Days, LocalResult, Months, NaiveDate, NaiveDateTime, NaiveTime, TimeDelta,
     TimeZone as ChronoTimeZone, Utc,
 };
 use serde::{Deserialize, Serialize};
 use snafu::{OptionExt, ResultExt};
 
+use crate::error;
 use crate::error::{ArithmeticOverflowSnafu, ParseTimestampSnafu, Result, TimestampOverflowSnafu};
+use crate::interval::{IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth};
 use crate::timezone::{get_timezone, Timezone};
 use crate::util::{datetime_to_utc, div_ceil};
-use crate::{error, Interval};
 
 /// Timestamp represents the value of units(seconds/milliseconds/microseconds/nanoseconds) elapsed
 /// since UNIX epoch. The valid value range of [Timestamp] depends on it's unit (all in UTC timezone):
@@ -140,40 +141,77 @@ impl Timestamp {
         })
     }
 
-    /// Adds given Interval to the current timestamp.
-    /// Returns None if the resulting timestamp would be out of range.
-    pub fn add_interval(&self, interval: Interval) -> Option<Timestamp> {
+    // FIXME(yingwen): remove add/sub intervals later
+    /// Adds given [IntervalYearMonth] to the current timestamp.
+    pub fn add_year_month(&self, interval: IntervalYearMonth) -> Option<Timestamp> {
         let naive_datetime = self.to_chrono_datetime()?;
-        let (months, days, nsecs) = interval.to_month_day_nano();
 
-        let naive_datetime = naive_datetime
-            .checked_add_months(Months::new(months as u32))?
-            .checked_add_days(Days::new(days as u64))?
-            + Duration::from_nanos(nsecs as u64);
+        let naive_datetime =
+            naive_datetime.checked_add_months(Months::new(interval.months as u32))?;
 
-        match Timestamp::from_chrono_datetime(naive_datetime) {
-            // Have to convert the new timestamp by the current unit.
-            Some(ts) => ts.convert_to(self.unit),
-            None => None,
-        }
+        // Have to convert the new timestamp by the current unit.
+        Timestamp::from_chrono_datetime(naive_datetime).and_then(|ts| ts.convert_to(self.unit))
     }
 
-    /// Subtracts given Interval to the current timestamp.
-    /// Returns None if the resulting timestamp would be out of range.
-    pub fn sub_interval(&self, interval: Interval) -> Option<Timestamp> {
+    /// Adds given [IntervalDayTime] to the current timestamp.
+    pub fn add_day_time(&self, interval: IntervalDayTime) -> Option<Timestamp> {
         let naive_datetime = self.to_chrono_datetime()?;
-        let (months, days, nsecs) = interval.to_month_day_nano();
 
         let naive_datetime = naive_datetime
-            .checked_sub_months(Months::new(months as u32))?
-            .checked_sub_days(Days::new(days as u64))?
-            - Duration::from_nanos(nsecs as u64);
+            .checked_add_days(Days::new(interval.days as u64))?
+            .checked_add_signed(TimeDelta::milliseconds(interval.milliseconds as i64))?;
 
-        match Timestamp::from_chrono_datetime(naive_datetime) {
-            // Have to convert the new timestamp by the current unit.
-            Some(ts) => ts.convert_to(self.unit),
-            None => None,
-        }
+        // Have to convert the new timestamp by the current unit.
+        Timestamp::from_chrono_datetime(naive_datetime).and_then(|ts| ts.convert_to(self.unit))
+    }
+
+    /// Adds given [IntervalMonthDayNano] to the current timestamp.
+    pub fn add_month_day_nano(&self, interval: IntervalMonthDayNano) -> Option<Timestamp> {
+        let naive_datetime = self.to_chrono_datetime()?;
+
+        let naive_datetime = naive_datetime
+            .checked_add_months(Months::new(interval.months as u32))?
+            .checked_add_days(Days::new(interval.days as u64))?
+            .checked_add_signed(TimeDelta::nanoseconds(interval.nanoseconds))?;
+
+        // Have to convert the new timestamp by the current unit.
+        Timestamp::from_chrono_datetime(naive_datetime).and_then(|ts| ts.convert_to(self.unit))
+    }
+
+    /// Subtracts given [IntervalYearMonth] to the current timestamp.
+    pub fn sub_year_month(&self, interval: IntervalYearMonth) -> Option<Timestamp> {
+        let naive_datetime = self.to_chrono_datetime()?;
+
+        let naive_datetime =
+            naive_datetime.checked_sub_months(Months::new(interval.months as u32))?;
+
+        // Have to convert the new timestamp by the current unit.
+        Timestamp::from_chrono_datetime(naive_datetime).and_then(|ts| ts.convert_to(self.unit))
+    }
+
+    /// Subtracts given [IntervalDayTime] to the current timestamp.
+    pub fn sub_day_time(&self, interval: IntervalDayTime) -> Option<Timestamp> {
+        let naive_datetime = self.to_chrono_datetime()?;
+
+        let naive_datetime = naive_datetime
+            .checked_sub_days(Days::new(interval.days as u64))?
+            .checked_sub_signed(TimeDelta::milliseconds(interval.milliseconds as i64))?;
+
+        // Have to convert the new timestamp by the current unit.
+        Timestamp::from_chrono_datetime(naive_datetime).and_then(|ts| ts.convert_to(self.unit))
+    }
+
+    /// Subtracts given [IntervalMonthDayNano] to the current timestamp.
+    pub fn sub_month_day_nano(&self, interval: IntervalMonthDayNano) -> Option<Timestamp> {
+        let naive_datetime = self.to_chrono_datetime()?;
+
+        let naive_datetime = naive_datetime
+            .checked_sub_months(Months::new(interval.months as u32))?
+            .checked_sub_days(Days::new(interval.days as u64))?
+            .checked_sub_signed(TimeDelta::nanoseconds(interval.nanoseconds))?;
+
+        // Have to convert the new timestamp by the current unit.
+        Timestamp::from_chrono_datetime(naive_datetime).and_then(|ts| ts.convert_to(self.unit))
     }
 
     /// Subtracts current timestamp with another timestamp, yielding a duration.
@@ -688,13 +726,13 @@ mod tests {
     fn test_add_sub_interval() {
         let ts = Timestamp::new(1000, TimeUnit::Millisecond);
 
-        let interval = Interval::from_day_time(1, 200);
+        let interval = IntervalDayTime::new(1, 200);
 
-        let new_ts = ts.add_interval(interval).unwrap();
+        let new_ts = ts.add_day_time(interval).unwrap();
         assert_eq!(new_ts.unit(), TimeUnit::Millisecond);
         assert_eq!(new_ts.value(), 1000 + 3600 * 24 * 1000 + 200);
 
-        assert_eq!(ts, new_ts.sub_interval(interval).unwrap());
+        assert_eq!(ts, new_ts.sub_day_time(interval).unwrap());
     }
 
     #[test]

src/common/wal/src/config/kafka/datanode.rs

@@ -46,6 +46,8 @@ pub struct DatanodeKafkaConfig {
     pub create_index: bool,
     #[serde(with = "humantime_serde")]
     pub dump_index_interval: Duration,
+    /// Ignore missing entries during read WAL.
+    pub overwrite_entry_start_id: bool,
 }
 
 impl Default for DatanodeKafkaConfig {
@@ -60,6 +62,7 @@ impl Default for DatanodeKafkaConfig {
             auto_create_topics: true,
             create_index: true,
             dump_index_interval: Duration::from_secs(60),
+            overwrite_entry_start_id: false,
         }
     }
 }

src/datanode/src/alive_keeper.rs

@@ -129,8 +129,10 @@ impl RegionAliveKeeper {
         let request = RegionRequest::Close(RegionCloseRequest {});
         if let Err(e) = self.region_server.handle_request(region_id, request).await {
             if e.status_code() != StatusCode::RegionNotFound {
-                let _ = self.region_server.set_writable(region_id, false);
-                error!(e; "Failed to close staled region {}, set region to readonly.",region_id);
+                let _ = self
+                    .region_server
+                    .set_region_role(region_id, RegionRole::Follower);
+                error!(e; "Failed to close staled region {}, convert region to follower.", region_id);
             }
         }
     }
@@ -378,7 +380,7 @@ impl CountdownTask {
                             }
                         },
                         Some(CountdownCommand::Reset((role, deadline))) => {
-                            let _ = self.region_server.set_writable(self.region_id, role.writable());
+                            let _ = self.region_server.set_region_role(self.region_id, role);
                             trace!(
                                 "Reset deadline of region {region_id} to approximately {} seconds later.",
                                 (deadline - Instant::now()).as_secs_f32(),
@@ -399,8 +401,8 @@ impl CountdownTask {
                     }
                 }
                 () = &mut countdown => {
-                    warn!("The region {region_id} lease is expired, set region to readonly.");
-                    let _ = self.region_server.set_writable(self.region_id, false);
+                    warn!("The region {region_id} lease is expired, convert region to follower.");
+                    let _ = self.region_server.set_region_role(self.region_id, RegionRole::Follower);
                     // resets the countdown.
                     let far_future = Instant::now() + Duration::from_secs(86400 * 365 * 30);
                     countdown.as_mut().reset(far_future);
@@ -436,7 +438,9 @@ mod test {
             .handle_request(region_id, RegionRequest::Create(builder.build()))
             .await
             .unwrap();
-        region_server.set_writable(region_id, true).unwrap();
+        region_server
+            .set_region_role(region_id, RegionRole::Leader)
+            .unwrap();
 
         // Register a region before starting.
         alive_keeper.register_region(region_id).await;

src/datanode/src/datanode.rs

@@ -47,7 +47,7 @@ use servers::server::ServerHandlers;
 use servers::Mode;
 use snafu::{ensure, OptionExt, ResultExt};
 use store_api::path_utils::{region_dir, WAL_DIR};
-use store_api::region_engine::RegionEngineRef;
+use store_api::region_engine::{RegionEngineRef, RegionRole};
 use store_api::region_request::RegionOpenRequest;
 use store_api::storage::RegionId;
 use tokio::fs;
@@ -546,9 +546,9 @@ async fn open_all_regions(
 
     for region_id in open_regions {
         if open_with_writable {
-            if let Err(e) = region_server.set_writable(region_id, true) {
+            if let Err(e) = region_server.set_region_role(region_id, RegionRole::Leader) {
                 error!(
-                    e; "failed to set writable for region {region_id}"
+                    e; "failed to convert region {region_id} to leader"
                 );
             }
         }

src/datanode/src/heartbeat.rs

@@ -126,7 +126,9 @@ impl HeartbeatTask {
                     let mut follower_region_lease_count = 0;
                     for lease in &lease.regions {
                         match lease.role() {
-                            RegionRole::Leader => leader_region_lease_count += 1,
+                            RegionRole::Leader | RegionRole::DowngradingLeader => {
+                                leader_region_lease_count += 1
+                            }
                             RegionRole::Follower => follower_region_lease_count += 1,
                         }
                     }

src/datanode/src/heartbeat/handler.rs

@@ -153,6 +153,7 @@ mod tests {
     use mito2::engine::MITO_ENGINE_NAME;
     use mito2::test_util::{CreateRequestBuilder, TestEnv};
     use store_api::path_utils::region_dir;
+    use store_api::region_engine::RegionRole;
     use store_api::region_request::{RegionCloseRequest, RegionRequest};
     use store_api::storage::RegionId;
     use tokio::sync::mpsc::{self, Receiver};
@@ -213,6 +214,7 @@ mod tests {
         let instruction = Instruction::DowngradeRegion(DowngradeRegion {
             region_id: RegionId::new(2048, 1),
             flush_timeout: Some(Duration::from_secs(1)),
+            reject_write: false,
         });
         assert!(heartbeat_handler
             .is_acceptable(&heartbeat_env.create_handler_ctx((meta.clone(), instruction))));
@@ -295,7 +297,9 @@ mod tests {
             }
 
             assert_matches!(
-                region_server.set_writable(region_id, true).unwrap_err(),
+                region_server
+                    .set_region_role(region_id, RegionRole::Leader)
+                    .unwrap_err(),
                 error::Error::RegionNotFound { .. }
             );
         }
@@ -411,6 +415,7 @@ mod tests {
             let instruction = Instruction::DowngradeRegion(DowngradeRegion {
                 region_id,
                 flush_timeout: Some(Duration::from_secs(1)),
+                reject_write: false,
             });
 
             let mut ctx = heartbeat_env.create_handler_ctx((meta, instruction));
@@ -433,6 +438,7 @@ mod tests {
         let instruction = Instruction::DowngradeRegion(DowngradeRegion {
             region_id: RegionId::new(2048, 1),
             flush_timeout: Some(Duration::from_secs(1)),
+            reject_write: false,
         });
         let mut ctx = heartbeat_env.create_handler_ctx((meta, instruction));
         let control = heartbeat_handler.handle(&mut ctx).await.unwrap();

src/datanode/src/heartbeat/handler/downgrade_region.rs

@@ -16,7 +16,7 @@ use common_meta::instruction::{DowngradeRegion, DowngradeRegionReply, Instructio
 use common_telemetry::tracing::info;
 use common_telemetry::warn;
 use futures_util::future::BoxFuture;
-use store_api::region_engine::SetReadonlyResponse;
+use store_api::region_engine::{SetRegionRoleStateResponse, SettableRegionRoleState};
 use store_api::region_request::{RegionFlushRequest, RegionRequest};
 use store_api::storage::RegionId;
 
@@ -24,16 +24,20 @@ use crate::heartbeat::handler::HandlerContext;
 use crate::heartbeat::task_tracker::WaitResult;
 
 impl HandlerContext {
-    async fn set_readonly_gracefully(&self, region_id: RegionId) -> InstructionReply {
-        match self.region_server.set_readonly_gracefully(region_id).await {
-            Ok(SetReadonlyResponse::Success { last_entry_id }) => {
+    async fn downgrade_to_follower_gracefully(&self, region_id: RegionId) -> InstructionReply {
+        match self
+            .region_server
+            .set_region_role_state_gracefully(region_id, SettableRegionRoleState::Follower)
+            .await
+        {
+            Ok(SetRegionRoleStateResponse::Success { last_entry_id }) => {
                 InstructionReply::DowngradeRegion(DowngradeRegionReply {
                     last_entry_id,
                     exists: true,
                     error: None,
                 })
             }
-            Ok(SetReadonlyResponse::NotFound) => {
+            Ok(SetRegionRoleStateResponse::NotFound) => {
                 InstructionReply::DowngradeRegion(DowngradeRegionReply {
                     last_entry_id: None,
                     exists: false,
@@ -53,10 +57,12 @@ impl HandlerContext {
         DowngradeRegion {
             region_id,
             flush_timeout,
+            reject_write,
         }: DowngradeRegion,
     ) -> BoxFuture<'static, InstructionReply> {
         Box::pin(async move {
-            let Some(writable) = self.region_server.is_writable(region_id) else {
+            let Some(writable) = self.region_server.is_region_leader(region_id) else {
+                warn!("Region: {region_id} is not found");
                 return InstructionReply::DowngradeRegion(DowngradeRegionReply {
                     last_entry_id: None,
                     exists: false,
@@ -64,61 +70,89 @@ impl HandlerContext {
                 });
             };
 
+            let region_server_moved = self.region_server.clone();
+
             // Ignores flush request
             if !writable {
-                return self.set_readonly_gracefully(region_id).await;
+                return self.downgrade_to_follower_gracefully(region_id).await;
             }
 
-            let region_server_moved = self.region_server.clone();
-            if let Some(flush_timeout) = flush_timeout {
-                let register_result = self
-                    .downgrade_tasks
-                    .try_register(
+            // If flush_timeout is not set, directly convert region to follower.
+            let Some(flush_timeout) = flush_timeout else {
+                return self.downgrade_to_follower_gracefully(region_id).await;
+            };
+
+            if reject_write {
+                // Sets region to downgrading, the downgrading region will reject all write requests.
+                match self
+                    .region_server
+                    .set_region_role_state_gracefully(
                         region_id,
-                        Box::pin(async move {
-                            info!("Flush region: {region_id} before downgrading region");
-                            region_server_moved
-                                .handle_request(
-                                    region_id,
-                                    RegionRequest::Flush(RegionFlushRequest {
-                                        row_group_size: None,
-                                    }),
-                                )
-                                .await?;
-
-                            Ok(())
-                        }),
+                        SettableRegionRoleState::DowngradingLeader,
                     )
-                    .await;
-
-                if register_result.is_busy() {
-                    warn!("Another flush task is running for the region: {region_id}");
-                }
-
-                let mut watcher = register_result.into_watcher();
-                let result = self.catchup_tasks.wait(&mut watcher, flush_timeout).await;
-
-                match result {
-                    WaitResult::Timeout => {
-                        InstructionReply::DowngradeRegion(DowngradeRegionReply {
+                    .await
+                {
+                    Ok(SetRegionRoleStateResponse::Success { .. }) => {}
+                    Ok(SetRegionRoleStateResponse::NotFound) => {
+                        warn!("Region: {region_id} is not found");
+                        return InstructionReply::DowngradeRegion(DowngradeRegionReply {
                             last_entry_id: None,
-                            exists: true,
-                            error: Some(format!(
-                                "Flush region: {region_id} before downgrading region is timeout"
-                            )),
-                        })
+                            exists: false,
+                            error: None,
+                        });
                     }
-                    WaitResult::Finish(Ok(_)) => self.set_readonly_gracefully(region_id).await,
-                    WaitResult::Finish(Err(err)) => {
-                        InstructionReply::DowngradeRegion(DowngradeRegionReply {
+                    Err(err) => {
+                        warn!(err; "Failed to convert region to downgrading leader");
+                        return InstructionReply::DowngradeRegion(DowngradeRegionReply {
                             last_entry_id: None,
                             exists: true,
                             error: Some(format!("{err:?}")),
-                        })
+                        });
                     }
                 }
-            } else {
-                self.set_readonly_gracefully(region_id).await
+            }
+
+            let register_result = self
+                .downgrade_tasks
+                .try_register(
+                    region_id,
+                    Box::pin(async move {
+                        info!("Flush region: {region_id} before converting region to follower");
+                        region_server_moved
+                            .handle_request(
+                                region_id,
+                                RegionRequest::Flush(RegionFlushRequest {
+                                    row_group_size: None,
+                                }),
+                            )
+                            .await?;
+
+                        Ok(())
+                    }),
+                )
+                .await;
+
+            if register_result.is_busy() {
+                warn!("Another flush task is running for the region: {region_id}");
+            }
+
+            let mut watcher = register_result.into_watcher();
+            let result = self.catchup_tasks.wait(&mut watcher, flush_timeout).await;
+
+            match result {
+                WaitResult::Timeout => InstructionReply::DowngradeRegion(DowngradeRegionReply {
+                    last_entry_id: None,
+                    exists: true,
+                    error: Some(format!("Flush region: {region_id} is timeout")),
+                }),
+                WaitResult::Finish(Ok(_)) => self.downgrade_to_follower_gracefully(region_id).await,
+                WaitResult::Finish(Err(err)) => {
+                    InstructionReply::DowngradeRegion(DowngradeRegionReply {
+                        last_entry_id: None,
+                        exists: true,
+                        error: Some(format!("{err:?}")),
+                    })
+                }
             }
         })
     }
@@ -131,7 +165,7 @@ mod tests {
 
     use common_meta::instruction::{DowngradeRegion, InstructionReply};
     use mito2::engine::MITO_ENGINE_NAME;
-    use store_api::region_engine::{RegionRole, SetReadonlyResponse};
+    use store_api::region_engine::{RegionRole, SetRegionRoleStateResponse};
     use store_api::region_request::RegionRequest;
     use store_api::storage::RegionId;
     use tokio::time::Instant;
@@ -155,6 +189,7 @@ mod tests {
                 .handle_downgrade_region_instruction(DowngradeRegion {
                     region_id,
                     flush_timeout,
+                    reject_write: false,
                 })
                 .await;
             assert_matches!(reply, InstructionReply::DowngradeRegion(_));
@@ -182,8 +217,9 @@ mod tests {
 
                     Ok(0)
                 }));
-                region_engine.handle_set_readonly_gracefully_mock_fn =
-                    Some(Box::new(|_| Ok(SetReadonlyResponse::success(Some(1024)))))
+                region_engine.handle_set_readonly_gracefully_mock_fn = Some(Box::new(|_| {
+                    Ok(SetRegionRoleStateResponse::success(Some(1024)))
+                }))
             });
         mock_region_server.register_test_region(region_id, mock_engine);
         let handler_context = HandlerContext::new_for_test(mock_region_server);
@@ -195,6 +231,7 @@ mod tests {
                 .handle_downgrade_region_instruction(DowngradeRegion {
                     region_id,
                     flush_timeout,
+                    reject_write: false,
                 })
                 .await;
             assert_matches!(reply, InstructionReply::DowngradeRegion(_));
@@ -215,8 +252,9 @@ mod tests {
             MockRegionEngine::with_custom_apply_fn(MITO_ENGINE_NAME, |region_engine| {
                 region_engine.mock_role = Some(Some(RegionRole::Leader));
                 region_engine.handle_request_delay = Some(Duration::from_secs(100));
-                region_engine.handle_set_readonly_gracefully_mock_fn =
-                    Some(Box::new(|_| Ok(SetReadonlyResponse::success(Some(1024)))))
+                region_engine.handle_set_readonly_gracefully_mock_fn = Some(Box::new(|_| {
+                    Ok(SetRegionRoleStateResponse::success(Some(1024)))
+                }))
             });
         mock_region_server.register_test_region(region_id, mock_engine);
         let handler_context = HandlerContext::new_for_test(mock_region_server);
@@ -227,6 +265,7 @@ mod tests {
             .handle_downgrade_region_instruction(DowngradeRegion {
                 region_id,
                 flush_timeout: Some(flush_timeout),
+                reject_write: false,
             })
             .await;
         assert_matches!(reply, InstructionReply::DowngradeRegion(_));
@@ -246,8 +285,9 @@ mod tests {
             MockRegionEngine::with_custom_apply_fn(MITO_ENGINE_NAME, |region_engine| {
                 region_engine.mock_role = Some(Some(RegionRole::Leader));
                 region_engine.handle_request_delay = Some(Duration::from_millis(300));
-                region_engine.handle_set_readonly_gracefully_mock_fn =
-                    Some(Box::new(|_| Ok(SetReadonlyResponse::success(Some(1024)))))
+                region_engine.handle_set_readonly_gracefully_mock_fn = Some(Box::new(|_| {
+                    Ok(SetRegionRoleStateResponse::success(Some(1024)))
+                }))
             });
         mock_region_server.register_test_region(region_id, mock_engine);
         let handler_context = HandlerContext::new_for_test(mock_region_server);
@@ -263,6 +303,7 @@ mod tests {
                 .handle_downgrade_region_instruction(DowngradeRegion {
                     region_id,
                     flush_timeout,
+                    reject_write: false,
                 })
                 .await;
             assert_matches!(reply, InstructionReply::DowngradeRegion(_));
@@ -277,6 +318,7 @@ mod tests {
             .handle_downgrade_region_instruction(DowngradeRegion {
                 region_id,
                 flush_timeout: Some(Duration::from_millis(500)),
+                reject_write: false,
             })
             .await;
         assert_matches!(reply, InstructionReply::DowngradeRegion(_));
@@ -304,8 +346,9 @@ mod tests {
                     }
                     .fail()
                 }));
-                region_engine.handle_set_readonly_gracefully_mock_fn =
-                    Some(Box::new(|_| Ok(SetReadonlyResponse::success(Some(1024)))))
+                region_engine.handle_set_readonly_gracefully_mock_fn = Some(Box::new(|_| {
+                    Ok(SetRegionRoleStateResponse::success(Some(1024)))
+                }))
             });
         mock_region_server.register_test_region(region_id, mock_engine);
         let handler_context = HandlerContext::new_for_test(mock_region_server);
@@ -321,6 +364,7 @@ mod tests {
                 .handle_downgrade_region_instruction(DowngradeRegion {
                     region_id,
                     flush_timeout,
+                    reject_write: false,
                 })
                 .await;
             assert_matches!(reply, InstructionReply::DowngradeRegion(_));
@@ -335,6 +379,7 @@ mod tests {
             .handle_downgrade_region_instruction(DowngradeRegion {
                 region_id,
                 flush_timeout: Some(Duration::from_millis(500)),
+                reject_write: false,
             })
             .await;
         assert_matches!(reply, InstructionReply::DowngradeRegion(_));
@@ -356,7 +401,7 @@ mod tests {
             MockRegionEngine::with_custom_apply_fn(MITO_ENGINE_NAME, |region_engine| {
                 region_engine.mock_role = Some(Some(RegionRole::Leader));
                 region_engine.handle_set_readonly_gracefully_mock_fn =
-                    Some(Box::new(|_| Ok(SetReadonlyResponse::NotFound)));
+                    Some(Box::new(|_| Ok(SetRegionRoleStateResponse::NotFound)));
             });
         mock_region_server.register_test_region(region_id, mock_engine);
         let handler_context = HandlerContext::new_for_test(mock_region_server);
@@ -365,6 +410,7 @@ mod tests {
             .handle_downgrade_region_instruction(DowngradeRegion {
                 region_id,
                 flush_timeout: None,
+                reject_write: false,
             })
             .await;
         assert_matches!(reply, InstructionReply::DowngradeRegion(_));
@@ -396,6 +442,7 @@ mod tests {
             .handle_downgrade_region_instruction(DowngradeRegion {
                 region_id,
                 flush_timeout: None,
+                reject_write: false,
             })
             .await;
         assert_matches!(reply, InstructionReply::DowngradeRegion(_));

src/datanode/src/heartbeat/handler/upgrade_region.rs

@@ -31,7 +31,7 @@ impl HandlerContext {
         }: UpgradeRegion,
     ) -> BoxFuture<'static, InstructionReply> {
         Box::pin(async move {
-            let Some(writable) = self.region_server.is_writable(region_id) else {
+            let Some(writable) = self.region_server.is_region_leader(region_id) else {
                 return InstructionReply::UpgradeRegion(UpgradeRegionReply {
                     ready: false,
                     exists: false,

src/datanode/src/region_server.rs

@@ -32,7 +32,7 @@ use common_recordbatch::SendableRecordBatchStream;
 use common_runtime::Runtime;
 use common_telemetry::tracing::{self, info_span};
 use common_telemetry::tracing_context::{FutureExt, TracingContext};
-use common_telemetry::{error, info, warn};
+use common_telemetry::{debug, error, info, warn};
 use dashmap::DashMap;
 use datafusion::datasource::{provider_as_source, TableProvider};
 use datafusion::error::Result as DfResult;
@@ -54,7 +54,10 @@ use snafu::{ensure, OptionExt, ResultExt};
 use store_api::metric_engine_consts::{
     FILE_ENGINE_NAME, LOGICAL_TABLE_METADATA_KEY, METRIC_ENGINE_NAME,
 };
-use store_api::region_engine::{RegionEngineRef, RegionRole, RegionStatistic, SetReadonlyResponse};
+use store_api::region_engine::{
+    RegionEngineRef, RegionRole, RegionStatistic, SetRegionRoleStateResponse,
+    SettableRegionRoleState,
+};
 use store_api::region_request::{
     AffectedRows, RegionCloseRequest, RegionOpenRequest, RegionRequest,
 };
@@ -274,37 +277,47 @@ impl RegionServer {
             .collect()
     }
 
-    pub fn is_writable(&self, region_id: RegionId) -> Option<bool> {
-        // TODO(weny): Finds a better way.
+    pub fn is_region_leader(&self, region_id: RegionId) -> Option<bool> {
         self.inner.region_map.get(&region_id).and_then(|engine| {
             engine.role(region_id).map(|role| match role {
                 RegionRole::Follower => false,
                 RegionRole::Leader => true,
+                RegionRole::DowngradingLeader => true,
             })
         })
     }
 
-    pub fn set_writable(&self, region_id: RegionId, writable: bool) -> Result<()> {
+    pub fn set_region_role(&self, region_id: RegionId, role: RegionRole) -> Result<()> {
         let engine = self
             .inner
             .region_map
             .get(&region_id)
             .with_context(|| RegionNotFoundSnafu { region_id })?;
         engine
-            .set_writable(region_id, writable)
+            .set_region_role(region_id, role)
             .with_context(|_| HandleRegionRequestSnafu { region_id })
     }
 
-    pub async fn set_readonly_gracefully(
+    /// Set region role state gracefully.
+    ///
+    /// For [SettableRegionRoleState::Follower]:
+    /// After the call returns, the engine ensures that
+    /// no **further** write or flush operations will succeed in this region.
+    ///
+    /// For [SettableRegionRoleState::DowngradingLeader]:
+    /// After the call returns, the engine ensures that
+    /// no **further** write operations will succeed in this region.
+    pub async fn set_region_role_state_gracefully(
         &self,
         region_id: RegionId,
-    ) -> Result<SetReadonlyResponse> {
+        state: SettableRegionRoleState,
+    ) -> Result<SetRegionRoleStateResponse> {
         match self.inner.region_map.get(&region_id) {
             Some(engine) => Ok(engine
-                .set_readonly_gracefully(region_id)
+                .set_region_role_state_gracefully(region_id, state)
                 .await
                 .with_context(|_| HandleRegionRequestSnafu { region_id })?),
-            None => Ok(SetReadonlyResponse::NotFound),
+            None => Ok(SetRegionRoleStateResponse::NotFound),
         }
     }
 
@@ -842,7 +855,7 @@ impl RegionServerInner {
                 info!("Region {region_id} is deregistered from engine {engine_type}");
                 self.region_map
                     .remove(&region_id)
-                    .map(|(id, engine)| engine.set_writable(id, false));
+                    .map(|(id, engine)| engine.set_region_role(id, RegionRole::Follower));
                 self.event_listener.on_region_deregistered(region_id);
             }
             RegionChange::Catchup => {
@@ -880,7 +893,7 @@ impl RegionServerInner {
         for region in logical_regions {
             self.region_map
                 .insert(region, RegionEngineWithStatus::Ready(engine.clone()));
-            info!("Logical region {} is registered!", region);
+            debug!("Logical region {} is registered!", region);
         }
         Ok(())
     }
@@ -922,17 +935,19 @@ impl RegionServerInner {
             .iter()
             .map(|x| (*x.key(), x.value().clone()))
             .collect::<Vec<_>>();
+        let num_regions = regions.len();
 
         for (region_id, engine) in regions {
             let closed = engine
                 .handle_request(region_id, RegionRequest::Close(RegionCloseRequest {}))
                 .await;
             match closed {
-                Ok(_) => info!("Region {region_id} is closed"),
+                Ok(_) => debug!("Region {region_id} is closed"),
                 Err(e) => warn!("Failed to close region {region_id}, err: {e}"),
             }
         }
         self.region_map.clear();
+        info!("closed {num_regions} regions");
 
         let engines = self.engines.write().unwrap().drain().collect::<Vec<_>>();
         for (engine_name, engine) in engines {

src/datanode/src/tests.rs

@@ -32,7 +32,8 @@ use query::{QueryEngine, QueryEngineContext};
 use session::context::QueryContextRef;
 use store_api::metadata::RegionMetadataRef;
 use store_api::region_engine::{
-    RegionEngine, RegionRole, RegionScannerRef, RegionStatistic, SetReadonlyResponse,
+    RegionEngine, RegionRole, RegionScannerRef, RegionStatistic, SetRegionRoleStateResponse,
+    SettableRegionRoleState,
 };
 use store_api::region_request::{AffectedRows, RegionRequest};
 use store_api::storage::{RegionId, ScanRequest};
@@ -106,7 +107,7 @@ pub type MockRequestHandler =
     Box<dyn Fn(RegionId, RegionRequest) -> Result<AffectedRows, Error> + Send + Sync>;
 
 pub type MockSetReadonlyGracefullyHandler =
-    Box<dyn Fn(RegionId) -> Result<SetReadonlyResponse, Error> + Send + Sync>;
+    Box<dyn Fn(RegionId) -> Result<SetRegionRoleStateResponse, Error> + Send + Sync>;
 
 pub struct MockRegionEngine {
     sender: Sender<(RegionId, RegionRequest)>,
@@ -220,14 +221,15 @@ impl RegionEngine for MockRegionEngine {
         Ok(())
     }
 
-    fn set_writable(&self, _region_id: RegionId, _writable: bool) -> Result<(), BoxedError> {
+    fn set_region_role(&self, _region_id: RegionId, _role: RegionRole) -> Result<(), BoxedError> {
         Ok(())
     }
 
-    async fn set_readonly_gracefully(
+    async fn set_region_role_state_gracefully(
         &self,
         region_id: RegionId,
-    ) -> Result<SetReadonlyResponse, BoxedError> {
+        _region_role_state: SettableRegionRoleState,
+    ) -> Result<SetRegionRoleStateResponse, BoxedError> {
         if let Some(mock_fn) = &self.handle_set_readonly_gracefully_mock_fn {
             return mock_fn(region_id).map_err(BoxedError::new);
         };

src/datatypes/src/data_type.rs

@@ -249,6 +249,15 @@ impl ConcreteDataType {
         ]
     }
 
+    pub fn timestamps() -> Vec<ConcreteDataType> {
+        vec![
+            ConcreteDataType::timestamp_second_datatype(),
+            ConcreteDataType::timestamp_millisecond_datatype(),
+            ConcreteDataType::timestamp_microsecond_datatype(),
+            ConcreteDataType::timestamp_nanosecond_datatype(),
+        ]
+    }
+
     /// Convert arrow data type to [ConcreteDataType].
     ///
     /// # Panics

src/datatypes/src/interval.rs

@@ -12,11 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use common_time::interval::Interval;
+use common_time::{IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth};
 use paste::paste;
-use serde::{Deserialize, Serialize};
 
-use crate::prelude::{Scalar, Value, ValueRef};
+use crate::prelude::Scalar;
 use crate::scalars::ScalarRef;
 use crate::types::{
     IntervalDayTimeType, IntervalMonthDayNanoType, IntervalYearMonthType, WrapperType,
@@ -26,39 +25,6 @@ use crate::vectors::{IntervalDayTimeVector, IntervalMonthDayNanoVector, Interval
 macro_rules! define_interval_with_unit {
     ($unit: ident, $native_ty: ty) => {
         paste! {
-            #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
-            pub struct [<Interval $unit>](pub Interval);
-
-            impl [<Interval $unit>] {
-                pub fn new(val: $native_ty) -> Self {
-                    Self(Interval:: [<from_ $native_ty>](val))
-                }
-            }
-
-            impl Default for [<Interval $unit>] {
-                fn default() -> Self {
-                    Self::new(0)
-                }
-            }
-
-            impl From<[<Interval $unit>]> for Value {
-                fn from(t: [<Interval $unit>]) -> Value {
-                    Value::Interval(t.0)
-                }
-            }
-
-            impl From<[<Interval $unit>]> for serde_json::Value {
-                fn from(t: [<Interval $unit>]) -> Self {
-                    t.0.into()
-                }
-            }
-
-            impl From<[<Interval $unit>]> for ValueRef<'static> {
-                fn from(t: [<Interval $unit>]) -> Self {
-                    ValueRef::Interval(t.0)
-                }
-            }
-
             impl Scalar for [<Interval $unit>] {
                 type VectorType = [<Interval $unit Vector>];
                 type RefType<'a> = [<Interval $unit>];
@@ -87,41 +53,11 @@ macro_rules! define_interval_with_unit {
                 type Native = $native_ty;
 
                 fn from_native(value: Self::Native) -> Self {
-                    Self::new(value)
+                    Self::[<from_ $native_ty>](value)
                 }
 
                 fn into_native(self) -> Self::Native {
-                    self.0.[<to_ $native_ty>]()
-                }
-            }
-
-            impl From<$native_ty> for [<Interval $unit>] {
-                fn from(val: $native_ty) -> Self {
-                    [<Interval $unit>]::from_native(val as $native_ty)
-                }
-            }
-
-            impl From<[<Interval $unit>]> for $native_ty {
-                fn from(val: [<Interval $unit>]) -> Self {
-                    val.0.[<to_ $native_ty>]()
-                }
-            }
-
-            impl TryFrom<Value> for Option<[<Interval $unit>]> {
-                type Error = $crate::error::Error;
-
-                #[inline]
-                fn try_from(from: Value) -> std::result::Result<Self, Self::Error> {
-                    match from {
-                        Value::Interval(v) if v.unit() == common_time::interval::IntervalUnit::$unit => {
-                            Ok(Some([<Interval $unit>](v)))
-                        },
-                        Value::Null => Ok(None),
-                        _ => $crate::error::TryFromValueSnafu {
-                            reason: format!("{:?} is not a {}", from, stringify!([<Interval $unit>])),
-                        }
-                        .fail(),
-                    }
+                    self.[<to_ $native_ty>]()
                 }
             }
         }
@@ -138,17 +74,17 @@ mod tests {
 
     #[test]
     fn test_interval_scalar() {
-        let interval = IntervalYearMonth::new(1000);
+        let interval = IntervalYearMonth::from(1000);
         assert_eq!(interval, interval.as_scalar_ref());
         assert_eq!(interval, interval.to_owned_scalar());
         assert_eq!(1000, interval.into_native());
 
-        let interval = IntervalDayTime::new(1000);
+        let interval = IntervalDayTime::from(1000);
         assert_eq!(interval, interval.as_scalar_ref());
         assert_eq!(interval, interval.to_owned_scalar());
         assert_eq!(1000, interval.into_native());
 
-        let interval = IntervalMonthDayNano::new(1000);
+        let interval = IntervalMonthDayNano::from(1000);
         assert_eq!(interval, interval.as_scalar_ref());
         assert_eq!(interval, interval.to_owned_scalar());
         assert_eq!(1000, interval.into_native());
@@ -156,15 +92,15 @@ mod tests {
 
     #[test]
     fn test_interval_convert_to_native_type() {
-        let interval = IntervalMonthDayNano::new(1000);
+        let interval = IntervalMonthDayNano::from(1000);
         let native_value: i128 = interval.into();
         assert_eq!(native_value, 1000);
 
-        let interval = IntervalDayTime::new(1000);
+        let interval = IntervalDayTime::from(1000);
         let native_interval: i64 = interval.into();
         assert_eq!(native_interval, 1000);
 
-        let interval = IntervalYearMonth::new(1000);
+        let interval = IntervalYearMonth::from(1000);
         let native_interval: i32 = interval.into();
         assert_eq!(native_interval, 1000);
     }

src/datatypes/src/types/interval_type.rs

@@ -17,8 +17,9 @@ use arrow::datatypes::{
     IntervalMonthDayNanoType as ArrowIntervalMonthDayNanoType, IntervalUnit as ArrowIntervalUnit,
     IntervalYearMonthType as ArrowIntervalYearMonthType,
 };
-use common_time::interval::IntervalUnit;
-use common_time::Interval;
+use common_time::interval::{
+    IntervalDayTime, IntervalMonthDayNano, IntervalUnit, IntervalYearMonth,
+};
 use enum_dispatch::enum_dispatch;
 use paste::paste;
 use serde::{Deserialize, Serialize};
@@ -26,7 +27,6 @@ use snafu::OptionExt;
 
 use crate::data_type::ConcreteDataType;
 use crate::error;
-use crate::interval::{IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth};
 use crate::prelude::{
     DataType, LogicalTypeId, MutableVector, ScalarVectorBuilder, Value, ValueRef, Vector,
 };
@@ -75,7 +75,7 @@ macro_rules! impl_data_type_for_interval {
                 }
 
                 fn default_value(&self) -> Value {
-                    Value::Interval(Interval::from_i128(0))
+                    Value::[<Interval $unit>]([<Interval $unit>]::default())
                 }
 
                 fn as_arrow_type(&self) -> ArrowDataType {
@@ -124,7 +124,7 @@ macro_rules! impl_data_type_for_interval {
                 fn cast_value_ref(value: ValueRef) -> crate::Result<Option<Self::Wrapper>> {
                     match value {
                         ValueRef::Null => Ok(None),
-                        ValueRef::Interval(t) => Ok(Some([<Interval $unit>](t))),
+                        ValueRef::[<Interval $unit>](t) => Ok(Some(t)),
                         other => error::CastTypeSnafu {
                             msg: format!("Failed to cast value {:?} to {}", other, stringify!([<Interval $unit>])),
                         }

src/datatypes/src/types/primitive_type.rs

@@ -16,7 +16,6 @@ use std::cmp::Ordering;
 use std::fmt;
 
 use arrow::datatypes::{ArrowNativeType, ArrowPrimitiveType, DataType as ArrowDataType};
-use common_time::interval::IntervalUnit;
 use common_time::{Date, DateTime};
 use serde::{Deserialize, Serialize};
 use snafu::OptionExt;
@@ -30,6 +29,7 @@ use crate::types::{DateTimeType, DateType};
 use crate::value::{Value, ValueRef};
 use crate::vectors::{MutableVector, PrimitiveVector, PrimitiveVectorBuilder, Vector};
 
+// TODO(yingwen): Can we remove `Into<serde_json::Value>`?
 /// Represents the wrapper type that wraps a native type using the `newtype pattern`,
 /// such as [Date](`common_time::Date`) is a wrapper type for the underlying native
 /// type `i32`.
@@ -364,11 +364,7 @@ impl DataType for Int64Type {
             Value::DateTime(v) => Some(Value::Int64(v.val())),
             Value::Timestamp(v) => Some(Value::Int64(v.value())),
             Value::Time(v) => Some(Value::Int64(v.value())),
-            Value::Interval(v) => match v.unit() {
-                IntervalUnit::DayTime => Some(Value::Int64(v.to_i64())),
-                IntervalUnit::YearMonth => None,
-                IntervalUnit::MonthDayNano => None,
-            },
+            // We don't allow casting interval type to int.
             _ => None,
         }
     }
@@ -410,11 +406,7 @@ impl DataType for Int32Type {
             Value::Float64(v) => num::cast::cast(v).map(Value::Int32),
             Value::String(v) => v.as_utf8().parse::<i32>().map(Value::Int32).ok(),
             Value::Date(v) => Some(Value::Int32(v.val())),
-            Value::Interval(v) => match v.unit() {
-                IntervalUnit::YearMonth => Some(Value::Int32(v.to_i32())),
-                IntervalUnit::DayTime => None,
-                IntervalUnit::MonthDayNano => None,
-            },
+            // We don't allow casting interval type to int.
             _ => None,
         }
     }

src/datatypes/src/types/string_type.rs

@@ -78,7 +78,15 @@ impl DataType for StringType {
             Value::DateTime(v) => Some(Value::String(StringBytes::from(v.to_string()))),
             Value::Timestamp(v) => Some(Value::String(StringBytes::from(v.to_iso8601_string()))),
             Value::Time(v) => Some(Value::String(StringBytes::from(v.to_iso8601_string()))),
-            Value::Interval(v) => Some(Value::String(StringBytes::from(v.to_iso8601_string()))),
+            Value::IntervalYearMonth(v) => {
+                Some(Value::String(StringBytes::from(v.to_iso8601_string())))
+            }
+            Value::IntervalDayTime(v) => {
+                Some(Value::String(StringBytes::from(v.to_iso8601_string())))
+            }
+            Value::IntervalMonthDayNano(v) => {
+                Some(Value::String(StringBytes::from(v.to_iso8601_string())))
+            }
             Value::Duration(v) => Some(Value::String(StringBytes::from(v.to_string()))),
             Value::Decimal128(v) => Some(Value::String(StringBytes::from(v.to_string()))),
 

src/datatypes/src/value.rs

@@ -28,7 +28,7 @@ use common_time::datetime::DateTime;
 use common_time::interval::IntervalUnit;
 use common_time::time::Time;
 use common_time::timestamp::{TimeUnit, Timestamp};
-use common_time::{Duration, Interval, Timezone};
+use common_time::{Duration, IntervalDayTime, IntervalMonthDayNano, IntervalYearMonth, Timezone};
 use datafusion_common::ScalarValue;
 use greptime_proto::v1::value::ValueData;
 pub use ordered_float::OrderedFloat;
@@ -38,6 +38,7 @@ use snafu::{ensure, ResultExt};
 
 use crate::error::{self, ConvertArrowArrayToScalarsSnafu, Error, Result, TryFromValueSnafu};
 use crate::prelude::*;
+use crate::schema::ColumnSchema;
 use crate::type_id::LogicalTypeId;
 use crate::types::{IntervalType, ListType};
 use crate::vectors::ListVector;
@@ -78,7 +79,10 @@ pub enum Value {
     Timestamp(Timestamp),
     Time(Time),
     Duration(Duration),
-    Interval(Interval),
+    // Interval types:
+    IntervalYearMonth(IntervalYearMonth),
+    IntervalDayTime(IntervalDayTime),
+    IntervalMonthDayNano(IntervalMonthDayNano),
 
     List(ListValue),
 }
@@ -111,7 +115,15 @@ impl Display for Value {
             Value::DateTime(v) => write!(f, "{v}"),
             Value::Timestamp(v) => write!(f, "{}", v.to_iso8601_string()),
             Value::Time(t) => write!(f, "{}", t.to_iso8601_string()),
-            Value::Interval(v) => write!(f, "{}", v.to_iso8601_string()),
+            Value::IntervalYearMonth(v) => {
+                write!(f, "{}", v.to_iso8601_string())
+            }
+            Value::IntervalDayTime(v) => {
+                write!(f, "{}", v.to_iso8601_string())
+            }
+            Value::IntervalMonthDayNano(v) => {
+                write!(f, "{}", v.to_iso8601_string())
+            }
             Value::Duration(d) => write!(f, "{d}"),
             Value::List(v) => {
                 let items = v
@@ -153,7 +165,15 @@ macro_rules! define_data_type_func {
                 $struct::DateTime(_) => ConcreteDataType::datetime_datatype(),
                 $struct::Time(t) => ConcreteDataType::time_datatype(*t.unit()),
                 $struct::Timestamp(v) => ConcreteDataType::timestamp_datatype(v.unit()),
-                $struct::Interval(v) => ConcreteDataType::interval_datatype(v.unit()),
+                $struct::IntervalYearMonth(_) => {
+                    ConcreteDataType::interval_datatype(IntervalUnit::YearMonth)
+                }
+                $struct::IntervalDayTime(_) => {
+                    ConcreteDataType::interval_datatype(IntervalUnit::DayTime)
+                }
+                $struct::IntervalMonthDayNano(_) => {
+                    ConcreteDataType::interval_datatype(IntervalUnit::MonthDayNano)
+                }
                 $struct::List(list) => ConcreteDataType::list_datatype(list.datatype().clone()),
                 $struct::Duration(d) => ConcreteDataType::duration_datatype(d.unit()),
                 $struct::Decimal128(d) => {
@@ -206,7 +226,9 @@ impl Value {
             Value::List(v) => ValueRef::List(ListValueRef::Ref { val: v }),
             Value::Timestamp(v) => ValueRef::Timestamp(*v),
             Value::Time(v) => ValueRef::Time(*v),
-            Value::Interval(v) => ValueRef::Interval(*v),
+            Value::IntervalYearMonth(v) => ValueRef::IntervalYearMonth(*v),
+            Value::IntervalDayTime(v) => ValueRef::IntervalDayTime(*v),
+            Value::IntervalMonthDayNano(v) => ValueRef::IntervalMonthDayNano(*v),
             Value::Duration(v) => ValueRef::Duration(*v),
             Value::Decimal128(v) => ValueRef::Decimal128(*v),
         }
@@ -220,14 +242,6 @@ impl Value {
         }
     }
 
-    /// Cast Value to Interval. Return None if value is not a valid interval data type.
-    pub fn as_interval(&self) -> Option<Interval> {
-        match self {
-            Value::Interval(i) => Some(*i),
-            _ => None,
-        }
-    }
-
     /// Cast Value to utf8 String. Return None if value is not a valid string data type.
     pub fn as_string(&self) -> Option<String> {
         match self {
@@ -255,12 +269,35 @@ impl Value {
     /// Cast Value to [Time]. Return None if value is not a valid time data type.
     pub fn as_time(&self) -> Option<Time> {
         match self {
-            Value::Int64(v) => Some(Time::new_millisecond(*v)),
             Value::Time(t) => Some(*t),
             _ => None,
         }
     }
 
+    /// Cast Value to [IntervalYearMonth]. Return None if value is not a valid interval year month data type.
+    pub fn as_interval_year_month(&self) -> Option<IntervalYearMonth> {
+        match self {
+            Value::IntervalYearMonth(v) => Some(*v),
+            _ => None,
+        }
+    }
+
+    /// Cast Value to [IntervalDayTime]. Return None if value is not a valid interval day time data type.
+    pub fn as_interval_day_time(&self) -> Option<IntervalDayTime> {
+        match self {
+            Value::IntervalDayTime(v) => Some(*v),
+            _ => None,
+        }
+    }
+
+    /// Cast Value to [IntervalMonthDayNano]. Return None if value is not a valid interval month day nano data type.
+    pub fn as_interval_month_day_nano(&self) -> Option<IntervalMonthDayNano> {
+        match self {
+            Value::IntervalMonthDayNano(v) => Some(*v),
+            _ => None,
+        }
+    }
+
     /// Cast Value to u64. Return None if value is not a valid uint64 data type.
     pub fn as_u64(&self) -> Option<u64> {
         match self {
@@ -321,11 +358,9 @@ impl Value {
                 TimeUnit::Microsecond => LogicalTypeId::TimeMicrosecond,
                 TimeUnit::Nanosecond => LogicalTypeId::TimeNanosecond,
             },
-            Value::Interval(v) => match v.unit() {
-                IntervalUnit::YearMonth => LogicalTypeId::IntervalYearMonth,
-                IntervalUnit::DayTime => LogicalTypeId::IntervalDayTime,
-                IntervalUnit::MonthDayNano => LogicalTypeId::IntervalMonthDayNano,
-            },
+            Value::IntervalYearMonth(_) => LogicalTypeId::IntervalYearMonth,
+            Value::IntervalDayTime(_) => LogicalTypeId::IntervalDayTime,
+            Value::IntervalMonthDayNano(_) => LogicalTypeId::IntervalMonthDayNano,
             Value::Duration(d) => match d.unit() {
                 TimeUnit::Second => LogicalTypeId::DurationSecond,
                 TimeUnit::Millisecond => LogicalTypeId::DurationMillisecond,
@@ -375,11 +410,9 @@ impl Value {
             }
             Value::Timestamp(t) => timestamp_to_scalar_value(t.unit(), Some(t.value())),
             Value::Time(t) => time_to_scalar_value(*t.unit(), Some(t.value()))?,
-            Value::Interval(v) => match v.unit() {
-                IntervalUnit::YearMonth => ScalarValue::IntervalYearMonth(Some(v.to_i32())),
-                IntervalUnit::DayTime => ScalarValue::IntervalDayTime(Some(v.to_i64())),
-                IntervalUnit::MonthDayNano => ScalarValue::IntervalMonthDayNano(Some(v.to_i128())),
-            },
+            Value::IntervalYearMonth(v) => ScalarValue::IntervalYearMonth(Some(v.to_i32())),
+            Value::IntervalDayTime(v) => ScalarValue::IntervalDayTime(Some(v.to_i64())),
+            Value::IntervalMonthDayNano(v) => ScalarValue::IntervalMonthDayNano(Some(v.to_i128())),
             Value::Duration(d) => duration_to_scalar_value(d.unit(), Some(d.value())),
             Value::Decimal128(d) => {
                 let (v, p, s) = d.to_scalar_value();
@@ -434,7 +467,9 @@ impl Value {
             Value::Timestamp(x) => Some(Value::Timestamp(x.negative())),
             Value::Time(x) => Some(Value::Time(x.negative())),
             Value::Duration(x) => Some(Value::Duration(x.negative())),
-            Value::Interval(x) => Some(Value::Interval(x.negative())),
+            Value::IntervalYearMonth(x) => Some(Value::IntervalYearMonth(x.negative())),
+            Value::IntervalDayTime(x) => Some(Value::IntervalDayTime(x.negative())),
+            Value::IntervalMonthDayNano(x) => Some(Value::IntervalMonthDayNano(x.negative())),
 
             Value::Binary(_) | Value::String(_) | Value::Boolean(_) | Value::List(_) => None,
         }
@@ -571,16 +606,6 @@ pub fn scalar_value_to_timestamp(
     }
 }
 
-/// Convert [ScalarValue] to [Interval].
-pub fn scalar_value_to_interval(scalar: &ScalarValue) -> Option<Interval> {
-    match scalar {
-        ScalarValue::IntervalYearMonth(v) => v.map(Interval::from_i32),
-        ScalarValue::IntervalDayTime(v) => v.map(Interval::from_i64),
-        ScalarValue::IntervalMonthDayNano(v) => v.map(Interval::from_i128),
-        _ => None,
-    }
-}
-
 macro_rules! impl_ord_for_value_like {
     ($Type: ident, $left: ident, $right: ident) => {
         if $left.is_null() && !$right.is_null() {
@@ -607,7 +632,9 @@ macro_rules! impl_ord_for_value_like {
                 ($Type::DateTime(v1), $Type::DateTime(v2)) => v1.cmp(v2),
                 ($Type::Timestamp(v1), $Type::Timestamp(v2)) => v1.cmp(v2),
                 ($Type::Time(v1), $Type::Time(v2)) => v1.cmp(v2),
-                ($Type::Interval(v1), $Type::Interval(v2)) => v1.cmp(v2),
+                ($Type::IntervalYearMonth(v1), $Type::IntervalYearMonth(v2)) => v1.cmp(v2),
+                ($Type::IntervalDayTime(v1), $Type::IntervalDayTime(v2)) => v1.cmp(v2),
+                ($Type::IntervalMonthDayNano(v1), $Type::IntervalMonthDayNano(v2)) => v1.cmp(v2),
                 ($Type::Duration(v1), $Type::Duration(v2)) => v1.cmp(v2),
                 ($Type::List(v1), $Type::List(v2)) => v1.cmp(v2),
                 _ => panic!(
@@ -685,7 +712,9 @@ impl_try_from_value!(Date, Date);
 impl_try_from_value!(Time, Time);
 impl_try_from_value!(DateTime, DateTime);
 impl_try_from_value!(Timestamp, Timestamp);
-impl_try_from_value!(Interval, Interval);
+impl_try_from_value!(IntervalYearMonth, IntervalYearMonth);
+impl_try_from_value!(IntervalDayTime, IntervalDayTime);
+impl_try_from_value!(IntervalMonthDayNano, IntervalMonthDayNano);
 impl_try_from_value!(Duration, Duration);
 impl_try_from_value!(Decimal128, Decimal128);
 
@@ -727,7 +756,9 @@ impl_value_from!(Date, Date);
 impl_value_from!(Time, Time);
 impl_value_from!(DateTime, DateTime);
 impl_value_from!(Timestamp, Timestamp);
-impl_value_from!(Interval, Interval);
+impl_value_from!(IntervalYearMonth, IntervalYearMonth);
+impl_value_from!(IntervalDayTime, IntervalDayTime);
+impl_value_from!(IntervalMonthDayNano, IntervalMonthDayNano);
 impl_value_from!(Duration, Duration);
 impl_value_from!(String, String);
 impl_value_from!(Decimal128, Decimal128);
@@ -774,7 +805,9 @@ impl TryFrom<Value> for serde_json::Value {
             Value::List(v) => serde_json::to_value(v)?,
             Value::Timestamp(v) => serde_json::to_value(v.value())?,
             Value::Time(v) => serde_json::to_value(v.value())?,
-            Value::Interval(v) => serde_json::to_value(v.to_i128())?,
+            Value::IntervalYearMonth(v) => serde_json::to_value(v.to_i32())?,
+            Value::IntervalDayTime(v) => serde_json::to_value(v.to_i64())?,
+            Value::IntervalMonthDayNano(v) => serde_json::to_value(v.to_i128())?,
             Value::Duration(v) => serde_json::to_value(v.value())?,
             Value::Decimal128(v) => serde_json::to_value(v.to_string())?,
         };
@@ -926,13 +959,13 @@ impl TryFrom<ScalarValue> for Value {
                 .unwrap_or(Value::Null),
 
             ScalarValue::IntervalYearMonth(t) => t
-                .map(|x| Value::Interval(Interval::from_i32(x)))
+                .map(|x| Value::IntervalYearMonth(IntervalYearMonth::from_i32(x)))
                 .unwrap_or(Value::Null),
             ScalarValue::IntervalDayTime(t) => t
-                .map(|x| Value::Interval(Interval::from_i64(x)))
+                .map(|x| Value::IntervalDayTime(IntervalDayTime::from_i64(x)))
                 .unwrap_or(Value::Null),
             ScalarValue::IntervalMonthDayNano(t) => t
-                .map(|x| Value::Interval(Interval::from_i128(x)))
+                .map(|x| Value::IntervalMonthDayNano(IntervalMonthDayNano::from_i128(x)))
                 .unwrap_or(Value::Null),
             ScalarValue::DurationSecond(d) => d
                 .map(|x| Value::Duration(Duration::new(x, TimeUnit::Second)))
@@ -987,7 +1020,9 @@ impl From<ValueRef<'_>> for Value {
             ValueRef::DateTime(v) => Value::DateTime(v),
             ValueRef::Timestamp(v) => Value::Timestamp(v),
             ValueRef::Time(v) => Value::Time(v),
-            ValueRef::Interval(v) => Value::Interval(v),
+            ValueRef::IntervalYearMonth(v) => Value::IntervalYearMonth(v),
+            ValueRef::IntervalDayTime(v) => Value::IntervalDayTime(v),
+            ValueRef::IntervalMonthDayNano(v) => Value::IntervalMonthDayNano(v),
             ValueRef::Duration(v) => Value::Duration(v),
             ValueRef::List(v) => v.to_value(),
             ValueRef::Decimal128(v) => Value::Decimal128(v),
@@ -1026,7 +1061,10 @@ pub enum ValueRef<'a> {
     Timestamp(Timestamp),
     Time(Time),
     Duration(Duration),
-    Interval(Interval),
+    // Interval types:
+    IntervalYearMonth(IntervalYearMonth),
+    IntervalDayTime(IntervalDayTime),
+    IntervalMonthDayNano(IntervalMonthDayNano),
 
     // Compound types:
     List(ListValueRef<'a>),
@@ -1150,9 +1188,19 @@ impl<'a> ValueRef<'a> {
         impl_as_for_value_ref!(self, Duration)
     }
 
-    /// Cast itself to [Interval].
-    pub fn as_interval(&self) -> Result<Option<Interval>> {
-        impl_as_for_value_ref!(self, Interval)
+    /// Cast itself to [IntervalYearMonth].
+    pub fn as_interval_year_month(&self) -> Result<Option<IntervalYearMonth>> {
+        impl_as_for_value_ref!(self, IntervalYearMonth)
+    }
+
+    /// Cast itself to [IntervalDayTime].
+    pub fn as_interval_day_time(&self) -> Result<Option<IntervalDayTime>> {
+        impl_as_for_value_ref!(self, IntervalDayTime)
+    }
+
+    /// Cast itself to [IntervalMonthDayNano].
+    pub fn as_interval_month_day_nano(&self) -> Result<Option<IntervalMonthDayNano>> {
+        impl_as_for_value_ref!(self, IntervalMonthDayNano)
     }
 
     /// Cast itself to [ListValueRef].
@@ -1212,7 +1260,9 @@ impl_value_ref_from!(Date, Date);
 impl_value_ref_from!(DateTime, DateTime);
 impl_value_ref_from!(Timestamp, Timestamp);
 impl_value_ref_from!(Time, Time);
-impl_value_ref_from!(Interval, Interval);
+impl_value_ref_from!(IntervalYearMonth, IntervalYearMonth);
+impl_value_ref_from!(IntervalDayTime, IntervalDayTime);
+impl_value_ref_from!(IntervalMonthDayNano, IntervalMonthDayNano);
 impl_value_ref_from!(Duration, Duration);
 impl_value_ref_from!(Decimal128, Decimal128);
 
@@ -1237,37 +1287,52 @@ impl<'a> From<Option<ListValueRef<'a>>> for ValueRef<'a> {
     }
 }
 
-impl<'a> TryFrom<ValueRef<'a>> for serde_json::Value {
-    type Error = serde_json::Error;
+/// transform a [ValueRef] to a [serde_json::Value].
+/// The json type will be handled specially
+pub fn transform_value_ref_to_json_value<'a>(
+    value: ValueRef<'a>,
+    schema: &'a ColumnSchema,
+) -> serde_json::Result<serde_json::Value> {
+    let json_value = match value {
+        ValueRef::Null => serde_json::Value::Null,
+        ValueRef::Boolean(v) => serde_json::Value::Bool(v),
+        ValueRef::UInt8(v) => serde_json::Value::from(v),
+        ValueRef::UInt16(v) => serde_json::Value::from(v),
+        ValueRef::UInt32(v) => serde_json::Value::from(v),
+        ValueRef::UInt64(v) => serde_json::Value::from(v),
+        ValueRef::Int8(v) => serde_json::Value::from(v),
+        ValueRef::Int16(v) => serde_json::Value::from(v),
+        ValueRef::Int32(v) => serde_json::Value::from(v),
+        ValueRef::Int64(v) => serde_json::Value::from(v),
+        ValueRef::Float32(v) => serde_json::Value::from(v.0),
+        ValueRef::Float64(v) => serde_json::Value::from(v.0),
+        ValueRef::String(bytes) => serde_json::Value::String(bytes.to_string()),
+        ValueRef::Binary(bytes) => {
+            if let ConcreteDataType::Json(_) = schema.data_type {
+                match jsonb::from_slice(bytes) {
+                    Ok(json) => json.into(),
+                    Err(e) => {
+                        error!(e; "Failed to parse jsonb");
+                        serde_json::Value::Null
+                    }
+                }
+            } else {
+                serde_json::to_value(bytes)?
+            }
+        }
+        ValueRef::Date(v) => serde_json::Value::Number(v.val().into()),
+        ValueRef::DateTime(v) => serde_json::Value::Number(v.val().into()),
+        ValueRef::List(v) => serde_json::to_value(v)?,
+        ValueRef::Timestamp(v) => serde_json::to_value(v.value())?,
+        ValueRef::Time(v) => serde_json::to_value(v.value())?,
+        ValueRef::IntervalYearMonth(v) => serde_json::Value::from(v),
+        ValueRef::IntervalDayTime(v) => serde_json::Value::from(v),
+        ValueRef::IntervalMonthDayNano(v) => serde_json::Value::from(v),
+        ValueRef::Duration(v) => serde_json::to_value(v.value())?,
+        ValueRef::Decimal128(v) => serde_json::to_value(v.to_string())?,
+    };
 
-    fn try_from(value: ValueRef<'a>) -> serde_json::Result<serde_json::Value> {
-        let json_value = match value {
-            ValueRef::Null => serde_json::Value::Null,
-            ValueRef::Boolean(v) => serde_json::Value::Bool(v),
-            ValueRef::UInt8(v) => serde_json::Value::from(v),
-            ValueRef::UInt16(v) => serde_json::Value::from(v),
-            ValueRef::UInt32(v) => serde_json::Value::from(v),
-            ValueRef::UInt64(v) => serde_json::Value::from(v),
-            ValueRef::Int8(v) => serde_json::Value::from(v),
-            ValueRef::Int16(v) => serde_json::Value::from(v),
-            ValueRef::Int32(v) => serde_json::Value::from(v),
-            ValueRef::Int64(v) => serde_json::Value::from(v),
-            ValueRef::Float32(v) => serde_json::Value::from(v.0),
-            ValueRef::Float64(v) => serde_json::Value::from(v.0),
-            ValueRef::String(bytes) => serde_json::Value::String(bytes.to_string()),
-            ValueRef::Binary(bytes) => serde_json::to_value(bytes)?,
-            ValueRef::Date(v) => serde_json::Value::Number(v.val().into()),
-            ValueRef::DateTime(v) => serde_json::Value::Number(v.val().into()),
-            ValueRef::List(v) => serde_json::to_value(v)?,
-            ValueRef::Timestamp(v) => serde_json::to_value(v.value())?,
-            ValueRef::Time(v) => serde_json::to_value(v.value())?,
-            ValueRef::Interval(v) => serde_json::to_value(v.to_i128())?,
-            ValueRef::Duration(v) => serde_json::to_value(v.value())?,
-            ValueRef::Decimal128(v) => serde_json::to_value(v.to_string())?,
-        };
-
-        Ok(json_value)
-    }
+    Ok(json_value)
 }
 
 /// Reference to a [ListValue].
@@ -1359,7 +1424,9 @@ impl<'a> ValueRef<'a> {
             ValueRef::Timestamp(_) => 16,
             ValueRef::Time(_) => 16,
             ValueRef::Duration(_) => 16,
-            ValueRef::Interval(_) => 24,
+            ValueRef::IntervalYearMonth(_) => 4,
+            ValueRef::IntervalDayTime(_) => 8,
+            ValueRef::IntervalMonthDayNano(_) => 16,
             ValueRef::Decimal128(_) => 32,
             ValueRef::List(v) => match v {
                 ListValueRef::Indexed { vector, .. } => vector.memory_size() / vector.len(),
@@ -1428,7 +1495,9 @@ pub fn column_data_to_json(data: ValueData) -> JsonValue {
 mod tests {
     use arrow::datatypes::DataType as ArrowDataType;
     use common_time::timezone::set_default_timezone;
-    use greptime_proto::v1::{Decimal128 as ProtoDecimal128, IntervalMonthDayNano};
+    use greptime_proto::v1::{
+        Decimal128 as ProtoDecimal128, IntervalMonthDayNano as ProtoIntervalMonthDayNano,
+    };
     use num_traits::Float;
 
     use super::*;
@@ -1525,11 +1594,13 @@ mod tests {
             JsonValue::String("interval year [12]".to_string())
         );
         assert_eq!(
-            column_data_to_json(ValueData::IntervalMonthDayNanoValue(IntervalMonthDayNano {
-                months: 1,
-                days: 2,
-                nanoseconds: 3,
-            })),
+            column_data_to_json(ValueData::IntervalMonthDayNanoValue(
+                ProtoIntervalMonthDayNano {
+                    months: 1,
+                    days: 2,
+                    nanoseconds: 3,
+                }
+            )),
             JsonValue::String("interval month [1][2][3]".to_string())
         );
         assert_eq!(
@@ -1740,12 +1811,10 @@ mod tests {
             ScalarValue::IntervalMonthDayNano(None).try_into().unwrap()
         );
         assert_eq!(
-            Value::Interval(Interval::from_month_day_nano(1, 1, 1)),
-            ScalarValue::IntervalMonthDayNano(Some(
-                Interval::from_month_day_nano(1, 1, 1).to_i128()
-            ))
-            .try_into()
-            .unwrap()
+            Value::IntervalMonthDayNano(IntervalMonthDayNano::new(1, 1, 1)),
+            ScalarValue::IntervalMonthDayNano(Some(IntervalMonthDayNano::new(1, 1, 1).to_i128()))
+                .try_into()
+                .unwrap()
         );
 
         assert_eq!(
@@ -1975,9 +2044,17 @@ mod tests {
             &ConcreteDataType::time_nanosecond_datatype(),
             &Value::Time(Time::new_nanosecond(1)),
         );
+        check_type_and_value(
+            &ConcreteDataType::interval_year_month_datatype(),
+            &Value::IntervalYearMonth(IntervalYearMonth::new(1)),
+        );
+        check_type_and_value(
+            &ConcreteDataType::interval_day_time_datatype(),
+            &Value::IntervalDayTime(IntervalDayTime::new(1, 2)),
+        );
         check_type_and_value(
             &ConcreteDataType::interval_month_day_nano_datatype(),
-            &Value::Interval(Interval::from_month_day_nano(1, 2, 3)),
+            &Value::IntervalMonthDayNano(IntervalMonthDayNano::new(1, 2, 3)),
         );
         check_type_and_value(
             &ConcreteDataType::duration_second_datatype(),
@@ -2160,7 +2237,9 @@ mod tests {
         check_as_value_ref!(Float64, OrderedF64::from(16.0));
         check_as_value_ref!(Timestamp, Timestamp::new_millisecond(1));
         check_as_value_ref!(Time, Time::new_millisecond(1));
-        check_as_value_ref!(Interval, Interval::from_month_day_nano(1, 2, 3));
+        check_as_value_ref!(IntervalYearMonth, IntervalYearMonth::new(1));
+        check_as_value_ref!(IntervalDayTime, IntervalDayTime::new(1, 2));
+        check_as_value_ref!(IntervalMonthDayNano, IntervalMonthDayNano::new(1, 2, 3));
         check_as_value_ref!(Duration, Duration::new_millisecond(1));
 
         assert_eq!(
@@ -2672,9 +2751,11 @@ mod tests {
         check_value_ref_size_eq(&ValueRef::DateTime(DateTime::new(1)), 8);
         check_value_ref_size_eq(&ValueRef::Timestamp(Timestamp::new_millisecond(1)), 16);
         check_value_ref_size_eq(&ValueRef::Time(Time::new_millisecond(1)), 16);
+        check_value_ref_size_eq(&ValueRef::IntervalYearMonth(IntervalYearMonth::new(1)), 4);
+        check_value_ref_size_eq(&ValueRef::IntervalDayTime(IntervalDayTime::new(1, 2)), 8);
         check_value_ref_size_eq(
-            &ValueRef::Interval(Interval::from_month_day_nano(1, 2, 3)),
-            24,
+            &ValueRef::IntervalMonthDayNano(IntervalMonthDayNano::new(1, 2, 3)),
+            16,
         );
         check_value_ref_size_eq(&ValueRef::Duration(Duration::new_millisecond(1)), 16);
         check_value_ref_size_eq(

src/datatypes/src/vectors/helper.rs

@@ -421,7 +421,7 @@ mod tests {
     use common_decimal::Decimal128;
     use common_time::time::Time;
     use common_time::timestamp::TimeUnit;
-    use common_time::{Date, DateTime, Duration, Interval};
+    use common_time::{Date, DateTime, Duration, IntervalMonthDayNano};
 
     use super::*;
     use crate::value::Value;
@@ -689,7 +689,10 @@ mod tests {
         );
         assert_eq!(3, vector.len());
         for i in 0..vector.len() {
-            assert_eq!(Value::Interval(Interval::from_i128(2000)), vector.get(i));
+            assert_eq!(
+                Value::IntervalMonthDayNano(IntervalMonthDayNano::from_i128(2000)),
+                vector.get(i)
+            );
         }
     }
 

src/file-engine/src/engine.rs

@@ -26,8 +26,8 @@ use object_store::ObjectStore;
 use snafu::{ensure, OptionExt};
 use store_api::metadata::RegionMetadataRef;
 use store_api::region_engine::{
-    RegionEngine, RegionRole, RegionScannerRef, RegionStatistic, SetReadonlyResponse,
-    SinglePartitionScanner,
+    RegionEngine, RegionRole, RegionScannerRef, RegionStatistic, SetRegionRoleStateResponse,
+    SettableRegionRoleState, SinglePartitionScanner,
 };
 use store_api::region_request::{
     AffectedRows, RegionCloseRequest, RegionCreateRequest, RegionDropRequest, RegionOpenRequest,
@@ -113,22 +113,23 @@ impl RegionEngine for FileRegionEngine {
         None
     }
 
-    fn set_writable(&self, region_id: RegionId, writable: bool) -> Result<(), BoxedError> {
+    fn set_region_role(&self, region_id: RegionId, role: RegionRole) -> Result<(), BoxedError> {
         self.inner
-            .set_writable(region_id, writable)
+            .set_region_role(region_id, role)
             .map_err(BoxedError::new)
     }
 
-    async fn set_readonly_gracefully(
+    async fn set_region_role_state_gracefully(
         &self,
         region_id: RegionId,
-    ) -> Result<SetReadonlyResponse, BoxedError> {
+        _region_role_state: SettableRegionRoleState,
+    ) -> Result<SetRegionRoleStateResponse, BoxedError> {
         let exists = self.inner.get_region(region_id).await.is_some();
 
         if exists {
-            Ok(SetReadonlyResponse::success(None))
+            Ok(SetRegionRoleStateResponse::success(None))
         } else {
-            Ok(SetReadonlyResponse::NotFound)
+            Ok(SetRegionRoleStateResponse::NotFound)
         }
     }
 
@@ -189,7 +190,7 @@ impl EngineInner {
         Ok(())
     }
 
-    fn set_writable(&self, _region_id: RegionId, _writable: bool) -> EngineResult<()> {
+    fn set_region_role(&self, _region_id: RegionId, _region_role: RegionRole) -> EngineResult<()> {
         // TODO(zhongzc): Improve the semantics and implementation of this API.
         Ok(())
     }

src/flow/src/compute/render/reduce.rs

@@ -560,7 +560,7 @@ fn reduce_batch_subgraph(
                     .get_mut(i)
                     .context(InternalSnafu{
                         reason: format!(
-                            "Output builder should have the same length as the row, expected at most {} but got {}", 
+                            "Output builder should have the same length as the row, expected at most {} but got {}",
                             column_cnt - 1,
                             i
                         )
@@ -1162,7 +1162,9 @@ fn from_val_to_slice_idx(
 #[cfg(test)]
 mod test {
 
-    use common_time::{DateTime, Interval, Timestamp};
+    use std::time::Duration;
+
+    use common_time::Timestamp;
     use datatypes::data_type::{ConcreteDataType, ConcreteDataType as CDT};
     use hydroflow::scheduled::graph::Hydroflow;
 
@@ -1214,8 +1216,8 @@ mod test {
         let expected = TypedPlan {
             schema: RelationType::new(vec![
                 ColumnType::new(CDT::uint64_datatype(), true), // sum(number)
-                ColumnType::new(CDT::datetime_datatype(), false), // window start
-                ColumnType::new(CDT::datetime_datatype(), false), // window end
+                ColumnType::new(CDT::timestamp_millisecond_datatype(), false), // window start
+                ColumnType::new(CDT::timestamp_millisecond_datatype(), false), // window end
             ])
             .into_unnamed(),
             // TODO(discord9): mfp indirectly ref to key columns
@@ -1232,7 +1234,10 @@ mod test {
                             .with_types(
                                 RelationType::new(vec![
                                     ColumnType::new(ConcreteDataType::uint32_datatype(), false),
-                                    ColumnType::new(ConcreteDataType::datetime_datatype(), false),
+                                    ColumnType::new(
+                                        ConcreteDataType::timestamp_millisecond_datatype(),
+                                        false,
+                                    ),
                                 ])
                                 .into_unnamed(),
                             ),
@@ -1242,22 +1247,18 @@ mod test {
                                 .map(vec![
                                     ScalarExpr::Column(1).call_unary(
                                         UnaryFunc::TumbleWindowFloor {
-                                            window_size: Interval::from_month_day_nano(
-                                                0,
-                                                0,
-                                                1_000_000_000,
-                                            ),
-                                            start_time: Some(DateTime::new(1625097600000)),
+                                            window_size: Duration::from_nanos(1_000_000_000),
+                                            start_time: Some(Timestamp::new_millisecond(
+                                                1625097600000,
+                                            )),
                                         },
                                     ),
                                     ScalarExpr::Column(1).call_unary(
                                         UnaryFunc::TumbleWindowCeiling {
-                                            window_size: Interval::from_month_day_nano(
-                                                0,
-                                                0,
-                                                1_000_000_000,
-                                            ),
-                                            start_time: Some(DateTime::new(1625097600000)),
+                                            window_size: Duration::from_nanos(1_000_000_000),
+                                            start_time: Some(Timestamp::new_millisecond(
+                                                1625097600000,
+                                            )),
                                         },
                                     ),
                                 ])
@@ -1278,9 +1279,9 @@ mod test {
                     }
                     .with_types(
                         RelationType::new(vec![
-                            ColumnType::new(CDT::datetime_datatype(), false), // window start
-                            ColumnType::new(CDT::datetime_datatype(), false), // window end
-                            ColumnType::new(CDT::uint64_datatype(), true),    //sum(number)
+                            ColumnType::new(CDT::timestamp_millisecond_datatype(), false), // window start
+                            ColumnType::new(CDT::timestamp_millisecond_datatype(), false), // window end
+                            ColumnType::new(CDT::uint64_datatype(), true), //sum(number)
                         ])
                         .with_key(vec![1])
                         .with_time_index(Some(0))

src/flow/src/expr/df_func.rs

@@ -171,9 +171,13 @@ impl DfScalarFunction {
     }
 }
 
-#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+#[derive(Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
 pub struct RawDfScalarFn {
-    /// The raw bytes encoded datafusion scalar function
+    /// The raw bytes encoded datafusion scalar function,
+    /// due to substrait have too many layers of nested struct and `ScalarFunction` 's derive is different
+    /// for simplicity's sake
+    /// so we store bytes instead of `ScalarFunction` here
+    /// but in unit test we will still compare decoded struct(using `f_decoded` field in Debug impl)
     pub(crate) f: bytes::BytesMut,
     /// The input schema of the function
     pub(crate) input_schema: RelationDesc,
@@ -181,6 +185,17 @@ pub struct RawDfScalarFn {
     pub(crate) extensions: FunctionExtensions,
 }
 
+impl std::fmt::Debug for RawDfScalarFn {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("RawDfScalarFn")
+            .field("f", &self.f)
+            .field("f_decoded", &ScalarFunction::decode(&mut self.f.as_ref()))
+            .field("df_schema", &self.input_schema)
+            .field("extensions", &self.extensions)
+            .finish()
+    }
+}
+
 impl RawDfScalarFn {
     pub fn from_proto(
         f: &substrait::substrait_proto_df::proto::expression::ScalarFunction,

src/flow/src/expr/func.rs

@@ -16,19 +16,18 @@
 
 use std::collections::HashMap;
 use std::sync::{Arc, OnceLock};
+use std::time::Duration;
 
 use arrow::array::{ArrayRef, BooleanArray};
 use common_error::ext::BoxedError;
 use common_time::timestamp::TimeUnit;
-use common_time::{DateTime, Timestamp};
+use common_time::Timestamp;
 use datafusion_expr::Operator;
 use datatypes::data_type::ConcreteDataType;
 use datatypes::prelude::DataType;
 use datatypes::types::cast;
 use datatypes::value::Value;
-use datatypes::vectors::{
-    BooleanVector, DateTimeVector, Helper, TimestampMillisecondVector, VectorRef,
-};
+use datatypes::vectors::{BooleanVector, Helper, TimestampMillisecondVector, VectorRef};
 use serde::{Deserialize, Serialize};
 use smallvec::smallvec;
 use snafu::{ensure, OptionExt, ResultExt};
@@ -52,8 +51,8 @@ pub enum UnmaterializableFunc {
     CurrentSchema,
     TumbleWindow {
         ts: Box<TypedExpr>,
-        window_size: common_time::Interval,
-        start_time: Option<DateTime>,
+        window_size: Duration,
+        start_time: Option<Timestamp>,
     },
 }
 
@@ -63,7 +62,8 @@ impl UnmaterializableFunc {
         match self {
             Self::Now => Signature {
                 input: smallvec![],
-                output: ConcreteDataType::datetime_datatype(),
+                // TODO(yingwen): Maybe return timestamp.
+                output: ConcreteDataType::timestamp_millisecond_datatype(),
                 generic_fn: GenericFn::Now,
             },
             Self::CurrentSchema => Signature {
@@ -110,12 +110,12 @@ pub enum UnaryFunc {
     StepTimestamp,
     Cast(ConcreteDataType),
     TumbleWindowFloor {
-        window_size: common_time::Interval,
-        start_time: Option<DateTime>,
+        window_size: Duration,
+        start_time: Option<Timestamp>,
     },
     TumbleWindowCeiling {
-        window_size: common_time::Interval,
-        start_time: Option<DateTime>,
+        window_size: Duration,
+        start_time: Option<Timestamp>,
     },
 }
 
@@ -139,8 +139,8 @@ impl UnaryFunc {
                 },
             },
             Self::StepTimestamp => Signature {
-                input: smallvec![ConcreteDataType::datetime_datatype()],
-                output: ConcreteDataType::datetime_datatype(),
+                input: smallvec![ConcreteDataType::timestamp_millisecond_datatype()],
+                output: ConcreteDataType::timestamp_millisecond_datatype(),
                 generic_fn: GenericFn::StepTimestamp,
             },
             Self::Cast(to) => Signature {
@@ -238,19 +238,19 @@ impl UnaryFunc {
                 }
             }
             Self::StepTimestamp => {
-                let datetime_array = get_datetime_array(&arg_col)?;
-                let date_array_ref = datetime_array
+                let timestamp_array = get_timestamp_array(&arg_col)?;
+                let timestamp_array_ref = timestamp_array
                     .as_any()
-                    .downcast_ref::<arrow::array::Date64Array>()
+                    .downcast_ref::<arrow::array::TimestampMillisecondArray>()
                     .context({
                         TypeMismatchSnafu {
                             expected: ConcreteDataType::boolean_datatype(),
-                            actual: ConcreteDataType::from_arrow_type(datetime_array.data_type()),
+                            actual: ConcreteDataType::from_arrow_type(timestamp_array.data_type()),
                         }
                     })?;
 
-                let ret = arrow::compute::unary(date_array_ref, |arr| arr + 1);
-                let ret = DateTimeVector::from(ret);
+                let ret = arrow::compute::unary(timestamp_array_ref, |arr| arr + 1);
+                let ret = TimestampMillisecondVector::from(ret);
                 Ok(Arc::new(ret))
             }
             Self::Cast(to) => {
@@ -266,19 +266,19 @@ impl UnaryFunc {
                 window_size,
                 start_time,
             } => {
-                let datetime_array = get_datetime_array(&arg_col)?;
-                let date_array_ref = datetime_array
+                let timestamp_array = get_timestamp_array(&arg_col)?;
+                let date_array_ref = timestamp_array
                     .as_any()
-                    .downcast_ref::<arrow::array::Date64Array>()
+                    .downcast_ref::<arrow::array::TimestampMillisecondArray>()
                     .context({
                         TypeMismatchSnafu {
                             expected: ConcreteDataType::boolean_datatype(),
-                            actual: ConcreteDataType::from_arrow_type(datetime_array.data_type()),
+                            actual: ConcreteDataType::from_arrow_type(timestamp_array.data_type()),
                         }
                     })?;
 
-                let start_time = start_time.map(|t| t.val());
-                let window_size = (window_size.to_nanosecond() / 1_000_000) as repr::Duration; // nanosecond to millisecond
+                let start_time = start_time.map(|t| t.value());
+                let window_size = window_size.as_millis() as repr::Duration;
 
                 let ret = arrow::compute::unary(date_array_ref, |ts| {
                     get_window_start(ts, window_size, start_time)
@@ -291,19 +291,19 @@ impl UnaryFunc {
                 window_size,
                 start_time,
             } => {
-                let datetime_array = get_datetime_array(&arg_col)?;
-                let date_array_ref = datetime_array
+                let timestamp_array = get_timestamp_array(&arg_col)?;
+                let date_array_ref = timestamp_array
                     .as_any()
-                    .downcast_ref::<arrow::array::Date64Array>()
+                    .downcast_ref::<arrow::array::TimestampMillisecondArray>()
                     .context({
                         TypeMismatchSnafu {
                             expected: ConcreteDataType::boolean_datatype(),
-                            actual: ConcreteDataType::from_arrow_type(datetime_array.data_type()),
+                            actual: ConcreteDataType::from_arrow_type(timestamp_array.data_type()),
                         }
                     })?;
 
-                let start_time = start_time.map(|t| t.val());
-                let window_size = (window_size.to_nanosecond() / 1_000_000) as repr::Duration; // nanosecond to millisecond
+                let start_time = start_time.map(|t| t.value());
+                let window_size = window_size.as_millis() as repr::Duration;
 
                 let ret = arrow::compute::unary(date_array_ref, |ts| {
                     get_window_start(ts, window_size, start_time) + window_size
@@ -330,19 +330,20 @@ impl UnaryFunc {
                     })?;
                     if let Some(window_size) = window_size_untyped.as_string() {
                         // cast as interval
-                        cast(
+                        let interval = cast(
                             Value::from(window_size),
-                            &ConcreteDataType::interval_month_day_nano_datatype(),
+                            &ConcreteDataType::interval_day_time_datatype(),
                         )
                         .map_err(BoxedError::new)
                         .context(ExternalSnafu)?
-                        .as_interval()
+                        .as_interval_day_time()
                         .context(UnexpectedSnafu {
                             reason: "Expect window size arg to be interval after successful cast"
                                 .to_string(),
-                        })?
-                    } else if let Some(interval) = window_size_untyped.as_interval() {
-                        interval
+                        })?;
+                        Duration::from_millis(interval.as_millis() as u64)
+                    } else if let Some(interval) = window_size_untyped.as_interval_day_time() {
+                        Duration::from_millis(interval.as_millis() as u64)
                     } else {
                         InvalidQuerySnafu {
                                 reason: format!(
@@ -357,16 +358,19 @@ impl UnaryFunc {
                 let start_time = match args.get(2) {
                     Some(start_time) => {
                         if let Some(value) = start_time.expr.as_literal() {
-                            // cast as DateTime
-                            let ret = cast(value, &ConcreteDataType::datetime_datatype())
-                                .map_err(BoxedError::new)
-                                .context(ExternalSnafu)?
-                                .as_datetime()
-                                .context(UnexpectedSnafu {
-                                    reason:
-                                        "Expect start time arg to be datetime after successful cast"
-                                            .to_string(),
-                                })?;
+                            // cast as timestamp
+                            let ret = cast(
+                                value,
+                                &ConcreteDataType::timestamp_millisecond_datatype(),
+                            )
+                            .map_err(BoxedError::new)
+                            .context(ExternalSnafu)?
+                            .as_timestamp()
+                            .context(UnexpectedSnafu {
+                                reason:
+                                    "Expect start time arg to be timestamp after successful cast"
+                                        .to_string(),
+                            })?;
                             Some(ret)
                         } else {
                             UnexpectedSnafu {
@@ -446,15 +450,15 @@ impl UnaryFunc {
             }
             Self::StepTimestamp => {
                 let ty = arg.data_type();
-                if let Value::DateTime(datetime) = arg {
-                    let datetime = DateTime::from(datetime.val() + 1);
-                    Ok(Value::from(datetime))
+                if let Value::Timestamp(timestamp) = arg {
+                    let timestamp = Timestamp::new_millisecond(timestamp.value() + 1);
+                    Ok(Value::from(timestamp))
                 } else if let Ok(v) = value_to_internal_ts(arg) {
-                    let datetime = DateTime::from(v + 1);
-                    Ok(Value::from(datetime))
+                    let timestamp = Timestamp::new_millisecond(v + 1);
+                    Ok(Value::from(timestamp))
                 } else {
                     TypeMismatchSnafu {
-                        expected: ConcreteDataType::datetime_datatype(),
+                        expected: ConcreteDataType::timestamp_millisecond_datatype(),
                         actual: ty,
                     }
                     .fail()?
@@ -474,8 +478,8 @@ impl UnaryFunc {
                 start_time,
             } => {
                 let ts = get_ts_as_millisecond(arg)?;
-                let start_time = start_time.map(|t| t.val());
-                let window_size = (window_size.to_nanosecond() / 1_000_000) as repr::Duration; // nanosecond to millisecond
+                let start_time = start_time.map(|t| t.value());
+                let window_size = window_size.as_millis() as repr::Duration;
                 let window_start = get_window_start(ts, window_size, start_time);
 
                 let ret = Timestamp::new_millisecond(window_start);
@@ -486,8 +490,8 @@ impl UnaryFunc {
                 start_time,
             } => {
                 let ts = get_ts_as_millisecond(arg)?;
-                let start_time = start_time.map(|t| t.val());
-                let window_size = (window_size.to_nanosecond() / 1_000_000) as repr::Duration; // nanosecond to millisecond
+                let start_time = start_time.map(|t| t.value());
+                let window_size = window_size.as_millis() as repr::Duration;
                 let window_start = get_window_start(ts, window_size, start_time);
 
                 let window_end = window_start + window_size;
@@ -498,21 +502,22 @@ impl UnaryFunc {
     }
 }
 
-fn get_datetime_array(vector: &VectorRef) -> Result<arrow::array::ArrayRef, EvalError> {
+fn get_timestamp_array(vector: &VectorRef) -> Result<arrow::array::ArrayRef, EvalError> {
     let arrow_array = vector.to_arrow_array();
-    let datetime_array =
-        if *arrow_array.data_type() == ConcreteDataType::datetime_datatype().as_arrow_type() {
-            arrow_array
-        } else {
-            arrow::compute::cast(
-                &arrow_array,
-                &ConcreteDataType::datetime_datatype().as_arrow_type(),
-            )
-            .context(ArrowSnafu {
-                context: "Trying to cast to datetime in StepTimestamp",
-            })?
-        };
-    Ok(datetime_array)
+    let timestamp_array = if *arrow_array.data_type()
+        == ConcreteDataType::timestamp_millisecond_datatype().as_arrow_type()
+    {
+        arrow_array
+    } else {
+        arrow::compute::cast(
+            &arrow_array,
+            &ConcreteDataType::timestamp_millisecond_datatype().as_arrow_type(),
+        )
+        .context(ArrowSnafu {
+            context: "Trying to cast to timestamp in StepTimestamp",
+        })?
+    };
+    Ok(timestamp_array)
 }
 
 fn get_window_start(
@@ -1284,7 +1289,6 @@ where
 mod test {
     use std::sync::Arc;
 
-    use common_time::Interval;
     use datatypes::vectors::Vector;
     use pretty_assertions::assert_eq;
 
@@ -1292,18 +1296,18 @@ mod test {
 
     #[test]
     fn test_tumble_batch() {
-        let datetime_vector = DateTimeVector::from_vec(vec![1, 2, 10, 13, 14, 20, 25]);
+        let timestamp_vector = TimestampMillisecondVector::from_vec(vec![1, 2, 10, 13, 14, 20, 25]);
         let tumble_start = UnaryFunc::TumbleWindowFloor {
-            window_size: Interval::from_day_time(0, 10),
+            window_size: Duration::from_millis(10),
             start_time: None,
         };
         let tumble_end = UnaryFunc::TumbleWindowCeiling {
-            window_size: Interval::from_day_time(0, 10),
+            window_size: Duration::from_millis(10),
             start_time: None,
         };
 
-        let len = datetime_vector.len();
-        let batch = Batch::try_new(vec![Arc::new(datetime_vector)], len).unwrap();
+        let len = timestamp_vector.len();
+        let batch = Batch::try_new(vec![Arc::new(timestamp_vector)], len).unwrap();
         let arg = ScalarExpr::Column(0);
 
         let start = tumble_start.eval_batch(&batch, &arg).unwrap();
@@ -1459,4 +1463,17 @@ mod test {
             Err(Error::InvalidQuery { .. })
         );
     }
+
+    #[test]
+    fn test_cast_int() {
+        let interval = cast(
+            Value::from("1 second"),
+            &ConcreteDataType::interval_day_time_datatype(),
+        )
+        .unwrap();
+        assert_eq!(
+            interval,
+            Value::from(common_time::IntervalDayTime::new(0, 1000))
+        );
+    }
 }