feat!: Divide flush and compaction job pool (#4871 )

* feat: divide flush/compact job pool * feat!: divide bg jobs config * docs: update config examples * test: fix tests
chore: graceful exit on bind fail (#4882 )
2026-01-04 20:32:56 +00:00 · 2024-10-25 23:36:16 +00:00 · 2024-10-25 09:29:39 +00:00 · 2024-10-25 07:09:51 +00:00 · 2024-10-25 03:45:24 +00:00 · 2024-10-25 03:13:34 +00:00
268 changed files with 10324 additions and 3736 deletions
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -1,6 +1,6 @@
 # This file is automatically @generated by Cargo.
 # It is not intended for manual editing.
-version = 3
+version = 4

 [[package]]
 name = "Inflector"
@@ -200,12 +200,6 @@ version = "1.0.89"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "86fdf8605db99b54d3cd748a44c6d04df638eb5dafb219b135d0149bd0db01f6"

-[[package]]
-name = "anymap"
-version = "1.0.0-beta.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8f1f8f5a6f3d50d89e3797d7593a50f96bb2aaa20ca0cc7be1fb673232c91d72"
-
 [[package]]
 name = "anymap2"
 version = "0.13.0"
@@ -214,7 +208,7 @@ checksum = "d301b3b94cb4b2f23d7917810addbbaff90738e0ca2be692bd027e70d7e0330c"

 [[package]]
 name = "api"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "common-base",
 "common-decimal",
@@ -230,6 +224,15 @@ dependencies = [
 "tonic-build",
 ]

+[[package]]
+name = "approx"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3f2a05fd1bd10b2527e20a2cd32d8873d115b8b39fe219ee25f42a8aca6ba278"
+dependencies = [
+ "num-traits",
+]
+
 [[package]]
 name = "approx"
 version = "0.5.1"
@@ -766,7 +769,7 @@ dependencies = [

 [[package]]
 name = "auth"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "async-trait",
@@ -985,6 +988,7 @@ dependencies = [
 "num-bigint",
 "num-integer",
 "num-traits",
+ "serde",
 ]

 [[package]]
@@ -1375,7 +1379,7 @@ dependencies = [

 [[package]]
 name = "cache"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "catalog",
 "common-error",
@@ -1383,7 +1387,7 @@ dependencies = [
 "common-meta",
 "moka",
 "snafu 0.8.5",
- "substrait 0.9.3",
+ "substrait 0.9.5",
 ]

 [[package]]
@@ -1410,7 +1414,7 @@ checksum = "37b2a672a2cb129a2e41c10b1224bb368f9f37a2b16b612598138befd7b37eb5"

 [[package]]
 name = "catalog"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arrow",
@@ -1548,6 +1552,16 @@ dependencies = [
 "vob",
 ]

+[[package]]
+name = "cgmath"
+version = "0.18.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1a98d30140e3296250832bbaaff83b27dcd6fa3cc70fb6f1f3e5c9c0023b5317"
+dependencies = [
+ "approx 0.4.0",
+ "num-traits",
+]
+
 [[package]]
 name = "chrono"
 version = "0.4.38"
@@ -1739,7 +1753,7 @@ checksum = "1462739cb27611015575c0c11df5df7601141071f07518d56fcc1be504cbec97"

 [[package]]
 name = "client"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arc-swap",
@@ -1769,7 +1783,7 @@ dependencies = [
 "serde_json",
 "snafu 0.8.5",
 "substrait 0.37.3",
- "substrait 0.9.3",
+ "substrait 0.9.5",
 "tokio",
 "tokio-stream",
 "tonic 0.11.0",
@@ -1788,6 +1802,17 @@ dependencies = [
 "winapi",
 ]

+[[package]]
+name = "clocksource"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "129026dd5a8a9592d96916258f3a5379589e513ea5e86aeb0bd2530286e44e9e"
+dependencies = [
+ "libc",
+ "time",
+ "winapi",
+]
+
 [[package]]
 name = "cmake"
 version = "0.1.51"
@@ -1799,7 +1824,7 @@ dependencies = [

 [[package]]
 name = "cmd"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-trait",
 "auth",
@@ -1856,7 +1881,7 @@ dependencies = [
 "similar-asserts",
 "snafu 0.8.5",
 "store-api",
- "substrait 0.9.3",
+ "substrait 0.9.5",
 "table",
 "temp-env",
 "tempfile",
@@ -1902,7 +1927,7 @@ checksum = "55b672471b4e9f9e95499ea597ff64941a309b2cdbffcc46f2cc5e2d971fd335"

 [[package]]
 name = "common-base"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "anymap2",
 "async-trait",
@@ -1920,7 +1945,7 @@ dependencies = [

 [[package]]
 name = "common-catalog"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "chrono",
 "common-error",
@@ -1931,7 +1956,7 @@ dependencies = [

 [[package]]
 name = "common-config"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "common-base",
 "common-error",
@@ -1954,7 +1979,7 @@ dependencies = [

 [[package]]
 name = "common-datasource"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "arrow",
 "arrow-schema",
@@ -1991,7 +2016,7 @@ dependencies = [

 [[package]]
 name = "common-decimal"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "bigdecimal 0.4.5",
 "common-error",
@@ -2004,7 +2029,7 @@ dependencies = [

 [[package]]
 name = "common-error"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "snafu 0.8.5",
 "strum 0.25.0",
@@ -2013,7 +2038,7 @@ dependencies = [

 [[package]]
 name = "common-frontend"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "async-trait",
@@ -2028,7 +2053,7 @@ dependencies = [

 [[package]]
 name = "common-function"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arc-swap",
@@ -2054,6 +2079,7 @@ dependencies = [
 "once_cell",
 "paste",
 "ron",
+ "s2",
 "serde",
 "serde_json",
 "session",
@@ -2067,7 +2093,7 @@ dependencies = [

 [[package]]
 name = "common-greptimedb-telemetry"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-trait",
 "common-runtime",
@@ -2084,7 +2110,7 @@ dependencies = [

 [[package]]
 name = "common-grpc"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arrow-flight",
@@ -2110,7 +2136,7 @@ dependencies = [

 [[package]]
 name = "common-grpc-expr"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "common-base",
@@ -2128,7 +2154,7 @@ dependencies = [

 [[package]]
 name = "common-macro"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "arc-swap",
 "common-query",
@@ -2142,7 +2168,7 @@ dependencies = [

 [[package]]
 name = "common-mem-prof"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "common-error",
 "common-macro",
@@ -2155,7 +2181,7 @@ dependencies = [

 [[package]]
 name = "common-meta"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "anymap2",
 "api",
@@ -2212,11 +2238,23 @@ dependencies = [

 [[package]]
 name = "common-plugins"
-version = "0.9.3"
+version = "0.9.5"
+
+[[package]]
+name = "common-pprof"
+version = "0.9.5"
+dependencies = [
+ "common-error",
+ "common-macro",
+ "pprof",
+ "prost 0.12.6",
+ "snafu 0.8.5",
+ "tokio",
+]

 [[package]]
 name = "common-procedure"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-stream",
 "async-trait",
@@ -2243,7 +2281,7 @@ dependencies = [

 [[package]]
 name = "common-procedure-test"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-trait",
 "common-procedure",
@@ -2251,7 +2289,7 @@ dependencies = [

 [[package]]
 name = "common-query"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "async-trait",
@@ -2277,7 +2315,7 @@ dependencies = [

 [[package]]
 name = "common-recordbatch"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "arc-swap",
 "common-error",
@@ -2296,19 +2334,27 @@ dependencies = [

 [[package]]
 name = "common-runtime"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-trait",
+ "clap 4.5.19",
 "common-error",
 "common-macro",
 "common-telemetry",
+ "futures",
 "lazy_static",
 "num_cpus",
 "once_cell",
+ "parking_lot 0.12.3",
 "paste",
+ "pin-project",
 "prometheus",
+ "rand",
+ "ratelimit",
 "serde",
+ "serde_json",
 "snafu 0.8.5",
+ "tempfile",
 "tokio",
 "tokio-metrics",
 "tokio-metrics-collector",
@@ -2318,7 +2364,7 @@ dependencies = [

 [[package]]
 name = "common-telemetry"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "atty",
 "backtrace",
@@ -2346,7 +2392,7 @@ dependencies = [

 [[package]]
 name = "common-test-util"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "client",
 "common-query",
@@ -2358,7 +2404,7 @@ dependencies = [

 [[package]]
 name = "common-time"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "arrow",
 "chrono",
@@ -2374,7 +2420,7 @@ dependencies = [

 [[package]]
 name = "common-version"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "build-data",
 "const_format",
@@ -2385,7 +2431,7 @@ dependencies = [

 [[package]]
 name = "common-wal"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "common-base",
 "common-error",
@@ -3194,7 +3240,7 @@ dependencies = [

 [[package]]
 name = "datanode"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arrow-flight",
@@ -3244,7 +3290,7 @@ dependencies = [
 "session",
 "snafu 0.8.5",
 "store-api",
- "substrait 0.9.3",
+ "substrait 0.9.5",
 "table",
 "tokio",
 "toml 0.8.19",
@@ -3253,7 +3299,7 @@ dependencies = [

 [[package]]
 name = "datatypes"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -3859,7 +3905,7 @@ dependencies = [

 [[package]]
 name = "file-engine"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "async-trait",
@@ -3959,9 +4005,18 @@ version = "1.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "28a80e3145d8ad11ba0995949bbcf48b9df2be62772b3d351ef017dff6ecb853"

+[[package]]
+name = "float_extras"
+version = "0.1.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b22b70f8649ea2315955f1a36d964b0e4da482dfaa5f0d04df0d1fb7c338ab7a"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "flow"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arrow",
@@ -4018,7 +4073,7 @@ dependencies = [
 "snafu 0.8.5",
 "store-api",
 "strum 0.25.0",
- "substrait 0.9.3",
+ "substrait 0.9.5",
 "table",
 "tokio",
 "tonic 0.11.0",
@@ -4080,7 +4135,7 @@ checksum = "6c2141d6d6c8512188a7891b4b01590a45f6dac67afb4f255c4124dbb86d4eaa"

 [[package]]
 name = "frontend"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arc-swap",
@@ -4389,7 +4444,7 @@ version = "0.7.13"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9ff16065e5720f376fbced200a5ae0f47ace85fd70b7e54269790281353b6d61"
 dependencies = [
- "approx",
+ "approx 0.5.1",
 "num-traits",
 "serde",
 ]
@@ -4476,7 +4531,7 @@ dependencies = [
 [[package]]
 name = "greptime-proto"
 version = "0.1.0"
-source = "git+https://github.com/GreptimeTeam/greptime-proto.git?rev=0b4f7c8ab06399f6b90e1626e8d5b9697cb33bb9#0b4f7c8ab06399f6b90e1626e8d5b9697cb33bb9"
+source = "git+https://github.com/GreptimeTeam/greptime-proto.git?rev=b4d301184eb0d01fd4d1042fcc7c5dfb54f3c1e3#b4d301184eb0d01fd4d1042fcc7c5dfb54f3c1e3"
 dependencies = [
 "prost 0.12.6",
 "serde",
@@ -5128,7 +5183,7 @@ dependencies = [

 [[package]]
 name = "index"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-trait",
 "asynchronous-codec",
@@ -5208,7 +5263,7 @@ dependencies = [
 [[package]]
 name = "influxdb_line_protocol"
 version = "0.1.0"
-source = "git+https://github.com/evenyag/influxdb_iox?branch=feat/line-protocol#10ef0d0b02705ac7518717390939fa3a9bcfcacc"
+source = "git+https://github.com/evenyag/influxdb_iox?branch=feat%2Fline-protocol#10ef0d0b02705ac7518717390939fa3a9bcfcacc"
 dependencies = [
 "bytes",
 "nom",
@@ -5959,7 +6014,7 @@ checksum = "a7a70ba024b9dc04c27ea2f0c0548feb474ec5c54bba33a7f72f873a39d07b24"

 [[package]]
 name = "log-store"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-stream",
 "async-trait",
@@ -6279,7 +6334,7 @@ dependencies = [

 [[package]]
 name = "meta-client"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "async-trait",
@@ -6305,7 +6360,7 @@ dependencies = [

 [[package]]
 name = "meta-srv"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "async-trait",
@@ -6366,9 +6421,9 @@ dependencies = [
 [[package]]
 name = "meter-core"
 version = "0.1.0"
-source = "git+https://github.com/GreptimeTeam/greptime-meter.git?rev=80eb97c24c88af4dd9a86f8bbaf50e741d4eb8cd#80eb97c24c88af4dd9a86f8bbaf50e741d4eb8cd"
+source = "git+https://github.com/GreptimeTeam/greptime-meter.git?rev=a10facb353b41460eeb98578868ebf19c2084fac#a10facb353b41460eeb98578868ebf19c2084fac"
 dependencies = [
- "anymap",
+ "anymap2",
 "once_cell",
 "parking_lot 0.12.3",
 ]
@@ -6376,14 +6431,14 @@ dependencies = [
 [[package]]
 name = "meter-macros"
 version = "0.1.0"
-source = "git+https://github.com/GreptimeTeam/greptime-meter.git?rev=80eb97c24c88af4dd9a86f8bbaf50e741d4eb8cd#80eb97c24c88af4dd9a86f8bbaf50e741d4eb8cd"
+source = "git+https://github.com/GreptimeTeam/greptime-meter.git?rev=a10facb353b41460eeb98578868ebf19c2084fac#a10facb353b41460eeb98578868ebf19c2084fac"
 dependencies = [
 "meter-core",
 ]

 [[package]]
 name = "metric-engine"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "aquamarine",
@@ -6486,7 +6541,7 @@ dependencies = [

 [[package]]
 name = "mito2"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "aquamarine",
@@ -6871,7 +6926,7 @@ version = "0.29.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d506eb7e08d6329505faa8a3a00a5dcc6de9f76e0c77e4b75763ae3c770831ff"
 dependencies = [
- "approx",
+ "approx 0.5.1",
 "matrixmultiply",
 "nalgebra-macros",
 "num-complex",
@@ -7222,7 +7277,7 @@ dependencies = [

 [[package]]
 name = "object-store"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "anyhow",
 "bytes",
@@ -7507,12 +7562,13 @@ dependencies = [
 "ordered-float 4.3.0",
 "percent-encoding",
 "rand",
+ "serde_json",
 "thiserror",
 ]

 [[package]]
 name = "operator"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "async-trait",
@@ -7557,7 +7613,7 @@ dependencies = [
 "sql",
 "sqlparser 0.45.0 (git+https://github.com/GreptimeTeam/sqlparser-rs.git?rev=54a267ac89c09b11c0c88934690530807185d3e7)",
 "store-api",
- "substrait 0.9.3",
+ "substrait 0.9.5",
 "table",
 "tokio",
 "tokio-util",
@@ -7807,7 +7863,7 @@ dependencies = [

 [[package]]
 name = "partition"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "async-trait",
@@ -8108,7 +8164,7 @@ checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"

 [[package]]
 name = "pipeline"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "ahash 0.8.11",
 "api",
@@ -8270,13 +8326,14 @@ dependencies = [

 [[package]]
 name = "plugins"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "auth",
 "common-base",
 "datanode",
 "frontend",
 "meta-srv",
+ "serde",
 "snafu 0.8.5",
 ]

@@ -8544,7 +8601,7 @@ dependencies = [

 [[package]]
 name = "promql"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "ahash 0.8.11",
 "async-trait",
@@ -8570,11 +8627,12 @@ dependencies = [

 [[package]]
 name = "promql-parser"
-version = "0.4.0"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "007a331efb31f6ddb644590ef22359c9469784931162aad92599e34bcfa66583"
+checksum = "0c1ad4a4cfa84ec4aa5831c82e57af0a3faf3f0af83bee13fa1390b2d0a32dc9"
 dependencies = [
 "cfgrammar",
+ "chrono",
 "lazy_static",
 "lrlex",
 "lrpar",
@@ -8779,7 +8837,7 @@ dependencies = [

 [[package]]
 name = "puffin"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-compression 0.4.13",
 "async-trait",
@@ -8901,7 +8959,7 @@ dependencies = [

 [[package]]
 name = "query"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "ahash 0.8.11",
 "api",
@@ -8936,6 +8994,7 @@ dependencies = [
 "datafusion-physical-expr",
 "datafusion-sql",
 "datatypes",
+ "fastrand",
 "format_num",
 "futures",
 "futures-util",
@@ -8950,12 +9009,15 @@ dependencies = [
 "object-store",
 "once_cell",
 "paste",
+ "pretty_assertions",
 "prometheus",
 "promql",
 "promql-parser",
 "prost 0.12.6",
 "rand",
 "regex",
+ "serde",
+ "serde_json",
 "session",
 "snafu 0.8.5",
 "sql",
@@ -8964,7 +9026,7 @@ dependencies = [
 "stats-cli",
 "store-api",
 "streaming-stats",
- "substrait 0.9.3",
+ "substrait 0.9.5",
 "table",
 "tokio",
 "tokio-stream",
@@ -9147,6 +9209,17 @@ dependencies = [
 "rand",
 ]

+[[package]]
+name = "ratelimit"
+version = "0.9.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6c1bb13e2dcfa2232ac6887157aad8d9b3fe4ca57f7c8d4938ff5ea9be742300"
+dependencies = [
+ "clocksource",
+ "parking_lot 0.12.3",
+ "thiserror",
+]
+
 [[package]]
 name = "raw-cpuid"
 version = "11.2.0"
@@ -10262,6 +10335,20 @@ version = "1.0.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f3cb5ba0dc43242ce17de99c180e96db90b235b8a9fdc9543c96d2209116bd9f"

+[[package]]
+name = "s2"
+version = "0.0.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cc7fbc04bb52c40b5f48c9bb2d2961375301916e0c25d9f373750654d588cd5c"
+dependencies = [
+ "bigdecimal 0.3.1",
+ "cgmath",
+ "float_extras",
+ "lazy_static",
+ "libm",
+ "serde",
+]
+
 [[package]]
 name = "safe-proc-macro2"
 version = "1.0.67"
@@ -10384,7 +10471,7 @@ checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"

 [[package]]
 name = "script"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arc-swap",
@@ -10678,8 +10765,9 @@ dependencies = [

 [[package]]
 name = "servers"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
+ "ahash 0.8.11",
 "aide",
 "api",
 "arrow",
@@ -10705,6 +10793,7 @@ dependencies = [
 "common-mem-prof",
 "common-meta",
 "common-plugins",
+ "common-pprof",
 "common-query",
 "common-recordbatch",
 "common-runtime",
@@ -10787,7 +10876,7 @@ dependencies = [

 [[package]]
 name = "session"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arc-swap",
@@ -10848,9 +10937,9 @@ dependencies = [

 [[package]]
 name = "shadow-rs"
-version = "0.31.1"
+version = "0.35.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "02c282402d25101f9c893e9cd7e4cae535fe7db18b81291de973026c219ddf1e"
+checksum = "2311e39772c00391875f40e34d43efef247b23930143a70ca5fbec9505937420"
 dependencies = [
 "const_format",
 "git2",
@@ -10899,7 +10988,7 @@ version = "0.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f0b7840f121a46d63066ee7a99fc81dcabbc6105e437cae43528cea199b5a05f"
 dependencies = [
- "approx",
+ "approx 0.5.1",
 "num-complex",
 "num-traits",
 "paste",
@@ -11108,7 +11197,7 @@ dependencies = [

 [[package]]
 name = "sql"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "chrono",
@@ -11169,7 +11258,7 @@ dependencies = [

 [[package]]
 name = "sqlness-runner"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-trait",
 "clap 4.5.19",
@@ -11370,7 +11459,7 @@ version = "0.16.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b35a062dbadac17a42e0fc64c27f419b25d6fae98572eb43c8814c9e873d7721"
 dependencies = [
- "approx",
+ "approx 0.5.1",
 "lazy_static",
 "nalgebra",
 "num-traits",
@@ -11389,7 +11478,7 @@ dependencies = [

 [[package]]
 name = "store-api"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "aquamarine",
@@ -11558,7 +11647,7 @@ dependencies = [

 [[package]]
 name = "substrait"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "async-trait",
 "bytes",
@@ -11757,7 +11846,7 @@ dependencies = [

 [[package]]
 name = "table"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "async-trait",
@@ -12023,7 +12112,7 @@ checksum = "3369f5ac52d5eb6ab48c6b4ffdc8efbcad6b89c765749064ba298f2c68a16a76"

 [[package]]
 name = "tests-fuzz"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "arbitrary",
 "async-trait",
@@ -12065,7 +12154,7 @@ dependencies = [

 [[package]]
 name = "tests-integration"
-version = "0.9.3"
+version = "0.9.5"
 dependencies = [
 "api",
 "arrow-flight",
@@ -12127,7 +12216,7 @@ dependencies = [
 "sql",
 "sqlx",
 "store-api",
- "substrait 0.9.3",
+ "substrait 0.9.5",
 "table",
 "tempfile",
 "time",
--- a/Cargo.toml
+++ b/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"

@@ -120,13 +121,13 @@ 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 = "0b4f7c8ab06399f6b90e1626e8d5b9697cb33bb9" }
+greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "b4d301184eb0d01fd4d1042fcc7c5dfb54f3c1e3" }
 humantime = "2.1"
 humantime-serde = "1.1"
 itertools = "0.10"
 jsonb = { git = "https://github.com/datafuselabs/jsonb.git", rev = "46ad50fc71cf75afbf98eec455f7892a6387c1fc", default-features = false }
 lazy_static = "1.4"
-meter-core = { git = "https://github.com/GreptimeTeam/greptime-meter.git", rev = "80eb97c24c88af4dd9a86f8bbaf50e741d4eb8cd" }
+meter-core = { git = "https://github.com/GreptimeTeam/greptime-meter.git", rev = "a10facb353b41460eeb98578868ebf19c2084fac" }
 mockall = "0.11.4"
 moka = "0.12"
 notify = "6.1"
@@ -137,15 +138,18 @@ opentelemetry-proto = { version = "0.5", features = [
    "metrics",
    "trace",
    "with-serde",
+    "logs",
 ] }
+parking_lot = "0.12"
 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"
+ratelimit = "0.9"
 regex = "1.8"
 regex-automata = { version = "0.4" }
 reqwest = { version = "0.12", default-features = false, features = [
@@ -165,7 +169,7 @@ schemars = "0.8"
 serde = { version = "1.0", features = ["derive"] }
 serde_json = { version = "1.0", features = ["float_roundtrip"] }
 serde_with = "3"
-shadow-rs = "0.31"
+shadow-rs = "0.35"
 similar-asserts = "1.6.0"
 smallvec = { version = "1", features = ["serde"] }
 snafu = "0.8"
@@ -208,6 +212,7 @@ 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" }
@@ -256,7 +261,7 @@ tokio-rustls = { git = "https://github.com/GreptimeTeam/tokio-rustls" }

 [workspace.dependencies.meter-macros]
 git = "https://github.com/GreptimeTeam/greptime-meter.git"
-rev = "80eb97c24c88af4dd9a86f8bbaf50e741d4eb8cd"
+rev = "a10facb353b41460eeb98578868ebf19c2084fac"

 [profile.release]
 debug = 1
--- a/2
+++ b/2
@@ -8,7 +8,7 @@ CARGO_BUILD_OPTS := --locked
 IMAGE_REGISTRY ?= docker.io
 IMAGE_NAMESPACE ?= greptime
 IMAGE_TAG ?= latest
-DEV_BUILDER_IMAGE_TAG ?= 2024-06-06-5674c14f-20240920110415
+DEV_BUILDER_IMAGE_TAG ?= 2024-10-19-a5c00e85-20241024184445
 BUILDX_MULTI_PLATFORM_BUILD ?= false
 BUILDX_BUILDER_NAME ?= gtbuilder
 BASE_IMAGE ?= ubuntu
--- a/config/config.md
+++ b/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. |
@@ -115,7 +116,9 @@
 | `region_engine.mito.worker_request_batch_size` | Integer | `64` | Max batch size for a worker to handle requests. |
 | `region_engine.mito.manifest_checkpoint_distance` | Integer | `10` | Number of meta action updated to trigger a new checkpoint for the manifest. |
 | `region_engine.mito.compress_manifest` | Bool | `false` | Whether to compress manifest and checkpoint file by gzip (default false). |
-| `region_engine.mito.max_background_jobs` | Integer | `4` | Max number of running background jobs |
+| `region_engine.mito.max_background_flushes` | Integer | Auto | Max number of running background flush jobs (default: 1/2 of cpu cores). |
+| `region_engine.mito.max_background_compactions` | Integer | Auto | Max number of running background compaction jobs (default: 1/4 of cpu cores). |
+| `region_engine.mito.max_background_purges` | Integer | Auto | Max number of running background purge jobs (default: number of cpu cores). |
 | `region_engine.mito.auto_flush_interval` | String | `1h` | Interval to auto flush a region if it has not flushed yet. |
 | `region_engine.mito.global_write_buffer_size` | String | Auto | Global write buffer size for all regions. If not set, it's default to 1/8 of OS memory with a max limitation of 1GB. |
 | `region_engine.mito.global_write_buffer_reject_size` | String | Auto | Global write buffer size threshold to reject write requests. If not set, it's default to 2 times of `global_write_buffer_size`. |
@@ -409,6 +412,7 @@
 | `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. |
@@ -435,7 +439,9 @@
 | `region_engine.mito.worker_request_batch_size` | Integer | `64` | Max batch size for a worker to handle requests. |
 | `region_engine.mito.manifest_checkpoint_distance` | Integer | `10` | Number of meta action updated to trigger a new checkpoint for the manifest. |
 | `region_engine.mito.compress_manifest` | Bool | `false` | Whether to compress manifest and checkpoint file by gzip (default false). |
-| `region_engine.mito.max_background_jobs` | Integer | `4` | Max number of running background jobs |
+| `region_engine.mito.max_background_flushes` | Integer | Auto | Max number of running background flush jobs (default: 1/2 of cpu cores). |
+| `region_engine.mito.max_background_compactions` | Integer | Auto | Max number of running background compaction jobs (default: 1/4 of cpu cores). |
+| `region_engine.mito.max_background_purges` | Integer | Auto | Max number of running background purge jobs (default: number of cpu cores). |
 | `region_engine.mito.auto_flush_interval` | String | `1h` | Interval to auto flush a region if it has not flushed yet. |
 | `region_engine.mito.global_write_buffer_size` | String | Auto | Global write buffer size for all regions. If not set, it's default to 1/8 of OS memory with a max limitation of 1GB. |
 | `region_engine.mito.global_write_buffer_reject_size` | String | Auto | Global write buffer size threshold to reject write requests. If not set, it's default to 2 times of `global_write_buffer_size` |
--- a/config/datanode.example.toml
+++ b/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:
@@ -405,8 +416,17 @@ manifest_checkpoint_distance = 10
 ## Whether to compress manifest and checkpoint file by gzip (default false).
 compress_manifest = false

-## Max number of running background jobs
-max_background_jobs = 4
+## Max number of running background flush jobs (default: 1/2 of cpu cores).
+## @toml2docs:none-default="Auto"
+#+ max_background_flushes = 4
+
+## Max number of running background compaction jobs (default: 1/4 of cpu cores).
+## @toml2docs:none-default="Auto"
+#+ max_background_compactions = 2
+
+## Max number of running background purge jobs (default: number of cpu cores).
+## @toml2docs:none-default="Auto"
+#+ max_background_purges = 8

 ## Interval to auto flush a region if it has not flushed yet.
 auto_flush_interval = "1h"
--- a/config/standalone.example.toml
+++ b/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:
@@ -443,8 +454,17 @@ manifest_checkpoint_distance = 10
 ## Whether to compress manifest and checkpoint file by gzip (default false).
 compress_manifest = false

-## Max number of running background jobs
-max_background_jobs = 4
+## Max number of running background flush jobs (default: 1/2 of cpu cores).
+## @toml2docs:none-default="Auto"
+#+ max_background_flushes = 4
+
+## Max number of running background compaction jobs (default: 1/4 of cpu cores).
+## @toml2docs:none-default="Auto"
+#+ max_background_compactions = 2
+
+## Max number of running background purge jobs (default: number of cpu cores).
+## @toml2docs:none-default="Auto"
+#+ max_background_purges = 8

 ## Interval to auto flush a region if it has not flushed yet.
 auto_flush_interval = "1h"
--- a/docs/benchmarks/log/README.md
+++ b/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).
--- a/docs/how-to/how-to-profile-cpu.md
+++ b/docs/how-to/how-to-profile-cpu.md
@@ -1,11 +1,5 @@
 # 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
--- a/docs/how-to/how-to-profile-memory.md
+++ b/docs/how-to/how-to-profile-memory.md
@@ -18,12 +18,6 @@ sudo apt install libjemalloc-dev
 curl https://raw.githubusercontent.com/brendangregg/FlameGraph/master/flamegraph.pl > ./flamegraph.pl
 ```

-### Build GreptimeDB with `mem-prof` feature.
-
-```bash
-cargo build --features=mem-prof
-```
-
 ## Profiling

 Start GreptimeDB instance with environment variables:
--- a/grafana/greptimedb.json
+++ b/grafana/greptimedb.json
@@ -409,7 +409,39 @@
      "fieldConfig": {
        "defaults": {
          "color": {
-            "mode": "thresholds"
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "axisBorderShow": false,
+            "axisCenteredZero": false,
+            "axisColorMode": "text",
+            "axisLabel": "",
+            "axisPlacement": "auto",
+            "barAlignment": 0,
+            "drawStyle": "line",
+            "fillOpacity": 0,
+            "gradientMode": "none",
+            "hideFrom": {
+              "legend": false,
+              "tooltip": false,
+              "viz": false
+            },
+            "insertNulls": false,
+            "lineInterpolation": "linear",
+            "lineWidth": 1,
+            "pointSize": 5,
+            "scaleDistribution": {
+              "type": "linear"
+            },
+            "showPoints": "auto",
+            "spanNulls": false,
+            "stacking": {
+              "group": "A",
+              "mode": "none"
+            },
+            "thresholdsStyle": {
+              "mode": "off"
+            }
          },
          "fieldMinMax": false,
          "mappings": [],
@@ -438,18 +470,16 @@
      },
      "id": 27,
      "options": {
-        "colorMode": "value",
-        "graphMode": "area",
-        "justifyMode": "auto",
-        "orientation": "auto",
-        "reduceOptions": {
-          "calcs": ["lastNotNull"],
-          "fields": "",
-          "values": false
+        "legend": {
+          "calcs": [],
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": true
        },
-        "text": {},
-        "textMode": "auto",
-        "wideLayout": true
+        "tooltip": {
+          "mode": "single",
+          "sort": "none"
+        }
      },
      "pluginVersion": "10.2.3",
      "targets": [
@@ -467,7 +497,7 @@
        }
      ],
      "title": "CPU",
-      "type": "stat"
+      "type": "timeseries"
    },
    {
      "datasource": {
@@ -477,7 +507,39 @@
      "fieldConfig": {
        "defaults": {
          "color": {
-            "mode": "thresholds"
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "axisBorderShow": false,
+            "axisCenteredZero": false,
+            "axisColorMode": "text",
+            "axisLabel": "",
+            "axisPlacement": "auto",
+            "barAlignment": 0,
+            "drawStyle": "line",
+            "fillOpacity": 0,
+            "gradientMode": "none",
+            "hideFrom": {
+              "legend": false,
+              "tooltip": false,
+              "viz": false
+            },
+            "insertNulls": false,
+            "lineInterpolation": "linear",
+            "lineWidth": 1,
+            "pointSize": 5,
+            "scaleDistribution": {
+              "type": "linear"
+            },
+            "showPoints": "auto",
+            "spanNulls": false,
+            "stacking": {
+              "group": "A",
+              "mode": "none"
+            },
+            "thresholdsStyle": {
+              "mode": "off"
+            }
          },
          "decimals": 0,
          "fieldMinMax": false,
@@ -503,18 +565,16 @@
      },
      "id": 28,
      "options": {
-        "colorMode": "value",
-        "graphMode": "area",
-        "justifyMode": "auto",
-        "orientation": "auto",
-        "reduceOptions": {
-          "calcs": ["lastNotNull"],
-          "fields": "",
-          "values": false
+        "legend": {
+          "calcs": [],
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": true
        },
-        "text": {},
-        "textMode": "auto",
-        "wideLayout": true
+        "tooltip": {
+          "mode": "single",
+          "sort": "none"
+        }
      },
      "pluginVersion": "10.2.3",
      "targets": [
@@ -532,7 +592,7 @@
        }
      ],
      "title": "Memory",
-      "type": "stat"
+      "type": "timeseries"
    },
    {
      "collapsed": false,
@@ -3335,6 +3395,6 @@
  "timezone": "",
  "title": "GreptimeDB",
  "uid": "e7097237-669b-4f8d-b751-13067afbfb68",
-  "version": 15,
+  "version": 16,
  "weekStart": ""
 }
--- a/rust-toolchain.toml
+++ b/rust-toolchain.toml
@@ -1,3 +1,2 @@
 [toolchain]
-channel = "nightly-2024-06-06"
-
+channel = "nightly-2024-10-19"
--- a/src/api/src/helper.rs
+++ b/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);
    }
--- a/src/auth/src/user_provider/static_user_provider.rs
+++ b/src/auth/src/user_provider/static_user_provider.rs
@@ -33,7 +33,7 @@ impl StaticUserProvider {
            value: value.to_string(),
            msg: "StaticUserProviderOption must be in format `<option>:<value>`",
        })?;
-        return match mode {
+        match mode {
            "file" => {
                let users = load_credential_from_file(content)?
                    .context(InvalidConfigSnafu {
@@ -58,7 +58,7 @@ impl StaticUserProvider {
                msg: "StaticUserProviderOption must be in format `file:<path>` or `cmd:<values>`",
            }
                .fail(),
-        };
+        }
    }
 }

--- a/src/catalog/src/system_schema/information_schema/table_names.rs
+++ b/src/catalog/src/system_schema/information_schema/table_names.rs
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-/// All table names in `information_schema`.
+//! All table names in `information_schema`.

 pub const TABLES: &str = "tables";
 pub const COLUMNS: &str = "columns";
--- a/src/catalog/src/system_schema/memory_table.rs
+++ b/src/catalog/src/system_schema/memory_table.rs
@@ -74,7 +74,7 @@ impl MemoryTableBuilder {
    /// Construct the `information_schema.{table_name}` virtual table
    pub async fn memory_records(&mut self) -> Result<RecordBatch> {
        if self.columns.is_empty() {
-            RecordBatch::new_empty(self.schema.clone()).context(CreateRecordBatchSnafu)
+            Ok(RecordBatch::new_empty(self.schema.clone()))
        } else {
            RecordBatch::new(self.schema.clone(), std::mem::take(&mut self.columns))
                .context(CreateRecordBatchSnafu)
--- a/src/catalog/src/system_schema/pg_catalog/pg_namespace.rs
+++ b/src/catalog/src/system_schema/pg_catalog/pg_namespace.rs
@@ -12,6 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

+//! The `pg_catalog.pg_namespace` table implementation.
+//! namespace is a schema in greptime
+
 pub(super) mod oid_map;

 use std::sync::{Arc, Weak};
@@ -40,9 +43,6 @@ use crate::system_schema::utils::tables::{string_column, u32_column};
 use crate::system_schema::SystemTable;
 use crate::CatalogManager;

-/// The `pg_catalog.pg_namespace` table implementation.
-/// namespace is a schema in greptime
-
 const NSPNAME: &str = "nspname";
 const INIT_CAPACITY: usize = 42;

--- a/src/client/Cargo.toml
+++ b/src/client/Cargo.toml
@@ -28,7 +28,7 @@ enum_dispatch = "0.3"
 futures-util.workspace = true
 lazy_static.workspace = true
 moka = { workspace = true, features = ["future"] }
-parking_lot = "0.12"
+parking_lot.workspace = true
 prometheus.workspace = true
 prost.workspace = true
 query.workspace = true
--- a/src/cmd/Cargo.toml
+++ b/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"]

--- a/src/cmd/src/datanode.rs
+++ b/src/cmd/src/datanode.rs
@@ -272,9 +272,10 @@ impl StartCommand {
        info!("Datanode start command: {:#?}", self);
        info!("Datanode options: {:#?}", opts);

+        let plugin_opts = opts.plugins;
        let opts = opts.component;
        let mut plugins = Plugins::new();
-        plugins::setup_datanode_plugins(&mut plugins, &opts)
+        plugins::setup_datanode_plugins(&mut plugins, &plugin_opts, &opts)
            .await
            .context(StartDatanodeSnafu)?;

--- a/src/cmd/src/frontend.rs
+++ b/src/cmd/src/frontend.rs
@@ -266,9 +266,10 @@ impl StartCommand {
        info!("Frontend start command: {:#?}", self);
        info!("Frontend options: {:#?}", opts);

+        let plugin_opts = opts.plugins;
        let opts = opts.component;
        let mut plugins = Plugins::new();
-        plugins::setup_frontend_plugins(&mut plugins, &opts)
+        plugins::setup_frontend_plugins(&mut plugins, &plugin_opts, &opts)
            .await
            .context(StartFrontendSnafu)?;

@@ -342,6 +343,8 @@ impl StartCommand {
        // Some queries are expected to take long time.
        let channel_config = ChannelConfig {
            timeout: None,
+            tcp_nodelay: opts.datanode.client.tcp_nodelay,
+            connect_timeout: Some(opts.datanode.client.connect_timeout),
            ..Default::default()
        };
        let client = NodeClients::new(channel_config);
@@ -472,7 +475,7 @@ mod tests {
        };

        let mut plugins = Plugins::new();
-        plugins::setup_frontend_plugins(&mut plugins, &fe_opts)
+        plugins::setup_frontend_plugins(&mut plugins, &[], &fe_opts)
            .await
            .unwrap();

--- a/src/cmd/src/lib.rs
+++ b/src/cmd/src/lib.rs
@@ -84,6 +84,7 @@ pub trait App: Send {
 }

 /// Log the versions of the application, and the arguments passed to the cli.
+///
 /// `version` should be the same as the output of cli "--version";
 /// and the `short_version` is the short version of the codes, often consist of git branch and commit.
 pub fn log_versions(version: &str, short_version: &str, app: &str) {
--- a/src/cmd/src/metasrv.rs
+++ b/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)]
@@ -249,9 +273,10 @@ impl StartCommand {
        info!("Metasrv start command: {:#?}", self);
        info!("Metasrv options: {:#?}", opts);

+        let plugin_opts = opts.plugins;
        let opts = opts.component;
        let mut plugins = Plugins::new();
-        plugins::setup_metasrv_plugins(&mut plugins, &opts)
+        plugins::setup_metasrv_plugins(&mut plugins, &plugin_opts, &opts)
            .await
            .context(StartMetaServerSnafu)?;

--- a/src/cmd/src/options.rs
+++ b/src/cmd/src/options.rs
@@ -15,6 +15,7 @@
 use clap::Parser;
 use common_config::Configurable;
 use common_runtime::global::RuntimeOptions;
+use plugins::PluginOptions;
 use serde::{Deserialize, Serialize};

 #[derive(Parser, Default, Debug, Clone)]
@@ -40,6 +41,8 @@ pub struct GlobalOptions {
 pub struct GreptimeOptions<T> {
    /// The runtime options.
    pub runtime: RuntimeOptions,
+    /// The plugin options.
+    pub plugins: Vec<PluginOptions>,

    /// The options of each component (like Datanode or Standalone) of GreptimeDB.
    #[serde(flatten)]
--- a/src/cmd/src/standalone.rs
+++ b/src/cmd/src/standalone.rs
@@ -445,15 +445,16 @@ impl StartCommand {
        info!("Standalone options: {opts:#?}");

        let mut plugins = Plugins::new();
+        let plugin_opts = opts.plugins;
        let opts = opts.component;
        let fe_opts = opts.frontend_options();
        let dn_opts = opts.datanode_options();

-        plugins::setup_frontend_plugins(&mut plugins, &fe_opts)
+        plugins::setup_frontend_plugins(&mut plugins, &plugin_opts, &fe_opts)
            .await
            .context(StartFrontendSnafu)?;

-        plugins::setup_datanode_plugins(&mut plugins, &dn_opts)
+        plugins::setup_datanode_plugins(&mut plugins, &plugin_opts, &dn_opts)
            .await
            .context(StartDatanodeSnafu)?;

@@ -653,7 +654,7 @@ impl StartCommand {
    }
 }

-struct StandaloneInformationExtension {
+pub struct StandaloneInformationExtension {
    region_server: RegionServer,
    procedure_manager: ProcedureManagerRef,
    start_time_ms: u64,
@@ -762,7 +763,7 @@ mod tests {
        };

        let mut plugins = Plugins::new();
-        plugins::setup_frontend_plugins(&mut plugins, &fe_opts)
+        plugins::setup_frontend_plugins(&mut plugins, &[], &fe_opts)
            .await
            .unwrap();

--- a/src/cmd/tests/load_config_test.rs
+++ b/src/cmd/tests/load_config_test.rs
@@ -20,7 +20,7 @@ use common_config::Configurable;
 use common_grpc::channel_manager::{
    DEFAULT_MAX_GRPC_RECV_MESSAGE_SIZE, DEFAULT_MAX_GRPC_SEND_MESSAGE_SIZE,
 };
-use common_telemetry::logging::{LoggingOptions, DEFAULT_OTLP_ENDPOINT};
+use common_telemetry::logging::{LoggingOptions, SlowQueryOptions, DEFAULT_OTLP_ENDPOINT};
 use common_wal::config::raft_engine::RaftEngineConfig;
 use common_wal::config::DatanodeWalConfig;
 use datanode::config::{DatanodeOptions, RegionEngineConfig, StorageConfig};
@@ -159,8 +159,20 @@ fn test_load_metasrv_example_config() {
                level: Some("info".to_string()),
                otlp_endpoint: Some(DEFAULT_OTLP_ENDPOINT.to_string()),
                tracing_sample_ratio: Some(Default::default()),
+                slow_query: SlowQueryOptions {
+                    enable: false,
+                    threshold: Some(Duration::from_secs(10)),
+                    sample_ratio: Some(1.0),
+                },
                ..Default::default()
            },
+            datanode: meta_srv::metasrv::DatanodeOptions {
+                client: meta_srv::metasrv::DatanodeClientOptions {
+                    timeout: Duration::from_secs(10),
+                    connect_timeout: Duration::from_secs(10),
+                    tcp_nodelay: true,
+                },
+            },
            export_metrics: ExportMetricsOption {
                self_import: Some(Default::default()),
                remote_write: Some(Default::default()),
--- a/src/common/base/src/plugins.rs
+++ b/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,
--- a/src/common/base/src/secrets.rs
+++ b/src/common/base/src/secrets.rs
@@ -46,8 +46,9 @@ impl From<String> for SecretString {
    }
 }

-/// Wrapper type for values that contains secrets, which attempts to limit
-/// accidental exposure and ensure secrets are wiped from memory when dropped.
+/// Wrapper type for values that contains secrets.
+///
+/// It attempts to limit accidental exposure and ensure secrets are wiped from memory when dropped.
 /// (e.g. passwords, cryptographic keys, access tokens or other credentials)
 ///
 /// Access to the secret inner value occurs through the [`ExposeSecret`]
--- a/src/common/catalog/src/consts.rs
+++ b/src/common/catalog/src/consts.rs
@@ -103,14 +103,15 @@ pub const INFORMATION_SCHEMA_PROCEDURE_INFO_TABLE_ID: u32 = 34;
 /// id for information_schema.region_statistics
 pub const INFORMATION_SCHEMA_REGION_STATISTICS_TABLE_ID: u32 = 35;

-/// ----- End of information_schema tables -----
+// ----- End of information_schema tables -----

 /// ----- Begin of pg_catalog tables -----
 pub const PG_CATALOG_PG_CLASS_TABLE_ID: u32 = 256;
 pub const PG_CATALOG_PG_TYPE_TABLE_ID: u32 = 257;
 pub const PG_CATALOG_PG_NAMESPACE_TABLE_ID: u32 = 258;

-/// ----- End of pg_catalog tables -----
+// ----- End of pg_catalog tables -----
+
 pub const MITO_ENGINE: &str = "mito";
 pub const MITO2_ENGINE: &str = "mito2";
 pub const METRIC_ENGINE: &str = "metric";
--- a/src/common/function/Cargo.toml
+++ b/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
--- a/src/common/function/src/scalars/aggregate.rs
+++ b/src/common/function/src/scalars/aggregate.rs
@@ -31,7 +31,6 @@ pub use polyval::PolyvalAccumulatorCreator;
 pub use scipy_stats_norm_cdf::ScipyStatsNormCdfAccumulatorCreator;
 pub use scipy_stats_norm_pdf::ScipyStatsNormPdfAccumulatorCreator;

-use super::geo::encoding::JsonPathEncodeFunctionCreator;
 use crate::function_registry::FunctionRegistry;

 /// A function creates `AggregateFunctionCreator`.
@@ -93,6 +92,11 @@ impl AggregateFunctions {
        register_aggr_func!("scipystatsnormcdf", 2, ScipyStatsNormCdfAccumulatorCreator);
        register_aggr_func!("scipystatsnormpdf", 2, ScipyStatsNormPdfAccumulatorCreator);

-        register_aggr_func!("json_encode_path", 3, JsonPathEncodeFunctionCreator);
+        #[cfg(feature = "geo")]
+        register_aggr_func!(
+            "json_encode_path",
+            3,
+            super::geo::encoding::JsonPathEncodeFunctionCreator
+        );
    }
 }
--- a/src/common/function/src/scalars/date/date_add.rs
+++ b/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!(),
-            }
-        }
-    }
 }
--- a/src/common/function/src/scalars/date/date_sub.rs
+++ b/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!(),
-            }
-        }
-    }
 }
--- a/src/common/function/src/scalars/geo.rs
+++ b/src/common/function/src/scalars/geo.rs
@@ -17,8 +17,7 @@ pub(crate) mod encoding;
 mod geohash;
 mod h3;
 mod helpers;
-
-use geohash::{GeohashFunction, GeohashNeighboursFunction};
+mod s2;

 use crate::function_registry::FunctionRegistry;

@@ -27,8 +26,8 @@ pub(crate) struct GeoFunctions;
 impl GeoFunctions {
    pub fn register(registry: &FunctionRegistry) {
        // geohash
-        registry.register(Arc::new(GeohashFunction));
-        registry.register(Arc::new(GeohashNeighboursFunction));
+        registry.register(Arc::new(geohash::GeohashFunction));
+        registry.register(Arc::new(geohash::GeohashNeighboursFunction));

        // h3 index
        registry.register(Arc::new(h3::H3LatLngToCell));
@@ -55,5 +54,11 @@ impl GeoFunctions {
        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));
    }
 }
--- a/src/common/function/src/scalars/geo/encoding.rs
+++ b/src/common/function/src/scalars/geo/encoding.rs
@@ -17,7 +17,7 @@ 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, InvalidFuncArgsSnafu, InvalidInputStateSnafu, Result};
+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;
--- a/src/common/function/src/scalars/geo/h3.rs
+++ b/src/common/function/src/scalars/geo/h3.rs
@@ -16,7 +16,7 @@ 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;
@@ -29,9 +29,9 @@ use datatypes::vectors::{
 use derive_more::Display;
 use h3o::{CellIndex, LatLng, Resolution};
 use once_cell::sync::Lazy;
-use snafu::{ensure, ResultExt};
+use snafu::ResultExt;

-use super::helpers::{ensure_columns_len, ensure_columns_n};
+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(|| {
@@ -382,15 +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();
@@ -982,18 +974,6 @@ fn value_to_resolution(v: Value) -> Result<Resolution> {
        .context(error::ExecuteSnafu)
 }

-macro_rules! ensure_and_coerce {
-    ($compare:expr, $coerce:expr) => {{
-        ensure!(
-            $compare,
-            InvalidFuncArgsSnafu {
-                err_msg: "Argument was outside of acceptable range "
-            }
-        );
-        Ok($coerce)
-    }};
-}
-
 fn value_to_position(v: Value) -> Result<u64> {
    match v {
        Value::Int8(v) => ensure_and_coerce!(v >= 0, v as u64),
--- a/src/common/function/src/scalars/geo/helpers.rs
+++ b/src/common/function/src/scalars/geo/helpers.rs
@@ -14,15 +14,15 @@

 macro_rules! ensure_columns_len {
    ($columns:ident) => {
-        ensure!(
+        snafu::ensure!(
            $columns.windows(2).all(|c| c[0].len() == c[1].len()),
-            InvalidFuncArgsSnafu {
+            common_query::error::InvalidFuncArgsSnafu {
                err_msg: "The length of input columns are in different size"
            }
        )
    };
    ($column_a:ident, $column_b:ident, $($column_n:ident),*) => {
-        ensure!(
+        snafu::ensure!(
            {
                let mut result = $column_a.len() == $column_b.len();
                $(
@@ -30,7 +30,7 @@ macro_rules! ensure_columns_len {
                )*
                result
            }
-            InvalidFuncArgsSnafu {
+            common_query::error::InvalidFuncArgsSnafu {
                err_msg: "The length of input columns are in different size"
            }
        )
@@ -41,9 +41,9 @@ pub(super) use ensure_columns_len;

 macro_rules! ensure_columns_n {
    ($columns:ident, $n:literal) => {
-        ensure!(
+        snafu::ensure!(
            $columns.len() == $n,
-            InvalidFuncArgsSnafu {
+            common_query::error::InvalidFuncArgsSnafu {
                err_msg: format!(
                    "The length of arguments is not correct, expect {}, provided : {}",
                    stringify!($n),
@@ -59,3 +59,17 @@ macro_rules! ensure_columns_n {
 }

 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;
--- a/src/common/function/src/scalars/geo/s2.rs
+++ b/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!(),
+    }
+}
--- a/src/common/function/src/scalars/json.rs
+++ b/src/common/function/src/scalars/json.rs
@@ -16,6 +16,7 @@ use std::sync::Arc;
 mod json_get;
 mod json_is;
 mod json_path_exists;
+mod json_path_match;
 mod json_to_string;
 mod parse_json;

@@ -49,5 +50,6 @@ impl JsonFunction {
        registry.register(Arc::new(JsonIsObject));

        registry.register(Arc::new(json_path_exists::JsonPathExistsFunction));
+        registry.register(Arc::new(json_path_match::JsonPathMatchFunction));
    }
 }
--- a/src/common/function/src/scalars/json/json_path_match.rs
+++ b/src/common/function/src/scalars/json/json_path_match.rs
@@ -0,0 +1,202 @@
+// 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 match the given JSON path's predicate.
+#[derive(Clone, Debug, Default)]
+pub struct JsonPathMatchFunction;
+
+const NAME: &str = "json_path_match";
+
+impl Function for JsonPathMatchFunction {
+    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 mut results = BooleanVectorBuilder::with_capacity(size);
+
+        for i in 0..size {
+            let json = jsons.get_ref(i);
+            let path = paths.get_ref(i);
+
+            match json.data_type() {
+                // JSON data type uses binary vector
+                ConcreteDataType::Binary(_) => {
+                    let json = json.as_binary();
+                    let path = path.as_string();
+                    let result = match (json, path) {
+                        (Ok(Some(json)), Ok(Some(path))) => {
+                            if !jsonb::is_null(json) {
+                                let json_path = jsonb::jsonpath::parse_json_path(path.as_bytes());
+                                match json_path {
+                                    Ok(json_path) => jsonb::path_match(json, json_path).ok(),
+                                    Err(_) => None,
+                                }
+                            } else {
+                                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 JsonPathMatchFunction {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "JSON_PATH_MATCH")
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::Arc;
+
+    use common_query::prelude::TypeSignature;
+    use datatypes::vectors::{BinaryVector, StringVector};
+
+    use super::*;
+
+    #[test]
+    fn test_json_path_match_function() {
+        let json_path_match = JsonPathMatchFunction;
+
+        assert_eq!("json_path_match", json_path_match.name());
+        assert_eq!(
+            ConcreteDataType::boolean_datatype(),
+            json_path_match
+                .return_type(&[ConcreteDataType::json_datatype()])
+                .unwrap()
+        );
+
+        assert!(matches!(json_path_match.signature(),
+                         Signature {
+                             type_signature: TypeSignature::Exact(valid_types),
+                             volatility: Volatility::Immutable
+                         } if valid_types == vec![ConcreteDataType::json_datatype(), ConcreteDataType::string_datatype()],
+        ));
+
+        let json_strings = [
+            Some(r#"{"a": {"b": 2}, "b": 2, "c": 3}"#.to_string()),
+            Some(r#"{"a": 1, "b": [1,2,3]}"#.to_string()),
+            Some(r#"{"a": 1 ,"b": [1,2,3]}"#.to_string()),
+            Some(r#"[1,2,3]"#.to_string()),
+            Some(r#"{"a":1,"b":[1,2,3]}"#.to_string()),
+            Some(r#"null"#.to_string()),
+            Some(r#"null"#.to_string()),
+        ];
+
+        let paths = vec![
+            Some("$.a.b == 2".to_string()),
+            Some("$.b[1 to last] >= 2".to_string()),
+            Some("$.c > 0".to_string()),
+            Some("$[0 to last] > 0".to_string()),
+            Some(r#"null"#.to_string()),
+            Some("$.c > 0".to_string()),
+            Some(r#"null"#.to_string()),
+        ];
+
+        let results = [
+            Some(true),
+            Some(true),
+            Some(false),
+            Some(true),
+            None,
+            None,
+            None,
+        ];
+
+        let jsonbs = json_strings
+            .into_iter()
+            .map(|s| s.map(|json| jsonb::parse_value(json.as_bytes()).unwrap().to_vec()))
+            .collect::<Vec<_>>();
+
+        let json_vector = BinaryVector::from(jsonbs);
+        let path_vector = StringVector::from(paths);
+        let args: Vec<VectorRef> = vec![Arc::new(json_vector), Arc::new(path_vector)];
+        let vector = json_path_match
+            .eval(FunctionContext::default(), &args)
+            .unwrap();
+
+        assert_eq!(7, vector.len());
+        for (i, expected) in results.iter().enumerate() {
+            let result = vector.get_ref(i);
+
+            match expected {
+                Some(expected_value) => {
+                    assert!(!result.is_null());
+                    let result_value = result.as_boolean().unwrap().unwrap();
+                    assert_eq!(*expected_value, result_value);
+                }
+                None => {
+                    assert!(result.is_null());
+                }
+            }
+        }
+    }
+}
--- a/src/common/greptimedb-telemetry/src/lib.rs
+++ b/src/common/greptimedb-telemetry/src/lib.rs
@@ -199,6 +199,7 @@ pub fn default_get_uuid(working_home: &Option<String>) -> Option<String> {
 }

 /// Report version info to GreptimeDB.
+///
 /// We do not collect any identity-sensitive information.
 /// This task is scheduled to run every 30 minutes.
 /// The task will be disabled default. It can be enabled by setting the build feature `greptimedb-telemetry`
@@ -324,7 +325,7 @@ mod tests {
            });
            let addr = ([127, 0, 0, 1], port).into();

-            let server = Server::bind(&addr).serve(make_svc);
+            let server = Server::try_bind(&addr).unwrap().serve(make_svc);
            let graceful = server.with_graceful_shutdown(async {
                rx.await.ok();
            });
--- a/src/common/grpc/src/select.rs
+++ b/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(_),
--- a/src/common/macro/src/lib.rs
+++ b/src/common/macro/src/lib.rs
@@ -35,7 +35,9 @@ pub fn aggr_func_type_store_derive(input: TokenStream) -> TokenStream {
 }

 /// A struct can be used as a creator for aggregate function if it has been annotated with this
-/// attribute first. This attribute add a necessary field which is intended to store the input
+/// attribute first.
+///
+/// This attribute add a necessary field which is intended to store the input
 /// data's types to the struct.
 /// This attribute is expected to be used along with derive macro [AggrFuncTypeStore].
 #[proc_macro_attribute]
@@ -44,9 +46,10 @@ pub fn as_aggr_func_creator(args: TokenStream, input: TokenStream) -> TokenStrea
 }

 /// Attribute macro to convert an arithimetic function to a range function. The annotated function
-/// should accept servaral arrays as input and return a single value as output. This procedure
-/// macro can works on any number of input parameters. Return type can be either primitive type
-/// or wrapped in `Option`.
+/// should accept servaral arrays as input and return a single value as output.
+///
+/// This procedure macro can works on any number of input parameters. Return type can be either
+/// primitive type or wrapped in `Option`.
 ///
 /// # Example
 /// Take `count_over_time()` in PromQL as an example:
--- a/src/common/meta/src/cluster.rs
+++ b/src/common/meta/src/cluster.rs
@@ -55,6 +55,7 @@ pub trait ClusterInfo {
 }

 /// The key of [NodeInfo] in the storage. The format is `__meta_cluster_node_info-{cluster_id}-{role}-{node_id}`.
+///
 /// This key cannot be used to describe the `Metasrv` because the `Metasrv` does not have
 /// a `cluster_id`, it serves multiple clusters.
 #[derive(Debug, Clone, Eq, Hash, PartialEq, Serialize, Deserialize)]
--- a/src/common/meta/src/key/catalog_name.rs
+++ b/src/common/meta/src/key/catalog_name.rs
@@ -35,7 +35,7 @@ pub struct CatalogNameKey<'a> {
    pub catalog: &'a str,
 }

-impl<'a> Default for CatalogNameKey<'a> {
+impl Default for CatalogNameKey<'_> {
    fn default() -> Self {
        Self {
            catalog: DEFAULT_CATALOG_NAME,
--- a/src/common/meta/src/key/datanode_table.rs
+++ b/src/common/meta/src/key/datanode_table.rs
@@ -77,7 +77,7 @@ impl DatanodeTableKey {
    }
 }

-impl<'a> MetadataKey<'a, DatanodeTableKey> for DatanodeTableKey {
+impl MetadataKey<'_, DatanodeTableKey> for DatanodeTableKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }
--- a/src/common/meta/src/key/flow/flow_info.rs
+++ b/src/common/meta/src/key/flow/flow_info.rs
@@ -42,6 +42,8 @@ lazy_static! {
 /// The layout: `__flow/info/{flow_id}`.
 pub struct FlowInfoKey(FlowScoped<FlowInfoKeyInner>);

+pub type FlowInfoDecodeResult = Result<Option<DeserializedValueWithBytes<FlowInfoValue>>>;
+
 impl<'a> MetadataKey<'a, FlowInfoKey> for FlowInfoKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.0.to_bytes()
@@ -203,9 +205,7 @@ impl FlowInfoManager {
        flow_value: &FlowInfoValue,
    ) -> Result<(
        Txn,
-        impl FnOnce(
-            &mut TxnOpGetResponseSet,
-        ) -> Result<Option<DeserializedValueWithBytes<FlowInfoValue>>>,
+        impl FnOnce(&mut TxnOpGetResponseSet) -> FlowInfoDecodeResult,
    )> {
        let key = FlowInfoKey::new(flow_id).to_bytes();
        let txn = Txn::put_if_not_exists(key.clone(), flow_value.try_as_raw_value()?);
--- a/src/common/meta/src/key/flow/flow_name.rs
+++ b/src/common/meta/src/key/flow/flow_name.rs
@@ -46,6 +46,8 @@ lazy_static! {
 /// The layout: `__flow/name/{catalog_name}/{flow_name}`.
 pub struct FlowNameKey<'a>(FlowScoped<FlowNameKeyInner<'a>>);

+pub type FlowNameDecodeResult = Result<Option<DeserializedValueWithBytes<FlowNameValue>>>;
+
 #[allow(dead_code)]
 impl<'a> FlowNameKey<'a> {
    /// Returns the [FlowNameKey]
@@ -104,7 +106,7 @@ impl<'a> MetadataKey<'a, FlowNameKeyInner<'a>> for FlowNameKeyInner<'_> {
        .into_bytes()
    }

-    fn from_bytes(bytes: &'a [u8]) -> Result<FlowNameKeyInner> {
+    fn from_bytes(bytes: &'a [u8]) -> Result<FlowNameKeyInner<'a>> {
        let key = std::str::from_utf8(bytes).map_err(|e| {
            error::InvalidMetadataSnafu {
                err_msg: format!(
@@ -223,9 +225,7 @@ impl FlowNameManager {
        flow_id: FlowId,
    ) -> Result<(
        Txn,
-        impl FnOnce(
-            &mut TxnOpGetResponseSet,
-        ) -> Result<Option<DeserializedValueWithBytes<FlowNameValue>>>,
+        impl FnOnce(&mut TxnOpGetResponseSet) -> FlowNameDecodeResult,
    )> {
        let key = FlowNameKey::new(catalog_name, flow_name);
        let raw_key = key.to_bytes();
--- a/src/common/meta/src/key/node_address.rs
+++ b/src/common/meta/src/key/node_address.rs
@@ -52,7 +52,7 @@ impl NodeAddressValue {
    }
 }

-impl<'a> MetadataKey<'a, NodeAddressKey> for NodeAddressKey {
+impl MetadataKey<'_, NodeAddressKey> for NodeAddressKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }
--- a/src/common/meta/src/key/schema_name.rs
+++ b/src/common/meta/src/key/schema_name.rs
@@ -41,7 +41,7 @@ pub struct SchemaNameKey<'a> {
    pub schema: &'a str,
 }

-impl<'a> Default for SchemaNameKey<'a> {
+impl Default for SchemaNameKey<'_> {
    fn default() -> Self {
        Self {
            catalog: DEFAULT_CATALOG_NAME,
--- a/src/common/meta/src/key/table_info.rs
+++ b/src/common/meta/src/key/table_info.rs
@@ -51,7 +51,7 @@ impl Display for TableInfoKey {
    }
 }

-impl<'a> MetadataKey<'a, TableInfoKey> for TableInfoKey {
+impl MetadataKey<'_, TableInfoKey> for TableInfoKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }
@@ -132,6 +132,7 @@ pub type TableInfoManagerRef = Arc<TableInfoManager>;
 pub struct TableInfoManager {
    kv_backend: KvBackendRef,
 }
+pub type TableInfoDecodeResult = Result<Option<DeserializedValueWithBytes<TableInfoValue>>>;

 impl TableInfoManager {
    pub fn new(kv_backend: KvBackendRef) -> Self {
@@ -145,9 +146,7 @@ impl TableInfoManager {
        table_info_value: &TableInfoValue,
    ) -> Result<(
        Txn,
-        impl FnOnce(
-            &mut TxnOpGetResponseSet,
-        ) -> Result<Option<DeserializedValueWithBytes<TableInfoValue>>>,
+        impl FnOnce(&mut TxnOpGetResponseSet) -> TableInfoDecodeResult,
    )> {
        let key = TableInfoKey::new(table_id);
        let raw_key = key.to_bytes();
@@ -169,9 +168,7 @@ impl TableInfoManager {
        new_table_info_value: &TableInfoValue,
    ) -> Result<(
        Txn,
-        impl FnOnce(
-            &mut TxnOpGetResponseSet,
-        ) -> Result<Option<DeserializedValueWithBytes<TableInfoValue>>>,
+        impl FnOnce(&mut TxnOpGetResponseSet) -> TableInfoDecodeResult,
    )> {
        let key = TableInfoKey::new(table_id);
        let raw_key = key.to_bytes();
--- a/src/common/meta/src/key/table_route.rs
+++ b/src/common/meta/src/key/table_route.rs
@@ -245,7 +245,7 @@ impl LogicalTableRouteValue {
    }
 }

-impl<'a> MetadataKey<'a, TableRouteKey> for TableRouteKey {
+impl MetadataKey<'_, TableRouteKey> for TableRouteKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }
@@ -472,6 +472,8 @@ pub struct TableRouteStorage {
    kv_backend: KvBackendRef,
 }

+pub type TableRouteValueDecodeResult = Result<Option<DeserializedValueWithBytes<TableRouteValue>>>;
+
 impl TableRouteStorage {
    pub fn new(kv_backend: KvBackendRef) -> Self {
        Self { kv_backend }
@@ -485,9 +487,7 @@ impl TableRouteStorage {
        table_route_value: &TableRouteValue,
    ) -> Result<(
        Txn,
-        impl FnOnce(
-            &mut TxnOpGetResponseSet,
-        ) -> Result<Option<DeserializedValueWithBytes<TableRouteValue>>>,
+        impl FnOnce(&mut TxnOpGetResponseSet) -> TableRouteValueDecodeResult,
    )> {
        let key = TableRouteKey::new(table_id);
        let raw_key = key.to_bytes();
@@ -510,9 +510,7 @@ impl TableRouteStorage {
        new_table_route_value: &TableRouteValue,
    ) -> Result<(
        Txn,
-        impl FnOnce(
-            &mut TxnOpGetResponseSet,
-        ) -> Result<Option<DeserializedValueWithBytes<TableRouteValue>>>,
+        impl FnOnce(&mut TxnOpGetResponseSet) -> TableRouteValueDecodeResult,
    )> {
        let key = TableRouteKey::new(table_id);
        let raw_key = key.to_bytes();
--- a/src/common/meta/src/key/view_info.rs
+++ b/src/common/meta/src/key/view_info.rs
@@ -53,7 +53,7 @@ impl Display for ViewInfoKey {
    }
 }

-impl<'a> MetadataKey<'a, ViewInfoKey> for ViewInfoKey {
+impl MetadataKey<'_, ViewInfoKey> for ViewInfoKey {
    fn to_bytes(&self) -> Vec<u8> {
        self.to_string().into_bytes()
    }
@@ -139,6 +139,8 @@ pub struct ViewInfoManager {

 pub type ViewInfoManagerRef = Arc<ViewInfoManager>;

+pub type ViewInfoValueDecodeResult = Result<Option<DeserializedValueWithBytes<ViewInfoValue>>>;
+
 impl ViewInfoManager {
    pub fn new(kv_backend: KvBackendRef) -> Self {
        Self { kv_backend }
@@ -151,9 +153,7 @@ impl ViewInfoManager {
        view_info_value: &ViewInfoValue,
    ) -> Result<(
        Txn,
-        impl FnOnce(
-            &mut TxnOpGetResponseSet,
-        ) -> Result<Option<DeserializedValueWithBytes<ViewInfoValue>>>,
+        impl FnOnce(&mut TxnOpGetResponseSet) -> ViewInfoValueDecodeResult,
    )> {
        let key = ViewInfoKey::new(view_id);
        let raw_key = key.to_bytes();
@@ -175,9 +175,7 @@ impl ViewInfoManager {
        new_view_info_value: &ViewInfoValue,
    ) -> Result<(
        Txn,
-        impl FnOnce(
-            &mut TxnOpGetResponseSet,
-        ) -> Result<Option<DeserializedValueWithBytes<ViewInfoValue>>>,
+        impl FnOnce(&mut TxnOpGetResponseSet) -> ViewInfoValueDecodeResult,
    )> {
        let key = ViewInfoKey::new(view_id);
        let raw_key = key.to_bytes();
--- a/src/common/meta/src/leadership_notifier.rs
+++ b/src/common/meta/src/leadership_notifier.rs
@@ -12,10 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use std::sync::Arc;
+use std::sync::{Arc, Mutex};

 use async_trait::async_trait;
-use common_telemetry::error;
+use common_telemetry::{error, info};

 use crate::error::Result;

@@ -24,6 +24,8 @@ pub type LeadershipChangeNotifierCustomizerRef = Arc<dyn LeadershipChangeNotifie
 /// 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.
@@ -45,6 +47,31 @@ 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>) {
--- a/src/common/meta/src/lock_key.rs
+++ b/src/common/meta/src/lock_key.rs
@@ -34,7 +34,7 @@ pub enum CatalogLock<'a> {
    Write(&'a str),
 }

-impl<'a> Display for CatalogLock<'a> {
+impl Display for CatalogLock<'_> {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        let key = match self {
            CatalogLock::Read(s) => s,
@@ -44,7 +44,7 @@ impl<'a> Display for CatalogLock<'a> {
    }
 }

-impl<'a> From<CatalogLock<'a>> for StringKey {
+impl From<CatalogLock<'_>> for StringKey {
    fn from(value: CatalogLock) -> Self {
        match value {
            CatalogLock::Write(_) => StringKey::Exclusive(value.to_string()),
--- a/src/common/meta/src/rpc/router.rs
+++ b/src/common/meta/src/rpc/router.rs
@@ -289,6 +289,7 @@ pub enum LeaderState {
    ///
    /// - The [`Region`] may be unavailable (e.g., Crashed, Network disconnected).
    /// - The [`Region`] was planned to migrate to another [`Peer`].
+    #[serde(alias = "Downgraded")]
    Downgrading,
 }

@@ -516,6 +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 {
--- a/src/common/pprof/Cargo.toml
+++ b/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
--- a/src/common/pprof/src/lib.rs
+++ b/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;
--- a/src/common/pprof/src/nix.rs
+++ b/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)
+    }
+}
--- a/src/common/procedure/src/store.rs
+++ b/src/common/procedure/src/store.rs
@@ -297,7 +297,7 @@ struct ParsedKey<'a> {
    key_type: KeyType,
 }

-impl<'a> fmt::Display for ParsedKey<'a> {
+impl fmt::Display for ParsedKey<'_> {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        write!(
            f,
--- a/src/common/recordbatch/src/recordbatch.rs
+++ b/src/common/recordbatch/src/recordbatch.rs
@@ -17,6 +17,7 @@ use std::slice;
 use std::sync::Arc;

 use datafusion::arrow::util::pretty::pretty_format_batches;
+use datatypes::prelude::DataType;
 use datatypes::schema::SchemaRef;
 use datatypes::value::Value;
 use datatypes::vectors::{Helper, VectorRef};
@@ -58,13 +59,18 @@ impl RecordBatch {
    }

    /// Create an empty [`RecordBatch`] from `schema`.
-    pub fn new_empty(schema: SchemaRef) -> Result<RecordBatch> {
+    pub fn new_empty(schema: SchemaRef) -> RecordBatch {
        let df_record_batch = DfRecordBatch::new_empty(schema.arrow_schema().clone());
-        Ok(RecordBatch {
+        let columns = schema
+            .column_schemas()
+            .iter()
+            .map(|col| col.data_type.create_mutable_vector(0).to_vector())
+            .collect();
+        RecordBatch {
            schema,
-            columns: vec![],
+            columns,
            df_record_batch,
-        })
+        }
    }

    pub fn try_project(&self, indices: &[usize]) -> Result<Self> {
@@ -220,7 +226,7 @@ pub struct RecordBatchRowIterator<'a> {
 }

 impl<'a> RecordBatchRowIterator<'a> {
-    fn new(record_batch: &'a RecordBatch) -> RecordBatchRowIterator {
+    fn new(record_batch: &'a RecordBatch) -> RecordBatchRowIterator<'a> {
        RecordBatchRowIterator {
            record_batch,
            rows: record_batch.df_record_batch.num_rows(),
@@ -230,7 +236,7 @@ impl<'a> RecordBatchRowIterator<'a> {
    }
 }

-impl<'a> Iterator for RecordBatchRowIterator<'a> {
+impl Iterator for RecordBatchRowIterator<'_> {
    type Item = Vec<Value>;

    fn next(&mut self) -> Option<Self::Item> {
--- a/src/common/runtime/Cargo.toml
+++ b/src/common/runtime/Cargo.toml
@@ -4,21 +4,36 @@ version.workspace = true
 edition.workspace = true
 license.workspace = true

+[lib]
+path = "src/lib.rs"
+
+[[bin]]
+name = "common-runtime-bin"
+path = "src/bin.rs"
+
 [lints]
 workspace = true

 [dependencies]
 async-trait.workspace = true
+clap.workspace = true
 common-error.workspace = true
 common-macro.workspace = true
 common-telemetry.workspace = true
+futures.workspace = true
 lazy_static.workspace = true
 num_cpus.workspace = true
 once_cell.workspace = true
+parking_lot.workspace = true
 paste.workspace = true
+pin-project.workspace = true
 prometheus.workspace = true
+rand.workspace = true
+ratelimit.workspace = true
 serde.workspace = true
+serde_json.workspace = true
 snafu.workspace = true
+tempfile.workspace = true
 tokio.workspace = true
 tokio-metrics = "0.3"
 tokio-metrics-collector = { git = "https://github.com/MichaelScofield/tokio-metrics-collector.git", rev = "89d692d5753d28564a7aac73c6ac5aba22243ba0" }
--- a/src/common/runtime/README.md
+++ b/src/common/runtime/README.md
@@ -0,0 +1,60 @@
+# Greptime Runtime
+
+## Run performance test for different priority & workload type
+
+```
+# workspace is at this subcrate
+cargo run --release -- --loop-cnt 500
+```
+
+## Related PRs & issues
+
+- Preliminary support cpu limitation
+  
+  ISSUE: https://github.com/GreptimeTeam/greptimedb/issues/3685
+  
+  PR: https://github.com/GreptimeTeam/greptimedb/pull/4782
+
+## CPU resource constraints (ThrottleableRuntime)
+
+
+To achieve CPU resource constraints, we adopt the concept of rate limiting. When creating a future, we first wrap it with another layer of future to intercept the poll operation during runtime. By using the ratelimit library, we can simply implement a mechanism that allows only a limited number of polls for a batch of tasks under a certain priority within a specific time frame (the current token generation interval is set to 10ms).
+
+The default used runtime can be switched by 
+``` rust
+pub type Runtime = DefaultRuntime;
+```
+in `runtime.rs`.
+
+We tested four type of workload with 5 priorities, whose setup are as follows:
+
+``` rust
+impl Priority {
+    fn ratelimiter_count(&self) -> Result<Option<Ratelimiter>> {
+        let max = 8000;
+        let gen_per_10ms = match self {
+            Priority::VeryLow => Some(2000),
+            Priority::Low => Some(4000),
+            Priority::Middle => Some(6000),
+            Priority::High => Some(8000),
+            Priority::VeryHigh => None,
+        };
+        if let Some(gen_per_10ms) = gen_per_10ms {
+            Ratelimiter::builder(gen_per_10ms, Duration::from_millis(10)) // generate poll count per 10ms
+                .max_tokens(max) // reserved token for batch request
+                .build()
+                .context(BuildRuntimeRateLimiterSnafu)
+                .map(Some)
+        } else {
+            Ok(None)
+        }
+    }
+}
+```
+
+This is the preliminary experimental effect so far:
+
+![](resources/rdme-exp.png)
+
+## TODO
+- Introduce PID to achieve more accurate limitation.
--- a/src/common/runtime/resources/rdme-exp.png
+++ b/src/common/runtime/resources/rdme-exp.png
--- a/src/common/runtime/src/bin.rs
+++ b/src/common/runtime/src/bin.rs
@@ -0,0 +1,205 @@
+// 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 clap::Parser;
+
+#[derive(Debug, Default, Parser)]
+pub struct Command {
+    #[clap(long)]
+    loop_cnt: usize,
+}
+
+fn main() {
+    common_telemetry::init_default_ut_logging();
+    let cmd = Command::parse();
+
+    test_diff_priority_cpu::test_diff_workload_priority(cmd.loop_cnt);
+}
+
+mod test_diff_priority_cpu {
+    use std::path::PathBuf;
+
+    use common_runtime::runtime::{BuilderBuild, Priority, RuntimeTrait};
+    use common_runtime::{Builder, Runtime};
+    use common_telemetry::debug;
+    use tempfile::TempDir;
+
+    fn compute_pi_str(precision: usize) -> String {
+        let mut pi = 0.0;
+        let mut sign = 1.0;
+
+        for i in 0..precision {
+            pi += sign / (2 * i + 1) as f64;
+            sign *= -1.0;
+        }
+
+        pi *= 4.0;
+        format!("{:.prec$}", pi, prec = precision)
+    }
+
+    macro_rules! def_workload_enum {
+        ($($variant:ident),+) => {
+            #[derive(Debug)]
+            enum WorkloadType {
+                $($variant),+
+            }
+
+            /// array of workloads for iteration
+            const WORKLOADS: &'static [WorkloadType] = &[
+                $( WorkloadType::$variant ),+
+            ];
+        };
+    }
+
+    def_workload_enum!(
+        ComputeHeavily,
+        ComputeHeavily2,
+        WriteFile,
+        SpawnBlockingWriteFile
+    );
+
+    async fn workload_compute_heavily() {
+        let prefix = 10;
+
+        for _ in 0..3000 {
+            let _ = compute_pi_str(prefix);
+            tokio::task::yield_now().await;
+        }
+    }
+    async fn workload_compute_heavily2() {
+        let prefix = 30;
+        for _ in 0..2000 {
+            let _ = compute_pi_str(prefix);
+            tokio::task::yield_now().await;
+        }
+    }
+    async fn workload_write_file(_idx: u64, tempdir: PathBuf) {
+        use tokio::io::AsyncWriteExt;
+        let prefix = 50;
+
+        let mut file = tokio::fs::OpenOptions::new()
+            .write(true)
+            .append(true)
+            .create(true)
+            .open(tempdir.join(format!("pi_{}", prefix)))
+            .await
+            .unwrap();
+        for i in 0..200 {
+            let pi = compute_pi_str(prefix);
+
+            if i % 2 == 0 {
+                file.write_all(pi.as_bytes()).await.unwrap();
+            }
+        }
+    }
+    async fn workload_spawn_blocking_write_file(tempdir: PathBuf) {
+        use std::io::Write;
+        let prefix = 100;
+        let mut file = Some(
+            std::fs::OpenOptions::new()
+                .append(true)
+                .create(true)
+                .open(tempdir.join(format!("pi_{}", prefix)))
+                .unwrap(),
+        );
+        for i in 0..100 {
+            let pi = compute_pi_str(prefix);
+            if i % 2 == 0 {
+                let mut file1 = file.take().unwrap();
+                file = Some(
+                    tokio::task::spawn_blocking(move || {
+                        file1.write_all(pi.as_bytes()).unwrap();
+                        file1
+                    })
+                    .await
+                    .unwrap(),
+                );
+            }
+        }
+    }
+
+    pub fn test_diff_workload_priority(loop_cnt: usize) {
+        let tempdir = tempfile::tempdir().unwrap();
+        let priorities = [
+            Priority::VeryLow,
+            Priority::Low,
+            Priority::Middle,
+            Priority::High,
+            Priority::VeryHigh,
+        ];
+        for wl in WORKLOADS {
+            for p in priorities.iter() {
+                let runtime: Runtime = Builder::default()
+                    .runtime_name("test")
+                    .thread_name("test")
+                    .worker_threads(8)
+                    .priority(*p)
+                    .build()
+                    .expect("Fail to create runtime");
+                let runtime2 = runtime.clone();
+                runtime.block_on(test_spec_priority_and_workload(
+                    *p, runtime2, wl, &tempdir, loop_cnt,
+                ));
+            }
+        }
+    }
+
+    async fn test_spec_priority_and_workload(
+        priority: Priority,
+        runtime: Runtime,
+        workload_id: &WorkloadType,
+        tempdir: &TempDir,
+        loop_cnt: usize,
+    ) {
+        tokio::time::sleep(tokio::time::Duration::from_millis(1000)).await;
+        debug!(
+            "testing cpu usage for priority {:?} workload_id {:?}",
+            priority, workload_id,
+        );
+        // start monitor thread
+        let mut tasks = vec![];
+        let start = std::time::Instant::now();
+        for i in 0..loop_cnt {
+            // persist cpu usage in json: {priority}.{workload_id}
+            match *workload_id {
+                WorkloadType::ComputeHeavily => {
+                    tasks.push(runtime.spawn(workload_compute_heavily()));
+                }
+                WorkloadType::ComputeHeavily2 => {
+                    tasks.push(runtime.spawn(workload_compute_heavily2()));
+                }
+                WorkloadType::SpawnBlockingWriteFile => {
+                    tasks.push(runtime.spawn(workload_spawn_blocking_write_file(
+                        tempdir.path().to_path_buf(),
+                    )));
+                }
+                WorkloadType::WriteFile => {
+                    tasks.push(
+                        runtime.spawn(workload_write_file(i as u64, tempdir.path().to_path_buf())),
+                    );
+                }
+            }
+        }
+        for task in tasks {
+            task.await.unwrap();
+        }
+        let elapsed = start.elapsed();
+        debug!(
+            "test cpu usage for priority {:?} workload_id {:?} elapsed {}ms",
+            priority,
+            workload_id,
+            elapsed.as_millis()
+        );
+    }
+}
--- a/src/common/runtime/src/error.rs
+++ b/src/common/runtime/src/error.rs
@@ -33,6 +33,14 @@ pub enum Error {
        location: Location,
    },

+    #[snafu(display("Failed to build runtime rate limiter"))]
+    BuildRuntimeRateLimiter {
+        #[snafu(implicit)]
+        location: Location,
+        #[snafu(source)]
+        error: ratelimit::Error,
+    },
+
    #[snafu(display("Repeated task {} is already started", name))]
    IllegalState {
        name: String,
--- a/src/common/runtime/src/global.rs
+++ b/src/common/runtime/src/global.rs
@@ -21,6 +21,7 @@ use once_cell::sync::Lazy;
 use paste::paste;
 use serde::{Deserialize, Serialize};

+use crate::runtime::{BuilderBuild, RuntimeTrait};
 use crate::{Builder, JoinHandle, Runtime};

 const GLOBAL_WORKERS: usize = 8;
--- a/src/common/runtime/src/lib.rs
+++ b/src/common/runtime/src/lib.rs
@@ -17,6 +17,8 @@ pub mod global;
 mod metrics;
 mod repeated_task;
 pub mod runtime;
+pub mod runtime_default;
+pub mod runtime_throttleable;

 pub use global::{
    block_on_compact, block_on_global, compact_runtime, create_runtime, global_runtime,
--- a/src/common/runtime/src/repeated_task.rs
+++ b/src/common/runtime/src/repeated_task.rs
@@ -23,6 +23,7 @@ use tokio::task::JoinHandle;
 use tokio_util::sync::CancellationToken;

 use crate::error::{IllegalStateSnafu, Result, WaitGcTaskStopSnafu};
+use crate::runtime::RuntimeTrait;
 use crate::Runtime;

 /// Task to execute repeatedly.
--- a/src/common/runtime/src/runtime.rs
+++ b/src/common/runtime/src/runtime.rs
@@ -19,24 +19,20 @@ use std::thread;
 use std::time::Duration;

 use snafu::ResultExt;
-use tokio::runtime::{Builder as RuntimeBuilder, Handle};
+use tokio::runtime::Builder as RuntimeBuilder;
 use tokio::sync::oneshot;
 pub use tokio::task::{JoinError, JoinHandle};

 use crate::error::*;
 use crate::metrics::*;
+use crate::runtime_default::DefaultRuntime;
+use crate::runtime_throttleable::ThrottleableRuntime;
+
+// configurations
+pub type Runtime = DefaultRuntime;

 static RUNTIME_ID: AtomicUsize = AtomicUsize::new(0);

-/// A runtime to run future tasks
-#[derive(Clone, Debug)]
-pub struct Runtime {
-    name: String,
-    handle: Handle,
-    // Used to receive a drop signal when dropper is dropped, inspired by databend
-    _dropper: Arc<Dropper>,
-}
-
 /// Dropping the dropper will cause runtime to shutdown.
 #[derive(Debug)]
 pub struct Dropper {
@@ -50,45 +46,42 @@ impl Drop for Dropper {
    }
 }

-impl Runtime {
-    pub fn builder() -> Builder {
+pub trait RuntimeTrait {
+    /// Get a runtime builder
+    fn builder() -> Builder {
        Builder::default()
    }

    /// Spawn a future and execute it in this thread pool
    ///
    /// Similar to tokio::runtime::Runtime::spawn()
-    pub fn spawn<F>(&self, future: F) -> JoinHandle<F::Output>
+    fn spawn<F>(&self, future: F) -> JoinHandle<F::Output>
    where
        F: Future + Send + 'static,
-        F::Output: Send + 'static,
-    {
-        self.handle.spawn(future)
-    }
+        F::Output: Send + 'static;

    /// Run the provided function on an executor dedicated to blocking
    /// operations.
-    pub fn spawn_blocking<F, R>(&self, func: F) -> JoinHandle<R>
+    fn spawn_blocking<F, R>(&self, func: F) -> JoinHandle<R>
    where
        F: FnOnce() -> R + Send + 'static,
-        R: Send + 'static,
-    {
-        self.handle.spawn_blocking(func)
-    }
+        R: Send + 'static;

    /// Run a future to complete, this is the runtime's entry point
-    pub fn block_on<F: Future>(&self, future: F) -> F::Output {
-        self.handle.block_on(future)
-    }
+    fn block_on<F: Future>(&self, future: F) -> F::Output;

-    pub fn name(&self) -> &str {
-        &self.name
-    }
+    /// Get the name of the runtime
+    fn name(&self) -> &str;
+}
+
+pub trait BuilderBuild<R: RuntimeTrait> {
+    fn build(&mut self) -> Result<R>;
 }

 pub struct Builder {
    runtime_name: String,
    thread_name: String,
+    priority: Priority,
    builder: RuntimeBuilder,
 }

@@ -98,11 +91,17 @@ impl Default for Builder {
            runtime_name: format!("runtime-{}", RUNTIME_ID.fetch_add(1, Ordering::Relaxed)),
            thread_name: "default-worker".to_string(),
            builder: RuntimeBuilder::new_multi_thread(),
+            priority: Priority::VeryHigh,
        }
    }
 }

 impl Builder {
+    pub fn priority(&mut self, priority: Priority) -> &mut Self {
+        self.priority = priority;
+        self
+    }
+
    /// Sets the number of worker threads the Runtime will use.
    ///
    /// This can be any number above 0. The default value is the number of cores available to the system.
@@ -139,8 +138,10 @@ impl Builder {
        self.thread_name = val.into();
        self
    }
+}

-    pub fn build(&mut self) -> Result<Runtime> {
+impl BuilderBuild<DefaultRuntime> for Builder {
+    fn build(&mut self) -> Result<DefaultRuntime> {
        let runtime = self
            .builder
            .enable_all()
@@ -163,18 +164,53 @@ impl Builder {
        #[cfg(tokio_unstable)]
        register_collector(name.clone(), &handle);

-        Ok(Runtime {
-            name,
+        Ok(DefaultRuntime::new(
+            &name,
            handle,
-            _dropper: Arc::new(Dropper {
+            Arc::new(Dropper {
                close: Some(send_stop),
            }),
-        })
+        ))
+    }
+}
+
+impl BuilderBuild<ThrottleableRuntime> for Builder {
+    fn build(&mut self) -> Result<ThrottleableRuntime> {
+        let runtime = self
+            .builder
+            .enable_all()
+            .thread_name(self.thread_name.clone())
+            .on_thread_start(on_thread_start(self.thread_name.clone()))
+            .on_thread_stop(on_thread_stop(self.thread_name.clone()))
+            .on_thread_park(on_thread_park(self.thread_name.clone()))
+            .on_thread_unpark(on_thread_unpark(self.thread_name.clone()))
+            .build()
+            .context(BuildRuntimeSnafu)?;
+
+        let name = self.runtime_name.clone();
+        let handle = runtime.handle().clone();
+        let (send_stop, recv_stop) = oneshot::channel();
+        // Block the runtime to shutdown.
+        let _ = thread::Builder::new()
+            .name(format!("{}-blocker", self.thread_name))
+            .spawn(move || runtime.block_on(recv_stop));
+
+        #[cfg(tokio_unstable)]
+        register_collector(name.clone(), &handle);
+
+        ThrottleableRuntime::new(
+            &name,
+            self.priority,
+            handle,
+            Arc::new(Dropper {
+                close: Some(send_stop),
+            }),
+        )
    }
 }

 #[cfg(tokio_unstable)]
-pub fn register_collector(name: String, handle: &Handle) {
+pub fn register_collector(name: String, handle: &tokio::runtime::Handle) {
    let name = name.replace("-", "_");
    let monitor = tokio_metrics::RuntimeMonitor::new(handle);
    let collector = tokio_metrics_collector::RuntimeCollector::new(monitor, name);
@@ -213,8 +249,18 @@ fn on_thread_unpark(thread_name: String) -> impl Fn() + 'static {
    }
 }

+#[derive(Clone, Copy, Debug, Hash, PartialEq, Eq)]
+pub enum Priority {
+    VeryLow = 0,
+    Low = 1,
+    Middle = 2,
+    High = 3,
+    VeryHigh = 4,
+}
+
 #[cfg(test)]
 mod tests {
+
    use std::sync::Arc;
    use std::thread;
    use std::time::Duration;
@@ -235,12 +281,12 @@ mod tests {

    #[test]
    fn test_metric() {
-        let runtime = Builder::default()
+        let runtime: Runtime = Builder::default()
            .worker_threads(5)
            .thread_name("test_runtime_metric")
            .build()
            .unwrap();
-        // wait threads created
+        // wait threads create
        thread::sleep(Duration::from_millis(50));

        let _handle = runtime.spawn(async {
--- a/src/common/runtime/src/runtime_default.rs
+++ b/src/common/runtime/src/runtime_default.rs
@@ -0,0 +1,77 @@
+// 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::future::Future;
+use std::sync::Arc;
+
+use tokio::runtime::Handle;
+pub use tokio::task::JoinHandle;
+
+use crate::runtime::{Dropper, RuntimeTrait};
+use crate::Builder;
+
+/// A runtime to run future tasks
+#[derive(Clone, Debug)]
+pub struct DefaultRuntime {
+    name: String,
+    handle: Handle,
+    // Used to receive a drop signal when dropper is dropped, inspired by databend
+    _dropper: Arc<Dropper>,
+}
+
+impl DefaultRuntime {
+    pub(crate) fn new(name: &str, handle: Handle, dropper: Arc<Dropper>) -> Self {
+        Self {
+            name: name.to_string(),
+            handle,
+            _dropper: dropper,
+        }
+    }
+}
+
+impl RuntimeTrait for DefaultRuntime {
+    fn builder() -> Builder {
+        Builder::default()
+    }
+
+    /// Spawn a future and execute it in this thread pool
+    ///
+    /// Similar to tokio::runtime::Runtime::spawn()
+    fn spawn<F>(&self, future: F) -> JoinHandle<F::Output>
+    where
+        F: Future + Send + 'static,
+        F::Output: Send + 'static,
+    {
+        self.handle.spawn(future)
+    }
+
+    /// Run the provided function on an executor dedicated to blocking
+    /// operations.
+    fn spawn_blocking<F, R>(&self, func: F) -> JoinHandle<R>
+    where
+        F: FnOnce() -> R + Send + 'static,
+        R: Send + 'static,
+    {
+        self.handle.spawn_blocking(func)
+    }
+
+    /// Run a future to complete, this is the runtime's entry point
+    fn block_on<F: Future>(&self, future: F) -> F::Output {
+        self.handle.block_on(future)
+    }
+
+    fn name(&self) -> &str {
+        &self.name
+    }
+}
--- a/src/common/runtime/src/runtime_throttleable.rs
+++ b/src/common/runtime/src/runtime_throttleable.rs
@@ -0,0 +1,285 @@
+// 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::Debug;
+use std::future::Future;
+use std::pin::Pin;
+use std::sync::Arc;
+use std::task::{Context, Poll};
+use std::time::Duration;
+
+use futures::FutureExt;
+use ratelimit::Ratelimiter;
+use snafu::ResultExt;
+use tokio::runtime::Handle;
+pub use tokio::task::JoinHandle;
+use tokio::time::Sleep;
+
+use crate::error::{BuildRuntimeRateLimiterSnafu, Result};
+use crate::runtime::{Dropper, Priority, RuntimeTrait};
+use crate::Builder;
+
+struct RuntimeRateLimiter {
+    pub ratelimiter: Option<Ratelimiter>,
+}
+
+impl Debug for RuntimeRateLimiter {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("RuntimeThrottleShareWithFuture")
+            .field(
+                "ratelimiter_max_tokens",
+                &self.ratelimiter.as_ref().map(|v| v.max_tokens()),
+            )
+            .field(
+                "ratelimiter_refill_amount",
+                &self.ratelimiter.as_ref().map(|v| v.refill_amount()),
+            )
+            .finish()
+    }
+}
+
+/// A runtime to run future tasks
+#[derive(Clone, Debug)]
+pub struct ThrottleableRuntime {
+    name: String,
+    handle: Handle,
+    shared_with_future: Arc<RuntimeRateLimiter>,
+    // Used to receive a drop signal when dropper is dropped, inspired by databend
+    _dropper: Arc<Dropper>,
+}
+
+impl ThrottleableRuntime {
+    pub(crate) fn new(
+        name: &str,
+        priority: Priority,
+        handle: Handle,
+        dropper: Arc<Dropper>,
+    ) -> Result<Self> {
+        Ok(Self {
+            name: name.to_string(),
+            handle,
+            shared_with_future: Arc::new(RuntimeRateLimiter {
+                ratelimiter: priority.ratelimiter_count()?,
+            }),
+            _dropper: dropper,
+        })
+    }
+}
+
+impl RuntimeTrait for ThrottleableRuntime {
+    fn builder() -> Builder {
+        Builder::default()
+    }
+
+    /// Spawn a future and execute it in this thread pool
+    ///
+    /// Similar to tokio::runtime::Runtime::spawn()
+    fn spawn<F>(&self, future: F) -> JoinHandle<F::Output>
+    where
+        F: Future + Send + 'static,
+        F::Output: Send + 'static,
+    {
+        self.handle
+            .spawn(ThrottleFuture::new(self.shared_with_future.clone(), future))
+    }
+
+    /// Run the provided function on an executor dedicated to blocking
+    /// operations.
+    fn spawn_blocking<F, R>(&self, func: F) -> JoinHandle<R>
+    where
+        F: FnOnce() -> R + Send + 'static,
+        R: Send + 'static,
+    {
+        self.handle.spawn_blocking(func)
+    }
+
+    /// Run a future to complete, this is the runtime's entry point
+    fn block_on<F: Future>(&self, future: F) -> F::Output {
+        self.handle.block_on(future)
+    }
+
+    fn name(&self) -> &str {
+        &self.name
+    }
+}
+
+enum State {
+    Pollable,
+    Throttled(Pin<Box<Sleep>>),
+}
+
+impl State {
+    fn unwrap_backoff(&mut self) -> &mut Pin<Box<Sleep>> {
+        match self {
+            State::Throttled(sleep) => sleep,
+            _ => panic!("unwrap_backoff failed"),
+        }
+    }
+}
+
+#[pin_project::pin_project]
+pub struct ThrottleFuture<F: Future + Send + 'static> {
+    #[pin]
+    future: F,
+
+    /// RateLimiter of this future
+    handle: Arc<RuntimeRateLimiter>,
+
+    state: State,
+}
+
+impl<F> ThrottleFuture<F>
+where
+    F: Future + Send + 'static,
+    F::Output: Send + 'static,
+{
+    fn new(handle: Arc<RuntimeRateLimiter>, future: F) -> Self {
+        Self {
+            future,
+            handle,
+            state: State::Pollable,
+        }
+    }
+}
+
+impl<F> Future for ThrottleFuture<F>
+where
+    F: Future + Send + 'static,
+    F::Output: Send + 'static,
+{
+    type Output = F::Output;
+
+    fn poll(self: std::pin::Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
+        let this = self.project();
+
+        match this.state {
+            State::Pollable => {}
+            State::Throttled(ref mut sleep) => match sleep.poll_unpin(cx) {
+                Poll::Ready(_) => {
+                    *this.state = State::Pollable;
+                }
+                Poll::Pending => return Poll::Pending,
+            },
+        };
+
+        if let Some(ratelimiter) = &this.handle.ratelimiter {
+            if let Err(wait) = ratelimiter.try_wait() {
+                *this.state = State::Throttled(Box::pin(tokio::time::sleep(wait)));
+                match this.state.unwrap_backoff().poll_unpin(cx) {
+                    Poll::Ready(_) => {
+                        *this.state = State::Pollable;
+                    }
+                    Poll::Pending => {
+                        return Poll::Pending;
+                    }
+                }
+            }
+        }
+
+        let poll_res = this.future.poll(cx);
+
+        match poll_res {
+            Poll::Ready(r) => Poll::Ready(r),
+            Poll::Pending => Poll::Pending,
+        }
+    }
+}
+
+impl Priority {
+    fn ratelimiter_count(&self) -> Result<Option<Ratelimiter>> {
+        let max = 8000;
+        let gen_per_10ms = match self {
+            Priority::VeryLow => Some(2000),
+            Priority::Low => Some(4000),
+            Priority::Middle => Some(6000),
+            Priority::High => Some(8000),
+            Priority::VeryHigh => None,
+        };
+        if let Some(gen_per_10ms) = gen_per_10ms {
+            Ratelimiter::builder(gen_per_10ms, Duration::from_millis(10)) // generate poll count per 10ms
+                .max_tokens(max) // reserved token for batch request
+                .build()
+                .context(BuildRuntimeRateLimiterSnafu)
+                .map(Some)
+        } else {
+            Ok(None)
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    use tokio::fs::File;
+    use tokio::io::AsyncWriteExt;
+    use tokio::time::Duration;
+
+    use super::*;
+    use crate::runtime::BuilderBuild;
+
+    #[tokio::test]
+    async fn test_throttleable_runtime_spawn_simple() {
+        for p in [
+            Priority::VeryLow,
+            Priority::Low,
+            Priority::Middle,
+            Priority::High,
+            Priority::VeryHigh,
+        ] {
+            let runtime: ThrottleableRuntime = Builder::default()
+                .runtime_name("test")
+                .thread_name("test")
+                .worker_threads(8)
+                .priority(p)
+                .build()
+                .expect("Fail to create runtime");
+
+            // Spawn a simple future that returns 42
+            let handle = runtime.spawn(async {
+                tokio::time::sleep(Duration::from_millis(10)).await;
+                42
+            });
+            let result = handle.await.expect("Task panicked");
+            assert_eq!(result, 42);
+        }
+    }
+
+    #[tokio::test]
+    async fn test_throttleable_runtime_spawn_complex() {
+        let tempdir = tempfile::tempdir().unwrap();
+        for p in [
+            Priority::VeryLow,
+            Priority::Low,
+            Priority::Middle,
+            Priority::High,
+            Priority::VeryHigh,
+        ] {
+            let runtime: ThrottleableRuntime = Builder::default()
+                .runtime_name("test")
+                .thread_name("test")
+                .worker_threads(8)
+                .priority(p)
+                .build()
+                .expect("Fail to create runtime");
+            let tempdirpath = tempdir.path().to_path_buf();
+            let handle = runtime.spawn(async move {
+                let mut file = File::create(tempdirpath.join("test.txt")).await.unwrap();
+                file.write_all(b"Hello, world!").await.unwrap();
+                42
+            });
+            let result = handle.await.expect("Task panicked");
+            assert_eq!(result, 42);
+        }
+    }
+}
--- a/src/common/telemetry/Cargo.toml
+++ b/src/common/telemetry/Cargo.toml
@@ -26,7 +26,7 @@ opentelemetry = { version = "0.21.0", default-features = false, features = [
 opentelemetry-otlp = { version = "0.14.0", features = ["tokio"] }
 opentelemetry-semantic-conventions = "0.13.0"
 opentelemetry_sdk = { version = "0.21.0", features = ["rt-tokio"] }
-parking_lot = { version = "0.12" }
+parking_lot.workspace = true
 prometheus.workspace = true
 serde.workspace = true
 serde_json.workspace = true
--- a/src/common/time/src/date.rs
+++ b/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]
--- a/src/common/time/src/datetime.rs
+++ b/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]
--- a/src/common/time/src/interval.rs
+++ b/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]
--- a/src/common/time/src/lib.rs
+++ b/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;
--- a/src/common/time/src/timestamp.rs
+++ b/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]
--- a/src/common/wal/src/config/kafka/datanode.rs
+++ b/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,
        }
    }
 }
--- a/src/datanode/src/region_server.rs
+++ b/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;
@@ -893,7 +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(())
    }
@@ -935,17 +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 {
--- a/src/datanode/src/tests.rs
+++ b/src/datanode/src/tests.rs
@@ -23,6 +23,7 @@ use common_function::function::FunctionRef;
 use common_function::scalars::aggregate::AggregateFunctionMetaRef;
 use common_query::prelude::ScalarUdf;
 use common_query::Output;
+use common_runtime::runtime::{BuilderBuild, RuntimeTrait};
 use common_runtime::Runtime;
 use datafusion_expr::LogicalPlan;
 use query::dataframe::DataFrame;
--- a/src/datatypes/src/interval.rs
+++ b/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);
    }
--- a/src/datatypes/src/scalars.rs
+++ b/src/datatypes/src/scalars.rs
@@ -203,7 +203,7 @@ impl Scalar for bool {
    }
 }

-impl<'a> ScalarRef<'a> for bool {
+impl ScalarRef<'_> for bool {
    type ScalarType = bool;

    #[inline]
@@ -273,7 +273,7 @@ impl Scalar for Date {
    }
 }

-impl<'a> ScalarRef<'a> for Date {
+impl ScalarRef<'_> for Date {
    type ScalarType = Date;

    fn to_owned_scalar(&self) -> Self::ScalarType {
@@ -294,7 +294,7 @@ impl Scalar for Decimal128 {
    }
 }

-impl<'a> ScalarRef<'a> for Decimal128 {
+impl ScalarRef<'_> for Decimal128 {
    type ScalarType = Decimal128;

    fn to_owned_scalar(&self) -> Self::ScalarType {
@@ -315,7 +315,7 @@ impl Scalar for DateTime {
    }
 }

-impl<'a> ScalarRef<'a> for DateTime {
+impl ScalarRef<'_> for DateTime {
    type ScalarType = DateTime;

    fn to_owned_scalar(&self) -> Self::ScalarType {
--- a/src/datatypes/src/types/cast.rs
+++ b/src/datatypes/src/types/cast.rs
@@ -82,8 +82,8 @@ pub fn cast_with_opt(
    }
 }

-/// Return true if the src_value can be casted to dest_type,
-/// Otherwise, return false.
+/// Return true if the src_value can be casted to dest_type, Otherwise, return false.
+///
 /// Notice: this function does not promise that the `cast_with_opt` will succeed,
 /// it only checks whether the src_value can be casted to dest_type.
 pub fn can_cast_type(src_value: &Value, dest_type: &ConcreteDataType) -> bool {
--- a/src/datatypes/src/types/interval_type.rs
+++ b/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>])),
                        }
--- a/src/datatypes/src/types/primitive_type.rs
+++ b/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`.
@@ -83,9 +83,10 @@ pub trait LogicalPrimitiveType: 'static + Sized {
    fn cast_value_ref(value: ValueRef) -> Result<Option<Self::Wrapper>>;
 }

-/// A new type for [WrapperType], complement the `Ord` feature for it. Wrapping non ordered
-/// primitive types like `f32` and `f64` in `OrdPrimitive` can make them be used in places that
-/// require `Ord`. For example, in `Median` UDAFs.
+/// A new type for [WrapperType], complement the `Ord` feature for it.
+///
+/// Wrapping non ordered primitive types like `f32` and `f64` in `OrdPrimitive`
+/// can make them be used in places that require `Ord`. For example, in `Median` UDAFs.
 #[derive(Debug, Clone, Copy, PartialEq)]
 pub struct OrdPrimitive<T: WrapperType>(pub T);

@@ -364,11 +365,7 @@ 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 +407,7 @@ 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,
        }
    }
--- a/src/datatypes/src/types/string_type.rs
+++ b/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()))),

--- a/src/datatypes/src/value.rs
+++ b/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>),
@@ -1049,7 +1087,7 @@ macro_rules! impl_as_for_value_ref {
    };
 }

-impl<'a> ValueRef<'a> {
+impl ValueRef<'_> {
    define_data_type_func!(ValueRef);

    /// Returns true if this is null.
@@ -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].
@@ -1166,13 +1214,13 @@ impl<'a> ValueRef<'a> {
    }
 }

-impl<'a> PartialOrd for ValueRef<'a> {
+impl PartialOrd for ValueRef<'_> {
    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
        Some(self.cmp(other))
    }
 }

-impl<'a> Ord for ValueRef<'a> {
+impl Ord for ValueRef<'_> {
    fn cmp(&self, other: &Self) -> Ordering {
        impl_ord_for_value_like!(ValueRef, self, other)
    }
@@ -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].
@@ -1282,7 +1347,7 @@ pub enum ListValueRef<'a> {
    Ref { val: &'a ListValue },
 }

-impl<'a> ListValueRef<'a> {
+impl ListValueRef<'_> {
    /// Convert self to [Value]. This method would clone the underlying data.
    fn to_value(self) -> Value {
        match self {
@@ -1300,7 +1365,7 @@ impl<'a> ListValueRef<'a> {
    }
 }

-impl<'a> Serialize for ListValueRef<'a> {
+impl Serialize for ListValueRef<'_> {
    fn serialize<S: Serializer>(&self, serializer: S) -> std::result::Result<S::Ok, S::Error> {
        match self {
            ListValueRef::Indexed { vector, idx } => match vector.get(*idx) {
@@ -1312,28 +1377,28 @@ impl<'a> Serialize for ListValueRef<'a> {
    }
 }

-impl<'a> PartialEq for ListValueRef<'a> {
+impl PartialEq for ListValueRef<'_> {
    fn eq(&self, other: &Self) -> bool {
        self.to_value().eq(&other.to_value())
    }
 }

-impl<'a> Eq for ListValueRef<'a> {}
+impl Eq for ListValueRef<'_> {}

-impl<'a> Ord for ListValueRef<'a> {
+impl Ord for ListValueRef<'_> {
    fn cmp(&self, other: &Self) -> Ordering {
        // Respect the order of `Value` by converting into value before comparison.
        self.to_value().cmp(&other.to_value())
    }
 }

-impl<'a> PartialOrd for ListValueRef<'a> {
+impl PartialOrd for ListValueRef<'_> {
    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
        Some(self.cmp(other))
    }
 }

-impl<'a> ValueRef<'a> {
+impl ValueRef<'_> {
    /// Returns the size of the underlying data in bytes,
    /// The size is estimated and only considers the data size.
    pub fn data_size(&self) -> usize {
@@ -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(
--- a/src/datatypes/src/vectors/decimal.rs
+++ b/src/datatypes/src/vectors/decimal.rs
@@ -247,7 +247,7 @@ pub struct Decimal128Iter<'a> {
    iter: ArrayIter<&'a Decimal128Array>,
 }

-impl<'a> Iterator for Decimal128Iter<'a> {
+impl Iterator for Decimal128Iter<'_> {
    type Item = Option<Decimal128>;

    fn next(&mut self) -> Option<Self::Item> {
--- a/src/datatypes/src/vectors/helper.rs
+++ b/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)
+            );
        }
    }

--- a/src/datatypes/src/vectors/list.rs
+++ b/src/datatypes/src/vectors/list.rs
@@ -157,7 +157,7 @@ pub struct ListIter<'a> {
 }

 impl<'a> ListIter<'a> {
-    fn new(vector: &'a ListVector) -> ListIter {
+    fn new(vector: &'a ListVector) -> ListIter<'a> {
        ListIter { vector, idx: 0 }
    }
 }
--- a/src/datatypes/src/vectors/primitive.rs
+++ b/src/datatypes/src/vectors/primitive.rs
@@ -207,7 +207,7 @@ pub struct PrimitiveIter<'a, T: LogicalPrimitiveType> {
    iter: ArrayIter<&'a PrimitiveArray<T::ArrowPrimitive>>,
 }

-impl<'a, T: LogicalPrimitiveType> Iterator for PrimitiveIter<'a, T> {
+impl<T: LogicalPrimitiveType> Iterator for PrimitiveIter<'_, T> {
    type Item = Option<T::Wrapper>;

    fn next(&mut self) -> Option<Option<T::Wrapper>> {
--- a/src/flow/src/adapter.rs
+++ b/src/flow/src/adapter.rs
@@ -271,10 +271,17 @@ impl FlowWorkerManager {
                        let rows_proto: Vec<v1::Row> = insert
                            .into_iter()
                            .map(|(mut row, _ts)| {
-                                // `update_at` col
-                                row.extend([Value::from(common_time::Timestamp::new_millisecond(
-                                    now,
-                                ))]);
+                                // extend `update_at` col if needed
+                                // if schema include a millisecond timestamp here, and result row doesn't have it, add it
+                                if row.len() < proto_schema.len()
+                                    && proto_schema[row.len()].datatype
+                                        == greptime_proto::v1::ColumnDataType::TimestampMillisecond
+                                            as i32
+                                {
+                                    row.extend([Value::from(
+                                        common_time::Timestamp::new_millisecond(now),
+                                    )]);
+                                }
                                // ts col, if auto create
                                if is_ts_placeholder {
                                    ensure!(
@@ -291,6 +298,17 @@ impl FlowWorkerManager {
                                        common_time::Timestamp::new_millisecond(0),
                                    )]);
                                }
+                                if row.len() != proto_schema.len() {
+                                    InternalSnafu {
+                                        reason: format!(
+                                            "Flow output row length mismatch, expect {} got {}, the columns in schema are: {:?}",
+                                            proto_schema.len(),
+                                            row.len(),
+                                            proto_schema.iter().map(|c|&c.column_name).collect_vec()
+                                        ),
+                                    }
+                                    .fail()?;
+                                }
                                Ok(row.into())
                            })
                            .collect::<Result<Vec<_>, Error>>()?;
--- a/src/flow/src/compute/render.rs
+++ b/src/flow/src/compute/render.rs
@@ -61,7 +61,7 @@ pub struct Context<'referred, 'df> {
    pub err_collector: ErrCollector,
 }

-impl<'referred, 'df> Drop for Context<'referred, 'df> {
+impl Drop for Context<'_, '_> {
    fn drop(&mut self) {
        for bundle in std::mem::take(&mut self.input_collection)
            .into_values()
@@ -92,7 +92,7 @@ impl<'referred, 'df> Drop for Context<'referred, 'df> {
    }
 }

-impl<'referred, 'df> Context<'referred, 'df> {
+impl Context<'_, '_> {
    pub fn insert_global(&mut self, id: GlobalId, collection: CollectionBundle) {
        self.input_collection.insert(id, collection);
    }
@@ -120,7 +120,7 @@ impl<'referred, 'df> Context<'referred, 'df> {
    }
 }

-impl<'referred, 'df> Context<'referred, 'df> {
+impl Context<'_, '_> {
    /// Like `render_plan` but in Batch Mode
    pub fn render_plan_batch(&mut self, plan: TypedPlan) -> Result<CollectionBundle<Batch>, Error> {
        match plan.plan {
--- a/src/flow/src/compute/render/map.rs
+++ b/src/flow/src/compute/render/map.rs
@@ -28,7 +28,7 @@ use crate::plan::TypedPlan;
 use crate::repr::{self, DiffRow, KeyValDiffRow, Row};
 use crate::utils::ArrangeHandler;

-impl<'referred, 'df> Context<'referred, 'df> {
+impl Context<'_, '_> {
    /// Like `render_mfp` but in batch mode
    pub fn render_mfp_batch(
        &mut self,
--- a/src/flow/src/compute/render/reduce.rs
+++ b/src/flow/src/compute/render/reduce.rs
@@ -34,7 +34,7 @@ use crate::plan::{AccumulablePlan, AggrWithIndex, KeyValPlan, ReducePlan, TypedP
 use crate::repr::{self, DiffRow, KeyValDiffRow, RelationType, Row};
 use crate::utils::{ArrangeHandler, ArrangeReader, ArrangeWriter, KeyExpiryManager};

-impl<'referred, 'df> Context<'referred, 'df> {
+impl Context<'_, '_> {
    const REDUCE_BATCH: &'static str = "reduce_batch";
    /// Like `render_reduce`, but for batch mode, and only barebone implementation
    /// no support for distinct aggregation for now
@@ -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))
--- a/src/flow/src/compute/render/src_sink.rs
+++ b/src/flow/src/compute/render/src_sink.rs
@@ -31,7 +31,7 @@ use crate::expr::{Batch, EvalError};
 use crate::repr::{DiffRow, Row, BROADCAST_CAP};

 #[allow(clippy::mutable_key_type)]
-impl<'referred, 'df> Context<'referred, 'df> {
+impl Context<'_, '_> {
    /// simply send the batch to downstream, without fancy features like buffering
    pub fn render_source_batch(
        &mut self,
--- a/src/flow/src/df_optimizer.rs
+++ b/src/flow/src/df_optimizer.rs
@@ -273,7 +273,7 @@ impl<'a> ExpandAvgRewriter<'a> {
    }
 }

-impl<'a> TreeNodeRewriter for ExpandAvgRewriter<'a> {
+impl TreeNodeRewriter for ExpandAvgRewriter<'_> {
    type Node = Expr;

    fn f_up(&mut self, expr: Expr) -> Result<Transformed<Expr>, DataFusionError> {
--- a/src/flow/src/expr/df_func.rs
+++ b/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,
--- a/src/flow/src/expr/func.rs
+++ b/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))
+        );
+    }
 }
--- a/src/flow/src/repr.rs
+++ b/src/flow/src/repr.rs
@@ -61,7 +61,7 @@ pub const BATCH_SIZE: usize = 32 * 16384;
 /// Convert a value that is or can be converted to Datetime to internal timestamp
 ///
 /// support types are: `Date`, `DateTime`, `TimeStamp`, `i64`
-pub fn value_to_internal_ts(value: Value) -> Result<Timestamp, EvalError> {
+pub fn value_to_internal_ts(value: Value) -> Result<i64, EvalError> {
    let is_supported_time_type = |arg: &Value| {
        let ty = arg.data_type();
        matches!(
@@ -76,14 +76,14 @@ pub fn value_to_internal_ts(value: Value) -> Result<Timestamp, EvalError> {
        Value::Int64(ts) => Ok(ts),
        arg if is_supported_time_type(&arg) => {
            let arg_ty = arg.data_type();
-            let res = cast(arg, &ConcreteDataType::datetime_datatype()).context({
+            let res = cast(arg, &ConcreteDataType::timestamp_millisecond_datatype()).context({
                CastValueSnafu {
                    from: arg_ty,
-                    to: ConcreteDataType::datetime_datatype(),
+                    to: ConcreteDataType::timestamp_millisecond_datatype(),
                }
            })?;
-            if let Value::DateTime(ts) = res {
-                Ok(ts.val())
+            if let Value::Timestamp(ts) = res {
+                Ok(ts.value())
            } else {
                unreachable!()
            }
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Yingwen	1008af5324	feat!: Divide flush and compaction job pool (#4871 ) * feat: divide flush/compact job pool * feat!: divide bg jobs config * docs: update config examples * test: fix tests	2024-10-25 23:36:16 +00:00
discord9	2485f66077	chore: graceful exit on bind fail (#4882 )	2024-10-25 09:29:39 +00:00
Weny Xu	4f3afb13b6	fix: fix broken import (#4880 )	2024-10-25 07:09:51 +00:00
shuiyisong	32a0023010	chore: add schema urls to otlp logs (#4876 ) * chore: add schema urls to otlp logs table * chore: update meter-macros version to remove anymap warning * chore: change span id and trace id to field	2024-10-25 03:45:24 +00:00
Kaifeng Zheng	4e9c251041	feat: add json_path_match udf (#4864 ) * add json_path_match udf * sql tests for json_path_match * fix clippy & comment * fix null value behavior * added null tests * adjust function's behavior on nulls * update test cases * fix null check of json	2024-10-25 03:13:34 +00:00
Lei, HUANG	e328c7067c	chore: udapte Rust toolchain to 2024-10-19 (#4857 ) * update rust toolchain * change toolchain to 2024-10-17 * fix: clippy * fix: ut * bump shadow-rs * fix: use nightly-2024-10-19 * fix: clippy * chore/udapte-toolchain-2024-10-17: Update DEV_BUILDER_IMAGE_TAG to 2024-10-19-a5c00e85-20241024184445 in Makefile	2024-10-25 00:23:32 +00:00
Weny Xu	8b307e4548	feat: introduce the `PluginOptions` (#4835 ) * feat: introduce the `PluginOptions` * chore: apply suggestions from CR	2024-10-24 12:02:10 +00:00
discord9	ff38abde2e	chore: better column schema check for flow (#4855 ) * chore: better column schema check for flow * chore: better msg * tests: clean up after tests * chore: better msg * chore: per review * tests: sqlness	2024-10-24 09:43:32 +00:00
jeremyhi	aa9a265984	chore: make pusher log easy to understand (#4841 ) * chore: make pusher log easy to understand * Update src/meta-srv/src/service/heartbeat.rs Co-authored-by: Yingwen <realevenyag@gmail.com> * Update src/meta-srv/src/service/heartbeat.rs Co-authored-by: Yingwen <realevenyag@gmail.com> * chore: by comment --------- Co-authored-by: Yingwen <realevenyag@gmail.com>	2024-10-24 07:44:16 +00:00
pa	9d3ee6384a	feat: Limit CPU in runtime (#3685 ) (#4782 ) feat: add throttle runtime (#3685)	2024-10-24 07:30:24 +00:00
localhost	fcde0a4874	feat: Add functionality to the Opentelemetry write interface to extract fields from attr to top-level data. (#4859 ) * chore: add otlp select * chore: change otlp select * chore: remove json path * chore: format toml * chore: change opentelemetry extract keys header name * chore: add some doc and remove useless code and lib * chore: make clippy happy * chore: fix by pr comment * chore: fix by pr comment * chore: opentelemetry logs select key change some type default semantic type	2024-10-24 05:55:57 +00:00
Weny Xu	5d42e63ab0	fix!: replace timeout_millis and connect_timeout_millis with Duration in DatanodeClientOptions (#4867 ) * fix: correct options struct * fix: fix unit test	2024-10-23 08:20:34 +00:00
discord9	0c01532a37	feat: Sort within each PartitionRange (#4847 ) * feat: PartSort * chore: rm unused * chore: typo * chore: mem pool df * chore: add location to arrow error * refactor: test_util * refactor: per review * chore: rm unused * chore: more cases * chore: test&buffer clear * fix: remove fetch * chore: fmt * chore: per review * chore: rm unused	2024-10-23 07:01:55 +00:00
ZonaHe	6d503b047a	feat: update dashboard to v0.6.0 (#4861 ) Co-authored-by: ZonaHex <ZonaHex@users.noreply.github.com>	2024-10-22 02:34:09 +00:00
Yingwen	5d28f7a912	feat: yields empty batch after reading a range (#4845 ) * feat: add empty batch to end of range stream * feat: add batch validation * fix: validate batch order * fix: not yield empty batch in compaction * fix: empty record batch * feat: add a flag to enable empty batch	2024-10-21 13:52:47 +00:00
Lei, HUANG	a50eea76a6	chore: bump greptime-meter (#4858 ) chore/bump-greptime-meter: Add meter-core package and update meter-core dependency across various packages to new git revision.	2024-10-21 08:18:30 +00:00
Yingwen	2ee1ce2ba1	docs: change cpu/mem panel to time-series (#4844 ) * docs: change cpu/mem panel to time-series * docs: update version	2024-10-18 08:42:01 +00:00
Weny Xu	c02b5dae93	chore: bump version to 0.9.5 (#4853 )	2024-10-18 08:07:13 +00:00
Weny Xu	081c6d9e74	fix: flush metric metadata region (#4852 ) * fix: flush metric metadata region * chore: apply suggestions from CR	2024-10-18 07:21:35 +00:00
Weny Xu	ca6e02980e	fix: overwrite entry_id if entry id is less than start_offset (#4842 ) * fix: overwrite entry_id if entry id is less than start_offset * feat: add `overwrite_entry_start_id` to options * chore: update config.md	2024-10-18 06:31:02 +00:00
Weny Xu	74bdba4613	fix: fix metadata forward compatibility issue (#4846 )	2024-10-18 06:26:41 +00:00
Weny Xu	2e0e82ddc8	chore: update greptime-proto to `b4d3011` (#4850 )	2024-10-18 04:10:22 +00:00
Yingwen	e0c4157ad8	feat: Seq scanner scans data by time range (#4809 ) * feat: seq scan by partition * feat: part metrics * chore: remove unused codes * chore: fmt stream * feat: build ranges returns smallvec * feat: move scan mem/file ranges to util and reuse * feat: log metrics * chore: correct some metrics * feat: get explain info from ranges * test: group test and remove unused codes * chore: fix clippy * feat: change PartitionRange end to exclusive * test: add tests	2024-10-17 11:05:12 +00:00
discord9	613e07afb4	feat: window sort physical plan (#4814 ) * WIP * feat: range split& tests * WIP: split range * add sort exprs * chore: typo * WIP * feat: find successive runs * WIP * READY FOR REVIEW PART ONE: more tests * refactor: break into smaller functions * feat: precompute working range(need testing) * tests: on working range * tests: on working range * feat: support rev working range * feat(to be tested): core logic of merge sort * fix: poll results * fix: find_slice_from_range&test * chore: remove some unused util func&fields * chore: typos * chore: impl exec plan for WindowedSortExec * test(WIP): window sort stream * test: window sort stream * chore: remove unused * fix: fetch * fix: WIP intersection remaining * test: fix and test! * chore: remove outdated comments * chore: rename test * chore: remove dbg line * chore: sorted runs * feat: handling unexpected data * chore: unused * chore: remove a print in test * chore: per review * docs: wrong comment * chore: more test cases	2024-10-16 11:50:25 +00:00
Weny Xu	0ce93f0b88	chore: add more metrics for region migration (#4838 )	2024-10-16 09:36:57 +00:00
Ning Sun	c231eee7c1	fix: respect feature flags for geo function (#4836 )	2024-10-16 07:46:31 +00:00
Yiran	176f2df5b3	fix: dead links (#4837 )	2024-10-16 07:43:14 +00:00
localhost	4622412dfe	feat: add API to write OpenTelemetry logs to GreptimeDB (#4755 ) * chore: otlp logs api * feat: add API to write OpenTelemetry logs to GreptimeDB * chore: fix test data schema error * chore: modify the underlying data structure of the pipeline value map type from hashmap to btremap to keep key order * chore: fix by pr comment * chore: resolve conflicts and add some test * chore: remove useless error * chore: change otlp header name * chore: fmt code * chore: fix integration test for otlp log write api * chore: fix by pr comment * chore: set otlp body with fulltext default	2024-10-16 04:36:08 +00:00
jeremyhi	59ec90299b	refactor: metasrv cannot be cloned (#4834 ) * refactor: metasrv cannot be cloned * chore: remove MetasrvInstance's clone	2024-10-15 11:36:48 +00:00
discord9	16b8cdc3d5	chore: bump version v0.9.4 (#4833 )	2024-10-15 10:48:03 +00:00
Weny Xu	3197b8b535	feat: introduce default customizers (#4831 ) * feat: introduce `DefaultHeartbeatHandlerGroupBuilderCustomizer` and `DefaultLeadershipChangeNotifierCustomizer` * chore: code styling	2024-10-15 09:48:13 +00:00
zyy17	972c2441af	chore: bump promql-parser to v0.4.1 and use `to_string()` for EvalStmt (#4832 ) chore: bump promql-parser to v0.4.1 and use to_string() for EvalStmt	2024-10-15 08:50:37 +00:00
Ning Sun	bb8b54b5d3	feat: add some s2 geo functions (#4823 ) * feat: add first batch of s2 functions * refactor: update reusable code from main * test: add sqlness tests for s2 * feat: add tostring function for s2 * Update src/common/function/src/scalars/geo/s2.rs Co-authored-by: Lei, HUANG <6406592+v0y4g3r@users.noreply.github.com> * Apply suggestions from code review * one more change Signed-off-by: Ruihang Xia <waynestxia@gmail.com> --------- Signed-off-by: Ruihang Xia <waynestxia@gmail.com> Co-authored-by: Ruihang Xia <waynestxia@gmail.com> Co-authored-by: Lei, HUANG <6406592+v0y4g3r@users.noreply.github.com>	2024-10-15 06:47:29 +00:00
Weny Xu	b5233e500b	feat: defer `HeartbeatHandlerGroup` construction and enhance `LeadershipChangeNotifier` (#4826 ) * feat: enhance `HeartbeatHandlerGroup` * chore: apply suggestions from CR * chore: minor refactoring * chore: code styling * chore: apply suggestions from CR	2024-10-15 03:35:31 +00:00
Ruihang Xia	b61a388d04	refactor: replace info logs with debug logs in region server (#4829 ) * refactor: replace info logs with debug logs in region server Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * fix: update error handling for closing and opening nonexistent regions Signed-off-by: Ruihang Xia <waynestxia@gmail.com> --------- Signed-off-by: Ruihang Xia <waynestxia@gmail.com>	2024-10-14 12:46:07 +00:00
Ruihang Xia	06e565d25a	feat: cache logical region's metadata (#4827 ) * feat: cache logical region's metadata Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * feat: implement logical region locking for metadata operations Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * fix: correct typo in comment for MetadataRegion struct Signed-off-by: Ruihang Xia <waynestxia@gmail.com> --------- Signed-off-by: Ruihang Xia <waynestxia@gmail.com>	2024-10-14 08:44:13 +00:00
Yingwen	3b2ce31a19	feat: enable prof features by default (#4815 ) * feat: enable prof by default * docs: don't need to build with features * feat: add common-pprof as optional dep for pprof feature * build: remove optional * feat: use dump_text	2024-10-14 03:32:47 +00:00
Ruihang Xia	a889ea88ca	fix: case sensitive for __field__ matcher (#4822 ) Signed-off-by: Ruihang Xia <waynestxia@gmail.com>	2024-10-14 03:18:59 +00:00
Yingwen	2f2b4b306c	feat!: implement interval type by multiple structs (#4772 ) * define structs and methods Signed-off-by: Ruihang Xia <waynestxia@gmail.com> * feat: re-implement interval types in time crate * feat: use new * feat: interval value * feat: query crate interval * feat: pg and mysql interval * chore: remove unused imports * chore: remove commented codes * feat: make flow compile but may not work * feat: flow datetime * test: fix some tests * test: fix some flow tests(WIP) * chore: some fix test&docs * fix: change interval order * chore: remove unused codes * chore: fix cilppy * chore: now signature change * chore: remove todo * feat: update error message --------- Signed-off-by: Ruihang Xia <waynestxia@gmail.com> Co-authored-by: Ruihang Xia <waynestxia@gmail.com> Co-authored-by: discord9 <discord9@163.com>	2024-10-14 03:09:03 +00:00
jeremyhi	856c0280f5	feat: remove the distributed lock (#4825 ) * feat: remove the distributed lock as we do not need it any more * chore: delete todo comment * chore: remove unused error	2024-10-12 09:04:22 +00:00