clippy

Signed-off-by: discord9 <discord9@163.com>
c
2026-01-06 05:12:54 +00:00 · 2025-12-22 21:07:24 +08:00 · 2025-12-22 21:05:23 +08:00 · 2025-12-22 21:05:13 +08:00 · 2025-12-22 18:39:14 +08:00 · 2025-12-22 18:35:19 +08:00
91 changed files with 5441 additions and 2228 deletions
--- a/.github/actions/setup-greptimedb-cluster/action.yml
+++ b/.github/actions/setup-greptimedb-cluster/action.yml
@@ -51,7 +51,7 @@ runs:
    run: |
      helm upgrade \
        --install my-greptimedb \
-        --set meta.backendStorage.etcd.endpoints=${{ inputs.etcd-endpoints }} \
+        --set 'meta.backendStorage.etcd.endpoints[0]=${{ inputs.etcd-endpoints }}' \
        --set meta.enableRegionFailover=${{ inputs.enable-region-failover }} \
        --set image.registry=${{ inputs.image-registry }} \
        --set image.repository=${{ inputs.image-repository }}  \
--- a/.github/scripts/create-version.sh
+++ b/.github/scripts/create-version.sh
@@ -49,6 +49,17 @@ function create_version() {
      echo "GITHUB_REF_NAME is empty in push event" >&2
      exit 1
    fi
+    
+    # For tag releases, ensure GITHUB_REF_NAME matches the version in Cargo.toml
+    CARGO_VERSION=$(grep '^version = ' Cargo.toml | cut -d '"' -f 2 | head -n 1)
+    EXPECTED_REF_NAME="v${CARGO_VERSION}"
+    
+    if [ "$GITHUB_REF_NAME" != "$EXPECTED_REF_NAME" ]; then
+      echo "Error: GITHUB_REF_NAME '$GITHUB_REF_NAME' does not match Cargo.toml version 'v${CARGO_VERSION}'" >&2
+      echo "Expected tag name: '$EXPECTED_REF_NAME'" >&2
+      exit 1
+    fi
+    
    echo "$GITHUB_REF_NAME"
  elif [ "$GITHUB_EVENT_NAME" = workflow_dispatch ]; then
    echo "$NEXT_RELEASE_VERSION-$(git rev-parse --short HEAD)-$(date "+%Y%m%d-%s")"
--- a/.github/scripts/deploy-greptimedb.sh
+++ b/.github/scripts/deploy-greptimedb.sh
@@ -81,7 +81,7 @@ function deploy_greptimedb_cluster() {
    --create-namespace \
    --set image.tag="$GREPTIMEDB_IMAGE_TAG" \
    --set initializer.tag="$GREPTIMEDB_INITIALIZER_IMAGE_TAG" \
-    --set meta.backendStorage.etcd.endpoints="etcd.$install_namespace:2379" \
+    --set "meta.backendStorage.etcd.endpoints[0]=etcd.$install_namespace.svc.cluster.local:2379" \
    --set meta.backendStorage.etcd.storeKeyPrefix="$cluster_name" \
    -n "$install_namespace"

@@ -119,7 +119,7 @@ function deploy_greptimedb_cluster_with_s3_storage() {
    --create-namespace \
    --set image.tag="$GREPTIMEDB_IMAGE_TAG" \
    --set initializer.tag="$GREPTIMEDB_INITIALIZER_IMAGE_TAG" \
-    --set meta.backendStorage.etcd.endpoints="etcd.$install_namespace:2379" \
+    --set "meta.backendStorage.etcd.endpoints[0]=etcd.$install_namespace.svc.cluster.local:2379" \
    --set meta.backendStorage.etcd.storeKeyPrefix="$cluster_name" \
    --set objectStorage.s3.bucket="$AWS_CI_TEST_BUCKET" \
    --set objectStorage.s3.region="$AWS_REGION" \
--- a/.github/workflows/check-git-deps.yml
+++ b/.github/workflows/check-git-deps.yml
@@ -0,0 +1,154 @@
+name: Check Git Dependencies on Main Branch
+
+on:
+    pull_request:
+        branches: [main]
+        paths:
+            - 'Cargo.toml'
+    push:
+        branches: [main]
+        paths:
+            - 'Cargo.toml'
+
+jobs:
+    check-git-deps:
+        runs-on: ubuntu-latest
+
+        steps:
+            - name: Checkout repository
+              uses: actions/checkout@v6
+
+            - name: Check git dependencies
+              env:
+                  WHITELIST_DEPS: "greptime-proto,meter-core,meter-macros"
+              run: |
+                  #!/bin/bash
+                  set -e
+
+                  echo "Checking whitelisted git dependencies..."
+
+                  # Function to check if a commit is on main branch
+                  check_commit_on_main() {
+                      local repo_url="$1"
+                      local commit="$2"
+                      local repo_name=$(basename "$repo_url" .git)
+
+                      echo "Checking $repo_name"
+                      echo "Repo: $repo_url"
+                      echo "Commit: $commit"
+
+                      # Create a temporary directory for cloning
+                      local temp_dir=$(mktemp -d)
+
+                      # Clone the repository
+                      if git clone "$repo_url" "$temp_dir" 2>/dev/null; then
+                          cd "$temp_dir"
+
+                          # Try to determine the main branch name
+                          local main_branch="main"
+                          if ! git rev-parse --verify origin/main >/dev/null 2>&1; then
+                              if git rev-parse --verify origin/master >/dev/null 2>&1; then
+                                  main_branch="master"
+                              else
+                                  # Try to get the default branch
+                                  main_branch=$(git symbolic-ref refs/remotes/origin/HEAD | sed 's@^refs/remotes/origin/@@')
+                              fi
+                          fi
+
+                          echo "Main branch: $main_branch"
+
+                          # Check if commit exists
+                          if git cat-file -e "$commit" 2>/dev/null; then
+                              # Check if commit is on main branch
+                              if git merge-base --is-ancestor "$commit" "origin/$main_branch" 2>/dev/null; then
+                                  echo "PASS: Commit $commit is on $main_branch branch"
+                                  cd - >/dev/null
+                                  rm -rf "$temp_dir"
+                                  return 0
+                              else
+                                  echo "FAIL: Commit $commit is NOT on $main_branch branch"
+
+                                  # Try to find which branch contains this commit
+                                  local branch_name=$(git branch -r --contains "$commit" 2>/dev/null | head -1 | sed 's/^[[:space:]]*origin\///' | sed 's/[[:space:]]*$//')
+                                  if [[ -n "$branch_name" ]]; then
+                                      echo "Found on branch: $branch_name"
+                                  fi
+                                  cd - >/dev/null
+                                  rm -rf "$temp_dir"
+                                  return 1
+                              fi
+                          else
+                              echo "FAIL: Commit $commit not found in repository"
+                              cd - >/dev/null
+                              rm -rf "$temp_dir"
+                              return 1
+                          fi
+                      else
+                          echo "FAIL: Failed to clone $repo_url"
+                          rm -rf "$temp_dir"
+                          return 1
+                      fi
+                  }
+
+                  # Extract whitelisted git dependencies from Cargo.toml
+                  echo "Extracting git dependencies from Cargo.toml..."
+
+                  # Create temporary array to store dependencies
+                  declare -a deps=()
+
+                  # Build awk pattern from whitelist
+                  IFS=',' read -ra WHITELIST <<< "$WHITELIST_DEPS"
+                  awk_pattern=""
+                  for dep in "${WHITELIST[@]}"; do
+                      if [[ -n "$awk_pattern" ]]; then
+                          awk_pattern="$awk_pattern|"
+                      fi
+                      awk_pattern="$awk_pattern$dep"
+                  done
+
+                  # Extract whitelisted dependencies
+                  while IFS= read -r line; do
+                      if [[ -n "$line" ]]; then
+                          deps+=("$line")
+                      fi
+                  done < <(awk -v pattern="$awk_pattern" '
+                  $0 ~ pattern ".*git = \"https:/" {
+                      match($0, /git = "([^"]+)"/, arr)
+                      git_url = arr[1]
+                      if (match($0, /rev = "([^"]+)"/, rev_arr)) {
+                          rev = rev_arr[1]
+                          print git_url " " rev
+                      } else {
+                          # Check next line for rev
+                          getline
+                          if (match($0, /rev = "([^"]+)"/, rev_arr)) {
+                              rev = rev_arr[1]
+                              print git_url " " rev
+                          }
+                      }
+                  }
+                  ' Cargo.toml)
+
+                  echo "Found ${#deps[@]} dependencies to check:"
+                  for dep in "${deps[@]}"; do
+                      echo "  $dep"
+                  done
+
+                  failed=0
+
+                  for dep in "${deps[@]}"; do
+                      read -r repo_url commit <<< "$dep"
+                      if ! check_commit_on_main "$repo_url" "$commit"; then
+                          failed=1
+                      fi
+                  done
+
+                  echo "Check completed."
+
+                  if [[ $failed -eq 1 ]]; then
+                      echo "ERROR: Some git dependencies are not on their main branches!"
+                      echo "Please update the commits to point to main branch commits."
+                      exit 1
+                  else
+                      echo "SUCCESS: All git dependencies are on their main branches!"
+                  fi
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -212,7 +212,7 @@ checksum = "d301b3b94cb4b2f23d7917810addbbaff90738e0ca2be692bd027e70d7e0330c"

 [[package]]
 name = "api"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arrow-schema",
 "common-base",
@@ -733,7 +733,7 @@ dependencies = [

 [[package]]
 name = "auth"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -1383,7 +1383,7 @@ dependencies = [

 [[package]]
 name = "cache"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "catalog",
 "common-error",
@@ -1418,7 +1418,7 @@ checksum = "37b2a672a2cb129a2e41c10b1224bb368f9f37a2b16b612598138befd7b37eb5"

 [[package]]
 name = "catalog"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow",
@@ -1763,7 +1763,7 @@ checksum = "b94f61472cee1439c0b966b47e3aca9ae07e45d070759512cd390ea2bebc6675"

 [[package]]
 name = "cli"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-stream",
 "async-trait",
@@ -1786,6 +1786,7 @@ dependencies = [
 "common-recordbatch",
 "common-runtime",
 "common-telemetry",
+ "common-test-util",
 "common-time",
 "common-version",
 "common-wal",
@@ -1816,7 +1817,7 @@ dependencies = [

 [[package]]
 name = "client"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arc-swap",
@@ -1849,7 +1850,7 @@ dependencies = [
 "snafu 0.8.6",
 "store-api",
 "substrait 0.37.3",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "tokio",
 "tokio-stream",
 "tonic 0.13.1",
@@ -1889,7 +1890,7 @@ dependencies = [

 [[package]]
 name = "cmd"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "auth",
@@ -1977,6 +1978,17 @@ dependencies = [
 "unicode-width 0.2.1",
 ]

+[[package]]
+name = "codespan-reporting"
+version = "0.13.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "af491d569909a7e4dee0ad7db7f5341fef5c614d5b8ec8cf765732aba3cff681"
+dependencies = [
+ "serde",
+ "termcolor",
+ "unicode-width 0.2.1",
+]
+
 [[package]]
 name = "colorchoice"
 version = "1.0.4"
@@ -2012,7 +2024,7 @@ checksum = "55b672471b4e9f9e95499ea597ff64941a309b2cdbffcc46f2cc5e2d971fd335"

 [[package]]
 name = "common-base"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "anymap2",
 "async-trait",
@@ -2036,14 +2048,14 @@ dependencies = [

 [[package]]
 name = "common-catalog"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "const_format",
 ]

 [[package]]
 name = "common-config"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-base",
 "common-error",
@@ -2068,7 +2080,7 @@ dependencies = [

 [[package]]
 name = "common-datasource"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arrow",
 "arrow-schema",
@@ -2103,7 +2115,7 @@ dependencies = [

 [[package]]
 name = "common-decimal"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "bigdecimal 0.4.8",
 "common-error",
@@ -2116,7 +2128,7 @@ dependencies = [

 [[package]]
 name = "common-error"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-macro",
 "http 1.3.1",
@@ -2127,7 +2139,7 @@ dependencies = [

 [[package]]
 name = "common-event-recorder"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -2149,7 +2161,7 @@ dependencies = [

 [[package]]
 name = "common-frontend"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -2171,7 +2183,7 @@ dependencies = [

 [[package]]
 name = "common-function"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -2231,7 +2243,7 @@ dependencies = [

 [[package]]
 name = "common-greptimedb-telemetry"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "common-runtime",
@@ -2248,7 +2260,7 @@ dependencies = [

 [[package]]
 name = "common-grpc"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow-flight",
@@ -2283,7 +2295,7 @@ dependencies = [

 [[package]]
 name = "common-grpc-expr"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "common-base",
@@ -2303,7 +2315,7 @@ dependencies = [

 [[package]]
 name = "common-macro"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "greptime-proto",
 "once_cell",
@@ -2314,7 +2326,7 @@ dependencies = [

 [[package]]
 name = "common-mem-prof"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "anyhow",
 "common-error",
@@ -2330,7 +2342,7 @@ dependencies = [

 [[package]]
 name = "common-memory-manager"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-error",
 "common-macro",
@@ -2343,7 +2355,7 @@ dependencies = [

 [[package]]
 name = "common-meta"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "anymap2",
 "api",
@@ -2415,7 +2427,7 @@ dependencies = [

 [[package]]
 name = "common-options"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-grpc",
 "humantime-serde",
@@ -2424,11 +2436,11 @@ dependencies = [

 [[package]]
 name = "common-plugins"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"

 [[package]]
 name = "common-pprof"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-error",
 "common-macro",
@@ -2440,7 +2452,7 @@ dependencies = [

 [[package]]
 name = "common-procedure"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-stream",
@@ -2469,7 +2481,7 @@ dependencies = [

 [[package]]
 name = "common-procedure-test"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "common-procedure",
@@ -2479,7 +2491,7 @@ dependencies = [

 [[package]]
 name = "common-query"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -2505,7 +2517,7 @@ dependencies = [

 [[package]]
 name = "common-recordbatch"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arc-swap",
 "common-base",
@@ -2529,7 +2541,7 @@ dependencies = [

 [[package]]
 name = "common-runtime"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "clap 4.5.40",
@@ -2558,7 +2570,7 @@ dependencies = [

 [[package]]
 name = "common-session"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "serde",
 "strum 0.27.1",
@@ -2566,7 +2578,7 @@ dependencies = [

 [[package]]
 name = "common-sql"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-base",
 "common-decimal",
@@ -2584,7 +2596,7 @@ dependencies = [

 [[package]]
 name = "common-stat"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-base",
 "common-runtime",
@@ -2599,7 +2611,7 @@ dependencies = [

 [[package]]
 name = "common-telemetry"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "backtrace",
 "common-base",
@@ -2628,7 +2640,7 @@ dependencies = [

 [[package]]
 name = "common-test-util"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "client",
 "common-grpc",
@@ -2641,7 +2653,7 @@ dependencies = [

 [[package]]
 name = "common-time"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arrow",
 "chrono",
@@ -2659,7 +2671,7 @@ dependencies = [

 [[package]]
 name = "common-version"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "build-data",
 "cargo-manifest",
@@ -2670,7 +2682,7 @@ dependencies = [

 [[package]]
 name = "common-wal"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-base",
 "common-error",
@@ -2693,7 +2705,7 @@ dependencies = [

 [[package]]
 name = "common-workload"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-telemetry",
 "serde",
@@ -3169,6 +3181,68 @@ dependencies = [
 "cipher",
 ]

+[[package]]
+name = "cxx"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a7620f6cfc4dcca21f2b085b7a890e16c60fd66f560cd69ee60594908dc72ab1"
+dependencies = [
+ "cc",
+ "cxx-build",
+ "cxxbridge-cmd",
+ "cxxbridge-flags",
+ "cxxbridge-macro",
+ "foldhash 0.2.0",
+ "link-cplusplus",
+]
+
+[[package]]
+name = "cxx-build"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7a9bc1a22964ff6a355fbec24cf68266a0ed28f8b84c0864c386474ea3d0e479"
+dependencies = [
+ "cc",
+ "codespan-reporting 0.13.1",
+ "indexmap 2.11.4",
+ "proc-macro2",
+ "quote",
+ "scratch",
+ "syn 2.0.106",
+]
+
+[[package]]
+name = "cxxbridge-cmd"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b1f29a879d35f7906e3c9b77d7a1005a6a0787d330c09dfe4ffb5f617728cb44"
+dependencies = [
+ "clap 4.5.40",
+ "codespan-reporting 0.13.1",
+ "indexmap 2.11.4",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.106",
+]
+
+[[package]]
+name = "cxxbridge-flags"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d67109015f93f683e364085aa6489a5b2118b4a40058482101d699936a7836d6"
+
+[[package]]
+name = "cxxbridge-macro"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d187e019e7b05a1f3e69a8396b70800ee867aa9fc2ab972761173ccee03742df"
+dependencies = [
+ "indexmap 2.11.4",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.106",
+]
+
 [[package]]
 name = "darling"
 version = "0.14.4"
@@ -3939,7 +4013,7 @@ dependencies = [

 [[package]]
 name = "datanode"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow-flight",
@@ -4003,7 +4077,7 @@ dependencies = [

 [[package]]
 name = "datatypes"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -4560,8 +4634,9 @@ dependencies = [

 [[package]]
 name = "etcd-client"
-version = "0.15.0"
-source = "git+https://github.com/GreptimeTeam/etcd-client?rev=f62df834f0cffda355eba96691fe1a9a332b75a7#f62df834f0cffda355eba96691fe1a9a332b75a7"
+version = "0.16.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "88365f1a5671eb2f7fc240adb216786bc6494b38ce15f1d26ad6eaa303d5e822"
 dependencies = [
 "http 1.3.1",
 "prost 0.13.5",
@@ -4677,7 +4752,7 @@ checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"

 [[package]]
 name = "file-engine"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -4809,7 +4884,7 @@ checksum = "8bf7cc16383c4b8d58b9905a8509f02926ce3058053c056376248d958c9df1e8"

 [[package]]
 name = "flow"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow",
@@ -4878,7 +4953,7 @@ dependencies = [
 "sql",
 "store-api",
 "strum 0.27.1",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "table",
 "tokio",
 "tonic 0.13.1",
@@ -4916,6 +4991,12 @@ version = "0.1.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d9c4f5dac5e15c24eb999c26181a6ca40b39fe946cbe4c263c7209467bc83af2"

+[[package]]
+name = "foldhash"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "77ce24cb58228fbb8aa041425bb1050850ac19177686ea6e0f41a70416f56fdb"
+
 [[package]]
 name = "form_urlencoded"
 version = "1.2.2"
@@ -4933,7 +5014,7 @@ checksum = "28dd6caf6059519a65843af8fe2a3ae298b14b80179855aeb4adc2c1934ee619"

 [[package]]
 name = "frontend"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arc-swap",
@@ -5380,7 +5461,7 @@ dependencies = [
 [[package]]
 name = "greptime-proto"
 version = "0.1.0"
-source = "git+https://github.com/GreptimeTeam/greptime-proto.git?rev=0423fa30203187c75e2937a668df1da699c8b96c#0423fa30203187c75e2937a668df1da699c8b96c"
+source = "git+https://github.com/GreptimeTeam/greptime-proto.git?rev=173efe5ec62722089db7c531c0b0d470a072b915#173efe5ec62722089db7c531c0b0d470a072b915"
 dependencies = [
 "prost 0.13.5",
 "prost-types 0.13.5",
@@ -5516,7 +5597,7 @@ checksum = "5971ac85611da7067dbfcabef3c70ebb5606018acd9e2a3903a0da507521e0d5"
 dependencies = [
 "allocator-api2",
 "equivalent",
- "foldhash",
+ "foldhash 0.1.5",
 ]

 [[package]]
@@ -6148,7 +6229,7 @@ dependencies = [

 [[package]]
 name = "index"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "asynchronous-codec",
@@ -6161,6 +6242,7 @@ dependencies = [
 "common-telemetry",
 "common-test-util",
 "criterion 0.4.0",
+ "datatypes",
 "fastbloom",
 "fst",
 "futures",
@@ -6169,6 +6251,7 @@ dependencies = [
 "jieba-rs",
 "lazy_static",
 "mockall",
+ "nalgebra",
 "pin-project",
 "prost 0.13.5",
 "puffin",
@@ -6186,6 +6269,7 @@ dependencies = [
 "tempfile",
 "tokio",
 "tokio-util",
+ "usearch",
 "uuid",
 ]

@@ -7017,6 +7101,15 @@ dependencies = [
 "vcpkg",
 ]

+[[package]]
+name = "link-cplusplus"
+version = "1.0.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f78c730aaa7d0b9336a299029ea49f9ee53b0ed06e9202e8cb7db9bae7b8c82"
+dependencies = [
+ "cc",
+]
+
 [[package]]
 name = "linked-hash-map"
 version = "0.5.6"
@@ -7077,7 +7170,7 @@ checksum = "13dc2df351e3202783a1fe0d44375f7295ffb4049267b0f3018346dc122a1d94"

 [[package]]
 name = "log-query"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "chrono",
 "common-error",
@@ -7089,7 +7182,7 @@ dependencies = [

 [[package]]
 name = "log-store"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-stream",
 "async-trait",
@@ -7390,7 +7483,7 @@ dependencies = [

 [[package]]
 name = "meta-client"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -7418,7 +7511,7 @@ dependencies = [

 [[package]]
 name = "meta-srv"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -7518,7 +7611,7 @@ dependencies = [

 [[package]]
 name = "metric-engine"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "aquamarine",
@@ -7615,7 +7708,7 @@ dependencies = [

 [[package]]
 name = "mito-codec"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "bytes",
@@ -7640,7 +7733,7 @@ dependencies = [

 [[package]]
 name = "mito2"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "aquamarine",
@@ -8380,7 +8473,7 @@ dependencies = [

 [[package]]
 name = "object-store"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "anyhow",
 "bytes",
@@ -8665,7 +8758,7 @@ dependencies = [

 [[package]]
 name = "operator"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -8725,7 +8818,7 @@ dependencies = [
 "sql",
 "sqlparser",
 "store-api",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "table",
 "tokio",
 "tokio-util",
@@ -9011,7 +9104,7 @@ dependencies = [

 [[package]]
 name = "partition"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -9368,7 +9461,7 @@ checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"

 [[package]]
 name = "pipeline"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -9524,7 +9617,7 @@ dependencies = [

 [[package]]
 name = "plugins"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "auth",
 "catalog",
@@ -9826,7 +9919,7 @@ dependencies = [

 [[package]]
 name = "promql"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "async-trait",
@@ -10109,7 +10202,7 @@ dependencies = [

 [[package]]
 name = "puffin"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-compression 0.4.19",
 "async-trait",
@@ -10151,7 +10244,7 @@ dependencies = [

 [[package]]
 name = "query"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -10218,7 +10311,7 @@ dependencies = [
 "sql",
 "sqlparser",
 "store-api",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "table",
 "tokio",
 "tokio-stream",
@@ -11290,6 +11383,12 @@ version = "1.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"

+[[package]]
+name = "scratch"
+version = "1.0.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d68f2ec51b097e4c1a75b681a8bec621909b5e91f15bb7b840c4f2f7b01148b2"
+
 [[package]]
 name = "scrypt"
 version = "0.11.0"
@@ -11554,7 +11653,7 @@ dependencies = [

 [[package]]
 name = "servers"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -11682,7 +11781,7 @@ dependencies = [

 [[package]]
 name = "session"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -12016,7 +12115,7 @@ dependencies = [

 [[package]]
 name = "sql"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow-buffer",
@@ -12076,7 +12175,7 @@ dependencies = [

 [[package]]
 name = "sqlness-runner"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "clap 4.5.40",
@@ -12353,7 +12452,7 @@ dependencies = [

 [[package]]
 name = "standalone"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "catalog",
@@ -12394,7 +12493,7 @@ checksum = "a2eb9349b6444b326872e140eb1cf5e7c522154d69e7a0ffb0fb81c06b37543f"

 [[package]]
 name = "store-api"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "aquamarine",
@@ -12607,7 +12706,7 @@ dependencies = [

 [[package]]
 name = "substrait"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "bytes",
@@ -12730,7 +12829,7 @@ dependencies = [

 [[package]]
 name = "table"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -12999,7 +13098,7 @@ checksum = "8f50febec83f5ee1df3015341d8bd429f2d1cc62bcba7ea2076759d315084683"

 [[package]]
 name = "tests-fuzz"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arbitrary",
 "async-trait",
@@ -13043,7 +13142,7 @@ dependencies = [

 [[package]]
 name = "tests-integration"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow-flight",
@@ -13118,7 +13217,7 @@ dependencies = [
 "sqlx",
 "standalone",
 "store-api",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "table",
 "tempfile",
 "time",
@@ -14143,6 +14242,16 @@ version = "2.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "daf8dba3b7eb870caf1ddeed7bc9d2a049f3cfdfae7cb521b087cc33ae4c49da"

+[[package]]
+name = "usearch"
+version = "2.21.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2cc9fc5f872a3a4f9081d5f42624d788231b763e1846c829b9968a3755ac884d"
+dependencies = [
+ "cxx",
+ "cxx-build",
+]
+
 [[package]]
 name = "utf8-ranges"
 version = "1.0.5"
@@ -14282,7 +14391,7 @@ dependencies = [
 "ciborium",
 "cidr",
 "clap 4.5.40",
- "codespan-reporting",
+ "codespan-reporting 0.12.0",
 "community-id",
 "convert_case 0.7.1",
 "crc",
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -75,7 +75,7 @@ members = [
 resolver = "2"

 [workspace.package]
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 edition = "2024"
 license = "Apache-2.0"

@@ -143,14 +143,14 @@ derive_builder = "0.20"
 derive_more = { version = "2.1", features = ["full"] }
 dotenv = "0.15"
 either = "1.15"
-etcd-client = { git = "https://github.com/GreptimeTeam/etcd-client", rev = "f62df834f0cffda355eba96691fe1a9a332b75a7", features = [
+etcd-client = { version = "0.16.1", features = [
    "tls",
    "tls-roots",
 ] }
 fst = "0.4.7"
 futures = "0.3"
 futures-util = "0.3"
-greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "0423fa30203187c75e2937a668df1da699c8b96c" }
+greptime-proto = { git = "https://github.com/GreptimeTeam/greptime-proto.git", rev = "173efe5ec62722089db7c531c0b0d470a072b915" }
 hex = "0.4"
 http = "1"
 humantime = "2.1"
--- a/config/config.md
+++ b/config/config.md
@@ -83,6 +83,8 @@
 | `wal.sync_period` | String | `10s` | Duration for fsyncing log files.<br/>**It's only used when the provider is `raft_engine`**. |
 | `wal.recovery_parallelism` | Integer | `2` | Parallelism during WAL recovery. |
 | `wal.broker_endpoints` | Array | -- | The Kafka broker endpoints.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.connect_timeout` | String | `3s` | The connect timeout for kafka client.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.timeout` | String | `3s` | The timeout for kafka client.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.auto_create_topics` | Bool | `true` | Automatically create topics for WAL.<br/>Set to `true` to automatically create topics for WAL.<br/>Otherwise, use topics named `topic_name_prefix_[0..num_topics)` |
 | `wal.num_topics` | Integer | `64` | Number of topics.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.selector_type` | String | `round_robin` | Topic selector type.<br/>Available selector types:<br/>- `round_robin` (default)<br/>**It's only used when the provider is `kafka`**. |
@@ -352,6 +354,7 @@
 | `region_failure_detector_initialization_delay` | String | `10m` | The delay before starting region failure detection.<br/>This delay helps prevent Metasrv from triggering unnecessary region failovers before all Datanodes are fully started.<br/>Especially useful when the cluster is not deployed with GreptimeDB Operator and maintenance mode is not enabled. |
 | `allow_region_failover_on_local_wal` | Bool | `false` | Whether to allow region failover on local WAL.<br/>**This option is not recommended to be set to true, because it may lead to data loss during failover.** |
 | `node_max_idle_time` | String | `24hours` | Max allowed idle time before removing node info from metasrv memory. |
+| `heartbeat_interval` | String | `3s` | Base heartbeat interval for calculating distributed time constants.<br/>The frontend heartbeat interval is 6 times of the base heartbeat interval.<br/>The flownode/datanode heartbeat interval is 1 times of the base heartbeat interval.<br/>e.g., If the base heartbeat interval is 3s, the frontend heartbeat interval is 18s, the flownode/datanode heartbeat interval is 3s.<br/>If you change this value, you need to change the heartbeat interval of the flownode/frontend/datanode accordingly. |
 | `enable_telemetry` | Bool | `true` | Whether to enable greptimedb telemetry. Enabled by default. |
 | `runtime` | -- | -- | The runtime options. |
 | `runtime.global_rt_size` | Integer | `8` | The number of threads to execute the runtime for global read operations. |
@@ -361,12 +364,18 @@
 | `backend_tls.cert_path` | String | `""` | Path to client certificate file (for client authentication)<br/>Like "/path/to/client.crt" |
 | `backend_tls.key_path` | String | `""` | Path to client private key file (for client authentication)<br/>Like "/path/to/client.key" |
 | `backend_tls.ca_cert_path` | String | `""` | Path to CA certificate file (for server certificate verification)<br/>Required when using custom CAs or self-signed certificates<br/>Leave empty to use system root certificates only<br/>Like "/path/to/ca.crt" |
+| `backend_client` | -- | -- | The backend client options.<br/>Currently, only applicable when using etcd as the metadata store. |
+| `backend_client.keep_alive_timeout` | String | `3s` | The keep alive timeout for backend client. |
+| `backend_client.keep_alive_interval` | String | `10s` | The keep alive interval for backend client. |
+| `backend_client.connect_timeout` | String | `3s` | The connect timeout for backend client. |
 | `grpc` | -- | -- | The gRPC server options. |
 | `grpc.bind_addr` | String | `127.0.0.1:3002` | The address to bind the gRPC server. |
 | `grpc.server_addr` | String | `127.0.0.1:3002` | The communication server address for the frontend and datanode to connect to metasrv.<br/>If left empty or unset, the server will automatically use the IP address of the first network interface<br/>on the host, with the same port number as the one specified in `bind_addr`. |
 | `grpc.runtime_size` | Integer | `8` | The number of server worker threads. |
 | `grpc.max_recv_message_size` | String | `512MB` | The maximum receive message size for gRPC server. |
 | `grpc.max_send_message_size` | String | `512MB` | The maximum send message size for gRPC server. |
+| `grpc.http2_keep_alive_interval` | String | `10s` | The server side HTTP/2 keep-alive interval |
+| `grpc.http2_keep_alive_timeout` | String | `3s` | The server side HTTP/2 keep-alive timeout. |
 | `http` | -- | -- | The HTTP server options. |
 | `http.addr` | String | `127.0.0.1:4000` | The address to bind the HTTP server. |
 | `http.timeout` | String | `0s` | HTTP request timeout. Set to 0 to disable timeout. |
@@ -476,6 +485,8 @@
 | `wal.sync_period` | String | `10s` | Duration for fsyncing log files.<br/>**It's only used when the provider is `raft_engine`**. |
 | `wal.recovery_parallelism` | Integer | `2` | Parallelism during WAL recovery. |
 | `wal.broker_endpoints` | Array | -- | The Kafka broker endpoints.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.connect_timeout` | String | `3s` | The connect timeout for kafka client.<br/>**It's only used when the provider is `kafka`**. |
+| `wal.timeout` | String | `3s` | The timeout for kafka client.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.max_batch_bytes` | String | `1MB` | The max size of a single producer batch.<br/>Warning: Kafka has a default limit of 1MB per message in a topic.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.consumer_wait_timeout` | String | `100ms` | The consumer wait timeout.<br/>**It's only used when the provider is `kafka`**. |
 | `wal.create_index` | Bool | `true` | Whether to enable WAL index creation.<br/>**It's only used when the provider is `kafka`**. |
--- a/config/datanode.example.toml
+++ b/config/datanode.example.toml
@@ -169,6 +169,14 @@ recovery_parallelism = 2
 ## **It's only used when the provider is `kafka`**.
 broker_endpoints = ["127.0.0.1:9092"]

+## The connect timeout for kafka client.
+## **It's only used when the provider is `kafka`**.
+#+ connect_timeout = "3s"
+
+## The timeout for kafka client.
+## **It's only used when the provider is `kafka`**.
+#+ timeout = "3s"
+
 ## The max size of a single producer batch.
 ## Warning: Kafka has a default limit of 1MB per message in a topic.
 ## **It's only used when the provider is `kafka`**.
@@ -225,6 +233,7 @@ overwrite_entry_start_id = false
 # endpoint = "https://s3.amazonaws.com"
 # region = "us-west-2"
 # enable_virtual_host_style = false
+# disable_ec2_metadata = false

 # Example of using Oss as the storage.
 # [storage]
--- a/config/frontend.example.toml
+++ b/config/frontend.example.toml
@@ -131,7 +131,6 @@ key_path = ""
 ## For now, gRPC tls config does not support auto reload.
 watch = false

-
 ## MySQL server options.
 [mysql]
 ## Whether to enable.
--- a/config/metasrv.example.toml
+++ b/config/metasrv.example.toml
@@ -71,6 +71,13 @@ allow_region_failover_on_local_wal = false
 ## Max allowed idle time before removing node info from metasrv memory.
 node_max_idle_time = "24hours"

+## Base heartbeat interval for calculating distributed time constants.
+## The frontend heartbeat interval is 6 times of the base heartbeat interval.
+## The flownode/datanode heartbeat interval is 1 times of the base heartbeat interval.
+## e.g., If the base heartbeat interval is 3s, the frontend heartbeat interval is 18s, the flownode/datanode heartbeat interval is 3s.
+## If you change this value, you need to change the heartbeat interval of the flownode/frontend/datanode accordingly.
+#+ heartbeat_interval = "3s"
+
 ## Whether to enable greptimedb telemetry. Enabled by default.
 #+ enable_telemetry = true

@@ -109,6 +116,16 @@ key_path = ""
 ## Like "/path/to/ca.crt"
 ca_cert_path = ""

+## The backend client options.
+## Currently, only applicable when using etcd as the metadata store.
+#+ [backend_client]
+## The keep alive timeout for backend client.
+#+ keep_alive_timeout = "3s"
+## The keep alive interval for backend client.
+#+ keep_alive_interval = "10s"
+## The connect timeout for backend client.
+#+ connect_timeout = "3s"
+
 ## The gRPC server options.
 [grpc]
 ## The address to bind the gRPC server.
@@ -123,6 +140,10 @@ runtime_size = 8
 max_recv_message_size = "512MB"
 ## The maximum send message size for gRPC server.
 max_send_message_size = "512MB"
+## The server side HTTP/2 keep-alive interval
+#+ http2_keep_alive_interval = "10s"
+## The server side HTTP/2 keep-alive timeout.
+#+ http2_keep_alive_timeout = "3s"

 ## The HTTP server options.
 [http]
--- a/config/standalone.example.toml
+++ b/config/standalone.example.toml
@@ -230,6 +230,14 @@ recovery_parallelism = 2
 ## **It's only used when the provider is `kafka`**.
 broker_endpoints = ["127.0.0.1:9092"]

+## The connect timeout for kafka client.
+## **It's only used when the provider is `kafka`**.
+#+ connect_timeout = "3s"
+
+## The timeout for kafka client.
+## **It's only used when the provider is `kafka`**.
+#+ timeout = "3s"
+
 ## Automatically create topics for WAL.
 ## Set to `true` to automatically create topics for WAL.
 ## Otherwise, use topics named `topic_name_prefix_[0..num_topics)`
@@ -332,6 +340,7 @@ max_running_procedures = 128
 # endpoint = "https://s3.amazonaws.com"
 # region = "us-west-2"
 # enable_virtual_host_style = false
+# disable_ec2_metadata = false

 # Example of using Oss as the storage.
 # [storage]
--- a/src/catalog/src/system_schema/information_schema.rs
+++ b/src/catalog/src/system_schema/information_schema.rs
@@ -428,7 +428,7 @@ pub trait InformationExtension {
 }

 /// The request to inspect the datanode.
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone, PartialEq)]
 pub struct DatanodeInspectRequest {
    /// Kind to fetch from datanode.
    pub kind: DatanodeInspectKind,
--- a/src/cli/Cargo.toml
+++ b/src/cli/Cargo.toml
@@ -67,6 +67,7 @@ tracing-appender.workspace = true

 [dev-dependencies]
 common-meta = { workspace = true, features = ["testing"] }
+common-test-util.workspace = true
 common-version.workspace = true
 serde.workspace = true
 tempfile.workspace = true
--- a/src/cli/src/common.rs
+++ b/src/cli/src/common.rs
@@ -15,5 +15,8 @@
 mod object_store;
 mod store;

-pub use object_store::{ObjectStoreConfig, new_fs_object_store};
+pub use object_store::{
+    ObjectStoreConfig, PrefixedAzblobConnection, PrefixedGcsConnection, PrefixedOssConnection,
+    PrefixedS3Connection, new_fs_object_store,
+};
 pub use store::StoreConfig;
--- a/src/cli/src/common/object_store.rs
+++ b/src/cli/src/common/object_store.rs
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use common_base::secrets::SecretString;
+use common_base::secrets::{ExposeSecret, SecretString};
 use common_error::ext::BoxedError;
 use object_store::services::{Azblob, Fs, Gcs, Oss, S3};
 use object_store::util::{with_instrument_layers, with_retry_layers};
@@ -22,9 +22,69 @@ use snafu::ResultExt;

 use crate::error::{self};

+/// Trait to convert CLI field types to target struct field types.
+/// This enables `Option<SecretString>` (CLI) -> `SecretString` (target) conversions,
+/// allowing us to distinguish "not provided" from "provided but empty".
+trait IntoField<T> {
+    fn into_field(self) -> T;
+}
+
+/// Identity conversion for types that are the same.
+impl<T> IntoField<T> for T {
+    fn into_field(self) -> T {
+        self
+    }
+}
+
+/// Convert `Option<SecretString>` to `SecretString`, using default for None.
+impl IntoField<SecretString> for Option<SecretString> {
+    fn into_field(self) -> SecretString {
+        self.unwrap_or_default()
+    }
+}
+
+/// Trait for checking if a field is effectively empty.
+///
+/// **`is_empty()`**: Checks if the field has no meaningful value
+/// - Used when backend is enabled to validate required fields
+/// - `None`, `Some("")`, `false`, or `""` are considered empty
+trait FieldValidator {
+    /// Check if the field is empty (has no meaningful value).
+    fn is_empty(&self) -> bool;
+}
+
+/// String fields: empty if the string is empty
+impl FieldValidator for String {
+    fn is_empty(&self) -> bool {
+        self.is_empty()
+    }
+}
+
+/// Bool fields: false is considered "empty", true is "provided"
+impl FieldValidator for bool {
+    fn is_empty(&self) -> bool {
+        !self
+    }
+}
+
+/// Option<String> fields: None or empty content is empty
+impl FieldValidator for Option<String> {
+    fn is_empty(&self) -> bool {
+        self.as_ref().is_none_or(|s| s.is_empty())
+    }
+}
+
+/// Option<SecretString> fields: None or empty secret is empty
+/// For secrets, Some("") is treated as "not provided" for both checks
+impl FieldValidator for Option<SecretString> {
+    fn is_empty(&self) -> bool {
+        self.as_ref().is_none_or(|s| s.expose_secret().is_empty())
+    }
+}
+
 macro_rules! wrap_with_clap_prefix {
    (
-        $new_name:ident, $prefix:literal, $base:ty, {
+        $new_name:ident, $prefix:literal, $enable_flag:literal, $base:ty, {
            $( $( #[doc = $doc:expr] )? $( #[alias = $alias:literal] )? $field:ident : $type:ty $( = $default:expr )? ),* $(,)?
        }
    ) => {
@@ -34,15 +94,16 @@ macro_rules! wrap_with_clap_prefix {
                $(
                    $( #[doc = $doc] )?
                    $( #[clap(alias = $alias)] )?
-                    #[clap(long $(, default_value_t = $default )? )]
-                    [<$prefix $field>]: $type,
+                    #[clap(long, requires = $enable_flag $(, default_value_t = $default )? )]
+                    pub [<$prefix $field>]: $type,
                )*
            }

            impl From<$new_name> for $base {
                fn from(w: $new_name) -> Self {
                    Self {
-                        $( $field: w.[<$prefix $field>] ),*
+                        // Use into_field() to handle Option<SecretString> -> SecretString conversion
+                        $( $field: w.[<$prefix $field>].into_field() ),*
                    }
                }
            }
@@ -50,9 +111,90 @@ macro_rules! wrap_with_clap_prefix {
    };
 }

+/// Macro for declarative backend validation.
+///
+/// # Validation Rules
+///
+/// For each storage backend (S3, OSS, GCS, Azblob), this function validates:
+/// **When backend is enabled** (e.g., `--s3`): All required fields must be non-empty
+///
+/// Note: When backend is disabled, clap's `requires` attribute ensures no configuration
+/// fields can be provided at parse time.
+///
+/// # Syntax
+///
+/// ```ignore
+/// validate_backend!(
+///     enable: self.enable_s3,
+///     name: "S3",
+///     required: [(field1, "name1"), (field2, "name2"), ...],
+///     custom_validator: |missing| { ... }  // optional
+/// )
+/// ```
+///
+/// # Arguments
+///
+/// - `enable`: Boolean expression indicating if backend is enabled
+/// - `name`: Human-readable backend name for error messages
+/// - `required`: Array of (field_ref, field_name) tuples for required fields
+/// - `custom_validator`: Optional closure for complex validation logic
+///
+/// # Example
+///
+/// ```ignore
+/// validate_backend!(
+///     enable: self.enable_s3,
+///     name: "S3",
+///     required: [
+///         (&self.s3.s3_bucket, "bucket"),
+///         (&self.s3.s3_access_key_id, "access key ID"),
+///     ]
+/// )
+/// ```
+macro_rules! validate_backend {
+    (
+        enable: $enable:expr,
+        name: $backend_name:expr,
+        required: [ $( ($field:expr, $field_name:expr) ),* $(,)? ]
+        $(, custom_validator: $custom_validator:expr)?
+    ) => {{
+        if $enable {
+            // Check required fields when backend is enabled
+            let mut missing = Vec::new();
+            $(
+                if FieldValidator::is_empty($field) {
+                    missing.push($field_name);
+                }
+            )*
+
+            // Run custom validation if provided
+            $(
+                $custom_validator(&mut missing);
+            )?
+
+            if !missing.is_empty() {
+                return Err(BoxedError::new(
+                    error::MissingConfigSnafu {
+                        msg: format!(
+                            "{} {} must be set when --{} is enabled.",
+                            $backend_name,
+                            missing.join(", "),
+                            $backend_name.to_lowercase()
+                        ),
+                    }
+                    .build(),
+                ));
+            }
+        }
+
+        Ok(())
+    }};
+}
+
 wrap_with_clap_prefix! {
    PrefixedAzblobConnection,
    "azblob-",
+    "enable_azblob",
    AzblobConnection,
    {
        #[doc = "The container of the object store."]
@@ -60,9 +202,9 @@ wrap_with_clap_prefix! {
        #[doc = "The root of the object store."]
        root: String = Default::default(),
        #[doc = "The account name of the object store."]
-        account_name: SecretString = Default::default(),
+        account_name: Option<SecretString>,
        #[doc = "The account key of the object store."]
-        account_key: SecretString = Default::default(),
+        account_key: Option<SecretString>,
        #[doc = "The endpoint of the object store."]
        endpoint: String = Default::default(),
        #[doc = "The SAS token of the object store."]
@@ -70,9 +212,33 @@ wrap_with_clap_prefix! {
    }
 }

+impl PrefixedAzblobConnection {
+    pub fn validate(&self) -> Result<(), BoxedError> {
+        validate_backend!(
+            enable: true,
+            name: "AzBlob",
+            required: [
+                (&self.azblob_container, "container"),
+                (&self.azblob_root, "root"),
+                (&self.azblob_account_name, "account name"),
+                (&self.azblob_endpoint, "endpoint"),
+            ],
+            custom_validator: |missing: &mut Vec<&str>| {
+                // account_key is only required if sas_token is not provided
+                if self.azblob_sas_token.is_none()
+                    && self.azblob_account_key.is_empty()
+                {
+                    missing.push("account key (when sas_token is not provided)");
+                }
+            }
+        )
+    }
+}
+
 wrap_with_clap_prefix! {
    PrefixedS3Connection,
    "s3-",
+    "enable_s3",
    S3Connection,
    {
        #[doc = "The bucket of the object store."]
@@ -80,21 +246,39 @@ wrap_with_clap_prefix! {
        #[doc = "The root of the object store."]
        root: String = Default::default(),
        #[doc = "The access key ID of the object store."]
-        access_key_id: SecretString = Default::default(),
+        access_key_id: Option<SecretString>,
        #[doc = "The secret access key of the object store."]
-        secret_access_key: SecretString = Default::default(),
+        secret_access_key: Option<SecretString>,
        #[doc = "The endpoint of the object store."]
        endpoint: Option<String>,
        #[doc = "The region of the object store."]
        region: Option<String>,
        #[doc = "Enable virtual host style for the object store."]
        enable_virtual_host_style: bool = Default::default(),
+        #[doc = "Disable EC2 metadata service for the object store."]
+        disable_ec2_metadata: bool = Default::default(),
+    }
+}
+
+impl PrefixedS3Connection {
+    pub fn validate(&self) -> Result<(), BoxedError> {
+        validate_backend!(
+            enable: true,
+            name: "S3",
+            required: [
+                (&self.s3_bucket, "bucket"),
+                (&self.s3_access_key_id, "access key ID"),
+                (&self.s3_secret_access_key, "secret access key"),
+                (&self.s3_region, "region"),
+            ]
+        )
    }
 }

 wrap_with_clap_prefix! {
    PrefixedOssConnection,
    "oss-",
+    "enable_oss",
    OssConnection,
    {
        #[doc = "The bucket of the object store."]
@@ -102,17 +286,33 @@ wrap_with_clap_prefix! {
        #[doc = "The root of the object store."]
        root: String = Default::default(),
        #[doc = "The access key ID of the object store."]
-        access_key_id: SecretString = Default::default(),
+        access_key_id: Option<SecretString>,
        #[doc = "The access key secret of the object store."]
-        access_key_secret: SecretString = Default::default(),
+        access_key_secret: Option<SecretString>,
        #[doc = "The endpoint of the object store."]
        endpoint: String = Default::default(),
    }
 }

+impl PrefixedOssConnection {
+    pub fn validate(&self) -> Result<(), BoxedError> {
+        validate_backend!(
+            enable: true,
+            name: "OSS",
+            required: [
+                (&self.oss_bucket, "bucket"),
+                (&self.oss_access_key_id, "access key ID"),
+                (&self.oss_access_key_secret, "access key secret"),
+                (&self.oss_endpoint, "endpoint"),
+            ]
+        )
+    }
+}
+
 wrap_with_clap_prefix! {
    PrefixedGcsConnection,
    "gcs-",
+    "enable_gcs",
    GcsConnection,
    {
        #[doc = "The root of the object store."]
@@ -122,40 +322,72 @@ wrap_with_clap_prefix! {
        #[doc = "The scope of the object store."]
        scope: String = Default::default(),
        #[doc = "The credential path of the object store."]
-        credential_path: SecretString = Default::default(),
+        credential_path: Option<SecretString>,
        #[doc = "The credential of the object store."]
-        credential: SecretString = Default::default(),
+        credential: Option<SecretString>,
        #[doc = "The endpoint of the object store."]
        endpoint: String = Default::default(),
    }
 }

-/// common config for object store.
+impl PrefixedGcsConnection {
+    pub fn validate(&self) -> Result<(), BoxedError> {
+        validate_backend!(
+            enable: true,
+            name: "GCS",
+            required: [
+                (&self.gcs_bucket, "bucket"),
+                (&self.gcs_root, "root"),
+                (&self.gcs_scope, "scope"),
+            ]
+            // No custom_validator needed: GCS supports Application Default Credentials (ADC)
+            // where neither credential_path nor credential is required.
+            // Endpoint is also optional (defaults to https://storage.googleapis.com).
+        )
+    }
+}
+
+/// Common config for object store.
+///
+/// # Dependency Enforcement
+///
+/// Each backend's configuration fields (e.g., `--s3-bucket`) requires its corresponding
+/// enable flag (e.g., `--s3`) to be present. This is enforced by `clap` at parse time
+/// using the `requires` attribute.
+///
+/// For example, attempting to use `--s3-bucket my-bucket` without `--s3` will result in:
+/// ```text
+/// error: The argument '--s3-bucket <BUCKET>' requires '--s3'
+/// ```
+///
+/// This ensures that users cannot accidentally provide backend-specific configuration
+/// without explicitly enabling that backend.
 #[derive(clap::Parser, Debug, Clone, PartialEq, Default)]
+#[clap(group(clap::ArgGroup::new("storage_backend").required(false).multiple(false)))]
 pub struct ObjectStoreConfig {
    /// Whether to use S3 object store.
-    #[clap(long, alias = "s3")]
+    #[clap(long = "s3", group = "storage_backend")]
    pub enable_s3: bool,

    #[clap(flatten)]
    pub s3: PrefixedS3Connection,

    /// Whether to use OSS.
-    #[clap(long, alias = "oss")]
+    #[clap(long = "oss", group = "storage_backend")]
    pub enable_oss: bool,

    #[clap(flatten)]
    pub oss: PrefixedOssConnection,

    /// Whether to use GCS.
-    #[clap(long, alias = "gcs")]
+    #[clap(long = "gcs", group = "storage_backend")]
    pub enable_gcs: bool,

    #[clap(flatten)]
    pub gcs: PrefixedGcsConnection,

    /// Whether to use Azure Blob.
-    #[clap(long, alias = "azblob")]
+    #[clap(long = "azblob", group = "storage_backend")]
    pub enable_azblob: bool,

    #[clap(flatten)]
@@ -173,52 +405,66 @@ pub fn new_fs_object_store(root: &str) -> std::result::Result<ObjectStore, Boxed
    Ok(with_instrument_layers(object_store, false))
 }

+macro_rules! gen_object_store_builder {
+    ($method:ident, $field:ident, $conn_type:ty, $service_type:ty) => {
+        pub fn $method(&self) -> Result<ObjectStore, BoxedError> {
+            let config = <$conn_type>::from(self.$field.clone());
+            common_telemetry::info!(
+                "Building object store with {}: {:?}",
+                stringify!($field),
+                config
+            );
+            let object_store = ObjectStore::new(<$service_type>::from(&config))
+                .context(error::InitBackendSnafu)
+                .map_err(BoxedError::new)?
+                .finish();
+            Ok(with_instrument_layers(
+                with_retry_layers(object_store),
+                false,
+            ))
+        }
+    };
+}
+
 impl ObjectStoreConfig {
+    gen_object_store_builder!(build_s3, s3, S3Connection, S3);
+
+    gen_object_store_builder!(build_oss, oss, OssConnection, Oss);
+
+    gen_object_store_builder!(build_gcs, gcs, GcsConnection, Gcs);
+
+    gen_object_store_builder!(build_azblob, azblob, AzblobConnection, Azblob);
+
+    pub fn validate(&self) -> Result<(), BoxedError> {
+        if self.enable_s3 {
+            self.s3.validate()?;
+        }
+        if self.enable_oss {
+            self.oss.validate()?;
+        }
+        if self.enable_gcs {
+            self.gcs.validate()?;
+        }
+        if self.enable_azblob {
+            self.azblob.validate()?;
+        }
+        Ok(())
+    }
+
    /// Builds the object store from the config.
    pub fn build(&self) -> Result<Option<ObjectStore>, BoxedError> {
-        let object_store = if self.enable_s3 {
-            let s3 = S3Connection::from(self.s3.clone());
-            common_telemetry::info!("Building object store with s3: {:?}", s3);
-            Some(
-                ObjectStore::new(S3::from(&s3))
-                    .context(error::InitBackendSnafu)
-                    .map_err(BoxedError::new)?
-                    .finish(),
-            )
+        self.validate()?;
+
+        if self.enable_s3 {
+            self.build_s3().map(Some)
        } else if self.enable_oss {
-            let oss = OssConnection::from(self.oss.clone());
-            common_telemetry::info!("Building object store with oss: {:?}", oss);
-            Some(
-                ObjectStore::new(Oss::from(&oss))
-                    .context(error::InitBackendSnafu)
-                    .map_err(BoxedError::new)?
-                    .finish(),
-            )
+            self.build_oss().map(Some)
        } else if self.enable_gcs {
-            let gcs = GcsConnection::from(self.gcs.clone());
-            common_telemetry::info!("Building object store with gcs: {:?}", gcs);
-            Some(
-                ObjectStore::new(Gcs::from(&gcs))
-                    .context(error::InitBackendSnafu)
-                    .map_err(BoxedError::new)?
-                    .finish(),
-            )
+            self.build_gcs().map(Some)
        } else if self.enable_azblob {
-            let azblob = AzblobConnection::from(self.azblob.clone());
-            common_telemetry::info!("Building object store with azblob: {:?}", azblob);
-            Some(
-                ObjectStore::new(Azblob::from(&azblob))
-                    .context(error::InitBackendSnafu)
-                    .map_err(BoxedError::new)?
-                    .finish(),
-            )
+            self.build_azblob().map(Some)
        } else {
-            None
-        };
-
-        let object_store = object_store
-            .map(|object_store| with_instrument_layers(with_retry_layers(object_store), false));
-
-        Ok(object_store)
+            Ok(None)
+        }
    }
 }
--- a/src/cli/src/common/store.rs
+++ b/src/cli/src/common/store.rs
@@ -19,7 +19,7 @@ use common_error::ext::BoxedError;
 use common_meta::kv_backend::KvBackendRef;
 use common_meta::kv_backend::chroot::ChrootKvBackend;
 use common_meta::kv_backend::etcd::EtcdStore;
-use meta_srv::metasrv::BackendImpl;
+use meta_srv::metasrv::{BackendClientOptions, BackendImpl};
 use meta_srv::utils::etcd::create_etcd_client_with_tls;
 use servers::tls::{TlsMode, TlsOption};

@@ -112,9 +112,13 @@ impl StoreConfig {
            let kvbackend = match self.backend {
                BackendImpl::EtcdStore => {
                    let tls_config = self.tls_config();
-                    let etcd_client = create_etcd_client_with_tls(store_addrs, tls_config.as_ref())
-                        .await
-                        .map_err(BoxedError::new)?;
+                    let etcd_client = create_etcd_client_with_tls(
+                        store_addrs,
+                        &BackendClientOptions::default(),
+                        tls_config.as_ref(),
+                    )
+                    .await
+                    .map_err(BoxedError::new)?;
                    Ok(EtcdStore::with_etcd_client(etcd_client, max_txn_ops))
                }
                #[cfg(feature = "pg_kvbackend")]
--- a/src/cli/src/data.rs
+++ b/src/cli/src/data.rs
@@ -14,6 +14,7 @@

 mod export;
 mod import;
+mod storage_export;

 use clap::Subcommand;
 use client::DEFAULT_CATALOG_NAME;
--- a/src/cli/src/data/export.rs
+++ b/src/cli/src/data/export.rs
--- a/src/cli/src/data/storage_export.rs
+++ b/src/cli/src/data/storage_export.rs
@@ -0,0 +1,373 @@
+// 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::path::PathBuf;
+
+use common_base::secrets::{ExposeSecret, SecretString};
+use common_error::ext::BoxedError;
+
+use crate::common::{
+    PrefixedAzblobConnection, PrefixedGcsConnection, PrefixedOssConnection, PrefixedS3Connection,
+};
+
+/// Helper function to extract secret string from Option<SecretString>.
+/// Returns empty string if None.
+fn expose_optional_secret(secret: &Option<SecretString>) -> &str {
+    secret
+        .as_ref()
+        .map(|s| s.expose_secret().as_str())
+        .unwrap_or("")
+}
+
+/// Helper function to format root path with leading slash if non-empty.
+fn format_root_path(root: &str) -> String {
+    if root.is_empty() {
+        String::new()
+    } else {
+        format!("/{}", root)
+    }
+}
+
+/// Helper function to mask multiple secrets in a string.
+fn mask_secrets(mut sql: String, secrets: &[&str]) -> String {
+    for secret in secrets {
+        if !secret.is_empty() {
+            sql = sql.replace(secret, "[REDACTED]");
+        }
+    }
+    sql
+}
+
+/// Helper function to format storage URI.
+fn format_uri(scheme: &str, bucket: &str, root: &str, path: &str) -> String {
+    let root = format_root_path(root);
+    format!("{}://{}{}/{}", scheme, bucket, root, path)
+}
+
+/// Trait for storage backends that can be used for data export.
+pub trait StorageExport: Send + Sync {
+    /// Generate the storage path for COPY DATABASE command.
+    /// Returns (path, connection_string) where connection_string includes CONNECTION clause.
+    fn get_storage_path(&self, catalog: &str, schema: &str) -> (String, String);
+
+    /// Format the output path for logging purposes.
+    fn format_output_path(&self, file_path: &str) -> String;
+
+    /// Mask sensitive information in SQL commands for safe logging.
+    fn mask_sensitive_info(&self, sql: &str) -> String;
+}
+
+macro_rules! define_backend {
+    ($name:ident, $config:ty) => {
+        #[derive(Clone)]
+        pub struct $name {
+            config: $config,
+        }
+
+        impl $name {
+            pub fn new(config: $config) -> Result<Self, BoxedError> {
+                config.validate()?;
+                Ok(Self { config })
+            }
+        }
+    };
+}
+
+/// Local file system storage backend.
+#[derive(Clone)]
+pub struct FsBackend {
+    output_dir: String,
+}
+
+impl FsBackend {
+    pub fn new(output_dir: String) -> Self {
+        Self { output_dir }
+    }
+}
+
+impl StorageExport for FsBackend {
+    fn get_storage_path(&self, catalog: &str, schema: &str) -> (String, String) {
+        if self.output_dir.is_empty() {
+            unreachable!("output_dir must be set when not using remote storage")
+        }
+        let path = PathBuf::from(&self.output_dir)
+            .join(catalog)
+            .join(format!("{schema}/"))
+            .to_string_lossy()
+            .to_string();
+        (path, String::new())
+    }
+
+    fn format_output_path(&self, file_path: &str) -> String {
+        format!("{}/{}", self.output_dir, file_path)
+    }
+
+    fn mask_sensitive_info(&self, sql: &str) -> String {
+        sql.to_string()
+    }
+}
+
+define_backend!(S3Backend, PrefixedS3Connection);
+
+impl StorageExport for S3Backend {
+    fn get_storage_path(&self, catalog: &str, schema: &str) -> (String, String) {
+        let s3_path = format_uri(
+            "s3",
+            &self.config.s3_bucket,
+            &self.config.s3_root,
+            &format!("{}/{}/", catalog, schema),
+        );
+
+        let mut connection_options = vec![
+            format!(
+                "ACCESS_KEY_ID='{}'",
+                expose_optional_secret(&self.config.s3_access_key_id)
+            ),
+            format!(
+                "SECRET_ACCESS_KEY='{}'",
+                expose_optional_secret(&self.config.s3_secret_access_key)
+            ),
+        ];
+
+        if let Some(region) = &self.config.s3_region {
+            connection_options.push(format!("REGION='{}'", region));
+        }
+
+        if let Some(endpoint) = &self.config.s3_endpoint {
+            connection_options.push(format!("ENDPOINT='{}'", endpoint));
+        }
+
+        let connection_str = format!(" CONNECTION ({})", connection_options.join(", "));
+        (s3_path, connection_str)
+    }
+
+    fn format_output_path(&self, file_path: &str) -> String {
+        format_uri(
+            "s3",
+            &self.config.s3_bucket,
+            &self.config.s3_root,
+            file_path,
+        )
+    }
+
+    fn mask_sensitive_info(&self, sql: &str) -> String {
+        mask_secrets(
+            sql.to_string(),
+            &[
+                expose_optional_secret(&self.config.s3_access_key_id),
+                expose_optional_secret(&self.config.s3_secret_access_key),
+            ],
+        )
+    }
+}
+
+define_backend!(OssBackend, PrefixedOssConnection);
+
+impl StorageExport for OssBackend {
+    fn get_storage_path(&self, catalog: &str, schema: &str) -> (String, String) {
+        let oss_path = format_uri(
+            "oss",
+            &self.config.oss_bucket,
+            &self.config.oss_root,
+            &format!("{}/{}/", catalog, schema),
+        );
+
+        let connection_options = [
+            format!(
+                "ACCESS_KEY_ID='{}'",
+                expose_optional_secret(&self.config.oss_access_key_id)
+            ),
+            format!(
+                "ACCESS_KEY_SECRET='{}'",
+                expose_optional_secret(&self.config.oss_access_key_secret)
+            ),
+        ];
+
+        let connection_str = format!(" CONNECTION ({})", connection_options.join(", "));
+        (oss_path, connection_str)
+    }
+
+    fn format_output_path(&self, file_path: &str) -> String {
+        format_uri(
+            "oss",
+            &self.config.oss_bucket,
+            &self.config.oss_root,
+            file_path,
+        )
+    }
+
+    fn mask_sensitive_info(&self, sql: &str) -> String {
+        mask_secrets(
+            sql.to_string(),
+            &[
+                expose_optional_secret(&self.config.oss_access_key_id),
+                expose_optional_secret(&self.config.oss_access_key_secret),
+            ],
+        )
+    }
+}
+
+define_backend!(GcsBackend, PrefixedGcsConnection);
+
+impl StorageExport for GcsBackend {
+    fn get_storage_path(&self, catalog: &str, schema: &str) -> (String, String) {
+        let gcs_path = format_uri(
+            "gcs",
+            &self.config.gcs_bucket,
+            &self.config.gcs_root,
+            &format!("{}/{}/", catalog, schema),
+        );
+
+        let mut connection_options = Vec::new();
+
+        let credential_path = expose_optional_secret(&self.config.gcs_credential_path);
+        if !credential_path.is_empty() {
+            connection_options.push(format!("CREDENTIAL_PATH='{}'", credential_path));
+        }
+
+        let credential = expose_optional_secret(&self.config.gcs_credential);
+        if !credential.is_empty() {
+            connection_options.push(format!("CREDENTIAL='{}'", credential));
+        }
+
+        if !self.config.gcs_endpoint.is_empty() {
+            connection_options.push(format!("ENDPOINT='{}'", self.config.gcs_endpoint));
+        }
+
+        let connection_str = if connection_options.is_empty() {
+            String::new()
+        } else {
+            format!(" CONNECTION ({})", connection_options.join(", "))
+        };
+
+        (gcs_path, connection_str)
+    }
+
+    fn format_output_path(&self, file_path: &str) -> String {
+        format_uri(
+            "gcs",
+            &self.config.gcs_bucket,
+            &self.config.gcs_root,
+            file_path,
+        )
+    }
+
+    fn mask_sensitive_info(&self, sql: &str) -> String {
+        mask_secrets(
+            sql.to_string(),
+            &[
+                expose_optional_secret(&self.config.gcs_credential_path),
+                expose_optional_secret(&self.config.gcs_credential),
+            ],
+        )
+    }
+}
+
+define_backend!(AzblobBackend, PrefixedAzblobConnection);
+
+impl StorageExport for AzblobBackend {
+    fn get_storage_path(&self, catalog: &str, schema: &str) -> (String, String) {
+        let azblob_path = format_uri(
+            "azblob",
+            &self.config.azblob_container,
+            &self.config.azblob_root,
+            &format!("{}/{}/", catalog, schema),
+        );
+
+        let mut connection_options = vec![
+            format!(
+                "ACCOUNT_NAME='{}'",
+                expose_optional_secret(&self.config.azblob_account_name)
+            ),
+            format!(
+                "ACCOUNT_KEY='{}'",
+                expose_optional_secret(&self.config.azblob_account_key)
+            ),
+        ];
+
+        if let Some(sas_token) = &self.config.azblob_sas_token {
+            connection_options.push(format!("SAS_TOKEN='{}'", sas_token));
+        }
+
+        let connection_str = format!(" CONNECTION ({})", connection_options.join(", "));
+        (azblob_path, connection_str)
+    }
+
+    fn format_output_path(&self, file_path: &str) -> String {
+        format_uri(
+            "azblob",
+            &self.config.azblob_container,
+            &self.config.azblob_root,
+            file_path,
+        )
+    }
+
+    fn mask_sensitive_info(&self, sql: &str) -> String {
+        mask_secrets(
+            sql.to_string(),
+            &[
+                expose_optional_secret(&self.config.azblob_account_name),
+                expose_optional_secret(&self.config.azblob_account_key),
+            ],
+        )
+    }
+}
+
+#[derive(Clone)]
+pub enum StorageType {
+    Fs(FsBackend),
+    S3(S3Backend),
+    Oss(OssBackend),
+    Gcs(GcsBackend),
+    Azblob(AzblobBackend),
+}
+
+impl StorageExport for StorageType {
+    fn get_storage_path(&self, catalog: &str, schema: &str) -> (String, String) {
+        match self {
+            StorageType::Fs(backend) => backend.get_storage_path(catalog, schema),
+            StorageType::S3(backend) => backend.get_storage_path(catalog, schema),
+            StorageType::Oss(backend) => backend.get_storage_path(catalog, schema),
+            StorageType::Gcs(backend) => backend.get_storage_path(catalog, schema),
+            StorageType::Azblob(backend) => backend.get_storage_path(catalog, schema),
+        }
+    }
+
+    fn format_output_path(&self, file_path: &str) -> String {
+        match self {
+            StorageType::Fs(backend) => backend.format_output_path(file_path),
+            StorageType::S3(backend) => backend.format_output_path(file_path),
+            StorageType::Oss(backend) => backend.format_output_path(file_path),
+            StorageType::Gcs(backend) => backend.format_output_path(file_path),
+            StorageType::Azblob(backend) => backend.format_output_path(file_path),
+        }
+    }
+
+    fn mask_sensitive_info(&self, sql: &str) -> String {
+        match self {
+            StorageType::Fs(backend) => backend.mask_sensitive_info(sql),
+            StorageType::S3(backend) => backend.mask_sensitive_info(sql),
+            StorageType::Oss(backend) => backend.mask_sensitive_info(sql),
+            StorageType::Gcs(backend) => backend.mask_sensitive_info(sql),
+            StorageType::Azblob(backend) => backend.mask_sensitive_info(sql),
+        }
+    }
+}
+
+impl StorageType {
+    /// Returns true if the storage backend is remote (not local filesystem).
+    pub fn is_remote_storage(&self) -> bool {
+        !matches!(self, StorageType::Fs(_))
+    }
+}
--- a/src/cli/src/error.rs
+++ b/src/cli/src/error.rs
@@ -253,12 +253,6 @@ pub enum Error {
        error: ObjectStoreError,
    },

-    #[snafu(display("S3 config need be set"))]
-    S3ConfigNotSet {
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Output directory not set"))]
    OutputDirNotSet {
        #[snafu(implicit)]
@@ -364,9 +358,9 @@ impl ErrorExt for Error {

            Error::Other { source, .. } => source.status_code(),
            Error::OpenDal { .. } | Error::InitBackend { .. } => StatusCode::Internal,
-            Error::S3ConfigNotSet { .. }
-            | Error::OutputDirNotSet { .. }
-            | Error::EmptyStoreAddrs { .. } => StatusCode::InvalidArguments,
+            Error::OutputDirNotSet { .. } | Error::EmptyStoreAddrs { .. } => {
+                StatusCode::InvalidArguments
+            }

            Error::BuildRuntime { source, .. } => source.status_code(),

--- a/src/cmd/src/metasrv.rs
+++ b/src/cmd/src/metasrv.rs
@@ -20,6 +20,7 @@ use async_trait::async_trait;
 use clap::Parser;
 use common_base::Plugins;
 use common_config::Configurable;
+use common_meta::distributed_time_constants::init_distributed_time_constants;
 use common_telemetry::info;
 use common_telemetry::logging::{DEFAULT_LOGGING_DIR, TracingOptions};
 use common_version::{short_version, verbose_version};
@@ -327,6 +328,7 @@ impl StartCommand {
        log_versions(verbose_version(), short_version(), APP_NAME);
        maybe_activate_heap_profile(&opts.component.memory);
        create_resource_limit_metrics(APP_NAME);
+        init_distributed_time_constants(opts.component.heartbeat_interval);

        info!("Metasrv start command: {:#?}", self);

--- a/src/cmd/src/standalone.rs
+++ b/src/cmd/src/standalone.rs
@@ -552,9 +552,8 @@ impl StartCommand {
        let grpc_handler = fe_instance.clone() as Arc<dyn GrpcQueryHandlerWithBoxedError>;
        let weak_grpc_handler = Arc::downgrade(&grpc_handler);
        frontend_instance_handler
-            .lock()
-            .unwrap()
-            .replace(weak_grpc_handler);
+            .set_handler(weak_grpc_handler)
+            .await;

        // set the frontend invoker for flownode
        let flow_streaming_engine = flownode.flow_engine().streaming_engine();
--- a/src/common/config/src/error.rs
+++ b/src/common/config/src/error.rs
@@ -59,15 +59,6 @@ pub enum Error {
        location: Location,
    },

-    #[snafu(display("Failed to canonicalize path: {}", path))]
-    CanonicalizePath {
-        path: String,
-        #[snafu(source)]
-        error: std::io::Error,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
    #[snafu(display("Invalid path '{}': expected a file, not a directory", path))]
    InvalidPath {
        path: String,
@@ -82,8 +73,7 @@ impl ErrorExt for Error {
            Error::TomlFormat { .. }
            | Error::LoadLayeredConfig { .. }
            | Error::FileWatch { .. }
-            | Error::InvalidPath { .. }
-            | Error::CanonicalizePath { .. } => StatusCode::InvalidArguments,
+            | Error::InvalidPath { .. } => StatusCode::InvalidArguments,
            Error::SerdeJson { .. } => StatusCode::Unexpected,
        }
    }
--- a/src/common/config/src/file_watcher.rs
+++ b/src/common/config/src/file_watcher.rs
@@ -30,7 +30,7 @@ use common_telemetry::{error, info, warn};
 use notify::{EventKind, RecursiveMode, Watcher};
 use snafu::ResultExt;

-use crate::error::{CanonicalizePathSnafu, FileWatchSnafu, InvalidPathSnafu, Result};
+use crate::error::{FileWatchSnafu, InvalidPathSnafu, Result};

 /// Configuration for the file watcher behavior.
 #[derive(Debug, Clone, Default)]
@@ -41,15 +41,10 @@ pub struct FileWatcherConfig {

 impl FileWatcherConfig {
    pub fn new() -> Self {
-        Self::default()
+        Default::default()
    }

-    pub fn with_modify_and_create(mut self) -> Self {
-        self.include_remove_events = false;
-        self
-    }
-
-    pub fn with_remove_events(mut self) -> Self {
+    pub fn include_remove_events(mut self) -> Self {
        self.include_remove_events = true;
        self
    }
@@ -93,11 +88,8 @@ impl FileWatcherBuilder {
                path: path.display().to_string(),
            }
        );
-        // Canonicalize the path for reliable comparison with event paths
-        let canonical = path.canonicalize().context(CanonicalizePathSnafu {
-            path: path.display().to_string(),
-        })?;
-        self.file_paths.push(canonical);
+
+        self.file_paths.push(path.to_path_buf());
        Ok(self)
    }

@@ -144,7 +136,6 @@ impl FileWatcherBuilder {
        }

        let config = self.config;
-        let watched_files: HashSet<PathBuf> = self.file_paths.iter().cloned().collect();

        info!(
            "Spawning file watcher for paths: {:?} (watching parent directories)",
@@ -165,25 +156,7 @@ impl FileWatcherBuilder {
                            continue;
                        }

-                        // Check if any of the event paths match our watched files
-                        let is_watched_file = event.paths.iter().any(|event_path| {
-                            // Try to canonicalize the event path for comparison
-                            // If the file was deleted, canonicalize will fail, so we also
-                            // compare the raw path
-                            if let Ok(canonical) = event_path.canonicalize()
-                                && watched_files.contains(&canonical)
-                            {
-                                return true;
-                            }
-                            // For deleted files, compare using the raw path
-                            watched_files.contains(event_path)
-                        });
-
-                        if !is_watched_file {
-                            continue;
-                        }
-
-                        info!(?event.kind, ?event.paths, "Detected file change");
+                        info!(?event.kind, ?event.paths, "Detected folder change");
                        callback();
                    }
                    Err(err) => {
@@ -301,55 +274,4 @@ mod tests {
            "Watcher should have detected file recreation"
        );
    }
-
-    #[test]
-    fn test_file_watcher_ignores_other_files() {
-        common_telemetry::init_default_ut_logging();
-
-        let dir = create_temp_dir("test_file_watcher_other");
-        let watched_file = dir.path().join("watched.txt");
-        let other_file = dir.path().join("other.txt");
-
-        // Create both files
-        std::fs::write(&watched_file, "watched content").unwrap();
-        std::fs::write(&other_file, "other content").unwrap();
-
-        let counter = Arc::new(AtomicUsize::new(0));
-        let counter_clone = counter.clone();
-
-        FileWatcherBuilder::new()
-            .watch_path(&watched_file)
-            .unwrap()
-            .config(FileWatcherConfig::new())
-            .spawn(move || {
-                counter_clone.fetch_add(1, Ordering::SeqCst);
-            })
-            .unwrap();
-
-        // Give watcher time to start
-        std::thread::sleep(Duration::from_millis(100));
-
-        // Modify the other file - should NOT trigger callback
-        std::fs::write(&other_file, "modified other content").unwrap();
-
-        // Wait for potential event
-        std::thread::sleep(Duration::from_millis(500));
-
-        assert_eq!(
-            counter.load(Ordering::SeqCst),
-            0,
-            "Watcher should not have detected changes to other files"
-        );
-
-        // Now modify the watched file - SHOULD trigger callback
-        std::fs::write(&watched_file, "modified watched content").unwrap();
-
-        // Wait for the event to be processed
-        std::thread::sleep(Duration::from_millis(500));
-
-        assert!(
-            counter.load(Ordering::SeqCst) >= 1,
-            "Watcher should have detected change to watched file"
-        );
-    }
 }
--- a/src/common/datasource/src/object_store/s3.rs
+++ b/src/common/datasource/src/object_store/s3.rs
@@ -27,6 +27,7 @@ const SECRET_ACCESS_KEY: &str = "secret_access_key";
 const SESSION_TOKEN: &str = "session_token";
 const REGION: &str = "region";
 const ENABLE_VIRTUAL_HOST_STYLE: &str = "enable_virtual_host_style";
+const DISABLE_EC2_METADATA: &str = "disable_ec2_metadata";

 pub fn is_supported_in_s3(key: &str) -> bool {
    [
@@ -36,6 +37,7 @@ pub fn is_supported_in_s3(key: &str) -> bool {
        SESSION_TOKEN,
        REGION,
        ENABLE_VIRTUAL_HOST_STYLE,
+        DISABLE_EC2_METADATA,
    ]
    .contains(&key)
 }
@@ -82,6 +84,21 @@ pub fn build_s3_backend(
        }
    }

+    if let Some(disable_str) = connection.get(DISABLE_EC2_METADATA) {
+        let disable = disable_str.as_str().parse::<bool>().map_err(|e| {
+            error::InvalidConnectionSnafu {
+                msg: format!(
+                    "failed to parse the option {}={}, {}",
+                    DISABLE_EC2_METADATA, disable_str, e
+                ),
+            }
+            .build()
+        })?;
+        if disable {
+            builder = builder.disable_ec2_metadata();
+        }
+    }
+
    // TODO(weny): Consider finding a better way to eliminate duplicate code.
    Ok(ObjectStore::new(builder)
        .context(error::BuildBackendSnafu)?
@@ -109,6 +126,7 @@ mod tests {
        assert!(is_supported_in_s3(SESSION_TOKEN));
        assert!(is_supported_in_s3(REGION));
        assert!(is_supported_in_s3(ENABLE_VIRTUAL_HOST_STYLE));
+        assert!(is_supported_in_s3(DISABLE_EC2_METADATA));
        assert!(!is_supported_in_s3("foo"))
    }
 }
--- a/src/common/function/Cargo.toml
+++ b/src/common/function/Cargo.toml
@@ -19,7 +19,7 @@ arc-swap = "1.0"
 arrow.workspace = true
 arrow-schema.workspace = true
 async-trait.workspace = true
-bincode = "1.3"
+bincode = "=1.3.3"
 catalog.workspace = true
 chrono.workspace = true
 common-base.workspace = true
--- a/src/common/memory-manager/src/error.rs
+++ b/src/common/memory-manager/src/error.rs
@@ -13,6 +13,7 @@
 // limitations under the License.

 use std::any::Any;
+use std::time::Duration;

 use common_error::ext::ErrorExt;
 use common_error::status_code::StatusCode;
@@ -35,6 +36,14 @@ pub enum Error {

    #[snafu(display("Memory semaphore unexpectedly closed"))]
    MemorySemaphoreClosed,
+
+    #[snafu(display(
+        "Timeout waiting for memory quota: requested {requested_bytes} bytes, waited {waited:?}"
+    ))]
+    MemoryAcquireTimeout {
+        requested_bytes: u64,
+        waited: Duration,
+    },
 }

 impl ErrorExt for Error {
@@ -44,6 +53,7 @@ impl ErrorExt for Error {
        match self {
            MemoryLimitExceeded { .. } => StatusCode::RuntimeResourcesExhausted,
            MemorySemaphoreClosed => StatusCode::Unexpected,
+            MemoryAcquireTimeout { .. } => StatusCode::RuntimeResourcesExhausted,
        }
    }

--- a/src/common/memory-manager/src/granularity.rs
+++ b/src/common/memory-manager/src/granularity.rs
@@ -0,0 +1,168 @@
+// 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;
+
+/// Memory permit granularity for different use cases.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
+pub enum PermitGranularity {
+    /// 1 KB per permit
+    ///
+    /// Use for:
+    /// - HTTP/gRPC request limiting (small, high-concurrency operations)
+    /// - Small batch operations
+    /// - Scenarios requiring fine-grained fairness
+    Kilobyte,
+
+    /// 1 MB per permit (default)
+    ///
+    /// Use for:
+    /// - Query execution memory management
+    /// - Compaction memory control
+    /// - Large, long-running operations
+    #[default]
+    Megabyte,
+}
+
+impl PermitGranularity {
+    /// Returns the number of bytes per permit.
+    #[inline]
+    pub const fn bytes(self) -> u64 {
+        match self {
+            Self::Kilobyte => 1024,
+            Self::Megabyte => 1024 * 1024,
+        }
+    }
+
+    /// Returns a human-readable string representation.
+    pub const fn as_str(self) -> &'static str {
+        match self {
+            Self::Kilobyte => "1KB",
+            Self::Megabyte => "1MB",
+        }
+    }
+
+    /// Converts bytes to permits based on this granularity.
+    ///
+    /// Rounds up to ensure the requested bytes are fully covered.
+    /// Clamped to Semaphore::MAX_PERMITS.
+    #[inline]
+    pub fn bytes_to_permits(self, bytes: u64) -> u32 {
+        use tokio::sync::Semaphore;
+
+        let granularity_bytes = self.bytes();
+        bytes
+            .saturating_add(granularity_bytes - 1)
+            .saturating_div(granularity_bytes)
+            .min(Semaphore::MAX_PERMITS as u64)
+            .min(u32::MAX as u64) as u32
+    }
+
+    /// Converts permits to bytes based on this granularity.
+    #[inline]
+    pub fn permits_to_bytes(self, permits: u32) -> u64 {
+        (permits as u64).saturating_mul(self.bytes())
+    }
+}
+
+impl fmt::Display for PermitGranularity {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "{}", self.as_str())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_bytes_to_permits_kilobyte() {
+        let granularity = PermitGranularity::Kilobyte;
+
+        // Exact multiples
+        assert_eq!(granularity.bytes_to_permits(1024), 1);
+        assert_eq!(granularity.bytes_to_permits(2048), 2);
+        assert_eq!(granularity.bytes_to_permits(10 * 1024), 10);
+
+        // Rounds up
+        assert_eq!(granularity.bytes_to_permits(1), 1);
+        assert_eq!(granularity.bytes_to_permits(1025), 2);
+        assert_eq!(granularity.bytes_to_permits(2047), 2);
+    }
+
+    #[test]
+    fn test_bytes_to_permits_megabyte() {
+        let granularity = PermitGranularity::Megabyte;
+
+        // Exact multiples
+        assert_eq!(granularity.bytes_to_permits(1024 * 1024), 1);
+        assert_eq!(granularity.bytes_to_permits(2 * 1024 * 1024), 2);
+
+        // Rounds up
+        assert_eq!(granularity.bytes_to_permits(1), 1);
+        assert_eq!(granularity.bytes_to_permits(1024), 1);
+        assert_eq!(granularity.bytes_to_permits(1024 * 1024 + 1), 2);
+    }
+
+    #[test]
+    fn test_bytes_to_permits_zero_bytes() {
+        assert_eq!(PermitGranularity::Kilobyte.bytes_to_permits(0), 0);
+        assert_eq!(PermitGranularity::Megabyte.bytes_to_permits(0), 0);
+    }
+
+    #[test]
+    fn test_bytes_to_permits_clamps_to_maximum() {
+        use tokio::sync::Semaphore;
+
+        let max_permits = (Semaphore::MAX_PERMITS as u64).min(u32::MAX as u64) as u32;
+
+        assert_eq!(
+            PermitGranularity::Kilobyte.bytes_to_permits(u64::MAX),
+            max_permits
+        );
+        assert_eq!(
+            PermitGranularity::Megabyte.bytes_to_permits(u64::MAX),
+            max_permits
+        );
+    }
+
+    #[test]
+    fn test_permits_to_bytes() {
+        assert_eq!(PermitGranularity::Kilobyte.permits_to_bytes(1), 1024);
+        assert_eq!(PermitGranularity::Kilobyte.permits_to_bytes(10), 10 * 1024);
+
+        assert_eq!(PermitGranularity::Megabyte.permits_to_bytes(1), 1024 * 1024);
+        assert_eq!(
+            PermitGranularity::Megabyte.permits_to_bytes(10),
+            10 * 1024 * 1024
+        );
+    }
+
+    #[test]
+    fn test_round_trip_conversion() {
+        // Kilobyte: bytes -> permits -> bytes (should round up)
+        let kb = PermitGranularity::Kilobyte;
+        let permits = kb.bytes_to_permits(1500);
+        let bytes = kb.permits_to_bytes(permits);
+        assert!(bytes >= 1500); // Must cover original request
+        assert_eq!(bytes, 2048); // 2KB
+
+        // Megabyte: bytes -> permits -> bytes (should round up)
+        let mb = PermitGranularity::Megabyte;
+        let permits = mb.bytes_to_permits(1500);
+        let bytes = mb.permits_to_bytes(permits);
+        assert!(bytes >= 1500);
+        assert_eq!(bytes, 1024 * 1024); // 1MB
+    }
+}
--- a/src/common/memory-manager/src/guard.rs
+++ b/src/common/memory-manager/src/guard.rs
@@ -17,7 +17,7 @@ use std::{fmt, mem};
 use common_telemetry::debug;
 use tokio::sync::{OwnedSemaphorePermit, TryAcquireError};

-use crate::manager::{MemoryMetrics, MemoryQuota, bytes_to_permits, permits_to_bytes};
+use crate::manager::{MemoryMetrics, MemoryQuota};

 /// Guard representing a slice of reserved memory.
 pub struct MemoryGuard<M: MemoryMetrics> {
@@ -49,7 +49,9 @@ impl<M: MemoryMetrics> MemoryGuard<M> {
    pub fn granted_bytes(&self) -> u64 {
        match &self.state {
            GuardState::Unlimited => 0,
-            GuardState::Limited { permit, .. } => permits_to_bytes(permit.num_permits() as u32),
+            GuardState::Limited { permit, quota } => {
+                quota.permits_to_bytes(permit.num_permits() as u32)
+            }
        }
    }

@@ -65,7 +67,7 @@ impl<M: MemoryMetrics> MemoryGuard<M> {
                    return true;
                }

-                let additional_permits = bytes_to_permits(bytes);
+                let additional_permits = quota.bytes_to_permits(bytes);

                match quota
                    .semaphore
@@ -99,11 +101,12 @@ impl<M: MemoryMetrics> MemoryGuard<M> {
                    return true;
                }

-                let release_permits = bytes_to_permits(bytes);
+                let release_permits = quota.bytes_to_permits(bytes);

                match permit.split(release_permits as usize) {
                    Some(released_permit) => {
-                        let released_bytes = permits_to_bytes(released_permit.num_permits() as u32);
+                        let released_bytes =
+                            quota.permits_to_bytes(released_permit.num_permits() as u32);
                        drop(released_permit);
                        quota.update_in_use_metric();
                        debug!("Early released {} bytes from memory guard", released_bytes);
@@ -121,7 +124,7 @@ impl<M: MemoryMetrics> Drop for MemoryGuard<M> {
        if let GuardState::Limited { permit, quota } =
            mem::replace(&mut self.state, GuardState::Unlimited)
        {
-            let bytes = permits_to_bytes(permit.num_permits() as u32);
+            let bytes = quota.permits_to_bytes(permit.num_permits() as u32);
            drop(permit);
            quota.update_in_use_metric();
            debug!("Released memory: {} bytes", bytes);
--- a/src/common/memory-manager/src/lib.rs
+++ b/src/common/memory-manager/src/lib.rs
@@ -19,6 +19,7 @@
 //! share the same allocation logic while using their own metrics.

 mod error;
+mod granularity;
 mod guard;
 mod manager;
 mod policy;
@@ -27,8 +28,9 @@ mod policy;
 mod tests;

 pub use error::{Error, Result};
+pub use granularity::PermitGranularity;
 pub use guard::MemoryGuard;
-pub use manager::{MemoryManager, MemoryMetrics, PERMIT_GRANULARITY_BYTES};
+pub use manager::{MemoryManager, MemoryMetrics};
 pub use policy::{DEFAULT_MEMORY_WAIT_TIMEOUT, OnExhaustedPolicy};

 /// No-op metrics implementation for testing.
--- a/src/common/memory-manager/src/manager.rs
+++ b/src/common/memory-manager/src/manager.rs
@@ -17,11 +17,12 @@ use std::sync::Arc;
 use snafu::ensure;
 use tokio::sync::{Semaphore, TryAcquireError};

-use crate::error::{MemoryLimitExceededSnafu, MemorySemaphoreClosedSnafu, Result};
+use crate::error::{
+    MemoryAcquireTimeoutSnafu, MemoryLimitExceededSnafu, MemorySemaphoreClosedSnafu, Result,
+};
+use crate::granularity::PermitGranularity;
 use crate::guard::MemoryGuard;
-
-/// Minimum bytes controlled by one semaphore permit.
-pub const PERMIT_GRANULARITY_BYTES: u64 = 1 << 20; // 1 MB
+use crate::policy::OnExhaustedPolicy;

 /// Trait for recording memory usage metrics.
 pub trait MemoryMetrics: Clone + Send + Sync + 'static {
@@ -40,6 +41,7 @@ pub struct MemoryManager<M: MemoryMetrics> {
 pub(crate) struct MemoryQuota<M: MemoryMetrics> {
    pub(crate) semaphore: Arc<Semaphore>,
    pub(crate) limit_permits: u32,
+    pub(crate) granularity: PermitGranularity,
    pub(crate) metrics: M,
 }

@@ -47,19 +49,25 @@ impl<M: MemoryMetrics> MemoryManager<M> {
    /// Creates a new memory manager with the given limit in bytes.
    /// `limit_bytes = 0` disables the limit.
    pub fn new(limit_bytes: u64, metrics: M) -> Self {
+        Self::with_granularity(limit_bytes, PermitGranularity::default(), metrics)
+    }
+
+    /// Creates a new memory manager with specified granularity.
+    pub fn with_granularity(limit_bytes: u64, granularity: PermitGranularity, metrics: M) -> Self {
        if limit_bytes == 0 {
            metrics.set_limit(0);
            return Self { quota: None };
        }

-        let limit_permits = bytes_to_permits(limit_bytes);
-        let limit_aligned_bytes = permits_to_bytes(limit_permits);
+        let limit_permits = granularity.bytes_to_permits(limit_bytes);
+        let limit_aligned_bytes = granularity.permits_to_bytes(limit_permits);
        metrics.set_limit(limit_aligned_bytes as i64);

        Self {
            quota: Some(MemoryQuota {
                semaphore: Arc::new(Semaphore::new(limit_permits as usize)),
                limit_permits,
+                granularity,
                metrics,
            }),
        }
@@ -69,7 +77,7 @@ impl<M: MemoryMetrics> MemoryManager<M> {
    pub fn limit_bytes(&self) -> u64 {
        self.quota
            .as_ref()
-            .map(|quota| permits_to_bytes(quota.limit_permits))
+            .map(|quota| quota.permits_to_bytes(quota.limit_permits))
            .unwrap_or(0)
    }

@@ -77,7 +85,7 @@ impl<M: MemoryMetrics> MemoryManager<M> {
    pub fn used_bytes(&self) -> u64 {
        self.quota
            .as_ref()
-            .map(|quota| permits_to_bytes(quota.used_permits()))
+            .map(|quota| quota.permits_to_bytes(quota.used_permits()))
            .unwrap_or(0)
    }

@@ -85,7 +93,7 @@ impl<M: MemoryMetrics> MemoryManager<M> {
    pub fn available_bytes(&self) -> u64 {
        self.quota
            .as_ref()
-            .map(|quota| permits_to_bytes(quota.available_permits_clamped()))
+            .map(|quota| quota.permits_to_bytes(quota.available_permits_clamped()))
            .unwrap_or(0)
    }

@@ -98,13 +106,13 @@ impl<M: MemoryMetrics> MemoryManager<M> {
        match &self.quota {
            None => Ok(MemoryGuard::unlimited()),
            Some(quota) => {
-                let permits = bytes_to_permits(bytes);
+                let permits = quota.bytes_to_permits(bytes);

                ensure!(
                    permits <= quota.limit_permits,
                    MemoryLimitExceededSnafu {
                        requested_bytes: bytes,
-                        limit_bytes: permits_to_bytes(quota.limit_permits),
+                        limit_bytes: self.limit_bytes()
                    }
                );

@@ -125,7 +133,7 @@ impl<M: MemoryMetrics> MemoryManager<M> {
        match &self.quota {
            None => Some(MemoryGuard::unlimited()),
            Some(quota) => {
-                let permits = bytes_to_permits(bytes);
+                let permits = quota.bytes_to_permits(bytes);

                match quota.semaphore.clone().try_acquire_many_owned(permits) {
                    Ok(permit) => {
@@ -140,9 +148,56 @@ impl<M: MemoryMetrics> MemoryManager<M> {
            }
        }
    }
+
+    /// Acquires memory based on the given policy.
+    ///
+    /// - For `OnExhaustedPolicy::Wait`: Waits up to the timeout duration for memory to become available
+    /// - For `OnExhaustedPolicy::Fail`: Returns immediately if memory is not available
+    ///
+    /// # Errors
+    /// - `MemoryLimitExceeded`: Requested bytes exceed the total limit (both policies), or memory is currently exhausted (Fail policy only)
+    /// - `MemoryAcquireTimeout`: Timeout elapsed while waiting for memory (Wait policy only)
+    /// - `MemorySemaphoreClosed`: The internal semaphore is unexpectedly closed (rare, indicates system issue)
+    pub async fn acquire_with_policy(
+        &self,
+        bytes: u64,
+        policy: OnExhaustedPolicy,
+    ) -> Result<MemoryGuard<M>> {
+        match policy {
+            OnExhaustedPolicy::Wait { timeout } => {
+                match tokio::time::timeout(timeout, self.acquire(bytes)).await {
+                    Ok(Ok(guard)) => Ok(guard),
+                    Ok(Err(e)) => Err(e),
+                    Err(_elapsed) => {
+                        // Timeout elapsed while waiting
+                        MemoryAcquireTimeoutSnafu {
+                            requested_bytes: bytes,
+                            waited: timeout,
+                        }
+                        .fail()
+                    }
+                }
+            }
+            OnExhaustedPolicy::Fail => self.try_acquire(bytes).ok_or_else(|| {
+                MemoryLimitExceededSnafu {
+                    requested_bytes: bytes,
+                    limit_bytes: self.limit_bytes(),
+                }
+                .build()
+            }),
+        }
+    }
 }

 impl<M: MemoryMetrics> MemoryQuota<M> {
+    pub(crate) fn bytes_to_permits(&self, bytes: u64) -> u32 {
+        self.granularity.bytes_to_permits(bytes)
+    }
+
+    pub(crate) fn permits_to_bytes(&self, permits: u32) -> u64 {
+        self.granularity.permits_to_bytes(permits)
+    }
+
    pub(crate) fn used_permits(&self) -> u32 {
        self.limit_permits
            .saturating_sub(self.available_permits_clamped())
@@ -155,19 +210,7 @@ impl<M: MemoryMetrics> MemoryQuota<M> {
    }

    pub(crate) fn update_in_use_metric(&self) {
-        let bytes = permits_to_bytes(self.used_permits());
+        let bytes = self.permits_to_bytes(self.used_permits());
        self.metrics.set_in_use(bytes as i64);
    }
 }
-
-pub(crate) fn bytes_to_permits(bytes: u64) -> u32 {
-    bytes
-        .saturating_add(PERMIT_GRANULARITY_BYTES - 1)
-        .saturating_div(PERMIT_GRANULARITY_BYTES)
-        .min(Semaphore::MAX_PERMITS as u64)
-        .min(u32::MAX as u64) as u32
-}
-
-pub(crate) fn permits_to_bytes(permits: u32) -> u64 {
-    (permits as u64).saturating_mul(PERMIT_GRANULARITY_BYTES)
-}
--- a/src/common/memory-manager/src/tests.rs
+++ b/src/common/memory-manager/src/tests.rs
@@ -14,7 +14,10 @@

 use tokio::time::{Duration, sleep};

-use crate::{MemoryManager, NoOpMetrics, PERMIT_GRANULARITY_BYTES};
+use crate::{MemoryManager, NoOpMetrics, PermitGranularity};
+
+// Helper constant for tests - use default Megabyte granularity
+const PERMIT_GRANULARITY_BYTES: u64 = PermitGranularity::Megabyte.bytes();

 #[test]
 fn test_try_acquire_unlimited() {
--- a/src/common/meta/src/distributed_time_constants.rs
+++ b/src/common/meta/src/distributed_time_constants.rs
@@ -12,27 +12,10 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

+use std::sync::OnceLock;
 use std::time::Duration;

-use etcd_client::ConnectOptions;
-
-/// Heartbeat interval time (is the basic unit of various time).
-pub const HEARTBEAT_INTERVAL_MILLIS: u64 = 3000;
-
-/// The frontend will also send heartbeats to Metasrv, sending an empty
-/// heartbeat every HEARTBEAT_INTERVAL_MILLIS * 6 seconds.
-pub const FRONTEND_HEARTBEAT_INTERVAL_MILLIS: u64 = HEARTBEAT_INTERVAL_MILLIS * 6;
-
-/// The lease seconds of a region. It's set by 3 heartbeat intervals
-/// (HEARTBEAT_INTERVAL_MILLIS × 3), plus some extra buffer (1 second).
-pub const REGION_LEASE_SECS: u64 =
-    Duration::from_millis(HEARTBEAT_INTERVAL_MILLIS * 3).as_secs() + 1;
-
-/// When creating table or region failover, a target node needs to be selected.
-/// If the node's lease has expired, the `Selector` will not select it.
-pub const DATANODE_LEASE_SECS: u64 = REGION_LEASE_SECS;
-
-pub const FLOWNODE_LEASE_SECS: u64 = DATANODE_LEASE_SECS;
+pub const BASE_HEARTBEAT_INTERVAL: Duration = Duration::from_secs(3);

 /// The lease seconds of metasrv leader.
 pub const META_LEASE_SECS: u64 = 5;
@@ -52,14 +35,6 @@ pub const HEARTBEAT_CHANNEL_KEEP_ALIVE_INTERVAL_SECS: Duration = Duration::from_
 /// The keep-alive timeout of the heartbeat channel.
 pub const HEARTBEAT_CHANNEL_KEEP_ALIVE_TIMEOUT_SECS: Duration = Duration::from_secs(5);

-/// The default options for the etcd client.
-pub fn default_etcd_client_options() -> ConnectOptions {
-    ConnectOptions::new()
-        .with_keep_alive_while_idle(true)
-        .with_keep_alive(Duration::from_secs(15), Duration::from_secs(5))
-        .with_connect_timeout(Duration::from_secs(10))
-}
-
 /// The default mailbox round-trip timeout.
 pub const MAILBOX_RTT_SECS: u64 = 1;

@@ -68,3 +43,60 @@ pub const TOPIC_STATS_REPORT_INTERVAL_SECS: u64 = 15;

 /// The retention seconds of topic stats.
 pub const TOPIC_STATS_RETENTION_SECS: u64 = TOPIC_STATS_REPORT_INTERVAL_SECS * 100;
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+/// The distributed time constants.
+pub struct DistributedTimeConstants {
+    pub heartbeat_interval: Duration,
+    pub frontend_heartbeat_interval: Duration,
+    pub region_lease: Duration,
+    pub datanode_lease: Duration,
+    pub flownode_lease: Duration,
+}
+
+/// The frontend heartbeat interval is 6 times of the base heartbeat interval.
+pub fn frontend_heartbeat_interval(base_heartbeat_interval: Duration) -> Duration {
+    base_heartbeat_interval * 6
+}
+
+impl DistributedTimeConstants {
+    /// Create a new DistributedTimeConstants from the heartbeat interval.
+    pub fn from_heartbeat_interval(heartbeat_interval: Duration) -> Self {
+        let region_lease = heartbeat_interval * 3 + Duration::from_secs(1);
+        let datanode_lease = region_lease;
+        let flownode_lease = datanode_lease;
+        Self {
+            heartbeat_interval,
+            frontend_heartbeat_interval: frontend_heartbeat_interval(heartbeat_interval),
+            region_lease,
+            datanode_lease,
+            flownode_lease,
+        }
+    }
+}
+
+impl Default for DistributedTimeConstants {
+    fn default() -> Self {
+        Self::from_heartbeat_interval(BASE_HEARTBEAT_INTERVAL)
+    }
+}
+
+static DEFAULT_DISTRIBUTED_TIME_CONSTANTS: OnceLock<DistributedTimeConstants> = OnceLock::new();
+
+/// Get the default distributed time constants.
+pub fn default_distributed_time_constants() -> &'static DistributedTimeConstants {
+    DEFAULT_DISTRIBUTED_TIME_CONSTANTS.get_or_init(Default::default)
+}
+
+/// Initialize the default distributed time constants.
+pub fn init_distributed_time_constants(base_heartbeat_interval: Duration) {
+    let distributed_time_constants =
+        DistributedTimeConstants::from_heartbeat_interval(base_heartbeat_interval);
+    DEFAULT_DISTRIBUTED_TIME_CONSTANTS
+        .set(distributed_time_constants)
+        .expect("Failed to set default distributed time constants");
+    common_telemetry::info!(
+        "Initialized default distributed time constants: {:#?}",
+        distributed_time_constants
+    );
+}
--- a/src/common/meta/src/wal_options_allocator/topic_creator.rs
+++ b/src/common/meta/src/wal_options_allocator/topic_creator.rs
@@ -14,7 +14,7 @@

 use common_telemetry::{debug, error, info};
 use common_wal::config::kafka::common::{
-    DEFAULT_BACKOFF_CONFIG, DEFAULT_CONNECT_TIMEOUT, KafkaConnectionConfig, KafkaTopicConfig,
+    DEFAULT_BACKOFF_CONFIG, KafkaConnectionConfig, KafkaTopicConfig,
 };
 use rskafka::client::error::Error as RsKafkaError;
 use rskafka::client::error::ProtocolError::TopicAlreadyExists;
@@ -211,7 +211,8 @@ pub async fn build_kafka_client(connection: &KafkaConnectionConfig) -> Result<Cl
    // Builds an kafka controller client for creating topics.
    let mut builder = ClientBuilder::new(connection.broker_endpoints.clone())
        .backoff_config(DEFAULT_BACKOFF_CONFIG)
-        .connect_timeout(Some(DEFAULT_CONNECT_TIMEOUT));
+        .connect_timeout(Some(connection.connect_timeout))
+        .timeout(Some(connection.timeout));
    if let Some(sasl) = &connection.sasl {
        builder = builder.sasl_config(sasl.config.clone().into_sasl_config());
    };
--- a/src/common/telemetry/src/metric.rs
+++ b/src/common/telemetry/src/metric.rs
@@ -71,6 +71,7 @@ pub fn convert_metric_to_write_request(
                        timestamp,
                    }],
                    exemplars: vec![],
+                    histograms: vec![],
                }),
                MetricType::GAUGE => timeseries.push(TimeSeries {
                    labels: convert_label(m.get_label(), mf_name, None),
@@ -79,6 +80,7 @@ pub fn convert_metric_to_write_request(
                        timestamp,
                    }],
                    exemplars: vec![],
+                    histograms: vec![],
                }),
                MetricType::HISTOGRAM => {
                    let h = m.get_histogram();
@@ -97,6 +99,7 @@ pub fn convert_metric_to_write_request(
                                timestamp,
                            }],
                            exemplars: vec![],
+                            histograms: vec![],
                        });
                        if upper_bound.is_sign_positive() && upper_bound.is_infinite() {
                            inf_seen = true;
@@ -114,6 +117,7 @@ pub fn convert_metric_to_write_request(
                                timestamp,
                            }],
                            exemplars: vec![],
+                            histograms: vec![],
                        });
                    }
                    timeseries.push(TimeSeries {
@@ -127,6 +131,7 @@ pub fn convert_metric_to_write_request(
                            timestamp,
                        }],
                        exemplars: vec![],
+                        histograms: vec![],
                    });
                    timeseries.push(TimeSeries {
                        labels: convert_label(
@@ -139,6 +144,7 @@ pub fn convert_metric_to_write_request(
                            timestamp,
                        }],
                        exemplars: vec![],
+                        histograms: vec![],
                    });
                }
                MetricType::SUMMARY => {
@@ -155,6 +161,7 @@ pub fn convert_metric_to_write_request(
                                timestamp,
                            }],
                            exemplars: vec![],
+                            histograms: vec![],
                        });
                    }
                    timeseries.push(TimeSeries {
@@ -168,6 +175,7 @@ pub fn convert_metric_to_write_request(
                            timestamp,
                        }],
                        exemplars: vec![],
+                        histograms: vec![],
                    });
                    timeseries.push(TimeSeries {
                        labels: convert_label(
@@ -180,6 +188,7 @@ pub fn convert_metric_to_write_request(
                            timestamp,
                        }],
                        exemplars: vec![],
+                        histograms: vec![],
                    });
                }
                MetricType::UNTYPED => {
@@ -274,7 +283,7 @@ mod test {

        assert_eq!(
            format!("{:?}", write_quest.timeseries),
-            r#"[TimeSeries { labels: [Label { name: "__name__", value: "test_counter" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }]"#
+            r#"[TimeSeries { labels: [Label { name: "__name__", value: "test_counter" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }]"#
        );

        let gauge_opts = Opts::new("test_gauge", "test help")
@@ -288,7 +297,7 @@ mod test {
        let write_quest = convert_metric_to_write_request(mf, None, 0);
        assert_eq!(
            format!("{:?}", write_quest.timeseries),
-            r#"[TimeSeries { labels: [Label { name: "__name__", value: "test_gauge" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 42.0, timestamp: 0 }], exemplars: [] }]"#
+            r#"[TimeSeries { labels: [Label { name: "__name__", value: "test_gauge" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 42.0, timestamp: 0 }], exemplars: [], histograms: [] }]"#
        );
    }

@@ -305,20 +314,20 @@ mod test {
            .iter()
            .map(|x| format!("{:?}", x))
            .collect();
-        let ans = r#"TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.005" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.01" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.025" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.05" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.1" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.25" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.5" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "1" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "2.5" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "5" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "10" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "+Inf" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_sum" }, Label { name: "a", value: "1" }], samples: [Sample { value: 0.25, timestamp: 0 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_count" }, Label { name: "a", value: "1" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }"#;
+        let ans = r#"TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.005" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.01" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.025" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.05" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.1" }], samples: [Sample { value: 0.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.25" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "0.5" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "1" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "2.5" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "5" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "10" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_bucket" }, Label { name: "a", value: "1" }, Label { name: "le", value: "+Inf" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_sum" }, Label { name: "a", value: "1" }], samples: [Sample { value: 0.25, timestamp: 0 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_count" }, Label { name: "a", value: "1" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }"#;
        assert_eq!(write_quest_str.join("\n"), ans);
    }

@@ -355,10 +364,10 @@ TimeSeries { labels: [Label { name: "__name__", value: "test_histogram_count" },
            .iter()
            .map(|x| format!("{:?}", x))
            .collect();
-        let ans = r#"TimeSeries { labels: [Label { name: "__name__", value: "test_summary" }, Label { name: "quantile", value: "50" }], samples: [Sample { value: 3.0, timestamp: 20 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_summary" }, Label { name: "quantile", value: "100" }], samples: [Sample { value: 5.0, timestamp: 20 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_summary_sum" }], samples: [Sample { value: 15.0, timestamp: 20 }], exemplars: [] }
-TimeSeries { labels: [Label { name: "__name__", value: "test_summary_count" }], samples: [Sample { value: 5.0, timestamp: 20 }], exemplars: [] }"#;
+        let ans = r#"TimeSeries { labels: [Label { name: "__name__", value: "test_summary" }, Label { name: "quantile", value: "50" }], samples: [Sample { value: 3.0, timestamp: 20 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_summary" }, Label { name: "quantile", value: "100" }], samples: [Sample { value: 5.0, timestamp: 20 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_summary_sum" }], samples: [Sample { value: 15.0, timestamp: 20 }], exemplars: [], histograms: [] }
+TimeSeries { labels: [Label { name: "__name__", value: "test_summary_count" }], samples: [Sample { value: 5.0, timestamp: 20 }], exemplars: [], histograms: [] }"#;
        assert_eq!(write_quest_str.join("\n"), ans);
    }

@@ -385,11 +394,11 @@ TimeSeries { labels: [Label { name: "__name__", value: "test_summary_count" }],
        let write_quest2 = convert_metric_to_write_request(mf, Some(&filter), 0);
        assert_eq!(
            format!("{:?}", write_quest1.timeseries),
-            r#"[TimeSeries { labels: [Label { name: "__name__", value: "filter_counter" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [] }, TimeSeries { labels: [Label { name: "__name__", value: "test_counter" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 2.0, timestamp: 0 }], exemplars: [] }]"#
+            r#"[TimeSeries { labels: [Label { name: "__name__", value: "filter_counter" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 1.0, timestamp: 0 }], exemplars: [], histograms: [] }, TimeSeries { labels: [Label { name: "__name__", value: "test_counter" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 2.0, timestamp: 0 }], exemplars: [], histograms: [] }]"#
        );
        assert_eq!(
            format!("{:?}", write_quest2.timeseries),
-            r#"[TimeSeries { labels: [Label { name: "__name__", value: "test_counter" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 2.0, timestamp: 0 }], exemplars: [] }]"#
+            r#"[TimeSeries { labels: [Label { name: "__name__", value: "test_counter" }, Label { name: "a", value: "1" }, Label { name: "b", value: "2" }], samples: [Sample { value: 2.0, timestamp: 0 }], exemplars: [], histograms: [] }]"#
        );
    }
 }
--- a/src/common/wal/src/config.rs
+++ b/src/common/wal/src/config.rs
@@ -206,6 +206,8 @@ mod tests {
                    client_cert_path: None,
                    client_key_path: None,
                }),
+                connect_timeout: Duration::from_secs(3),
+                timeout: Duration::from_secs(3),
            },
            kafka_topic: KafkaTopicConfig {
                num_topics: 32,
@@ -239,6 +241,8 @@ mod tests {
                    client_cert_path: None,
                    client_key_path: None,
                }),
+                connect_timeout: Duration::from_secs(3),
+                timeout: Duration::from_secs(3),
            },
            max_batch_bytes: ReadableSize::mb(1),
            consumer_wait_timeout: Duration::from_millis(100),
--- a/src/common/wal/src/config/kafka/common.rs
+++ b/src/common/wal/src/config/kafka/common.rs
@@ -36,9 +36,6 @@ pub const DEFAULT_BACKOFF_CONFIG: BackoffConfig = BackoffConfig {
    deadline: Some(Duration::from_secs(3)),
 };

-/// The default connect timeout for kafka client.
-pub const DEFAULT_CONNECT_TIMEOUT: Duration = Duration::from_secs(10);
-
 /// Default interval for auto WAL pruning.
 pub const DEFAULT_AUTO_PRUNE_INTERVAL: Duration = Duration::from_mins(30);
 /// Default limit for concurrent auto pruning tasks.
@@ -167,6 +164,12 @@ pub struct KafkaConnectionConfig {
    pub sasl: Option<KafkaClientSasl>,
    /// Client TLS config
    pub tls: Option<KafkaClientTls>,
+    /// The connect timeout for kafka client.
+    #[serde(with = "humantime_serde")]
+    pub connect_timeout: Duration,
+    /// The timeout for kafka client.
+    #[serde(with = "humantime_serde")]
+    pub timeout: Duration,
 }

 impl Default for KafkaConnectionConfig {
@@ -175,6 +178,8 @@ impl Default for KafkaConnectionConfig {
            broker_endpoints: vec![BROKER_ENDPOINT.to_string()],
            sasl: None,
            tls: None,
+            connect_timeout: Duration::from_secs(3),
+            timeout: Duration::from_secs(3),
        }
    }
 }
--- a/src/datatypes/src/schema.rs
+++ b/src/datatypes/src/schema.rs
@@ -33,7 +33,8 @@ pub use crate::schema::column_schema::{
    COLUMN_SKIPPING_INDEX_OPT_KEY_FALSE_POSITIVE_RATE, COLUMN_SKIPPING_INDEX_OPT_KEY_GRANULARITY,
    COLUMN_SKIPPING_INDEX_OPT_KEY_TYPE, COMMENT_KEY, ColumnExtType, ColumnSchema, FULLTEXT_KEY,
    FulltextAnalyzer, FulltextBackend, FulltextOptions, INVERTED_INDEX_KEY, Metadata,
-    SKIPPING_INDEX_KEY, SkippingIndexOptions, SkippingIndexType, TIME_INDEX_KEY,
+    SKIPPING_INDEX_KEY, SkippingIndexOptions, SkippingIndexType, TIME_INDEX_KEY, VECTOR_INDEX_KEY,
+    VectorDistanceMetric, VectorIndexEngineType, VectorIndexOptions,
 };
 pub use crate::schema::constraint::ColumnDefaultConstraint;
 pub use crate::schema::raw::RawSchema;
--- a/src/datatypes/src/schema/column_schema.rs
+++ b/src/datatypes/src/schema/column_schema.rs
@@ -46,6 +46,8 @@ pub const FULLTEXT_KEY: &str = "greptime:fulltext";
 pub const INVERTED_INDEX_KEY: &str = "greptime:inverted_index";
 /// Key used to store skip options in arrow field's metadata.
 pub const SKIPPING_INDEX_KEY: &str = "greptime:skipping_index";
+/// Key used to store vector index options in arrow field's metadata.
+pub const VECTOR_INDEX_KEY: &str = "greptime:vector_index";

 /// Keys used in fulltext options
 pub const COLUMN_FULLTEXT_CHANGE_OPT_KEY_ENABLE: &str = "enable";
@@ -216,6 +218,53 @@ impl ColumnSchema {
        self.metadata.contains_key(INVERTED_INDEX_KEY)
    }

+    /// Checks if this column has a vector index.
+    pub fn is_vector_indexed(&self) -> bool {
+        match self.vector_index_options() {
+            Ok(opts) => opts.is_some(),
+            Err(e) => {
+                common_telemetry::warn!(
+                    "Failed to deserialize vector_index_options for column '{}': {}",
+                    self.name,
+                    e
+                );
+                false
+            }
+        }
+    }
+
+    /// Gets the vector index options.
+    pub fn vector_index_options(&self) -> Result<Option<VectorIndexOptions>> {
+        match self.metadata.get(VECTOR_INDEX_KEY) {
+            None => Ok(None),
+            Some(json) => {
+                let options =
+                    serde_json::from_str(json).context(error::DeserializeSnafu { json })?;
+                Ok(Some(options))
+            }
+        }
+    }
+
+    /// Sets the vector index options.
+    pub fn set_vector_index_options(&mut self, options: &VectorIndexOptions) -> Result<()> {
+        self.metadata.insert(
+            VECTOR_INDEX_KEY.to_string(),
+            serde_json::to_string(options).context(error::SerializeSnafu)?,
+        );
+        Ok(())
+    }
+
+    /// Removes the vector index options.
+    pub fn unset_vector_index_options(&mut self) {
+        self.metadata.remove(VECTOR_INDEX_KEY);
+    }
+
+    /// Sets vector index options and returns self for chaining.
+    pub fn with_vector_index_options(mut self, options: &VectorIndexOptions) -> Result<Self> {
+        self.set_vector_index_options(options)?;
+        Ok(self)
+    }
+
    /// Set default constraint.
    ///
    /// If a default constraint exists for the column, this method will
@@ -964,6 +1013,181 @@ impl TryFrom<HashMap<String, String>> for SkippingIndexOptions {
    }
 }

+/// Distance metric for vector similarity search.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, Default, Visit, VisitMut)]
+#[serde(rename_all = "lowercase")]
+pub enum VectorDistanceMetric {
+    /// Squared Euclidean distance (L2^2).
+    #[default]
+    L2sq,
+    /// Cosine distance (1 - cosine similarity).
+    Cosine,
+    /// Inner product (negative, for maximum inner product search).
+    #[serde(alias = "ip")]
+    InnerProduct,
+}
+
+impl fmt::Display for VectorDistanceMetric {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            VectorDistanceMetric::L2sq => write!(f, "l2sq"),
+            VectorDistanceMetric::Cosine => write!(f, "cosine"),
+            VectorDistanceMetric::InnerProduct => write!(f, "ip"),
+        }
+    }
+}
+
+impl std::str::FromStr for VectorDistanceMetric {
+    type Err = String;
+
+    fn from_str(s: &str) -> std::result::Result<Self, Self::Err> {
+        match s.to_lowercase().as_str() {
+            "l2sq" | "l2" | "euclidean" => Ok(VectorDistanceMetric::L2sq),
+            "cosine" | "cos" => Ok(VectorDistanceMetric::Cosine),
+            "inner_product" | "ip" | "dot" => Ok(VectorDistanceMetric::InnerProduct),
+            _ => Err(format!(
+                "Unknown distance metric: {}. Expected: l2sq, cosine, or ip",
+                s
+            )),
+        }
+    }
+}
+
+impl VectorDistanceMetric {
+    /// Returns the metric as u8 for blob serialization.
+    pub fn as_u8(&self) -> u8 {
+        match self {
+            Self::L2sq => 0,
+            Self::Cosine => 1,
+            Self::InnerProduct => 2,
+        }
+    }
+
+    /// Parses metric from u8 (used when reading blob).
+    pub fn try_from_u8(v: u8) -> Option<Self> {
+        match v {
+            0 => Some(Self::L2sq),
+            1 => Some(Self::Cosine),
+            2 => Some(Self::InnerProduct),
+            _ => None,
+        }
+    }
+}
+
+/// Default HNSW connectivity parameter.
+const DEFAULT_VECTOR_INDEX_CONNECTIVITY: u32 = 16;
+/// Default expansion factor during index construction.
+const DEFAULT_VECTOR_INDEX_EXPANSION_ADD: u32 = 128;
+/// Default expansion factor during search.
+const DEFAULT_VECTOR_INDEX_EXPANSION_SEARCH: u32 = 64;
+
+fn default_vector_index_connectivity() -> u32 {
+    DEFAULT_VECTOR_INDEX_CONNECTIVITY
+}
+
+fn default_vector_index_expansion_add() -> u32 {
+    DEFAULT_VECTOR_INDEX_EXPANSION_ADD
+}
+
+fn default_vector_index_expansion_search() -> u32 {
+    DEFAULT_VECTOR_INDEX_EXPANSION_SEARCH
+}
+
+/// Supported vector index engine types.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default, Serialize, Deserialize, Visit, VisitMut)]
+#[serde(rename_all = "lowercase")]
+pub enum VectorIndexEngineType {
+    /// USearch HNSW implementation.
+    #[default]
+    Usearch,
+    // Future: Vsag,
+}
+
+impl VectorIndexEngineType {
+    /// Returns the engine type as u8 for blob serialization.
+    pub fn as_u8(&self) -> u8 {
+        match self {
+            Self::Usearch => 0,
+        }
+    }
+
+    /// Parses engine type from u8 (used when reading blob).
+    pub fn try_from_u8(v: u8) -> Option<Self> {
+        match v {
+            0 => Some(Self::Usearch),
+            _ => None,
+        }
+    }
+}
+
+impl fmt::Display for VectorIndexEngineType {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            Self::Usearch => write!(f, "usearch"),
+        }
+    }
+}
+
+impl std::str::FromStr for VectorIndexEngineType {
+    type Err = String;
+
+    fn from_str(s: &str) -> std::result::Result<Self, Self::Err> {
+        match s.to_lowercase().as_str() {
+            "usearch" => Ok(Self::Usearch),
+            _ => Err(format!(
+                "Unknown vector index engine: {}. Expected: usearch",
+                s
+            )),
+        }
+    }
+}
+
+/// Options for vector index (HNSW).
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Visit, VisitMut)]
+#[serde(rename_all = "kebab-case")]
+pub struct VectorIndexOptions {
+    /// Vector index engine type (default: usearch).
+    #[serde(default)]
+    pub engine: VectorIndexEngineType,
+    /// Distance metric for similarity search.
+    #[serde(default)]
+    pub metric: VectorDistanceMetric,
+    /// HNSW connectivity parameter (M in the paper).
+    /// Higher values improve recall but increase memory usage.
+    #[serde(default = "default_vector_index_connectivity")]
+    pub connectivity: u32,
+    /// Expansion factor during index construction (ef_construction).
+    /// Higher values improve index quality but slow down construction.
+    #[serde(default = "default_vector_index_expansion_add")]
+    pub expansion_add: u32,
+    /// Expansion factor during search (ef_search).
+    /// Higher values improve recall but slow down search.
+    #[serde(default = "default_vector_index_expansion_search")]
+    pub expansion_search: u32,
+}
+
+impl Default for VectorIndexOptions {
+    fn default() -> Self {
+        Self {
+            engine: VectorIndexEngineType::default(),
+            metric: VectorDistanceMetric::default(),
+            connectivity: DEFAULT_VECTOR_INDEX_CONNECTIVITY,
+            expansion_add: DEFAULT_VECTOR_INDEX_EXPANSION_ADD,
+            expansion_search: DEFAULT_VECTOR_INDEX_EXPANSION_SEARCH,
+        }
+    }
+}
+
+impl fmt::Display for VectorIndexOptions {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(
+            f,
+            "engine={}, metric={}, connectivity={}, expansion_add={}, expansion_search={}",
+            self.engine, self.metric, self.connectivity, self.expansion_add, self.expansion_search
+        )
+    }
+}
+
 #[cfg(test)]
 mod tests {
    use std::sync::Arc;
--- a/src/flow/src/batching_mode/frontend_client.rs
+++ b/src/flow/src/batching_mode/frontend_client.rs
@@ -15,7 +15,7 @@
 //! Frontend client to run flow as batching task which is time-window-aware normal query triggered every tick set by user

 use std::collections::HashMap;
-use std::sync::{Arc, Weak};
+use std::sync::{Arc, Mutex, Weak};
 use std::time::SystemTime;

 use api::v1::greptime_request::Request;
@@ -38,6 +38,7 @@ use servers::query_handler::grpc::GrpcQueryHandler;
 use session::context::{QueryContextBuilder, QueryContextRef};
 use session::hints::READ_PREFERENCE_HINT;
 use snafu::{OptionExt, ResultExt};
+use tokio::sync::SetOnce;

 use crate::batching_mode::BatchingModeOptions;
 use crate::error::{
@@ -75,7 +76,19 @@ impl<E: ErrorExt + Send + Sync + 'static, T: GrpcQueryHandler<Error = E> + Send
    }
 }

-type HandlerMutable = Arc<std::sync::Mutex<Option<Weak<dyn GrpcQueryHandlerWithBoxedError>>>>;
+#[derive(Debug, Clone)]
+pub struct HandlerMutable {
+    handler: Arc<Mutex<Option<Weak<dyn GrpcQueryHandlerWithBoxedError>>>>,
+    is_initialized: Arc<SetOnce<()>>,
+}
+
+impl HandlerMutable {
+    pub async fn set_handler(&self, handler: Weak<dyn GrpcQueryHandlerWithBoxedError>) {
+        *self.handler.lock().unwrap() = Some(handler);
+        // Ignore the error, as we allow the handler to be set multiple times.
+        let _ = self.is_initialized.set(());
+    }
+}

 /// A simple frontend client able to execute sql using grpc protocol
 ///
@@ -100,7 +113,11 @@ pub enum FrontendClient {
 impl FrontendClient {
    /// Create a new empty frontend client, with a `HandlerMutable` to set the grpc handler later
    pub fn from_empty_grpc_handler(query: QueryOptions) -> (Self, HandlerMutable) {
-        let handler = Arc::new(std::sync::Mutex::new(None));
+        let is_initialized = Arc::new(SetOnce::new());
+        let handler = HandlerMutable {
+            handler: Arc::new(Mutex::new(None)),
+            is_initialized,
+        };
        (
            Self::Standalone {
                database_client: handler.clone(),
@@ -110,23 +127,13 @@ impl FrontendClient {
        )
    }

-    /// Check if the frontend client is initialized.
-    ///
-    /// In distributed mode, it is always initialized.
-    /// In standalone mode, it checks if the database client is set.
-    pub fn is_initialized(&self) -> bool {
-        match self {
-            FrontendClient::Distributed { .. } => true,
-            FrontendClient::Standalone {
-                database_client, ..
-            } => {
-                let guard = database_client.lock();
-                if let Ok(guard) = guard {
-                    guard.is_some()
-                } else {
-                    false
-                }
-            }
+    /// Waits until the frontend client is initialized.
+    pub async fn wait_initialized(&self) {
+        if let FrontendClient::Standalone {
+            database_client, ..
+        } = self
+        {
+            database_client.is_initialized.wait().await;
        }
    }

@@ -158,8 +165,14 @@ impl FrontendClient {
        grpc_handler: Weak<dyn GrpcQueryHandlerWithBoxedError>,
        query: QueryOptions,
    ) -> Self {
+        let is_initialized = Arc::new(SetOnce::new_with(Some(())));
+        let handler = HandlerMutable {
+            handler: Arc::new(Mutex::new(Some(grpc_handler))),
+            is_initialized: is_initialized.clone(),
+        };
+
        Self::Standalone {
-            database_client: Arc::new(std::sync::Mutex::new(Some(grpc_handler))),
+            database_client: handler,
            query,
        }
    }
@@ -341,6 +354,7 @@ impl FrontendClient {
                {
                    let database_client = {
                        database_client
+                            .handler
                            .lock()
                            .map_err(|e| {
                                UnexpectedSnafu {
@@ -418,6 +432,7 @@ impl FrontendClient {
                {
                    let database_client = {
                        database_client
+                            .handler
                            .lock()
                            .map_err(|e| {
                                UnexpectedSnafu {
@@ -480,3 +495,73 @@ impl std::fmt::Display for PeerDesc {
        }
    }
 }
+
+#[cfg(test)]
+mod tests {
+    use std::time::Duration;
+
+    use common_query::Output;
+    use tokio::time::timeout;
+
+    use super::*;
+
+    #[derive(Debug)]
+    struct NoopHandler;
+
+    #[async_trait::async_trait]
+    impl GrpcQueryHandlerWithBoxedError for NoopHandler {
+        async fn do_query(
+            &self,
+            _query: Request,
+            _ctx: QueryContextRef,
+        ) -> std::result::Result<Output, BoxedError> {
+            Ok(Output::new_with_affected_rows(0))
+        }
+    }
+
+    #[tokio::test]
+    async fn wait_initialized() {
+        let (client, handler_mut) =
+            FrontendClient::from_empty_grpc_handler(QueryOptions::default());
+
+        assert!(
+            timeout(Duration::from_millis(50), client.wait_initialized())
+                .await
+                .is_err()
+        );
+
+        let handler: Arc<dyn GrpcQueryHandlerWithBoxedError> = Arc::new(NoopHandler);
+        handler_mut.set_handler(Arc::downgrade(&handler)).await;
+
+        timeout(Duration::from_secs(1), client.wait_initialized())
+            .await
+            .expect("wait_initialized should complete after handler is set");
+
+        timeout(Duration::from_millis(10), client.wait_initialized())
+            .await
+            .expect("wait_initialized should be a no-op once initialized");
+
+        let handler: Arc<dyn GrpcQueryHandlerWithBoxedError> = Arc::new(NoopHandler);
+        let client =
+            FrontendClient::from_grpc_handler(Arc::downgrade(&handler), QueryOptions::default());
+        assert!(
+            timeout(Duration::from_millis(10), client.wait_initialized())
+                .await
+                .is_ok()
+        );
+
+        let meta_client = Arc::new(MetaClient::default());
+        let client = FrontendClient::from_meta_client(
+            meta_client,
+            None,
+            QueryOptions::default(),
+            BatchingModeOptions::default(),
+        )
+        .unwrap();
+        assert!(
+            timeout(Duration::from_millis(10), client.wait_initialized())
+                .await
+                .is_ok()
+        );
+    }
+}
--- a/src/frontend/src/frontend.rs
+++ b/src/frontend/src/frontend.rs
@@ -157,7 +157,6 @@ mod tests {
    use common_error::from_header_to_err_code_msg;
    use common_error::status_code::StatusCode;
    use common_grpc::channel_manager::ChannelManager;
-    use common_meta::distributed_time_constants::FRONTEND_HEARTBEAT_INTERVAL_MILLIS;
    use common_meta::heartbeat::handler::HandlerGroupExecutor;
    use common_meta::heartbeat::handler::parse_mailbox_message::ParseMailboxMessageHandler;
    use common_meta::heartbeat::handler::suspend::SuspendHandler;
@@ -400,6 +399,10 @@ mod tests {
                ..Default::default()
            },
            meta_client: Some(meta_client_options.clone()),
+            heartbeat: HeartbeatOptions {
+                interval: Duration::from_secs(1),
+                ..Default::default()
+            },
            ..Default::default()
        };

@@ -409,7 +412,8 @@ mod tests {
        let meta_client = create_meta_client(&meta_client_options, server.clone()).await;
        let frontend = create_frontend(&options, meta_client).await?;

-        tokio::time::sleep(Duration::from_millis(FRONTEND_HEARTBEAT_INTERVAL_MILLIS)).await;
+        let frontend_heartbeat_interval = options.heartbeat.interval;
+        tokio::time::sleep(frontend_heartbeat_interval).await;
        // initial state: not suspend:
        assert!(!frontend.instance.is_suspended());
        verify_suspend_state_by_http(&frontend, Ok(r#"[{"records":{"schema":{"column_schemas":[{"name":"Int64(1)","data_type":"Int64"}]},"rows":[[1]],"total_rows":1}}]"#)).await;
@@ -426,7 +430,7 @@ mod tests {

        // make heartbeat server returned "suspend" instruction,
        server.suspend.store(true, Ordering::Relaxed);
-        tokio::time::sleep(Duration::from_millis(FRONTEND_HEARTBEAT_INTERVAL_MILLIS)).await;
+        tokio::time::sleep(frontend_heartbeat_interval).await;
        // ... then the frontend is suspended:
        assert!(frontend.instance.is_suspended());
        verify_suspend_state_by_http(
@@ -442,7 +446,7 @@ mod tests {

        // make heartbeat server NOT returned "suspend" instruction,
        server.suspend.store(false, Ordering::Relaxed);
-        tokio::time::sleep(Duration::from_millis(FRONTEND_HEARTBEAT_INTERVAL_MILLIS)).await;
+        tokio::time::sleep(frontend_heartbeat_interval).await;
        // ... then frontend's suspend state is cleared:
        assert!(!frontend.instance.is_suspended());
        verify_suspend_state_by_http(&frontend, Ok(r#"[{"records":{"schema":{"column_schemas":[{"name":"Int64(1)","data_type":"Int64"}]},"rows":[[1]],"total_rows":1}}]"#)).await;
--- a/src/index/Cargo.toml
+++ b/src/index/Cargo.toml
@@ -7,6 +7,9 @@ license.workspace = true
 [lints]
 workspace = true

+[features]
+vector_index = ["dep:usearch"]
+
 [dependencies]
 async-trait.workspace = true
 asynchronous-codec = "0.7.0"
@@ -17,6 +20,7 @@ common-error.workspace = true
 common-macro.workspace = true
 common-runtime.workspace = true
 common-telemetry.workspace = true
+datatypes.workspace = true
 fastbloom = "0.8"
 fst.workspace = true
 futures.workspace = true
@@ -25,6 +29,7 @@ itertools.workspace = true
 jieba-rs = "0.8"
 lazy_static.workspace = true
 mockall.workspace = true
+nalgebra.workspace = true
 pin-project.workspace = true
 prost.workspace = true
 puffin.workspace = true
@@ -39,6 +44,7 @@ tantivy = { version = "0.24", features = ["zstd-compression"] }
 tantivy-jieba = "0.16"
 tokio.workspace = true
 tokio-util.workspace = true
+usearch = { version = "2.21", default-features = false, features = ["fp16lib"], optional = true }
 uuid.workspace = true

 [dev-dependencies]
--- a/src/index/src/lib.rs
+++ b/src/index/src/lib.rs
@@ -22,6 +22,8 @@ pub mod external_provider;
 pub mod fulltext_index;
 pub mod inverted_index;
 pub mod target;
+#[cfg(feature = "vector_index")]
+pub mod vector;

 pub type Bytes = Vec<u8>;
 pub type BytesRef<'a> = &'a [u8];
--- a/src/index/src/vector.rs
+++ b/src/index/src/vector.rs
@@ -0,0 +1,163 @@
+// 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.
+
+//! Vector index types and options.
+//!
+//! This module re-exports types from `datatypes` and provides conversions
+//! to USearch types, as well as distance computation functions.
+
+pub use datatypes::schema::{VectorDistanceMetric, VectorIndexOptions};
+use nalgebra::DVectorView;
+pub use usearch::MetricKind;
+
+/// Converts a VectorDistanceMetric to a USearch MetricKind.
+pub fn distance_metric_to_usearch(metric: VectorDistanceMetric) -> MetricKind {
+    match metric {
+        VectorDistanceMetric::L2sq => MetricKind::L2sq,
+        VectorDistanceMetric::Cosine => MetricKind::Cos,
+        VectorDistanceMetric::InnerProduct => MetricKind::IP,
+    }
+}
+
+/// Computes distance between two vectors using the specified metric.
+///
+/// Uses SIMD-optimized implementations via nalgebra.
+///
+/// **Note:** The caller must ensure that the two vectors have the same length
+/// and are non-empty. Empty vectors return 0.0 for all metrics.
+pub fn compute_distance(v1: &[f32], v2: &[f32], metric: VectorDistanceMetric) -> f32 {
+    // Empty vectors are degenerate; return 0.0 uniformly across all metrics.
+    if v1.is_empty() || v2.is_empty() {
+        return 0.0;
+    }
+
+    match metric {
+        VectorDistanceMetric::L2sq => l2sq(v1, v2),
+        VectorDistanceMetric::Cosine => cosine(v1, v2),
+        VectorDistanceMetric::InnerProduct => -dot(v1, v2),
+    }
+}
+
+/// Calculates the squared L2 distance between two vectors.
+fn l2sq(lhs: &[f32], rhs: &[f32]) -> f32 {
+    let lhs = DVectorView::from_slice(lhs, lhs.len());
+    let rhs = DVectorView::from_slice(rhs, rhs.len());
+    (lhs - rhs).norm_squared()
+}
+
+/// Calculates the cosine distance between two vectors.
+///
+/// Returns a value in `[0.0, 2.0]` where 0.0 means identical direction and 2.0 means
+/// opposite direction. For degenerate cases (zero or near-zero magnitude vectors),
+/// returns 1.0 (maximum uncertainty) to avoid NaN and ensure safe index operations.
+fn cosine(lhs: &[f32], rhs: &[f32]) -> f32 {
+    let lhs_vec = DVectorView::from_slice(lhs, lhs.len());
+    let rhs_vec = DVectorView::from_slice(rhs, rhs.len());
+
+    let dot_product = lhs_vec.dot(&rhs_vec);
+    let lhs_norm = lhs_vec.norm();
+    let rhs_norm = rhs_vec.norm();
+
+    // Zero-magnitude vectors have undefined direction; return max distance as safe fallback.
+    if dot_product.abs() < f32::EPSILON
+        || lhs_norm.abs() < f32::EPSILON
+        || rhs_norm.abs() < f32::EPSILON
+    {
+        return 1.0;
+    }
+
+    let cos_similar = dot_product / (lhs_norm * rhs_norm);
+    let res = 1.0 - cos_similar;
+    // Clamp near-zero results to exactly 0.0 to avoid floating-point artifacts.
+    if res.abs() < f32::EPSILON { 0.0 } else { res }
+}
+
+/// Calculates the dot product between two vectors.
+fn dot(lhs: &[f32], rhs: &[f32]) -> f32 {
+    let lhs = DVectorView::from_slice(lhs, lhs.len());
+    let rhs = DVectorView::from_slice(rhs, rhs.len());
+    lhs.dot(&rhs)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_distance_metric_to_usearch() {
+        assert_eq!(
+            distance_metric_to_usearch(VectorDistanceMetric::L2sq),
+            MetricKind::L2sq
+        );
+        assert_eq!(
+            distance_metric_to_usearch(VectorDistanceMetric::Cosine),
+            MetricKind::Cos
+        );
+        assert_eq!(
+            distance_metric_to_usearch(VectorDistanceMetric::InnerProduct),
+            MetricKind::IP
+        );
+    }
+
+    #[test]
+    fn test_vector_index_options_default() {
+        let options = VectorIndexOptions::default();
+        assert_eq!(options.metric, VectorDistanceMetric::L2sq);
+        assert_eq!(options.connectivity, 16);
+        assert_eq!(options.expansion_add, 128);
+        assert_eq!(options.expansion_search, 64);
+    }
+
+    #[test]
+    fn test_compute_distance_l2sq() {
+        let v1 = vec![1.0, 2.0, 3.0];
+        let v2 = vec![4.0, 5.0, 6.0];
+        // L2sq = (4-1)^2 + (5-2)^2 + (6-3)^2 = 9 + 9 + 9 = 27
+        let dist = compute_distance(&v1, &v2, VectorDistanceMetric::L2sq);
+        assert!((dist - 27.0).abs() < 1e-6);
+    }
+
+    #[test]
+    fn test_compute_distance_cosine() {
+        let v1 = vec![1.0, 0.0, 0.0];
+        let v2 = vec![0.0, 1.0, 0.0];
+        // Orthogonal vectors have cosine similarity of 0, distance of 1
+        let dist = compute_distance(&v1, &v2, VectorDistanceMetric::Cosine);
+        assert!((dist - 1.0).abs() < 1e-6);
+    }
+
+    #[test]
+    fn test_compute_distance_inner_product() {
+        let v1 = vec![1.0, 2.0, 3.0];
+        let v2 = vec![4.0, 5.0, 6.0];
+        // Inner product = 1*4 + 2*5 + 3*6 = 4 + 10 + 18 = 32
+        // Distance is negated: -32
+        let dist = compute_distance(&v1, &v2, VectorDistanceMetric::InnerProduct);
+        assert!((dist - (-32.0)).abs() < 1e-6);
+    }
+
+    #[test]
+    fn test_compute_distance_empty_vectors() {
+        // Empty vectors should return 0.0 uniformly for all metrics
+        assert_eq!(compute_distance(&[], &[], VectorDistanceMetric::L2sq), 0.0);
+        assert_eq!(
+            compute_distance(&[], &[], VectorDistanceMetric::Cosine),
+            0.0
+        );
+        assert_eq!(
+            compute_distance(&[], &[], VectorDistanceMetric::InnerProduct),
+            0.0
+        );
+    }
+}
--- a/src/log-store/src/kafka/client_manager.rs
+++ b/src/log-store/src/kafka/client_manager.rs
@@ -16,7 +16,7 @@ use std::collections::HashMap;
 use std::sync::Arc;

 use common_wal::config::kafka::DatanodeKafkaConfig;
-use common_wal::config::kafka::common::{DEFAULT_BACKOFF_CONFIG, DEFAULT_CONNECT_TIMEOUT};
+use common_wal::config::kafka::common::DEFAULT_BACKOFF_CONFIG;
 use dashmap::DashMap;
 use rskafka::client::ClientBuilder;
 use rskafka::client::partition::{Compression, PartitionClient, UnknownTopicHandling};
@@ -79,7 +79,8 @@ impl ClientManager {
        // Sets backoff config for the top-level kafka client and all clients constructed by it.
        let mut builder = ClientBuilder::new(config.connection.broker_endpoints.clone())
            .backoff_config(DEFAULT_BACKOFF_CONFIG)
-            .connect_timeout(Some(DEFAULT_CONNECT_TIMEOUT));
+            .connect_timeout(Some(config.connection.connect_timeout))
+            .timeout(Some(config.connection.timeout));
        if let Some(sasl) = &config.connection.sasl {
            builder = builder.sasl_config(sasl.config.clone().into_sasl_config());
        };
--- a/src/meta-srv/src/bootstrap.rs
+++ b/src/meta-srv/src/bootstrap.rs
@@ -14,7 +14,6 @@

 use std::net::SocketAddr;
 use std::sync::Arc;
-use std::time::Duration;

 use api::v1::meta::cluster_server::ClusterServer;
 use api::v1::meta::heartbeat_server::HeartbeatServer;
@@ -60,11 +59,6 @@ use crate::service::admin::admin_axum_router;
 use crate::utils::etcd::create_etcd_client_with_tls;
 use crate::{Result, error};

-/// The default keep-alive interval for gRPC.
-const DEFAULT_GRPC_KEEP_ALIVE_INTERVAL: Duration = Duration::from_secs(10);
-/// The default keep-alive timeout for gRPC.
-const DEFAULT_GRPC_KEEP_ALIVE_TIMEOUT: Duration = Duration::from_secs(10);
-
 pub struct MetasrvInstance {
    metasrv: Arc<Metasrv>,

@@ -255,8 +249,8 @@ pub fn router(metasrv: Arc<Metasrv>) -> Router {
        // for admin services
        .accept_http1(true)
        // For quick network failures detection.
-        .http2_keepalive_interval(Some(DEFAULT_GRPC_KEEP_ALIVE_INTERVAL))
-        .http2_keepalive_timeout(Some(DEFAULT_GRPC_KEEP_ALIVE_TIMEOUT));
+        .http2_keepalive_interval(Some(metasrv.options().grpc.http2_keep_alive_interval))
+        .http2_keepalive_timeout(Some(metasrv.options().grpc.http2_keep_alive_timeout));
    let router = add_compressed_service!(router, HeartbeatServer::from_arc(metasrv.clone()));
    let router = add_compressed_service!(router, StoreServer::from_arc(metasrv.clone()));
    let router = add_compressed_service!(router, ClusterServer::from_arc(metasrv.clone()));
@@ -273,8 +267,12 @@ pub async fn metasrv_builder(
        (Some(kv_backend), _) => (kv_backend, None),
        (None, BackendImpl::MemoryStore) => (Arc::new(MemoryKvBackend::new()) as _, None),
        (None, BackendImpl::EtcdStore) => {
-            let etcd_client =
-                create_etcd_client_with_tls(&opts.store_addrs, opts.backend_tls.as_ref()).await?;
+            let etcd_client = create_etcd_client_with_tls(
+                &opts.store_addrs,
+                &opts.backend_client,
+                opts.backend_tls.as_ref(),
+            )
+            .await?;
            let kv_backend = EtcdStore::with_etcd_client(etcd_client.clone(), opts.max_txn_ops);
            let election = EtcdElection::with_etcd_client(
                &opts.grpc.server_addr,
--- a/src/meta-srv/src/discovery.rs
+++ b/src/meta-srv/src/discovery.rs
@@ -16,13 +16,9 @@ pub mod lease;
 pub mod node_info;
 pub mod utils;

-use std::time::Duration;
-
 use api::v1::meta::heartbeat_request::NodeWorkloads;
 use common_error::ext::BoxedError;
-use common_meta::distributed_time_constants::{
-    DATANODE_LEASE_SECS, FLOWNODE_LEASE_SECS, FRONTEND_HEARTBEAT_INTERVAL_MILLIS,
-};
+use common_meta::distributed_time_constants::default_distributed_time_constants;
 use common_meta::error::Result;
 use common_meta::peer::{Peer, PeerDiscovery, PeerResolver};
 use common_meta::{DatanodeId, FlownodeId};
@@ -38,7 +34,7 @@ impl PeerDiscovery for MetaPeerClient {
        utils::alive_frontends(
            &DefaultSystemTimer,
            self,
-            Duration::from_millis(FRONTEND_HEARTBEAT_INTERVAL_MILLIS),
+            default_distributed_time_constants().frontend_heartbeat_interval,
        )
        .await
        .map_err(BoxedError::new)
@@ -52,7 +48,7 @@ impl PeerDiscovery for MetaPeerClient {
        utils::alive_datanodes(
            &DefaultSystemTimer,
            self,
-            Duration::from_secs(DATANODE_LEASE_SECS),
+            default_distributed_time_constants().datanode_lease,
            filter,
        )
        .await
@@ -67,7 +63,7 @@ impl PeerDiscovery for MetaPeerClient {
        utils::alive_flownodes(
            &DefaultSystemTimer,
            self,
-            Duration::from_secs(FLOWNODE_LEASE_SECS),
+            default_distributed_time_constants().flownode_lease,
            filter,
        )
        .await
--- a/src/meta-srv/src/discovery/lease.rs
+++ b/src/meta-srv/src/discovery/lease.rs
@@ -102,7 +102,7 @@ mod tests {
    use api::v1::meta::heartbeat_request::NodeWorkloads;
    use api::v1::meta::{DatanodeWorkloads, FlownodeWorkloads};
    use common_meta::cluster::{FrontendStatus, NodeInfo, NodeInfoKey, NodeStatus, Role};
-    use common_meta::distributed_time_constants::FRONTEND_HEARTBEAT_INTERVAL_MILLIS;
+    use common_meta::distributed_time_constants::default_distributed_time_constants;
    use common_meta::kv_backend::ResettableKvBackendRef;
    use common_meta::peer::{Peer, PeerDiscovery};
    use common_meta::rpc::store::PutRequest;
@@ -473,8 +473,10 @@ mod tests {
        let client = create_meta_peer_client();
        let in_memory = client.memory_backend();

+        let frontend_heartbeat_interval =
+            default_distributed_time_constants().frontend_heartbeat_interval;
        let last_activity_ts =
-            current_time_millis() - FRONTEND_HEARTBEAT_INTERVAL_MILLIS as i64 - 1000;
+            current_time_millis() - frontend_heartbeat_interval.as_millis() as i64 - 1000;
        let active_frontend_node = NodeInfo {
            peer: Peer {
                id: 0,
--- a/src/meta-srv/src/failure_detector.rs
+++ b/src/meta-srv/src/failure_detector.rs
@@ -15,7 +15,6 @@
 use std::collections::VecDeque;
 use std::time::Duration;

-use common_meta::distributed_time_constants;
 use serde::{Deserialize, Serialize};

 const FIRST_HEARTBEAT_ESTIMATE_MILLIS: i64 = 1000;
@@ -79,9 +78,7 @@ impl Default for PhiAccrualFailureDetectorOptions {
        Self {
            threshold: 8_f32,
            min_std_deviation: Duration::from_millis(100),
-            acceptable_heartbeat_pause: Duration::from_secs(
-                distributed_time_constants::DATANODE_LEASE_SECS,
-            ),
+            acceptable_heartbeat_pause: Duration::from_secs(10),
        }
    }
 }
--- a/src/meta-srv/src/handler/region_lease_handler.rs
+++ b/src/meta-srv/src/handler/region_lease_handler.rs
@@ -134,7 +134,7 @@ mod test {
    use std::sync::Arc;

    use common_meta::datanode::{RegionManifestInfo, RegionStat, Stat};
-    use common_meta::distributed_time_constants;
+    use common_meta::distributed_time_constants::default_distributed_time_constants;
    use common_meta::key::TableMetadataManager;
    use common_meta::key::table_route::TableRouteValue;
    use common_meta::key::test_utils::new_test_table_info;
@@ -236,7 +236,7 @@ mod test {
        let opening_region_keeper = Arc::new(MemoryRegionKeeper::default());

        let handler = RegionLeaseHandler::new(
-            distributed_time_constants::REGION_LEASE_SECS,
+            default_distributed_time_constants().region_lease.as_secs(),
            table_metadata_manager.clone(),
            opening_region_keeper.clone(),
            None,
@@ -266,7 +266,7 @@ mod test {

        assert_eq!(
            acc.region_lease.as_ref().unwrap().lease_seconds,
-            distributed_time_constants::REGION_LEASE_SECS
+            default_distributed_time_constants().region_lease.as_secs()
        );

        assert_region_lease(
@@ -300,7 +300,7 @@ mod test {

        assert_eq!(
            acc.region_lease.as_ref().unwrap().lease_seconds,
-            distributed_time_constants::REGION_LEASE_SECS
+            default_distributed_time_constants().region_lease.as_secs()
        );

        assert_region_lease(
@@ -379,7 +379,7 @@ mod test {
        });

        let handler = RegionLeaseHandler::new(
-            distributed_time_constants::REGION_LEASE_SECS,
+            default_distributed_time_constants().region_lease.as_secs(),
            table_metadata_manager.clone(),
            Default::default(),
            None,
@@ -461,7 +461,7 @@ mod test {
            ..Default::default()
        });
        let handler = RegionLeaseHandler::new(
-            distributed_time_constants::REGION_LEASE_SECS,
+            default_distributed_time_constants().region_lease.as_secs(),
            table_metadata_manager.clone(),
            Default::default(),
            None,
--- a/src/meta-srv/src/metasrv.rs
+++ b/src/meta-srv/src/metasrv.rs
@@ -27,7 +27,7 @@ use common_event_recorder::EventRecorderOptions;
 use common_greptimedb_telemetry::GreptimeDBTelemetryTask;
 use common_meta::cache_invalidator::CacheInvalidatorRef;
 use common_meta::ddl_manager::DdlManagerRef;
-use common_meta::distributed_time_constants;
+use common_meta::distributed_time_constants::{self, default_distributed_time_constants};
 use common_meta::key::TableMetadataManagerRef;
 use common_meta::key::runtime_switch::RuntimeSwitchManagerRef;
 use common_meta::kv_backend::{KvBackendRef, ResettableKvBackend, ResettableKvBackendRef};
@@ -121,6 +121,27 @@ impl Default for StatsPersistenceOptions {
    }
 }

+#[derive(Clone, PartialEq, Serialize, Deserialize, Debug)]
+#[serde(default)]
+pub struct BackendClientOptions {
+    #[serde(with = "humantime_serde")]
+    pub keep_alive_timeout: Duration,
+    #[serde(with = "humantime_serde")]
+    pub keep_alive_interval: Duration,
+    #[serde(with = "humantime_serde")]
+    pub connect_timeout: Duration,
+}
+
+impl Default for BackendClientOptions {
+    fn default() -> Self {
+        Self {
+            keep_alive_interval: Duration::from_secs(10),
+            keep_alive_timeout: Duration::from_secs(3),
+            connect_timeout: Duration::from_secs(3),
+        }
+    }
+}
+
 #[derive(Clone, PartialEq, Serialize, Deserialize)]
 #[serde(default)]
 pub struct MetasrvOptions {
@@ -136,12 +157,22 @@ pub struct MetasrvOptions {
    /// Only applicable when using PostgreSQL or MySQL as the metadata store
    #[serde(default)]
    pub backend_tls: Option<TlsOption>,
+    /// The backend client options.
+    /// Currently, only applicable when using etcd as the metadata store.
+    #[serde(default)]
+    pub backend_client: BackendClientOptions,
    /// The type of selector.
    pub selector: SelectorType,
    /// Whether to use the memory store.
    pub use_memory_store: bool,
    /// Whether to enable region failover.
    pub enable_region_failover: bool,
+    /// The base heartbeat interval.
+    ///
+    /// This value is used to calculate the distributed time constants for components.
+    /// e.g., the region lease time is `heartbeat_interval * 3 + Duration::from_secs(1)`.
+    #[serde(with = "humantime_serde")]
+    pub heartbeat_interval: Duration,
    /// The delay before starting region failure detection.
    /// This delay helps prevent Metasrv from triggering unnecessary region failovers before all Datanodes are fully started.
    /// Especially useful when the cluster is not deployed with GreptimeDB Operator and maintenance mode is not enabled.
@@ -240,7 +271,9 @@ impl fmt::Debug for MetasrvOptions {
            .field("tracing", &self.tracing)
            .field("backend", &self.backend)
            .field("event_recorder", &self.event_recorder)
-            .field("stats_persistence", &self.stats_persistence);
+            .field("stats_persistence", &self.stats_persistence)
+            .field("heartbeat_interval", &self.heartbeat_interval)
+            .field("backend_client", &self.backend_client);

        #[cfg(any(feature = "pg_kvbackend", feature = "mysql_kvbackend"))]
        debug_struct.field("meta_table_name", &self.meta_table_name);
@@ -270,6 +303,7 @@ impl Default for MetasrvOptions {
            selector: SelectorType::default(),
            use_memory_store: false,
            enable_region_failover: false,
+            heartbeat_interval: distributed_time_constants::BASE_HEARTBEAT_INTERVAL,
            region_failure_detector_initialization_delay: Duration::from_secs(10 * 60),
            allow_region_failover_on_local_wal: false,
            grpc: GrpcOptions {
@@ -307,6 +341,7 @@ impl Default for MetasrvOptions {
            event_recorder: EventRecorderOptions::default(),
            stats_persistence: StatsPersistenceOptions::default(),
            gc: GcSchedulerOptions::default(),
+            backend_client: BackendClientOptions::default(),
        }
    }
 }
@@ -747,7 +782,7 @@ impl Metasrv {
            &DefaultSystemTimer,
            self.meta_peer_client.as_ref(),
            peer_id,
-            Duration::from_secs(distributed_time_constants::DATANODE_LEASE_SECS),
+            default_distributed_time_constants().datanode_lease,
        )
        .await
    }
--- a/src/meta-srv/src/metasrv/builder.rs
+++ b/src/meta-srv/src/metasrv/builder.rs
@@ -29,7 +29,7 @@ use common_meta::ddl::{
    DdlContext, NoopRegionFailureDetectorControl, RegionFailureDetectorControllerRef,
 };
 use common_meta::ddl_manager::{DdlManager, DdlManagerConfiguratorRef};
-use common_meta::distributed_time_constants::{self};
+use common_meta::distributed_time_constants::default_distributed_time_constants;
 use common_meta::key::TableMetadataManager;
 use common_meta::key::flow::FlowMetadataManager;
 use common_meta::key::flow::flow_state::FlowStateManager;
@@ -513,7 +513,7 @@ impl MetasrvBuilder {
            Some(handler_group_builder) => handler_group_builder,
            None => {
                let region_lease_handler = RegionLeaseHandler::new(
-                    distributed_time_constants::REGION_LEASE_SECS,
+                    default_distributed_time_constants().region_lease.as_secs(),
                    table_metadata_manager.clone(),
                    memory_region_keeper.clone(),
                    customized_region_lease_renewer,
--- a/src/meta-srv/src/procedure/region_migration.rs
+++ b/src/meta-srv/src/procedure/region_migration.rs
@@ -921,7 +921,7 @@ mod tests {
    use std::assert_matches::assert_matches;
    use std::sync::Arc;

-    use common_meta::distributed_time_constants::REGION_LEASE_SECS;
+    use common_meta::distributed_time_constants::default_distributed_time_constants;
    use common_meta::instruction::Instruction;
    use common_meta::key::test_utils::new_test_table_info;
    use common_meta::rpc::router::{Region, RegionRoute};
@@ -1192,8 +1192,10 @@ mod tests {
            .run_once()
            .await;

+        let region_lease = default_distributed_time_constants().region_lease.as_secs();
+
        // Ensure it didn't run into the slow path.
-        assert!(timer.elapsed().as_secs() < REGION_LEASE_SECS / 2);
+        assert!(timer.elapsed().as_secs() < region_lease / 2);

        runner.suite.verify_table_metadata().await;
    }
@@ -1539,8 +1541,9 @@ mod tests {
            .run_once()
            .await;

+        let region_lease = default_distributed_time_constants().region_lease.as_secs();
        // Ensure it didn't run into the slow path.
-        assert!(timer.elapsed().as_secs() < REGION_LEASE_SECS);
+        assert!(timer.elapsed().as_secs() < region_lease);
        runner.suite.verify_table_metadata().await;
    }
 }
--- a/src/meta-srv/src/procedure/region_migration/close_downgraded_region.rs
+++ b/src/meta-srv/src/procedure/region_migration/close_downgraded_region.rs
@@ -13,11 +13,10 @@
 // limitations under the License.

 use std::any::Any;
-use std::time::Duration;

 use api::v1::meta::MailboxMessage;
 use common_meta::RegionIdent;
-use common_meta::distributed_time_constants::REGION_LEASE_SECS;
+use common_meta::distributed_time_constants::default_distributed_time_constants;
 use common_meta::instruction::{Instruction, InstructionReply, SimpleReply};
 use common_procedure::{Context as ProcedureContext, Status};
 use common_telemetry::{info, warn};
@@ -30,9 +29,6 @@ use crate::procedure::region_migration::migration_end::RegionMigrationEnd;
 use crate::procedure::region_migration::{Context, State};
 use crate::service::mailbox::Channel;

-/// Uses lease time of a region as the timeout of closing a downgraded region.
-const CLOSE_DOWNGRADED_REGION_TIMEOUT: Duration = Duration::from_secs(REGION_LEASE_SECS);
-
 #[derive(Debug, Serialize, Deserialize)]
 pub struct CloseDowngradedRegion;

@@ -112,7 +108,7 @@ impl CloseDowngradedRegion {
        let ch = Channel::Datanode(downgrade_leader_datanode.id);
        let receiver = ctx
            .mailbox
-            .send(&ch, msg, CLOSE_DOWNGRADED_REGION_TIMEOUT)
+            .send(&ch, msg, default_distributed_time_constants().region_lease)
            .await?;

        match receiver.await {
--- a/src/meta-srv/src/procedure/region_migration/downgrade_leader_region.rs
+++ b/src/meta-srv/src/procedure/region_migration/downgrade_leader_region.rs
@@ -17,7 +17,7 @@ use std::time::Duration;

 use api::v1::meta::MailboxMessage;
 use common_error::ext::BoxedError;
-use common_meta::distributed_time_constants::REGION_LEASE_SECS;
+use common_meta::distributed_time_constants::default_distributed_time_constants;
 use common_meta::instruction::{
    DowngradeRegion, DowngradeRegionReply, DowngradeRegionsReply, Instruction, InstructionReply,
 };
@@ -64,7 +64,7 @@ impl State for DowngradeLeaderRegion {
        let now = Instant::now();
        // Ensures the `leader_region_lease_deadline` must exist after recovering.
        ctx.volatile_ctx
-            .set_leader_region_lease_deadline(Duration::from_secs(REGION_LEASE_SECS));
+            .set_leader_region_lease_deadline(default_distributed_time_constants().region_lease);

        match self.downgrade_region_with_retry(ctx).await {
            Ok(_) => {
@@ -277,14 +277,14 @@ impl DowngradeLeaderRegion {
        if let Some(last_connection_at) = last_connection_at {
            let now = current_time_millis();
            let elapsed = now - last_connection_at;
-            let region_lease = Duration::from_secs(REGION_LEASE_SECS);
+            let region_lease = default_distributed_time_constants().region_lease;

            // It's safe to update the region leader lease deadline here because:
            // 1. The old region leader has already been marked as downgraded in metadata,
            //    which means any attempts to renew its lease will be rejected.
            // 2. The pusher disconnect time record only gets removed when the datanode (from_peer)
            //    establishes a new heartbeat connection stream.
-            if elapsed >= (REGION_LEASE_SECS * 1000) as i64 {
+            if elapsed >= (region_lease.as_secs() * 1000) as i64 {
                ctx.volatile_ctx.reset_leader_region_lease_deadline();
                info!(
                    "Datanode {}({}) has been disconnected for longer than the region lease period ({:?}), reset leader region lease deadline to None, region: {:?}",
@@ -697,7 +697,8 @@ mod tests {
        let procedure_ctx = new_procedure_context();
        let (next, _) = state.next(&mut ctx, &procedure_ctx).await.unwrap();
        let elapsed = timer.elapsed().as_secs();
-        assert!(elapsed < REGION_LEASE_SECS / 2);
+        let region_lease = default_distributed_time_constants().region_lease.as_secs();
+        assert!(elapsed < region_lease / 2);
        assert_eq!(
            ctx.volatile_ctx
                .leader_region_last_entry_ids
--- a/src/meta-srv/src/procedure/region_migration/open_candidate_region.rs
+++ b/src/meta-srv/src/procedure/region_migration/open_candidate_region.rs
@@ -14,11 +14,10 @@

 use std::any::Any;
 use std::ops::Div;
-use std::time::Duration;

 use api::v1::meta::MailboxMessage;
 use common_meta::RegionIdent;
-use common_meta::distributed_time_constants::REGION_LEASE_SECS;
+use common_meta::distributed_time_constants::default_distributed_time_constants;
 use common_meta::instruction::{Instruction, InstructionReply, OpenRegion, SimpleReply};
 use common_meta::key::datanode_table::RegionInfo;
 use common_procedure::{Context as ProcedureContext, Status};
@@ -33,9 +32,6 @@ use crate::procedure::region_migration::flush_leader_region::PreFlushRegion;
 use crate::procedure::region_migration::{Context, State};
 use crate::service::mailbox::Channel;

-/// Uses lease time of a region as the timeout of opening a candidate region.
-const OPEN_CANDIDATE_REGION_TIMEOUT: Duration = Duration::from_secs(REGION_LEASE_SECS);
-
 #[derive(Debug, Serialize, Deserialize)]
 pub struct OpenCandidateRegion;

@@ -157,7 +153,9 @@ impl OpenCandidateRegion {
                .context(error::ExceededDeadlineSnafu {
                    operation: "Open candidate region",
                })?;
-        let operation_timeout = operation_timeout.div(2).max(OPEN_CANDIDATE_REGION_TIMEOUT);
+        let operation_timeout = operation_timeout
+            .div(2)
+            .max(default_distributed_time_constants().region_lease);
        let ch = Channel::Datanode(candidate.id);
        let now = Instant::now();
        let receiver = ctx.mailbox.send(&ch, msg, operation_timeout).await?;
--- a/src/meta-srv/src/service/heartbeat.rs
+++ b/src/meta-srv/src/service/heartbeat.rs
@@ -99,6 +99,7 @@ impl heartbeat_server::Heartbeat for Metasrv {
                            error!("Client disconnected: broken pipe");
                            break;
                        }
+                        error!(err; "Sending heartbeat response error");

                        if tx.send(Err(err)).await.is_err() {
                            info!("ReceiverStream was dropped; shutting down");
--- a/src/meta-srv/src/utils/etcd.rs
+++ b/src/meta-srv/src/utils/etcd.rs
@@ -12,17 +12,18 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use common_meta::distributed_time_constants::default_etcd_client_options;
 use common_meta::kv_backend::etcd::create_etcd_tls_options;
-use etcd_client::Client;
+use etcd_client::{Client, ConnectOptions};
 use servers::tls::{TlsMode, TlsOption};
 use snafu::ResultExt;

 use crate::error::{self, BuildTlsOptionsSnafu, Result};
+use crate::metasrv::BackendClientOptions;

 /// Creates an etcd client with TLS configuration.
 pub async fn create_etcd_client_with_tls(
    store_addrs: &[String],
+    client_options: &BackendClientOptions,
    tls_config: Option<&TlsOption>,
 ) -> Result<Client> {
    let etcd_endpoints = store_addrs
@@ -31,7 +32,12 @@ pub async fn create_etcd_client_with_tls(
        .filter(|x| !x.is_empty())
        .collect::<Vec<_>>();

-    let mut connect_options = default_etcd_client_options();
+    let mut connect_options = ConnectOptions::new()
+        .with_keep_alive_while_idle(true)
+        .with_keep_alive(
+            client_options.keep_alive_interval,
+            client_options.keep_alive_timeout,
+        );
    if let Some(tls_config) = tls_config
        && let Some(tls_options) = create_etcd_tls_options(&convert_tls_option(tls_config))
            .context(BuildTlsOptionsSnafu)?
--- a/src/mito-codec/src/index.rs
+++ b/src/mito-codec/src/index.rs
@@ -48,7 +48,7 @@ impl IndexValueCodec {
    ) -> Result<()> {
        ensure!(!value.is_null(), IndexEncodeNullSnafu);

-        if field.data_type().is_string() {
+        if field.encode_data_type().is_string() {
            let value = value
                .try_into_string()
                .context(FieldTypeMismatchSnafu)?
--- a/src/mito-codec/src/row_converter/dense.rs
+++ b/src/mito-codec/src/row_converter/dense.rs
@@ -57,15 +57,20 @@ impl SortField {
        &self.data_type
    }

-    pub fn estimated_size(&self) -> usize {
+    /// Returns the physical data type to encode of the field.
+    ///
+    /// For example, a dictionary field will be encoded as its value type.
+    pub fn encode_data_type(&self) -> &ConcreteDataType {
        match &self.data_type {
-            ConcreteDataType::Dictionary(dict_type) => {
-                Self::estimated_size_by_type(dict_type.value_type())
-            }
-            data_type => Self::estimated_size_by_type(data_type),
+            ConcreteDataType::Dictionary(dict_type) => dict_type.value_type(),
+            _ => &self.data_type,
        }
    }

+    pub fn estimated_size(&self) -> usize {
+        Self::estimated_size_by_type(self.encode_data_type())
+    }
+
    fn estimated_size_by_type(data_type: &ConcreteDataType) -> usize {
        match data_type {
            ConcreteDataType::Boolean(_) => 2,
@@ -98,12 +103,7 @@ impl SortField {
        serializer: &mut Serializer<&mut Vec<u8>>,
        value: &ValueRef,
    ) -> Result<()> {
-        match self.data_type() {
-            ConcreteDataType::Dictionary(dict_type) => {
-                Self::serialize_by_type(dict_type.value_type(), serializer, value)
-            }
-            data_type => Self::serialize_by_type(data_type, serializer, value),
-        }
+        Self::serialize_by_type(self.encode_data_type(), serializer, value)
    }

    fn serialize_by_type(
@@ -194,12 +194,7 @@ impl SortField {

    /// Deserialize a value from the deserializer.
    pub fn deserialize<B: Buf>(&self, deserializer: &mut Deserializer<B>) -> Result<Value> {
-        match &self.data_type {
-            ConcreteDataType::Dictionary(dict_type) => {
-                Self::deserialize_by_type(dict_type.value_type(), deserializer)
-            }
-            data_type => Self::deserialize_by_type(data_type, deserializer),
-        }
+        Self::deserialize_by_type(self.encode_data_type(), deserializer)
    }

    fn deserialize_by_type<B: Buf>(
@@ -301,12 +296,7 @@ impl SortField {
            return Ok(1);
        }

-        match &self.data_type {
-            ConcreteDataType::Dictionary(dict_type) => {
-                Self::skip_deserialize_by_type(dict_type.value_type(), bytes, deserializer)
-            }
-            data_type => Self::skip_deserialize_by_type(data_type, bytes, deserializer),
-        }
+        Self::skip_deserialize_by_type(self.encode_data_type(), bytes, deserializer)
    }

    fn skip_deserialize_by_type(
--- a/src/mito2/src/compaction/task.rs
+++ b/src/mito2/src/compaction/task.rs
@@ -25,7 +25,7 @@ use tokio::sync::mpsc;
 use crate::compaction::compactor::{CompactionRegion, Compactor};
 use crate::compaction::memory_manager::{CompactionMemoryGuard, CompactionMemoryManager};
 use crate::compaction::picker::{CompactionTask, PickerOutput};
-use crate::error::{CompactRegionSnafu, CompactionMemoryExhaustedSnafu, MemoryAcquireFailedSnafu};
+use crate::error::{CompactRegionSnafu, CompactionMemoryExhaustedSnafu};
 use crate::manifest::action::{RegionEdit, RegionMetaAction, RegionMetaActionList};
 use crate::metrics::{COMPACTION_FAILURE_COUNT, COMPACTION_MEMORY_WAIT, COMPACTION_STAGE_ELAPSED};
 use crate::region::RegionRoleState;
@@ -95,80 +95,16 @@ impl CompactionTaskImpl {
    async fn acquire_memory_with_policy(&self) -> error::Result<CompactionMemoryGuard> {
        let region_id = self.compaction_region.region_id;
        let requested_bytes = self.estimated_memory_bytes;
-        let limit_bytes = self.memory_manager.limit_bytes();
+        let policy = self.memory_policy;

-        if limit_bytes > 0 && requested_bytes > limit_bytes {
-            warn!(
-                "Compaction for region {} requires {} bytes but limit is {} bytes; cannot satisfy request",
-                region_id, requested_bytes, limit_bytes
-            );
-            return Err(CompactionMemoryExhaustedSnafu {
+        let _timer = COMPACTION_MEMORY_WAIT.start_timer();
+        self.memory_manager
+            .acquire_with_policy(requested_bytes, policy)
+            .await
+            .context(CompactionMemoryExhaustedSnafu {
                region_id,
-                required_bytes: requested_bytes,
-                limit_bytes,
-                policy: "exceed_limit".to_string(),
-            }
-            .build());
-        }
-
-        match self.memory_policy {
-            OnExhaustedPolicy::Wait {
-                timeout: wait_timeout,
-            } => {
-                let timer = COMPACTION_MEMORY_WAIT.start_timer();
-
-                match tokio::time::timeout(
-                    wait_timeout,
-                    self.memory_manager.acquire(requested_bytes),
-                )
-                .await
-                {
-                    Ok(Ok(guard)) => {
-                        timer.observe_duration();
-                        Ok(guard)
-                    }
-                    Ok(Err(e)) => {
-                        timer.observe_duration();
-                        Err(e).with_context(|_| MemoryAcquireFailedSnafu {
-                            region_id,
-                            policy: format!("wait_timeout({}ms)", wait_timeout.as_millis()),
-                        })
-                    }
-                    Err(_) => {
-                        timer.observe_duration();
-                        warn!(
-                            "Compaction for region {} waited {:?} for {} bytes but timed out",
-                            region_id, wait_timeout, requested_bytes
-                        );
-                        CompactionMemoryExhaustedSnafu {
-                            region_id,
-                            required_bytes: requested_bytes,
-                            limit_bytes,
-                            policy: format!("wait_timeout({}ms)", wait_timeout.as_millis()),
-                        }
-                        .fail()
-                    }
-                }
-            }
-            OnExhaustedPolicy::Fail => {
-                // Try to acquire, fail immediately if not available
-                self.memory_manager
-                    .try_acquire(requested_bytes)
-                    .ok_or_else(|| {
-                        warn!(
-                            "Compaction memory exhausted for region {} (policy=fail, need {} bytes, limit {} bytes)",
-                            region_id, requested_bytes, limit_bytes
-                        );
-                        CompactionMemoryExhaustedSnafu {
-                            region_id,
-                            required_bytes: requested_bytes,
-                            limit_bytes,
-                            policy: "fail".to_string(),
-                        }
-                        .build()
-                    })
-            }
-        }
+                policy: format!("{policy:?}"),
+            })
    }

    /// Remove expired ssts files, update manifest immediately
--- a/src/mito2/src/engine/basic_test.rs
+++ b/src/mito2/src/engine/basic_test.rs
@@ -872,9 +872,9 @@ StorageSstEntry { file_path: "test/11_0000000002/index/<file_id>.puffin", file_s
 StorageSstEntry { file_path: "test/22_0000000042/<file_id>.parquet", file_size: None, last_modified_ms: None, node_id: None }
 StorageSstEntry { file_path: "test/22_0000000042/index/<file_id>.puffin", file_size: None, last_modified_ms: None, node_id: None }"#).await;
    test_list_ssts_with_format(true, r#"
-ManifestSstEntry { table_dir: "test/", region_id: 47244640257(11, 1), table_id: 11, region_number: 1, region_group: 0, region_sequence: 1, file_id: "<file_id>", index_version: 0, level: 0, file_path: "test/11_0000000001/<file_id>.parquet", file_size: 2837, index_file_path: Some("test/11_0000000001/index/<file_id>.puffin"), index_file_size: Some(292), num_rows: 10, num_row_groups: 1, num_series: Some(1), min_ts: 0::Millisecond, max_ts: 9000::Millisecond, sequence: Some(10), origin_region_id: 47244640257(11, 1), node_id: None, visible: true }
-ManifestSstEntry { table_dir: "test/", region_id: 47244640258(11, 2), table_id: 11, region_number: 2, region_group: 0, region_sequence: 2, file_id: "<file_id>", index_version: 0, level: 0, file_path: "test/11_0000000002/<file_id>.parquet", file_size: 2837, index_file_path: Some("test/11_0000000002/index/<file_id>.puffin"), index_file_size: Some(292), num_rows: 10, num_row_groups: 1, num_series: Some(1), min_ts: 0::Millisecond, max_ts: 9000::Millisecond, sequence: Some(10), origin_region_id: 47244640258(11, 2), node_id: None, visible: true }
-ManifestSstEntry { table_dir: "test/", region_id: 94489280554(22, 42), table_id: 22, region_number: 42, region_group: 0, region_sequence: 42, file_id: "<file_id>", index_version: 0, level: 0, file_path: "test/22_0000000042/<file_id>.parquet", file_size: 2837, index_file_path: Some("test/22_0000000042/index/<file_id>.puffin"), index_file_size: Some(292), num_rows: 10, num_row_groups: 1, num_series: Some(1), min_ts: 0::Millisecond, max_ts: 9000::Millisecond, sequence: Some(10), origin_region_id: 94489280554(22, 42), node_id: None, visible: true }"#, 
+ManifestSstEntry { table_dir: "test/", region_id: 47244640257(11, 1), table_id: 11, region_number: 1, region_group: 0, region_sequence: 1, file_id: "<file_id>", index_version: 0, level: 0, file_path: "test/11_0000000001/<file_id>.parquet", file_size: 2837, index_file_path: Some("test/11_0000000001/index/<file_id>.puffin"), index_file_size: Some(250), num_rows: 10, num_row_groups: 1, num_series: Some(1), min_ts: 0::Millisecond, max_ts: 9000::Millisecond, sequence: Some(10), origin_region_id: 47244640257(11, 1), node_id: None, visible: true }
+ManifestSstEntry { table_dir: "test/", region_id: 47244640258(11, 2), table_id: 11, region_number: 2, region_group: 0, region_sequence: 2, file_id: "<file_id>", index_version: 0, level: 0, file_path: "test/11_0000000002/<file_id>.parquet", file_size: 2837, index_file_path: Some("test/11_0000000002/index/<file_id>.puffin"), index_file_size: Some(250), num_rows: 10, num_row_groups: 1, num_series: Some(1), min_ts: 0::Millisecond, max_ts: 9000::Millisecond, sequence: Some(10), origin_region_id: 47244640258(11, 2), node_id: None, visible: true }
+ManifestSstEntry { table_dir: "test/", region_id: 94489280554(22, 42), table_id: 22, region_number: 42, region_group: 0, region_sequence: 42, file_id: "<file_id>", index_version: 0, level: 0, file_path: "test/22_0000000042/<file_id>.parquet", file_size: 2837, index_file_path: Some("test/22_0000000042/index/<file_id>.puffin"), index_file_size: Some(250), num_rows: 10, num_row_groups: 1, num_series: Some(1), min_ts: 0::Millisecond, max_ts: 9000::Millisecond, sequence: Some(10), origin_region_id: 94489280554(22, 42), node_id: None, visible: true }"#, 
 r#"
 StorageSstEntry { file_path: "test/11_0000000001/<file_id>.parquet", file_size: None, last_modified_ms: None, node_id: None }
 StorageSstEntry { file_path: "test/11_0000000001/index/<file_id>.puffin", file_size: None, last_modified_ms: None, node_id: None }
--- a/src/mito2/src/error.rs
+++ b/src/mito2/src/error.rs
@@ -1042,20 +1042,8 @@ pub enum Error {
    #[snafu(display("Manual compaction is override by following operations."))]
    ManualCompactionOverride {},

-    #[snafu(display(
-        "Compaction memory limit exceeded for region {region_id}: required {required_bytes} bytes, limit {limit_bytes} bytes (policy: {policy})",
-    ))]
+    #[snafu(display("Compaction memory exhausted for region {region_id} (policy: {policy})",))]
    CompactionMemoryExhausted {
-        region_id: RegionId,
-        required_bytes: u64,
-        limit_bytes: u64,
-        policy: String,
-        #[snafu(implicit)]
-        location: Location,
-    },
-
-    #[snafu(display("Failed to acquire memory for region {region_id} (policy: {policy})"))]
-    MemoryAcquireFailed {
        region_id: RegionId,
        policy: String,
        #[snafu(source)]
@@ -1359,9 +1347,7 @@ impl ErrorExt for Error {

            ManualCompactionOverride {} => StatusCode::Cancelled,

-            CompactionMemoryExhausted { .. } => StatusCode::RuntimeResourcesExhausted,
-
-            MemoryAcquireFailed { source, .. } => source.status_code(),
+            CompactionMemoryExhausted { source, .. } => source.status_code(),

            IncompatibleWalProviderChange { .. } => StatusCode::InvalidArguments,

--- a/src/mito2/src/flush.rs
+++ b/src/mito2/src/flush.rs
@@ -774,7 +774,12 @@ fn memtable_flat_sources(
            let iter = only_range.build_record_batch_iter(None)?;
            // Dedup according to append mode and merge mode.
            // Even single range may have duplicate rows.
-            let iter = maybe_dedup_one(options, field_column_start, iter);
+            let iter = maybe_dedup_one(
+                options.append_mode,
+                options.merge_mode(),
+                field_column_start,
+                iter,
+            );
            flat_sources.sources.push(FlatSource::Iter(iter));
        };
    } else {
@@ -842,17 +847,18 @@ fn merge_and_dedup(
    Ok(maybe_dedup)
 }

-fn maybe_dedup_one(
-    options: &RegionOptions,
+pub fn maybe_dedup_one(
+    append_mode: bool,
+    merge_mode: MergeMode,
    field_column_start: usize,
    input_iter: BoxedRecordBatchIterator,
 ) -> BoxedRecordBatchIterator {
-    if options.append_mode {
+    if append_mode {
        // No dedup in append mode
        input_iter
    } else {
        // Dedup according to merge mode.
-        match options.merge_mode() {
+        match merge_mode {
            MergeMode::LastRow => {
                Box::new(FlatDedupIterator::new(input_iter, FlatLastRow::new(false)))
            }
--- a/src/mito2/src/gc.rs
+++ b/src/mito2/src/gc.rs
@@ -540,7 +540,7 @@ impl LocalGcWorker {
    fn filter_deletable_files(
        &self,
        entries: Vec<Entry>,
-        in_use_filenames: &HashSet<&FileId>,
+        in_use_filenames: &HashSet<FileId>,
        may_linger_filenames: &HashSet<&FileId>,
        eligible_for_removal: &HashSet<&FileId>,
        unknown_file_may_linger_until: chrono::DateTime<chrono::Utc>,
@@ -641,9 +641,6 @@ impl LocalGcWorker {
            .flatten()
            .collect::<HashSet<_>>();

-        // in use filenames, include sst and index files
-        let in_use_filenames = in_used.iter().collect::<HashSet<_>>();
-
        // When full_file_listing is false, skip expensive list operations and only delete
        // files that are tracked in recently_removed_files
        if !self.full_file_listing {
@@ -653,7 +650,7 @@ impl LocalGcWorker {
            // 3. Have passed the lingering time
            let files_to_delete: Vec<FileId> = eligible_for_removal
                .iter()
-                .filter(|file_id| !in_use_filenames.contains(*file_id))
+                .filter(|file_id| !in_used.contains(*file_id))
                .map(|&f| *f)
                .collect();

@@ -672,7 +669,7 @@ impl LocalGcWorker {
        let (all_unused_files_ready_for_delete, all_in_exist_linger_files) = self
            .filter_deletable_files(
                all_entries,
-                &in_use_filenames,
+                in_used,
                &may_linger_filenames,
                &eligible_for_removal,
                unknown_file_may_linger_until,
--- a/src/mito2/src/memtable.rs
+++ b/src/mito2/src/memtable.rs
@@ -55,10 +55,8 @@ pub mod time_partition;
 pub mod time_series;
 pub(crate) mod version;

-#[cfg(any(test, feature = "test"))]
-pub use bulk::part::BulkPart;
 pub use bulk::part::{
-    BulkPartEncoder, BulkPartMeta, UnorderedPart, record_batch_estimated_size,
+    BulkPart, BulkPartEncoder, BulkPartMeta, UnorderedPart, record_batch_estimated_size,
    sort_primary_key_record_batch,
 };
 #[cfg(any(test, feature = "test"))]
--- a/src/mito2/src/memtable/bulk.rs
+++ b/src/mito2/src/memtable/bulk.rs
@@ -668,10 +668,10 @@ impl BulkMemtable {
 }

 /// Iterator builder for bulk range
-struct BulkRangeIterBuilder {
-    part: BulkPart,
-    context: Arc<BulkIterContext>,
-    sequence: Option<SequenceRange>,
+pub struct BulkRangeIterBuilder {
+    pub part: BulkPart,
+    pub context: Arc<BulkIterContext>,
+    pub sequence: Option<SequenceRange>,
 }

 impl IterBuilder for BulkRangeIterBuilder {
@@ -1188,7 +1188,6 @@ impl MemtableBuilder for BulkMemtableBuilder {

 #[cfg(test)]
 mod tests {
-
    use mito_codec::row_converter::build_primary_key_codec;

    use super::*;
--- a/src/mito2/src/memtable/bulk/part.rs
+++ b/src/mito2/src/memtable/bulk/part.rs
@@ -1211,343 +1211,24 @@ impl BulkPartEncoder {
    }
 }

-/// Converts mutations to record batches.
-fn mutations_to_record_batch(
-    mutations: &[Mutation],
-    metadata: &RegionMetadataRef,
-    pk_encoder: &DensePrimaryKeyCodec,
-    dedup: bool,
-) -> Result<Option<(RecordBatch, i64, i64)>> {
-    let total_rows: usize = mutations
-        .iter()
-        .map(|m| m.rows.as_ref().map(|r| r.rows.len()).unwrap_or(0))
-        .sum();
-
-    if total_rows == 0 {
-        return Ok(None);
-    }
-
-    let mut pk_builder = BinaryBuilder::with_capacity(total_rows, 0);
-
-    let mut ts_vector: Box<dyn MutableVector> = metadata
-        .time_index_column()
-        .column_schema
-        .data_type
-        .create_mutable_vector(total_rows);
-    let mut sequence_builder = UInt64Builder::with_capacity(total_rows);
-    let mut op_type_builder = UInt8Builder::with_capacity(total_rows);
-
-    let mut field_builders: Vec<Box<dyn MutableVector>> = metadata
-        .field_columns()
-        .map(|f| f.column_schema.data_type.create_mutable_vector(total_rows))
-        .collect();
-
-    let mut pk_buffer = vec![];
-    for m in mutations {
-        let Some(key_values) = KeyValuesRef::new(metadata, m) else {
-            continue;
-        };
-
-        for row in key_values.iter() {
-            pk_buffer.clear();
-            pk_encoder
-                .encode_to_vec(row.primary_keys(), &mut pk_buffer)
-                .context(EncodeSnafu)?;
-            pk_builder.append_value(pk_buffer.as_bytes());
-            ts_vector.push_value_ref(&row.timestamp());
-            sequence_builder.append_value(row.sequence());
-            op_type_builder.append_value(row.op_type() as u8);
-            for (builder, field) in field_builders.iter_mut().zip(row.fields()) {
-                builder.push_value_ref(&field);
-            }
-        }
-    }
-
-    let arrow_schema = to_sst_arrow_schema(metadata);
-    // safety: timestamp column must be valid, and values must not be None.
-    let timestamp_unit = metadata
-        .time_index_column()
-        .column_schema
-        .data_type
-        .as_timestamp()
-        .unwrap()
-        .unit();
-    let sorter = ArraysSorter {
-        encoded_primary_keys: pk_builder.finish(),
-        timestamp_unit,
-        timestamp: ts_vector.to_vector().to_arrow_array(),
-        sequence: sequence_builder.finish(),
-        op_type: op_type_builder.finish(),
-        fields: field_builders
-            .iter_mut()
-            .map(|f| f.to_vector().to_arrow_array()),
-        dedup,
-        arrow_schema,
-    };
-
-    sorter.sort().map(Some)
-}
-
-struct ArraysSorter<I> {
-    encoded_primary_keys: BinaryArray,
-    timestamp_unit: TimeUnit,
-    timestamp: ArrayRef,
-    sequence: UInt64Array,
-    op_type: UInt8Array,
-    fields: I,
-    dedup: bool,
-    arrow_schema: SchemaRef,
-}
-
-impl<I> ArraysSorter<I>
-where
-    I: Iterator<Item = ArrayRef>,
-{
-    /// Converts arrays to record batch.
-    fn sort(self) -> Result<(RecordBatch, i64, i64)> {
-        debug_assert!(!self.timestamp.is_empty());
-        debug_assert!(self.timestamp.len() == self.sequence.len());
-        debug_assert!(self.timestamp.len() == self.op_type.len());
-        debug_assert!(self.timestamp.len() == self.encoded_primary_keys.len());
-
-        let timestamp_iter = timestamp_array_to_iter(self.timestamp_unit, &self.timestamp);
-        let (mut min_timestamp, mut max_timestamp) = (i64::MAX, i64::MIN);
-        let mut to_sort = self
-            .encoded_primary_keys
-            .iter()
-            .zip(timestamp_iter)
-            .zip(self.sequence.iter())
-            .map(|((pk, timestamp), sequence)| {
-                max_timestamp = max_timestamp.max(*timestamp);
-                min_timestamp = min_timestamp.min(*timestamp);
-                (pk, timestamp, sequence)
-            })
-            .enumerate()
-            .collect::<Vec<_>>();
-
-        to_sort.sort_unstable_by(|(_, (l_pk, l_ts, l_seq)), (_, (r_pk, r_ts, r_seq))| {
-            l_pk.cmp(r_pk)
-                .then(l_ts.cmp(r_ts))
-                .then(l_seq.cmp(r_seq).reverse())
-        });
-
-        if self.dedup {
-            // Dedup by timestamps while ignore sequence.
-            to_sort.dedup_by(|(_, (l_pk, l_ts, _)), (_, (r_pk, r_ts, _))| {
-                l_pk == r_pk && l_ts == r_ts
-            });
-        }
-
-        let indices = UInt32Array::from_iter_values(to_sort.iter().map(|v| v.0 as u32));
-
-        let pk_dictionary = Arc::new(binary_array_to_dictionary(
-            // safety: pk must be BinaryArray
-            arrow::compute::take(
-                &self.encoded_primary_keys,
-                &indices,
-                Some(TakeOptions {
-                    check_bounds: false,
-                }),
-            )
-            .context(ComputeArrowSnafu)?
-            .as_any()
-            .downcast_ref::<BinaryArray>()
-            .unwrap(),
-        )?) as ArrayRef;
-
-        let mut arrays = Vec::with_capacity(self.arrow_schema.fields.len());
-        for arr in self.fields {
-            arrays.push(
-                arrow::compute::take(
-                    &arr,
-                    &indices,
-                    Some(TakeOptions {
-                        check_bounds: false,
-                    }),
-                )
-                .context(ComputeArrowSnafu)?,
-            );
-        }
-
-        let timestamp = arrow::compute::take(
-            &self.timestamp,
-            &indices,
-            Some(TakeOptions {
-                check_bounds: false,
-            }),
-        )
-        .context(ComputeArrowSnafu)?;
-
-        arrays.push(timestamp);
-        arrays.push(pk_dictionary);
-        arrays.push(
-            arrow::compute::take(
-                &self.sequence,
-                &indices,
-                Some(TakeOptions {
-                    check_bounds: false,
-                }),
-            )
-            .context(ComputeArrowSnafu)?,
-        );
-
-        arrays.push(
-            arrow::compute::take(
-                &self.op_type,
-                &indices,
-                Some(TakeOptions {
-                    check_bounds: false,
-                }),
-            )
-            .context(ComputeArrowSnafu)?,
-        );
-
-        let batch = RecordBatch::try_new(self.arrow_schema, arrays).context(NewRecordBatchSnafu)?;
-        Ok((batch, min_timestamp, max_timestamp))
-    }
-}
-
-/// Converts timestamp array to an iter of i64 values.
-fn timestamp_array_to_iter(
-    timestamp_unit: TimeUnit,
-    timestamp: &ArrayRef,
-) -> impl Iterator<Item = &i64> {
-    match timestamp_unit {
-        // safety: timestamp column must be valid.
-        TimeUnit::Second => timestamp
-            .as_any()
-            .downcast_ref::<TimestampSecondArray>()
-            .unwrap()
-            .values()
-            .iter(),
-        TimeUnit::Millisecond => timestamp
-            .as_any()
-            .downcast_ref::<TimestampMillisecondArray>()
-            .unwrap()
-            .values()
-            .iter(),
-        TimeUnit::Microsecond => timestamp
-            .as_any()
-            .downcast_ref::<TimestampMicrosecondArray>()
-            .unwrap()
-            .values()
-            .iter(),
-        TimeUnit::Nanosecond => timestamp
-            .as_any()
-            .downcast_ref::<TimestampNanosecondArray>()
-            .unwrap()
-            .values()
-            .iter(),
-    }
-}
-
-/// Converts a **sorted** [BinaryArray] to [DictionaryArray].
-fn binary_array_to_dictionary(input: &BinaryArray) -> Result<PrimaryKeyArray> {
-    if input.is_empty() {
-        return Ok(DictionaryArray::new(
-            UInt32Array::from(Vec::<u32>::new()),
-            Arc::new(BinaryArray::from_vec(vec![])) as ArrayRef,
-        ));
-    }
-    let mut keys = Vec::with_capacity(16);
-    let mut values = BinaryBuilder::new();
-    let mut prev: usize = 0;
-    keys.push(prev as u32);
-    values.append_value(input.value(prev));
-
-    for current_bytes in input.iter().skip(1) {
-        // safety: encoded pk must present.
-        let current_bytes = current_bytes.unwrap();
-        let prev_bytes = input.value(prev);
-        if current_bytes != prev_bytes {
-            values.append_value(current_bytes);
-            prev += 1;
-        }
-        keys.push(prev as u32);
-    }
-
-    Ok(DictionaryArray::new(
-        UInt32Array::from(keys),
-        Arc::new(values.finish()) as ArrayRef,
-    ))
-}
-
 #[cfg(test)]
 mod tests {
-    use std::collections::VecDeque;
-
    use api::v1::{Row, SemanticType, WriteHint};
    use datafusion_common::ScalarValue;
    use datatypes::arrow::array::Float64Array;
-    use datatypes::prelude::{ConcreteDataType, ScalarVector, Value};
+    use datatypes::prelude::{ConcreteDataType, Value};
    use datatypes::schema::ColumnSchema;
-    use datatypes::vectors::{Float64Vector, TimestampMillisecondVector};
    use store_api::metadata::{ColumnMetadata, RegionMetadataBuilder};
    use store_api::storage::RegionId;
    use store_api::storage::consts::ReservedColumnId;

    use super::*;
    use crate::memtable::bulk::context::BulkIterContext;
-    use crate::sst::parquet::format::{PrimaryKeyReadFormat, ReadFormat};
    use crate::sst::{FlatSchemaOptions, to_flat_sst_arrow_schema};
    use crate::test_util::memtable_util::{
        build_key_values_with_ts_seq_values, metadata_for_test, region_metadata_to_row_schema,
    };

-    fn check_binary_array_to_dictionary(
-        input: &[&[u8]],
-        expected_keys: &[u32],
-        expected_values: &[&[u8]],
-    ) {
-        let input = BinaryArray::from_iter_values(input.iter());
-        let array = binary_array_to_dictionary(&input).unwrap();
-        assert_eq!(
-            &expected_keys,
-            &array.keys().iter().map(|v| v.unwrap()).collect::<Vec<_>>()
-        );
-        assert_eq!(
-            expected_values,
-            &array
-                .values()
-                .as_any()
-                .downcast_ref::<BinaryArray>()
-                .unwrap()
-                .iter()
-                .map(|v| v.unwrap())
-                .collect::<Vec<_>>()
-        );
-    }
-
-    #[test]
-    fn test_binary_array_to_dictionary() {
-        check_binary_array_to_dictionary(&[], &[], &[]);
-
-        check_binary_array_to_dictionary(&["a".as_bytes()], &[0], &["a".as_bytes()]);
-
-        check_binary_array_to_dictionary(
-            &["a".as_bytes(), "a".as_bytes()],
-            &[0, 0],
-            &["a".as_bytes()],
-        );
-
-        check_binary_array_to_dictionary(
-            &["a".as_bytes(), "a".as_bytes(), "b".as_bytes()],
-            &[0, 0, 1],
-            &["a".as_bytes(), "b".as_bytes()],
-        );
-
-        check_binary_array_to_dictionary(
-            &[
-                "a".as_bytes(),
-                "a".as_bytes(),
-                "b".as_bytes(),
-                "c".as_bytes(),
-            ],
-            &[0, 0, 1, 2],
-            &["a".as_bytes(), "b".as_bytes(), "c".as_bytes()],
-        );
-    }
-
    struct MutationInput<'a> {
        k0: &'a str,
        k1: u32,
@@ -1563,232 +1244,6 @@ mod tests {
        v1: &'a [Option<f64>],
    }

-    fn check_mutations_to_record_batches(
-        input: &[MutationInput],
-        expected: &[BatchOutput],
-        expected_timestamp: (i64, i64),
-        dedup: bool,
-    ) {
-        let metadata = metadata_for_test();
-        let mutations = input
-            .iter()
-            .map(|m| {
-                build_key_values_with_ts_seq_values(
-                    &metadata,
-                    m.k0.to_string(),
-                    m.k1,
-                    m.timestamps.iter().copied(),
-                    m.v1.iter().copied(),
-                    m.sequence,
-                )
-                .mutation
-            })
-            .collect::<Vec<_>>();
-        let total_rows: usize = mutations
-            .iter()
-            .flat_map(|m| m.rows.iter())
-            .map(|r| r.rows.len())
-            .sum();
-
-        let pk_encoder = DensePrimaryKeyCodec::new(&metadata);
-
-        let (batch, _, _) = mutations_to_record_batch(&mutations, &metadata, &pk_encoder, dedup)
-            .unwrap()
-            .unwrap();
-        let read_format = PrimaryKeyReadFormat::new_with_all_columns(metadata.clone());
-        let mut batches = VecDeque::new();
-        read_format
-            .convert_record_batch(&batch, None, &mut batches)
-            .unwrap();
-        if !dedup {
-            assert_eq!(
-                total_rows,
-                batches.iter().map(|b| { b.num_rows() }).sum::<usize>()
-            );
-        }
-        let batch_values = batches
-            .into_iter()
-            .map(|b| {
-                let pk_values = pk_encoder.decode(b.primary_key()).unwrap().into_dense();
-                let timestamps = b
-                    .timestamps()
-                    .as_any()
-                    .downcast_ref::<TimestampMillisecondVector>()
-                    .unwrap()
-                    .iter_data()
-                    .map(|v| v.unwrap().0.value())
-                    .collect::<Vec<_>>();
-                let float_values = b.fields()[1]
-                    .data
-                    .as_any()
-                    .downcast_ref::<Float64Vector>()
-                    .unwrap()
-                    .iter_data()
-                    .collect::<Vec<_>>();
-
-                (pk_values, timestamps, float_values)
-            })
-            .collect::<Vec<_>>();
-        assert_eq!(expected.len(), batch_values.len());
-
-        for idx in 0..expected.len() {
-            assert_eq!(expected[idx].pk_values, &batch_values[idx].0);
-            assert_eq!(expected[idx].timestamps, &batch_values[idx].1);
-            assert_eq!(expected[idx].v1, &batch_values[idx].2);
-        }
-    }
-
-    #[test]
-    fn test_mutations_to_record_batch() {
-        check_mutations_to_record_batches(
-            &[MutationInput {
-                k0: "a",
-                k1: 0,
-                timestamps: &[0],
-                v1: &[Some(0.1)],
-                sequence: 0,
-            }],
-            &[BatchOutput {
-                pk_values: &[Value::String("a".into()), Value::UInt32(0)],
-                timestamps: &[0],
-                v1: &[Some(0.1)],
-            }],
-            (0, 0),
-            true,
-        );
-
-        check_mutations_to_record_batches(
-            &[
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.1)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "b",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[1],
-                    v1: &[Some(0.2)],
-                    sequence: 1,
-                },
-                MutationInput {
-                    k0: "a",
-                    k1: 1,
-                    timestamps: &[1],
-                    v1: &[Some(0.3)],
-                    sequence: 2,
-                },
-            ],
-            &[
-                BatchOutput {
-                    pk_values: &[Value::String("a".into()), Value::UInt32(0)],
-                    timestamps: &[0, 1],
-                    v1: &[Some(0.1), Some(0.2)],
-                },
-                BatchOutput {
-                    pk_values: &[Value::String("a".into()), Value::UInt32(1)],
-                    timestamps: &[1],
-                    v1: &[Some(0.3)],
-                },
-                BatchOutput {
-                    pk_values: &[Value::String("b".into()), Value::UInt32(0)],
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                },
-            ],
-            (0, 1),
-            true,
-        );
-
-        check_mutations_to_record_batches(
-            &[
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.1)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "b",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.2)],
-                    sequence: 1,
-                },
-            ],
-            &[
-                BatchOutput {
-                    pk_values: &[Value::String("a".into()), Value::UInt32(0)],
-                    timestamps: &[0],
-                    v1: &[Some(0.2)],
-                },
-                BatchOutput {
-                    pk_values: &[Value::String("b".into()), Value::UInt32(0)],
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                },
-            ],
-            (0, 0),
-            true,
-        );
-        check_mutations_to_record_batches(
-            &[
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.1)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "b",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.2)],
-                    sequence: 1,
-                },
-            ],
-            &[
-                BatchOutput {
-                    pk_values: &[Value::String("a".into()), Value::UInt32(0)],
-                    timestamps: &[0, 0],
-                    v1: &[Some(0.2), Some(0.1)],
-                },
-                BatchOutput {
-                    pk_values: &[Value::String("b".into()), Value::UInt32(0)],
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                },
-            ],
-            (0, 0),
-            false,
-        );
-    }
-
    fn encode(input: &[MutationInput]) -> EncodedBulkPart {
        let metadata = metadata_for_test();
        let kvs = input
--- a/src/mito2/src/sst.rs
+++ b/src/mito2/src/sst.rs
@@ -121,7 +121,7 @@ impl FlatSchemaOptions {
 ///
 /// The schema is:
 /// ```text
-/// primary key columns, field columns, time index, __prmary_key, __sequence, __op_type
+/// primary key columns, field columns, time index, __primary_key, __sequence, __op_type
 /// ```
 ///
 /// # Panics
--- a/src/mito2/src/sst/parquet.rs
+++ b/src/mito2/src/sst/parquet.rs
@@ -95,21 +95,32 @@ mod tests {
    use std::collections::HashSet;
    use std::sync::Arc;

-    use api::v1::OpType;
+    use api::v1::{OpType, SemanticType};
+    use common_function::function::FunctionRef;
+    use common_function::function_factory::ScalarFunctionFactory;
+    use common_function::scalars::matches::MatchesFunction;
+    use common_function::scalars::matches_term::MatchesTermFunction;
    use common_time::Timestamp;
    use datafusion_common::{Column, ScalarValue};
+    use datafusion_expr::expr::ScalarFunction;
    use datafusion_expr::{BinaryExpr, Expr, Literal, Operator, col, lit};
    use datatypes::arrow;
    use datatypes::arrow::array::{
-        ArrayRef, BinaryDictionaryBuilder, RecordBatch, StringDictionaryBuilder,
+        ArrayRef, BinaryDictionaryBuilder, RecordBatch, StringArray, StringDictionaryBuilder,
        TimestampMillisecondArray, UInt8Array, UInt64Array,
    };
    use datatypes::arrow::datatypes::{DataType, Field, Schema, UInt32Type};
+    use datatypes::prelude::ConcreteDataType;
+    use datatypes::schema::{FulltextAnalyzer, FulltextBackend, FulltextOptions};
+    use object_store::ObjectStore;
    use parquet::arrow::AsyncArrowWriter;
    use parquet::basic::{Compression, Encoding, ZstdLevel};
    use parquet::file::metadata::KeyValue;
    use parquet::file::properties::WriterProperties;
+    use store_api::codec::PrimaryKeyEncoding;
+    use store_api::metadata::{ColumnMetadata, RegionMetadata, RegionMetadataBuilder};
    use store_api::region_request::PathType;
+    use store_api::storage::{ColumnSchema, RegionId};
    use table::predicate::Predicate;
    use tokio_util::compat::FuturesAsyncWriteCompatExt;

@@ -122,6 +133,7 @@ mod tests {
    use crate::sst::file::{FileHandle, FileMeta, RegionFileId, RegionIndexId};
    use crate::sst::file_purger::NoopFilePurger;
    use crate::sst::index::bloom_filter::applier::BloomFilterIndexApplierBuilder;
+    use crate::sst::index::fulltext_index::applier::builder::FulltextIndexApplierBuilder;
    use crate::sst::index::inverted_index::applier::builder::InvertedIndexApplierBuilder;
    use crate::sst::index::{IndexBuildType, Indexer, IndexerBuilder, IndexerBuilderImpl};
    use crate::sst::parquet::format::PrimaryKeyWriteFormat;
@@ -133,11 +145,13 @@ mod tests {
    use crate::test_util::sst_util::{
        assert_parquet_metadata_eq, build_test_binary_test_region_metadata, new_batch_by_range,
        new_batch_with_binary, new_batch_with_custom_sequence, new_primary_key, new_source,
-        sst_file_handle, sst_file_handle_with_file_id, sst_region_metadata,
+        new_sparse_primary_key, sst_file_handle, sst_file_handle_with_file_id, sst_region_metadata,
+        sst_region_metadata_with_encoding,
    };
    use crate::test_util::{TestEnv, check_reader_result};

    const FILE_DIR: &str = "/";
+    const REGION_ID: RegionId = RegionId::new(0, 0);

    #[derive(Clone)]
    struct FixedPathProvider {
@@ -1064,6 +1078,154 @@ mod tests {
        FlatSource::Iter(Box::new(batches.into_iter().map(Ok)))
    }

+    /// Creates a flat format RecordBatch for testing with sparse primary key encoding.
+    /// Similar to `new_record_batch_by_range` but without individual primary key columns.
+    fn new_record_batch_by_range_sparse(
+        tags: &[&str],
+        start: usize,
+        end: usize,
+        metadata: &Arc<RegionMetadata>,
+    ) -> RecordBatch {
+        assert!(end >= start);
+        let flat_schema = to_flat_sst_arrow_schema(
+            metadata,
+            &FlatSchemaOptions::from_encoding(PrimaryKeyEncoding::Sparse),
+        );
+
+        let num_rows = end - start;
+        let mut columns: Vec<ArrayRef> = Vec::new();
+
+        // NOTE: Individual primary key columns (tag_0, tag_1) are NOT included in sparse format
+
+        // Add field column (field_0)
+        let field_values: Vec<u64> = (start..end).map(|v| v as u64).collect();
+        columns.push(Arc::new(UInt64Array::from(field_values)) as ArrayRef);
+
+        // Add time index column (ts)
+        let timestamps: Vec<i64> = (start..end).map(|v| v as i64).collect();
+        columns.push(Arc::new(TimestampMillisecondArray::from(timestamps)) as ArrayRef);
+
+        // Add encoded primary key column using sparse encoding
+        let table_id = 1u32; // Test table ID
+        let tsid = 100u64; // Base TSID
+        let pk = new_sparse_primary_key(tags, metadata, table_id, tsid);
+
+        let mut pk_builder = BinaryDictionaryBuilder::<UInt32Type>::new();
+        for _ in 0..num_rows {
+            pk_builder.append(&pk).unwrap();
+        }
+        columns.push(Arc::new(pk_builder.finish()) as ArrayRef);
+
+        // Add sequence column
+        columns.push(Arc::new(UInt64Array::from_value(1000, num_rows)) as ArrayRef);
+
+        // Add op_type column
+        columns.push(Arc::new(UInt8Array::from_value(OpType::Put as u8, num_rows)) as ArrayRef);
+
+        RecordBatch::try_new(flat_schema, columns).unwrap()
+    }
+
+    /// Helper function to create IndexerBuilderImpl for tests.
+    fn create_test_indexer_builder(
+        env: &TestEnv,
+        object_store: ObjectStore,
+        file_path: RegionFilePathFactory,
+        metadata: Arc<RegionMetadata>,
+        row_group_size: usize,
+    ) -> IndexerBuilderImpl {
+        let puffin_manager = env.get_puffin_manager().build(object_store, file_path);
+        let intermediate_manager = env.get_intermediate_manager();
+
+        IndexerBuilderImpl {
+            build_type: IndexBuildType::Flush,
+            metadata,
+            row_group_size,
+            puffin_manager,
+            write_cache_enabled: false,
+            intermediate_manager,
+            index_options: IndexOptions {
+                inverted_index: InvertedIndexOptions {
+                    segment_row_count: 1,
+                    ..Default::default()
+                },
+            },
+            inverted_index_config: Default::default(),
+            fulltext_index_config: Default::default(),
+            bloom_filter_index_config: Default::default(),
+        }
+    }
+
+    /// Helper function to write flat SST and return SstInfo.
+    async fn write_flat_sst(
+        object_store: ObjectStore,
+        metadata: Arc<RegionMetadata>,
+        indexer_builder: IndexerBuilderImpl,
+        file_path: RegionFilePathFactory,
+        flat_source: FlatSource,
+        write_opts: &WriteOptions,
+    ) -> SstInfo {
+        let mut metrics = Metrics::new(WriteType::Flush);
+        let mut writer = ParquetWriter::new_with_object_store(
+            object_store,
+            metadata,
+            IndexConfig::default(),
+            indexer_builder,
+            file_path,
+            &mut metrics,
+        )
+        .await;
+
+        writer
+            .write_all_flat(flat_source, write_opts)
+            .await
+            .unwrap()
+            .remove(0)
+    }
+
+    /// Helper function to create FileHandle from SstInfo.
+    fn create_file_handle_from_sst_info(
+        info: &SstInfo,
+        metadata: &Arc<RegionMetadata>,
+    ) -> FileHandle {
+        FileHandle::new(
+            FileMeta {
+                region_id: metadata.region_id,
+                file_id: info.file_id,
+                time_range: info.time_range,
+                level: 0,
+                file_size: info.file_size,
+                max_row_group_uncompressed_size: info.max_row_group_uncompressed_size,
+                available_indexes: info.index_metadata.build_available_indexes(),
+                indexes: info.index_metadata.build_indexes(),
+                index_file_size: info.index_metadata.file_size,
+                index_version: 0,
+                num_row_groups: info.num_row_groups,
+                num_rows: info.num_rows as u64,
+                sequence: None,
+                partition_expr: match &metadata.partition_expr {
+                    Some(json_str) => partition::expr::PartitionExpr::from_json_str(json_str)
+                        .expect("partition expression should be valid JSON"),
+                    None => None,
+                },
+                num_series: 0,
+            },
+            Arc::new(NoopFilePurger),
+        )
+    }
+
+    /// Helper function to create test cache with standard settings.
+    fn create_test_cache() -> Arc<CacheManager> {
+        Arc::new(
+            CacheManager::builder()
+                .index_result_cache_size(1024 * 1024)
+                .index_metadata_size(1024 * 1024)
+                .index_content_page_size(1024 * 1024)
+                .index_content_size(1024 * 1024)
+                .puffin_metadata_size(1024 * 1024)
+                .build(),
+        )
+    }
+
    #[tokio::test]
    async fn test_write_flat_with_index() {
        let mut env = TestEnv::new().await;
@@ -1238,4 +1400,709 @@ mod tests {
            assert_eq!(*override_batch, expected_batch);
        }
    }
+
+    #[tokio::test]
+    async fn test_write_flat_read_with_inverted_index() {
+        let mut env = TestEnv::new().await;
+        let object_store = env.init_object_store_manager();
+        let file_path = RegionFilePathFactory::new(FILE_DIR.to_string(), PathType::Bare);
+        let metadata = Arc::new(sst_region_metadata());
+        let row_group_size = 100;
+
+        // Create flat format RecordBatches with non-overlapping timestamp ranges
+        // Each batch becomes one row group (row_group_size = 100)
+        // Data: ts tag_0 tag_1
+        // RG 0:   0-50  [a, d]
+        // RG 0:  50-100 [b, d]
+        // RG 1: 100-150 [c, d]
+        // RG 1: 150-200 [c, f]
+        let flat_batches = vec![
+            new_record_batch_by_range(&["a", "d"], 0, 50),
+            new_record_batch_by_range(&["b", "d"], 50, 100),
+            new_record_batch_by_range(&["c", "d"], 100, 150),
+            new_record_batch_by_range(&["c", "f"], 150, 200),
+        ];
+
+        let flat_source = new_flat_source_from_record_batches(flat_batches);
+
+        let write_opts = WriteOptions {
+            row_group_size,
+            ..Default::default()
+        };
+
+        let indexer_builder = create_test_indexer_builder(
+            &env,
+            object_store.clone(),
+            file_path.clone(),
+            metadata.clone(),
+            row_group_size,
+        );
+
+        let info = write_flat_sst(
+            object_store.clone(),
+            metadata.clone(),
+            indexer_builder,
+            file_path.clone(),
+            flat_source,
+            &write_opts,
+        )
+        .await;
+        assert_eq!(200, info.num_rows);
+        assert!(info.file_size > 0);
+        assert!(info.index_metadata.file_size > 0);
+
+        let handle = create_file_handle_from_sst_info(&info, &metadata);
+
+        let cache = create_test_cache();
+
+        // Test 1: Filter by tag_0 = "b"
+        // Expected: Only rows with tag_0="b"
+        let preds = vec![col("tag_0").eq(lit("b"))];
+        let inverted_index_applier = InvertedIndexApplierBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            object_store.clone(),
+            &metadata,
+            HashSet::from_iter([0]),
+            env.get_puffin_manager(),
+        )
+        .with_puffin_metadata_cache(cache.puffin_metadata_cache().cloned())
+        .with_inverted_index_cache(cache.inverted_index_cache().cloned())
+        .build(&preds)
+        .unwrap()
+        .map(Arc::new);
+
+        let builder = ParquetReaderBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            handle.clone(),
+            object_store.clone(),
+        )
+        .flat_format(true)
+        .predicate(Some(Predicate::new(preds)))
+        .inverted_index_appliers([inverted_index_applier.clone(), None])
+        .cache(CacheStrategy::EnableAll(cache.clone()));
+
+        let mut metrics = ReaderMetrics::default();
+        let (_context, selection) = builder.build_reader_input(&mut metrics).await.unwrap();
+
+        // Verify selection contains only RG 0 (tag_0="b", ts 0-100)
+        assert_eq!(selection.row_group_count(), 1);
+        assert_eq!(50, selection.get(0).unwrap().row_count());
+
+        // Verify filtering metrics
+        assert_eq!(metrics.filter_metrics.rg_total, 2);
+        assert_eq!(metrics.filter_metrics.rg_minmax_filtered, 1);
+        assert_eq!(metrics.filter_metrics.rg_inverted_filtered, 0);
+        assert_eq!(metrics.filter_metrics.rows_inverted_filtered, 50);
+    }
+
+    #[tokio::test]
+    async fn test_write_flat_read_with_bloom_filter() {
+        let mut env = TestEnv::new().await;
+        let object_store = env.init_object_store_manager();
+        let file_path = RegionFilePathFactory::new(FILE_DIR.to_string(), PathType::Bare);
+        let metadata = Arc::new(sst_region_metadata());
+        let row_group_size = 100;
+
+        // Create flat format RecordBatches with non-overlapping timestamp ranges
+        // Each batch becomes one row group (row_group_size = 100)
+        // Data: ts tag_0 tag_1
+        // RG 0:   0-50  [a, d]
+        // RG 0:  50-100 [b, e]
+        // RG 1: 100-150 [c, d]
+        // RG 1: 150-200 [c, f]
+        let flat_batches = vec![
+            new_record_batch_by_range(&["a", "d"], 0, 50),
+            new_record_batch_by_range(&["b", "e"], 50, 100),
+            new_record_batch_by_range(&["c", "d"], 100, 150),
+            new_record_batch_by_range(&["c", "f"], 150, 200),
+        ];
+
+        let flat_source = new_flat_source_from_record_batches(flat_batches);
+
+        let write_opts = WriteOptions {
+            row_group_size,
+            ..Default::default()
+        };
+
+        let indexer_builder = create_test_indexer_builder(
+            &env,
+            object_store.clone(),
+            file_path.clone(),
+            metadata.clone(),
+            row_group_size,
+        );
+
+        let info = write_flat_sst(
+            object_store.clone(),
+            metadata.clone(),
+            indexer_builder,
+            file_path.clone(),
+            flat_source,
+            &write_opts,
+        )
+        .await;
+        assert_eq!(200, info.num_rows);
+        assert!(info.file_size > 0);
+        assert!(info.index_metadata.file_size > 0);
+
+        let handle = create_file_handle_from_sst_info(&info, &metadata);
+
+        let cache = create_test_cache();
+
+        // Filter by ts >= 50 AND ts < 200 AND tag_1 = "d"
+        // Expected: RG 0 (ts 0-100) and RG 1 (ts 100-200), both have tag_1="d"
+        let preds = vec![
+            col("ts").gt_eq(lit(ScalarValue::TimestampMillisecond(Some(50), None))),
+            col("ts").lt(lit(ScalarValue::TimestampMillisecond(Some(200), None))),
+            col("tag_1").eq(lit("d")),
+        ];
+        let bloom_filter_applier = BloomFilterIndexApplierBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            object_store.clone(),
+            &metadata,
+            env.get_puffin_manager(),
+        )
+        .with_puffin_metadata_cache(cache.puffin_metadata_cache().cloned())
+        .with_bloom_filter_index_cache(cache.bloom_filter_index_cache().cloned())
+        .build(&preds)
+        .unwrap()
+        .map(Arc::new);
+
+        let builder = ParquetReaderBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            handle.clone(),
+            object_store.clone(),
+        )
+        .flat_format(true)
+        .predicate(Some(Predicate::new(preds)))
+        .bloom_filter_index_appliers([None, bloom_filter_applier.clone()])
+        .cache(CacheStrategy::EnableAll(cache.clone()));
+
+        let mut metrics = ReaderMetrics::default();
+        let (_context, selection) = builder.build_reader_input(&mut metrics).await.unwrap();
+
+        // Verify selection contains RG 0 and RG 1
+        assert_eq!(selection.row_group_count(), 2);
+        assert_eq!(50, selection.get(0).unwrap().row_count());
+        assert_eq!(50, selection.get(1).unwrap().row_count());
+
+        // Verify filtering metrics
+        assert_eq!(metrics.filter_metrics.rg_total, 2);
+        assert_eq!(metrics.filter_metrics.rg_minmax_filtered, 0);
+        assert_eq!(metrics.filter_metrics.rg_bloom_filtered, 0);
+        assert_eq!(metrics.filter_metrics.rows_bloom_filtered, 100);
+    }
+
+    #[tokio::test]
+    async fn test_write_flat_read_with_inverted_index_sparse() {
+        common_telemetry::init_default_ut_logging();
+
+        let mut env = TestEnv::new().await;
+        let object_store = env.init_object_store_manager();
+        let file_path = RegionFilePathFactory::new(FILE_DIR.to_string(), PathType::Bare);
+        let metadata = Arc::new(sst_region_metadata_with_encoding(
+            PrimaryKeyEncoding::Sparse,
+        ));
+        let row_group_size = 100;
+
+        // Create flat format RecordBatches with non-overlapping timestamp ranges
+        // Each batch becomes one row group (row_group_size = 100)
+        // Data: ts tag_0 tag_1
+        // RG 0:   0-50  [a, d]
+        // RG 0:  50-100 [b, d]
+        // RG 1: 100-150 [c, d]
+        // RG 1: 150-200 [c, f]
+        let flat_batches = vec![
+            new_record_batch_by_range_sparse(&["a", "d"], 0, 50, &metadata),
+            new_record_batch_by_range_sparse(&["b", "d"], 50, 100, &metadata),
+            new_record_batch_by_range_sparse(&["c", "d"], 100, 150, &metadata),
+            new_record_batch_by_range_sparse(&["c", "f"], 150, 200, &metadata),
+        ];
+
+        let flat_source = new_flat_source_from_record_batches(flat_batches);
+
+        let write_opts = WriteOptions {
+            row_group_size,
+            ..Default::default()
+        };
+
+        let indexer_builder = create_test_indexer_builder(
+            &env,
+            object_store.clone(),
+            file_path.clone(),
+            metadata.clone(),
+            row_group_size,
+        );
+
+        let info = write_flat_sst(
+            object_store.clone(),
+            metadata.clone(),
+            indexer_builder,
+            file_path.clone(),
+            flat_source,
+            &write_opts,
+        )
+        .await;
+        assert_eq!(200, info.num_rows);
+        assert!(info.file_size > 0);
+        assert!(info.index_metadata.file_size > 0);
+
+        let handle = create_file_handle_from_sst_info(&info, &metadata);
+
+        let cache = create_test_cache();
+
+        // Test 1: Filter by tag_0 = "b"
+        // Expected: Only rows with tag_0="b"
+        let preds = vec![col("tag_0").eq(lit("b"))];
+        let inverted_index_applier = InvertedIndexApplierBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            object_store.clone(),
+            &metadata,
+            HashSet::from_iter([0]),
+            env.get_puffin_manager(),
+        )
+        .with_puffin_metadata_cache(cache.puffin_metadata_cache().cloned())
+        .with_inverted_index_cache(cache.inverted_index_cache().cloned())
+        .build(&preds)
+        .unwrap()
+        .map(Arc::new);
+
+        let builder = ParquetReaderBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            handle.clone(),
+            object_store.clone(),
+        )
+        .flat_format(true)
+        .predicate(Some(Predicate::new(preds)))
+        .inverted_index_appliers([inverted_index_applier.clone(), None])
+        .cache(CacheStrategy::EnableAll(cache.clone()));
+
+        let mut metrics = ReaderMetrics::default();
+        let (_context, selection) = builder.build_reader_input(&mut metrics).await.unwrap();
+
+        // RG 0 has 50 matching rows (tag_0="b")
+        assert_eq!(selection.row_group_count(), 1);
+        assert_eq!(50, selection.get(0).unwrap().row_count());
+
+        // Verify filtering metrics
+        // Note: With sparse encoding, tag columns aren't stored separately,
+        // so minmax filtering on tags doesn't work (only inverted index)
+        assert_eq!(metrics.filter_metrics.rg_total, 2);
+        assert_eq!(metrics.filter_metrics.rg_minmax_filtered, 0); // No minmax stats for tags in sparse format
+        assert_eq!(metrics.filter_metrics.rg_inverted_filtered, 1);
+        assert_eq!(metrics.filter_metrics.rows_inverted_filtered, 150);
+    }
+
+    #[tokio::test]
+    async fn test_write_flat_read_with_bloom_filter_sparse() {
+        let mut env = TestEnv::new().await;
+        let object_store = env.init_object_store_manager();
+        let file_path = RegionFilePathFactory::new(FILE_DIR.to_string(), PathType::Bare);
+        let metadata = Arc::new(sst_region_metadata_with_encoding(
+            PrimaryKeyEncoding::Sparse,
+        ));
+        let row_group_size = 100;
+
+        // Create flat format RecordBatches with non-overlapping timestamp ranges
+        // Each batch becomes one row group (row_group_size = 100)
+        // Data: ts tag_0 tag_1
+        // RG 0:   0-50  [a, d]
+        // RG 0:  50-100 [b, e]
+        // RG 1: 100-150 [c, d]
+        // RG 1: 150-200 [c, f]
+        let flat_batches = vec![
+            new_record_batch_by_range_sparse(&["a", "d"], 0, 50, &metadata),
+            new_record_batch_by_range_sparse(&["b", "e"], 50, 100, &metadata),
+            new_record_batch_by_range_sparse(&["c", "d"], 100, 150, &metadata),
+            new_record_batch_by_range_sparse(&["c", "f"], 150, 200, &metadata),
+        ];
+
+        let flat_source = new_flat_source_from_record_batches(flat_batches);
+
+        let write_opts = WriteOptions {
+            row_group_size,
+            ..Default::default()
+        };
+
+        let indexer_builder = create_test_indexer_builder(
+            &env,
+            object_store.clone(),
+            file_path.clone(),
+            metadata.clone(),
+            row_group_size,
+        );
+
+        let info = write_flat_sst(
+            object_store.clone(),
+            metadata.clone(),
+            indexer_builder,
+            file_path.clone(),
+            flat_source,
+            &write_opts,
+        )
+        .await;
+        assert_eq!(200, info.num_rows);
+        assert!(info.file_size > 0);
+        assert!(info.index_metadata.file_size > 0);
+
+        let handle = create_file_handle_from_sst_info(&info, &metadata);
+
+        let cache = create_test_cache();
+
+        // Filter by ts >= 50 AND ts < 200 AND tag_1 = "d"
+        // Expected: RG 0 (ts 0-100) and RG 1 (ts 100-200), both have tag_1="d"
+        let preds = vec![
+            col("ts").gt_eq(lit(ScalarValue::TimestampMillisecond(Some(50), None))),
+            col("ts").lt(lit(ScalarValue::TimestampMillisecond(Some(200), None))),
+            col("tag_1").eq(lit("d")),
+        ];
+        let bloom_filter_applier = BloomFilterIndexApplierBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            object_store.clone(),
+            &metadata,
+            env.get_puffin_manager(),
+        )
+        .with_puffin_metadata_cache(cache.puffin_metadata_cache().cloned())
+        .with_bloom_filter_index_cache(cache.bloom_filter_index_cache().cloned())
+        .build(&preds)
+        .unwrap()
+        .map(Arc::new);
+
+        let builder = ParquetReaderBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            handle.clone(),
+            object_store.clone(),
+        )
+        .flat_format(true)
+        .predicate(Some(Predicate::new(preds)))
+        .bloom_filter_index_appliers([None, bloom_filter_applier.clone()])
+        .cache(CacheStrategy::EnableAll(cache.clone()));
+
+        let mut metrics = ReaderMetrics::default();
+        let (_context, selection) = builder.build_reader_input(&mut metrics).await.unwrap();
+
+        // Verify selection contains RG 0 and RG 1
+        assert_eq!(selection.row_group_count(), 2);
+        assert_eq!(50, selection.get(0).unwrap().row_count());
+        assert_eq!(50, selection.get(1).unwrap().row_count());
+
+        // Verify filtering metrics
+        assert_eq!(metrics.filter_metrics.rg_total, 2);
+        assert_eq!(metrics.filter_metrics.rg_minmax_filtered, 0);
+        assert_eq!(metrics.filter_metrics.rg_bloom_filtered, 0);
+        assert_eq!(metrics.filter_metrics.rows_bloom_filtered, 100);
+    }
+
+    /// Creates region metadata for testing fulltext indexes.
+    /// Schema: tag_0, text_bloom, text_tantivy, field_0, ts
+    fn fulltext_region_metadata() -> RegionMetadata {
+        let mut builder = RegionMetadataBuilder::new(REGION_ID);
+        builder
+            .push_column_metadata(ColumnMetadata {
+                column_schema: ColumnSchema::new(
+                    "tag_0".to_string(),
+                    ConcreteDataType::string_datatype(),
+                    true,
+                ),
+                semantic_type: SemanticType::Tag,
+                column_id: 0,
+            })
+            .push_column_metadata(ColumnMetadata {
+                column_schema: ColumnSchema::new(
+                    "text_bloom".to_string(),
+                    ConcreteDataType::string_datatype(),
+                    true,
+                )
+                .with_fulltext_options(FulltextOptions {
+                    enable: true,
+                    analyzer: FulltextAnalyzer::English,
+                    case_sensitive: false,
+                    backend: FulltextBackend::Bloom,
+                    granularity: 1,
+                    false_positive_rate_in_10000: 50,
+                })
+                .unwrap(),
+                semantic_type: SemanticType::Field,
+                column_id: 1,
+            })
+            .push_column_metadata(ColumnMetadata {
+                column_schema: ColumnSchema::new(
+                    "text_tantivy".to_string(),
+                    ConcreteDataType::string_datatype(),
+                    true,
+                )
+                .with_fulltext_options(FulltextOptions {
+                    enable: true,
+                    analyzer: FulltextAnalyzer::English,
+                    case_sensitive: false,
+                    backend: FulltextBackend::Tantivy,
+                    granularity: 1,
+                    false_positive_rate_in_10000: 50,
+                })
+                .unwrap(),
+                semantic_type: SemanticType::Field,
+                column_id: 2,
+            })
+            .push_column_metadata(ColumnMetadata {
+                column_schema: ColumnSchema::new(
+                    "field_0".to_string(),
+                    ConcreteDataType::uint64_datatype(),
+                    true,
+                ),
+                semantic_type: SemanticType::Field,
+                column_id: 3,
+            })
+            .push_column_metadata(ColumnMetadata {
+                column_schema: ColumnSchema::new(
+                    "ts".to_string(),
+                    ConcreteDataType::timestamp_millisecond_datatype(),
+                    false,
+                ),
+                semantic_type: SemanticType::Timestamp,
+                column_id: 4,
+            })
+            .primary_key(vec![0]);
+        builder.build().unwrap()
+    }
+
+    /// Creates a flat format RecordBatch with string fields for fulltext testing.
+    fn new_fulltext_record_batch_by_range(
+        tag: &str,
+        text_bloom: &str,
+        text_tantivy: &str,
+        start: usize,
+        end: usize,
+    ) -> RecordBatch {
+        assert!(end >= start);
+        let metadata = Arc::new(fulltext_region_metadata());
+        let flat_schema = to_flat_sst_arrow_schema(&metadata, &FlatSchemaOptions::default());
+
+        let num_rows = end - start;
+        let mut columns = Vec::new();
+
+        // Add primary key column (tag_0) as dictionary array
+        let mut tag_builder = StringDictionaryBuilder::<UInt32Type>::new();
+        for _ in 0..num_rows {
+            tag_builder.append_value(tag);
+        }
+        columns.push(Arc::new(tag_builder.finish()) as ArrayRef);
+
+        // Add text_bloom field (fulltext with bloom backend)
+        let text_bloom_values: Vec<_> = (0..num_rows).map(|_| text_bloom).collect();
+        columns.push(Arc::new(StringArray::from(text_bloom_values)));
+
+        // Add text_tantivy field (fulltext with tantivy backend)
+        let text_tantivy_values: Vec<_> = (0..num_rows).map(|_| text_tantivy).collect();
+        columns.push(Arc::new(StringArray::from(text_tantivy_values)));
+
+        // Add field column (field_0)
+        let field_values: Vec<u64> = (start..end).map(|v| v as u64).collect();
+        columns.push(Arc::new(UInt64Array::from(field_values)));
+
+        // Add time index column (ts)
+        let timestamps: Vec<i64> = (start..end).map(|v| v as i64).collect();
+        columns.push(Arc::new(TimestampMillisecondArray::from(timestamps)));
+
+        // Add encoded primary key column
+        let pk = new_primary_key(&[tag]);
+        let mut pk_builder = BinaryDictionaryBuilder::<UInt32Type>::new();
+        for _ in 0..num_rows {
+            pk_builder.append(&pk).unwrap();
+        }
+        columns.push(Arc::new(pk_builder.finish()));
+
+        // Add sequence column
+        columns.push(Arc::new(UInt64Array::from_value(1000, num_rows)));
+
+        // Add op_type column
+        columns.push(Arc::new(UInt8Array::from_value(
+            OpType::Put as u8,
+            num_rows,
+        )));
+
+        RecordBatch::try_new(flat_schema, columns).unwrap()
+    }
+
+    #[tokio::test]
+    async fn test_write_flat_read_with_fulltext_index() {
+        let mut env = TestEnv::new().await;
+        let object_store = env.init_object_store_manager();
+        let file_path = RegionFilePathFactory::new(FILE_DIR.to_string(), PathType::Bare);
+        let metadata = Arc::new(fulltext_region_metadata());
+        let row_group_size = 50;
+
+        // Create flat format RecordBatches with different text content
+        // RG 0:   0-50  tag="a", bloom="hello world", tantivy="quick brown fox"
+        // RG 1:  50-100 tag="b", bloom="hello world", tantivy="quick brown fox"
+        // RG 2: 100-150 tag="c", bloom="goodbye world", tantivy="lazy dog"
+        // RG 3: 150-200 tag="d", bloom="goodbye world", tantivy="lazy dog"
+        let flat_batches = vec![
+            new_fulltext_record_batch_by_range("a", "hello world", "quick brown fox", 0, 50),
+            new_fulltext_record_batch_by_range("b", "hello world", "quick brown fox", 50, 100),
+            new_fulltext_record_batch_by_range("c", "goodbye world", "lazy dog", 100, 150),
+            new_fulltext_record_batch_by_range("d", "goodbye world", "lazy dog", 150, 200),
+        ];
+
+        let flat_source = new_flat_source_from_record_batches(flat_batches);
+
+        let write_opts = WriteOptions {
+            row_group_size,
+            ..Default::default()
+        };
+
+        let indexer_builder = create_test_indexer_builder(
+            &env,
+            object_store.clone(),
+            file_path.clone(),
+            metadata.clone(),
+            row_group_size,
+        );
+
+        let mut info = write_flat_sst(
+            object_store.clone(),
+            metadata.clone(),
+            indexer_builder,
+            file_path.clone(),
+            flat_source,
+            &write_opts,
+        )
+        .await;
+        assert_eq!(200, info.num_rows);
+        assert!(info.file_size > 0);
+        assert!(info.index_metadata.file_size > 0);
+
+        // Verify fulltext indexes were created
+        assert!(info.index_metadata.fulltext_index.index_size > 0);
+        assert_eq!(info.index_metadata.fulltext_index.row_count, 200);
+        // text_bloom (column_id 1) and text_tantivy (column_id 2)
+        info.index_metadata.fulltext_index.columns.sort_unstable();
+        assert_eq!(info.index_metadata.fulltext_index.columns, vec![1, 2]);
+
+        assert_eq!(
+            (
+                Timestamp::new_millisecond(0),
+                Timestamp::new_millisecond(199)
+            ),
+            info.time_range
+        );
+
+        let handle = create_file_handle_from_sst_info(&info, &metadata);
+
+        let cache = create_test_cache();
+
+        // Helper functions to create fulltext function expressions
+        let matches_func = || {
+            Arc::new(
+                ScalarFunctionFactory::from(Arc::new(MatchesFunction::default()) as FunctionRef)
+                    .provide(Default::default()),
+            )
+        };
+
+        let matches_term_func = || {
+            Arc::new(
+                ScalarFunctionFactory::from(
+                    Arc::new(MatchesTermFunction::default()) as FunctionRef,
+                )
+                .provide(Default::default()),
+            )
+        };
+
+        // Test 1: Filter by text_bloom field using matches_term (bloom backend)
+        // Expected: RG 0 and RG 1 (rows 0-100) which have "hello" term
+        let preds = vec![Expr::ScalarFunction(ScalarFunction {
+            args: vec![col("text_bloom"), "hello".lit()],
+            func: matches_term_func(),
+        })];
+
+        let fulltext_applier = FulltextIndexApplierBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            object_store.clone(),
+            env.get_puffin_manager(),
+            &metadata,
+        )
+        .with_puffin_metadata_cache(cache.puffin_metadata_cache().cloned())
+        .with_bloom_filter_cache(cache.bloom_filter_index_cache().cloned())
+        .build(&preds)
+        .unwrap()
+        .map(Arc::new);
+
+        let builder = ParquetReaderBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            handle.clone(),
+            object_store.clone(),
+        )
+        .flat_format(true)
+        .predicate(Some(Predicate::new(preds)))
+        .fulltext_index_appliers([None, fulltext_applier.clone()])
+        .cache(CacheStrategy::EnableAll(cache.clone()));
+
+        let mut metrics = ReaderMetrics::default();
+        let (_context, selection) = builder.build_reader_input(&mut metrics).await.unwrap();
+
+        // Verify selection contains RG 0 and RG 1 (text_bloom="hello world")
+        assert_eq!(selection.row_group_count(), 2);
+        assert_eq!(50, selection.get(0).unwrap().row_count());
+        assert_eq!(50, selection.get(1).unwrap().row_count());
+
+        // Verify filtering metrics
+        assert_eq!(metrics.filter_metrics.rg_total, 4);
+        assert_eq!(metrics.filter_metrics.rg_minmax_filtered, 0);
+        assert_eq!(metrics.filter_metrics.rg_fulltext_filtered, 2);
+        assert_eq!(metrics.filter_metrics.rows_fulltext_filtered, 100);
+
+        // Test 2: Filter by text_tantivy field using matches (tantivy backend)
+        // Expected: RG 2 and RG 3 (rows 100-200) which have "lazy" in query
+        let preds = vec![Expr::ScalarFunction(ScalarFunction {
+            args: vec![col("text_tantivy"), "lazy".lit()],
+            func: matches_func(),
+        })];
+
+        let fulltext_applier = FulltextIndexApplierBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            object_store.clone(),
+            env.get_puffin_manager(),
+            &metadata,
+        )
+        .with_puffin_metadata_cache(cache.puffin_metadata_cache().cloned())
+        .with_bloom_filter_cache(cache.bloom_filter_index_cache().cloned())
+        .build(&preds)
+        .unwrap()
+        .map(Arc::new);
+
+        let builder = ParquetReaderBuilder::new(
+            FILE_DIR.to_string(),
+            PathType::Bare,
+            handle.clone(),
+            object_store.clone(),
+        )
+        .flat_format(true)
+        .predicate(Some(Predicate::new(preds)))
+        .fulltext_index_appliers([None, fulltext_applier.clone()])
+        .cache(CacheStrategy::EnableAll(cache.clone()));
+
+        let mut metrics = ReaderMetrics::default();
+        let (_context, selection) = builder.build_reader_input(&mut metrics).await.unwrap();
+
+        // Verify selection contains RG 2 and RG 3 (text_tantivy="lazy dog")
+        assert_eq!(selection.row_group_count(), 2);
+        assert_eq!(50, selection.get(2).unwrap().row_count());
+        assert_eq!(50, selection.get(3).unwrap().row_count());
+
+        // Verify filtering metrics
+        assert_eq!(metrics.filter_metrics.rg_total, 4);
+        assert_eq!(metrics.filter_metrics.rg_minmax_filtered, 0);
+        assert_eq!(metrics.filter_metrics.rg_fulltext_filtered, 2);
+        assert_eq!(metrics.filter_metrics.rows_fulltext_filtered, 100);
+    }
 }
--- a/src/mito2/src/test_util/scheduler_util.rs
+++ b/src/mito2/src/test_util/scheduler_util.rs
@@ -29,7 +29,7 @@ use tokio::sync::mpsc::Sender;
 use crate::access_layer::{AccessLayer, AccessLayerRef};
 use crate::cache::CacheManager;
 use crate::compaction::CompactionScheduler;
-use crate::compaction::memory_manager::{CompactionMemoryManager, new_compaction_memory_manager};
+use crate::compaction::memory_manager::new_compaction_memory_manager;
 use crate::config::MitoConfig;
 use crate::error::Result;
 use crate::flush::FlushScheduler;
--- a/src/mito2/src/test_util/sst_util.rs
+++ b/src/mito2/src/test_util/sst_util.rs
@@ -27,6 +27,10 @@ use parquet::file::metadata::ParquetMetaData;
 use store_api::metadata::{
    ColumnMetadata, RegionMetadata, RegionMetadataBuilder, RegionMetadataRef,
 };
+use store_api::metric_engine_consts::{
+    DATA_SCHEMA_TABLE_ID_COLUMN_NAME, DATA_SCHEMA_TSID_COLUMN_NAME,
+};
+use store_api::storage::consts::ReservedColumnId;
 use store_api::storage::{FileId, RegionId};

 use crate::read::{Batch, BatchBuilder, Source};
@@ -36,11 +40,44 @@ use crate::test_util::{VecBatchReader, new_batch_builder, new_noop_file_purger};
 /// Test region id.
 const REGION_ID: RegionId = RegionId::new(0, 0);

-/// Creates a new region metadata for testing SSTs.
+/// Creates a new region metadata for testing SSTs with specified encoding.
 ///
-/// Schema: tag_0, tag_1, field_0, ts
-pub fn sst_region_metadata() -> RegionMetadata {
+/// Dense schema: tag_0, tag_1, field_0, ts
+/// Sparse schema: __table_id, __tsid, tag_0, tag_1, field_0, ts
+pub fn sst_region_metadata_with_encoding(
+    encoding: store_api::codec::PrimaryKeyEncoding,
+) -> RegionMetadata {
    let mut builder = RegionMetadataBuilder::new(REGION_ID);
+
+    // For sparse encoding, add internal columns first
+    if encoding == store_api::codec::PrimaryKeyEncoding::Sparse {
+        builder
+            .push_column_metadata(ColumnMetadata {
+                column_schema: ColumnSchema::new(
+                    DATA_SCHEMA_TABLE_ID_COLUMN_NAME.to_string(),
+                    ConcreteDataType::uint32_datatype(),
+                    false,
+                )
+                .with_skipping_options(SkippingIndexOptions {
+                    granularity: 1,
+                    ..Default::default()
+                })
+                .unwrap(),
+                semantic_type: SemanticType::Tag,
+                column_id: ReservedColumnId::table_id(),
+            })
+            .push_column_metadata(ColumnMetadata {
+                column_schema: ColumnSchema::new(
+                    DATA_SCHEMA_TSID_COLUMN_NAME.to_string(),
+                    ConcreteDataType::uint64_datatype(),
+                    false,
+                ),
+                semantic_type: SemanticType::Tag,
+                column_id: ReservedColumnId::tsid(),
+            });
+    }
+
+    // Add user-defined columns (tag_0, tag_1, field_0, ts)
    builder
        .push_column_metadata(ColumnMetadata {
            column_schema: ColumnSchema::new(
@@ -83,12 +120,32 @@ pub fn sst_region_metadata() -> RegionMetadata {
            ),
            semantic_type: SemanticType::Timestamp,
            column_id: 3,
-        })
-        .primary_key(vec![0, 1]);
+        });
+
+    // Set primary key based on encoding
+    if encoding == store_api::codec::PrimaryKeyEncoding::Sparse {
+        builder.primary_key(vec![
+            ReservedColumnId::table_id(),
+            ReservedColumnId::tsid(),
+            0, // tag_0
+            1, // tag_1
+        ]);
+    } else {
+        builder.primary_key(vec![0, 1]); // Dense: just user tags
+    }
+
+    builder.primary_key_encoding(encoding);
    builder.build().unwrap()
 }

-/// Encodes a primary key for specific tags.
+/// Creates a new region metadata for testing SSTs.
+///
+/// Schema: tag_0, tag_1, field_0, ts
+pub fn sst_region_metadata() -> RegionMetadata {
+    sst_region_metadata_with_encoding(store_api::codec::PrimaryKeyEncoding::Dense)
+}
+
+/// Encodes a primary key for specific tags using dense encoding.
 pub fn new_primary_key(tags: &[&str]) -> Vec<u8> {
    let fields = (0..tags.len())
        .map(|idx| {
@@ -104,6 +161,31 @@ pub fn new_primary_key(tags: &[&str]) -> Vec<u8> {
        .unwrap()
 }

+/// Encodes a primary key for specific tags using sparse encoding.
+/// Includes internal columns (table_id, tsid) required by sparse format.
+pub fn new_sparse_primary_key(
+    tags: &[&str],
+    metadata: &Arc<RegionMetadata>,
+    table_id: u32,
+    tsid: u64,
+) -> Vec<u8> {
+    use mito_codec::row_converter::PrimaryKeyCodec;
+
+    let codec = mito_codec::row_converter::SparsePrimaryKeyCodec::new(metadata);
+
+    // Sparse encoding requires internal columns first, then user tags
+    let values = vec![
+        (ReservedColumnId::table_id(), ValueRef::UInt32(table_id)),
+        (ReservedColumnId::tsid(), ValueRef::UInt64(tsid)),
+        (0, ValueRef::String(tags[0])), // tag_0
+        (1, ValueRef::String(tags[1])), // tag_1
+    ];
+
+    let mut buffer = Vec::new();
+    codec.encode_value_refs(&values, &mut buffer).unwrap();
+    buffer
+}
+
 /// Creates a [Source] from `batches`.
 pub fn new_source(batches: &[Batch]) -> Source {
    let reader = VecBatchReader::new(batches);
--- a/src/object-store/src/config.rs
+++ b/src/object-store/src/config.rs
@@ -117,6 +117,11 @@ pub struct S3Connection {
    /// By default, opendal will send API to https://s3.us-east-1.amazonaws.com/bucket_name
    /// Enabled, opendal will send API to https://bucket_name.s3.us-east-1.amazonaws.com
    pub enable_virtual_host_style: bool,
+    /// Disable EC2 metadata service.
+    /// By default, opendal will use EC2 metadata service to load credentials from the instance metadata,
+    /// when access key id and secret access key are not provided.
+    /// If enabled, opendal will *NOT* use EC2 metadata service.
+    pub disable_ec2_metadata: bool,
 }

 impl From<&S3Connection> for S3 {
@@ -129,6 +134,10 @@ impl From<&S3Connection> for S3 {
            .access_key_id(connection.access_key_id.expose_secret())
            .secret_access_key(connection.secret_access_key.expose_secret());

+        if connection.disable_ec2_metadata {
+            builder = builder.disable_ec2_metadata();
+        }
+
        if let Some(endpoint) = &connection.endpoint {
            builder = builder.endpoint(endpoint);
        }
--- a/src/query/src/part_sort.rs
+++ b/src/query/src/part_sort.rs
--- a/src/servers/src/grpc.rs
+++ b/src/servers/src/grpc.rs
@@ -81,6 +81,12 @@ pub struct GrpcOptions {
    /// Default to `None`, means infinite.
    #[serde(with = "humantime_serde")]
    pub max_connection_age: Option<Duration>,
+    /// The HTTP/2 keep-alive interval.
+    #[serde(with = "humantime_serde")]
+    pub http2_keep_alive_interval: Duration,
+    /// The HTTP/2 keep-alive timeout.
+    #[serde(with = "humantime_serde")]
+    pub http2_keep_alive_timeout: Duration,
 }

 impl GrpcOptions {
@@ -144,6 +150,8 @@ impl Default for GrpcOptions {
            runtime_size: 8,
            tls: TlsOption::default(),
            max_connection_age: None,
+            http2_keep_alive_interval: Duration::from_secs(10),
+            http2_keep_alive_timeout: Duration::from_secs(3),
        }
    }
 }
@@ -164,6 +172,8 @@ impl GrpcOptions {
            runtime_size: 8,
            tls: TlsOption::default(),
            max_connection_age: None,
+            http2_keep_alive_interval: Duration::from_secs(10),
+            http2_keep_alive_timeout: Duration::from_secs(3),
        }
    }

--- a/src/servers/src/heartbeat_options.rs
+++ b/src/servers/src/heartbeat_options.rs
@@ -34,12 +34,10 @@ impl HeartbeatOptions {
    pub fn frontend_default() -> Self {
        Self {
            // Frontend can send heartbeat with a longer interval.
-            interval: Duration::from_millis(
-                distributed_time_constants::FRONTEND_HEARTBEAT_INTERVAL_MILLIS,
-            ),
-            retry_interval: Duration::from_millis(
-                distributed_time_constants::HEARTBEAT_INTERVAL_MILLIS,
+            interval: distributed_time_constants::frontend_heartbeat_interval(
+                distributed_time_constants::BASE_HEARTBEAT_INTERVAL,
            ),
+            retry_interval: distributed_time_constants::BASE_HEARTBEAT_INTERVAL,
        }
    }
 }
@@ -47,10 +45,8 @@ impl HeartbeatOptions {
 impl Default for HeartbeatOptions {
    fn default() -> Self {
        Self {
-            interval: Duration::from_millis(distributed_time_constants::HEARTBEAT_INTERVAL_MILLIS),
-            retry_interval: Duration::from_millis(
-                distributed_time_constants::HEARTBEAT_INTERVAL_MILLIS,
-            ),
+            interval: distributed_time_constants::BASE_HEARTBEAT_INTERVAL,
+            retry_interval: distributed_time_constants::BASE_HEARTBEAT_INTERVAL,
        }
    }
 }
--- a/src/servers/src/prom_store.rs
+++ b/src/servers/src/prom_store.rs
@@ -26,7 +26,7 @@ use arrow::datatypes::{Float64Type, TimestampMillisecondType};
 use common_grpc::precision::Precision;
 use common_query::prelude::{greptime_timestamp, greptime_value};
 use common_recordbatch::{RecordBatch, RecordBatches};
-use common_telemetry::tracing;
+use common_telemetry::{tracing, warn};
 use datafusion::dataframe::DataFrame;
 use datafusion::prelude::{Expr, col, lit, regexp_match};
 use datafusion_common::ScalarValue;
@@ -415,6 +415,10 @@ pub fn to_grpc_row_insert_requests(request: &WriteRequest) -> Result<(RowInsertR
                table_data.add_row(one_row);
            }
        }
+
+        if !series.histograms.is_empty() {
+            warn!("Native histograms are not supported yet, data ignored");
+        }
    }

    Ok(multi_table_data.into_row_insert_requests())
--- a/src/servers/src/tls.rs
+++ b/src/servers/src/tls.rs
@@ -362,13 +362,13 @@ mod tests {
                cert_path: "/path/to/cert_path".to_string(),
                key_path: "/path/to/key_path".to_string(),
                ca_cert_path: String::new(),
-                watch: false
+                watch: false,
            },
            TlsOption::new(
                Some(Disable),
                Some("/path/to/cert_path".to_string()),
                Some("/path/to/key_path".to_string()),
-                false
+                false,
            )
        );
    }
--- a/src/sql/src/error.rs
+++ b/src/sql/src/error.rs
@@ -285,6 +285,13 @@ pub enum Error {
        location: Location,
    },

+    #[snafu(display("Failed to set VECTOR index option"))]
+    SetVectorIndexOption {
+        source: datatypes::error::Error,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
    #[snafu(display(
        "Invalid partition number: {}, should be in range [2, 65536]",
        partition_num
@@ -394,7 +401,9 @@ impl ErrorExt for Error {
            ConvertValue { .. } => StatusCode::Unsupported,

            PermissionDenied { .. } => StatusCode::PermissionDenied,
-            SetFulltextOption { .. } | SetSkippingIndexOption { .. } => StatusCode::Unexpected,
+            SetFulltextOption { .. }
+            | SetSkippingIndexOption { .. }
+            | SetVectorIndexOption { .. } => StatusCode::Unexpected,
        }
    }

--- a/src/sql/src/parsers/create_parser.rs
+++ b/src/sql/src/parsers/create_parser.rs
@@ -43,6 +43,7 @@ use crate::parser::{FLOW, ParserContext};
 use crate::parsers::tql_parser;
 use crate::parsers::utils::{
    self, validate_column_fulltext_create_option, validate_column_skipping_index_create_option,
+    validate_column_vector_index_create_option,
 };
 use crate::statements::create::{
    Column, ColumnExtensions, CreateDatabase, CreateExternalTable, CreateFlow, CreateTable,
@@ -60,6 +61,7 @@ pub const EXPIRE: &str = "EXPIRE";
 pub const AFTER: &str = "AFTER";
 pub const INVERTED: &str = "INVERTED";
 pub const SKIPPING: &str = "SKIPPING";
+pub const VECTOR: &str = "VECTOR";

 pub type RawIntervalExpr = String;

@@ -928,6 +930,61 @@ impl<'a> ParserContext<'a> {
            is_index_declared |= true;
        }

+        // vector index
+        if let Token::Word(word) = parser.peek_token().token
+            && word.value.eq_ignore_ascii_case(VECTOR)
+        {
+            parser.next_token();
+            // Consume `INDEX` keyword
+            ensure!(
+                parser.parse_keyword(Keyword::INDEX),
+                InvalidColumnOptionSnafu {
+                    name: column_name.to_string(),
+                    msg: "expect INDEX after VECTOR keyword",
+                }
+            );
+
+            ensure!(
+                column_extensions.vector_index_options.is_none(),
+                InvalidColumnOptionSnafu {
+                    name: column_name.to_string(),
+                    msg: "duplicated VECTOR INDEX option",
+                }
+            );
+
+            // Check that column is a vector type
+            let column_type = get_unalias_type(column_type);
+            let data_type = sql_data_type_to_concrete_data_type(&column_type, column_extensions)?;
+            ensure!(
+                matches!(data_type, ConcreteDataType::Vector(_)),
+                InvalidColumnOptionSnafu {
+                    name: column_name.to_string(),
+                    msg: "VECTOR INDEX only supports Vector type columns",
+                }
+            );
+
+            let options = parser
+                .parse_options(Keyword::WITH)
+                .context(error::SyntaxSnafu)?
+                .into_iter()
+                .map(parse_option_string)
+                .collect::<Result<Vec<_>>>()?;
+
+            for (key, _) in options.iter() {
+                ensure!(
+                    validate_column_vector_index_create_option(key),
+                    InvalidColumnOptionSnafu {
+                        name: column_name.to_string(),
+                        msg: format!("invalid VECTOR INDEX option: {key}"),
+                    }
+                );
+            }
+
+            let options = OptionMap::new(options);
+            column_extensions.vector_index_options = Some(options);
+            is_index_declared |= true;
+        }
+
        Ok(is_index_declared)
    }

@@ -2714,7 +2771,8 @@ CREATE TABLE log (

    #[test]
    fn test_parse_column_extensions_vector() {
-        let sql = "VECTOR(128)";
+        // Test that vector options are parsed from data_type (no additional SQL needed)
+        let sql = "";
        let dialect = GenericDialect {};
        let mut tokenizer = Tokenizer::new(&dialect, sql);
        let tokens = tokenizer.tokenize().unwrap();
@@ -2734,7 +2792,8 @@ CREATE TABLE log (

    #[test]
    fn test_parse_column_extensions_vector_invalid() {
-        let sql = "VECTOR()";
+        // Test that vector with no dimension fails
+        let sql = "";
        let dialect = GenericDialect {};
        let mut tokenizer = Tokenizer::new(&dialect, sql);
        let tokens = tokenizer.tokenize().unwrap();
@@ -2912,4 +2971,174 @@ CREATE TABLE log (
                .unwrap();
        assert_eq!("SELECT '10 seconds'::INTERVAL", &stmts[0].to_string());
    }
+
+    #[test]
+    fn test_parse_create_table_vector_index_options() {
+        // Test basic vector index
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    vec VECTOR(128) VECTOR INDEX,
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default())
+                .unwrap();
+
+        if let Statement::CreateTable(c) = &result[0] {
+            c.columns.iter().for_each(|col| {
+                if col.name().value == "vec" {
+                    assert!(
+                        col.extensions
+                            .vector_index_options
+                            .as_ref()
+                            .unwrap()
+                            .is_empty()
+                    );
+                }
+            });
+        } else {
+            panic!("should be create_table statement");
+        }
+
+        // Test vector index with options
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    vec VECTOR(128) VECTOR INDEX WITH (metric='cosine', connectivity='32', expansion_add='256', expansion_search='128')
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default())
+                .unwrap();
+
+        if let Statement::CreateTable(c) = &result[0] {
+            c.columns.iter().for_each(|col| {
+                if col.name().value == "vec" {
+                    let options = col.extensions.vector_index_options.as_ref().unwrap();
+                    assert_eq!(options.len(), 4);
+                    assert_eq!(options.get("metric").unwrap(), "cosine");
+                    assert_eq!(options.get("connectivity").unwrap(), "32");
+                    assert_eq!(options.get("expansion_add").unwrap(), "256");
+                    assert_eq!(options.get("expansion_search").unwrap(), "128");
+                }
+            });
+        } else {
+            panic!("should be create_table statement");
+        }
+    }
+
+    #[test]
+    fn test_parse_create_table_vector_index_invalid_type() {
+        // Test vector index on non-vector type (should fail)
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    col INT VECTOR INDEX,
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default());
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("VECTOR INDEX only supports Vector type columns")
+        );
+    }
+
+    #[test]
+    fn test_parse_create_table_vector_index_duplicate() {
+        // Test duplicate vector index (should fail)
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    vec VECTOR(128) VECTOR INDEX VECTOR INDEX,
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default());
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("duplicated VECTOR INDEX option")
+        );
+    }
+
+    #[test]
+    fn test_parse_create_table_vector_index_invalid_option() {
+        // Test invalid option key (should fail)
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    vec VECTOR(128) VECTOR INDEX WITH (metric='l2sq', invalid_option='foo')
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default());
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("invalid VECTOR INDEX option")
+        );
+    }
+
+    #[test]
+    fn test_parse_column_extensions_vector_index() {
+        // Test vector index on vector type
+        {
+            let sql = "VECTOR INDEX WITH (metric = 'l2sq')";
+            let dialect = GenericDialect {};
+            let mut tokenizer = Tokenizer::new(&dialect, sql);
+            let tokens = tokenizer.tokenize().unwrap();
+            let mut parser = Parser::new(&dialect).with_tokens(tokens);
+            let name = Ident::new("vec_col");
+            let data_type =
+                DataType::Custom(vec![Ident::new("VECTOR")].into(), vec!["128".to_string()]);
+            // First, parse the vector type to set vector_options
+            let mut extensions = ColumnExtensions {
+                vector_options: Some(OptionMap::from([(
+                    VECTOR_OPT_DIM.to_string(),
+                    "128".to_string(),
+                )])),
+                ..Default::default()
+            };
+
+            let result = ParserContext::parse_column_extensions(
+                &mut parser,
+                &name,
+                &data_type,
+                &mut extensions,
+            );
+            assert!(result.is_ok());
+            assert!(extensions.vector_index_options.is_some());
+            let vi_options = extensions.vector_index_options.unwrap();
+            assert_eq!(vi_options.get("metric"), Some("l2sq"));
+        }
+
+        // Test vector index on non-vector type (should fail)
+        {
+            let sql = "VECTOR INDEX";
+            let dialect = GenericDialect {};
+            let mut tokenizer = Tokenizer::new(&dialect, sql);
+            let tokens = tokenizer.tokenize().unwrap();
+            let mut parser = Parser::new(&dialect).with_tokens(tokens);
+            let name = Ident::new("num_col");
+            let data_type = DataType::Int(None); // Non-vector type
+            let mut extensions = ColumnExtensions::default();
+            let result = ParserContext::parse_column_extensions(
+                &mut parser,
+                &name,
+                &data_type,
+                &mut extensions,
+            );
+            assert!(result.is_err());
+            assert!(
+                result
+                    .unwrap_err()
+                    .to_string()
+                    .contains("VECTOR INDEX only supports Vector type columns")
+            );
+        }
+    }
 }
--- a/src/sql/src/parsers/utils.rs
+++ b/src/sql/src/parsers/utils.rs
@@ -222,6 +222,29 @@ pub fn validate_column_skipping_index_create_option(key: &str) -> bool {
    .contains(&key)
 }

+/// Valid options for VECTOR INDEX:
+/// - engine: Vector index engine (usearch)
+/// - metric: Distance metric (l2sq, cosine, inner_product)
+/// - connectivity: HNSW M parameter
+/// - expansion_add: ef_construction parameter
+/// - expansion_search: ef_search parameter
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_ENGINE: &str = "engine";
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_METRIC: &str = "metric";
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_CONNECTIVITY: &str = "connectivity";
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_EXPANSION_ADD: &str = "expansion_add";
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_EXPANSION_SEARCH: &str = "expansion_search";
+
+pub fn validate_column_vector_index_create_option(key: &str) -> bool {
+    [
+        COLUMN_VECTOR_INDEX_OPT_KEY_ENGINE,
+        COLUMN_VECTOR_INDEX_OPT_KEY_METRIC,
+        COLUMN_VECTOR_INDEX_OPT_KEY_CONNECTIVITY,
+        COLUMN_VECTOR_INDEX_OPT_KEY_EXPANSION_ADD,
+        COLUMN_VECTOR_INDEX_OPT_KEY_EXPANSION_SEARCH,
+    ]
+    .contains(&key)
+}
+
 /// Convert an [`IntervalMonthDayNano`] to a [`Duration`].
 #[cfg(feature = "enterprise")]
 pub fn convert_month_day_nano_to_duration(
--- a/src/sql/src/statements.rs
+++ b/src/sql/src/statements.rs
@@ -55,7 +55,7 @@ use crate::ast::{
 use crate::error::{
    self, ConvertToGrpcDataTypeSnafu, ConvertValueSnafu, Result,
    SerializeColumnDefaultConstraintSnafu, SetFulltextOptionSnafu, SetJsonStructureSettingsSnafu,
-    SetSkippingIndexOptionSnafu, SqlCommonSnafu,
+    SetSkippingIndexOptionSnafu, SetVectorIndexOptionSnafu, SqlCommonSnafu,
 };
 use crate::statements::create::{Column, ColumnExtensions};
 pub use crate::statements::option_map::OptionMap;
@@ -147,6 +147,12 @@ pub fn column_to_schema(
            .context(SetSkippingIndexOptionSnafu)?;
    }

+    if let Some(options) = column.extensions.build_vector_index_options()? {
+        column_schema = column_schema
+            .with_vector_index_options(&options)
+            .context(SetVectorIndexOptionSnafu)?;
+    }
+
    column_schema.set_inverted_index(column.extensions.inverted_index_options.is_some());

    if matches!(column.data_type(), SqlDataType::JSON) {
@@ -710,6 +716,7 @@ mod tests {
                skipping_index_options: None,
                inverted_index_options: None,
                json_datatype_options: None,
+                vector_index_options: None,
            },
        };

@@ -720,4 +727,82 @@ mod tests {
        assert_eq!(fulltext_options.analyzer, FulltextAnalyzer::English);
        assert!(fulltext_options.case_sensitive);
    }
+
+    #[test]
+    fn test_column_to_schema_with_vector_index() {
+        use datatypes::schema::{VectorDistanceMetric, VectorIndexEngineType};
+
+        // Test with custom metric and parameters
+        let column = Column {
+            column_def: ColumnDef {
+                name: "embedding".into(),
+                data_type: SqlDataType::Custom(
+                    vec![Ident::new(VECTOR_TYPE_NAME)].into(),
+                    vec!["128".to_string()],
+                ),
+                options: vec![],
+            },
+            extensions: ColumnExtensions {
+                fulltext_index_options: None,
+                vector_options: None,
+                skipping_index_options: None,
+                inverted_index_options: None,
+                json_datatype_options: None,
+                vector_index_options: Some(OptionMap::from([
+                    ("metric".to_string(), "cosine".to_string()),
+                    ("connectivity".to_string(), "32".to_string()),
+                    ("expansion_add".to_string(), "200".to_string()),
+                    ("expansion_search".to_string(), "100".to_string()),
+                ])),
+            },
+        };
+
+        let column_schema = column_to_schema(&column, "ts", None).unwrap();
+        assert_eq!("embedding", column_schema.name);
+        assert!(column_schema.is_vector_indexed());
+
+        let vector_options = column_schema.vector_index_options().unwrap().unwrap();
+        assert_eq!(vector_options.engine, VectorIndexEngineType::Usearch);
+        assert_eq!(vector_options.metric, VectorDistanceMetric::Cosine);
+        assert_eq!(vector_options.connectivity, 32);
+        assert_eq!(vector_options.expansion_add, 200);
+        assert_eq!(vector_options.expansion_search, 100);
+    }
+
+    #[test]
+    fn test_column_to_schema_with_vector_index_defaults() {
+        use datatypes::schema::{VectorDistanceMetric, VectorIndexEngineType};
+
+        // Test with default values (empty options map)
+        let column = Column {
+            column_def: ColumnDef {
+                name: "vec".into(),
+                data_type: SqlDataType::Custom(
+                    vec![Ident::new(VECTOR_TYPE_NAME)].into(),
+                    vec!["64".to_string()],
+                ),
+                options: vec![],
+            },
+            extensions: ColumnExtensions {
+                fulltext_index_options: None,
+                vector_options: None,
+                skipping_index_options: None,
+                inverted_index_options: None,
+                json_datatype_options: None,
+                vector_index_options: Some(OptionMap::default()),
+            },
+        };
+
+        let column_schema = column_to_schema(&column, "ts", None).unwrap();
+        assert_eq!("vec", column_schema.name);
+        assert!(column_schema.is_vector_indexed());
+
+        let vector_options = column_schema.vector_index_options().unwrap().unwrap();
+        // Verify defaults
+        assert_eq!(vector_options.engine, VectorIndexEngineType::Usearch);
+        assert_eq!(vector_options.metric, VectorDistanceMetric::L2sq);
+        assert_eq!(vector_options.connectivity, 16);
+        assert_eq!(vector_options.expansion_add, 128);
+        assert_eq!(vector_options.expansion_search, 64);
+    }
 }
--- a/src/sql/src/statements/create.rs
+++ b/src/sql/src/statements/create.rs
@@ -17,7 +17,10 @@ use std::fmt::{Display, Formatter};

 use common_catalog::consts::FILE_ENGINE;
 use datatypes::json::JsonStructureSettings;
-use datatypes::schema::{FulltextOptions, SkippingIndexOptions};
+use datatypes::schema::{
+    FulltextOptions, SkippingIndexOptions, VectorDistanceMetric, VectorIndexEngineType,
+    VectorIndexOptions,
+};
 use itertools::Itertools;
 use serde::Serialize;
 use snafu::ResultExt;
@@ -133,6 +136,8 @@ pub struct ColumnExtensions {
    ///
    /// Inverted index doesn't have options at present. There won't be any options in that map.
    pub inverted_index_options: Option<OptionMap>,
+    /// Vector index options for HNSW-based vector similarity search.
+    pub vector_index_options: Option<OptionMap>,
    pub json_datatype_options: Option<OptionMap>,
 }

@@ -208,6 +213,15 @@ impl Display for Column {
                write!(f, " INVERTED INDEX")?;
            }
        }
+
+        if let Some(vector_index_options) = &self.extensions.vector_index_options {
+            if !vector_index_options.is_empty() {
+                let options = vector_index_options.kv_pairs();
+                write!(f, " VECTOR INDEX WITH({})", format_list_comma!(options))?;
+            } else {
+                write!(f, " VECTOR INDEX")?;
+            }
+        }
        Ok(())
    }
 }
@@ -233,6 +247,89 @@ impl ColumnExtensions {
        ))
    }

+    pub fn build_vector_index_options(&self) -> Result<Option<VectorIndexOptions>> {
+        let Some(options) = self.vector_index_options.as_ref() else {
+            return Ok(None);
+        };
+
+        let options_map: HashMap<String, String> = options.clone().into_map();
+        let mut result = VectorIndexOptions::default();
+
+        if let Some(s) = options_map.get("engine") {
+            result.engine = s.parse::<VectorIndexEngineType>().map_err(|e| {
+                InvalidSqlSnafu {
+                    msg: format!("invalid VECTOR INDEX engine: {e}"),
+                }
+                .build()
+            })?;
+        }
+
+        if let Some(s) = options_map.get("metric") {
+            result.metric = s.parse::<VectorDistanceMetric>().map_err(|e| {
+                InvalidSqlSnafu {
+                    msg: format!("invalid VECTOR INDEX metric: {e}"),
+                }
+                .build()
+            })?;
+        }
+
+        if let Some(s) = options_map.get("connectivity") {
+            let value = s.parse::<u32>().map_err(|_| {
+                InvalidSqlSnafu {
+                    msg: format!(
+                        "invalid VECTOR INDEX connectivity: {s}, expected positive integer"
+                    ),
+                }
+                .build()
+            })?;
+            if !(2..=2048).contains(&value) {
+                return InvalidSqlSnafu {
+                    msg: "VECTOR INDEX connectivity must be in the range [2, 2048].".to_string(),
+                }
+                .fail();
+            }
+            result.connectivity = value;
+        }
+
+        if let Some(s) = options_map.get("expansion_add") {
+            let value = s.parse::<u32>().map_err(|_| {
+                InvalidSqlSnafu {
+                    msg: format!(
+                        "invalid VECTOR INDEX expansion_add: {s}, expected positive integer"
+                    ),
+                }
+                .build()
+            })?;
+            if value == 0 {
+                return InvalidSqlSnafu {
+                    msg: "VECTOR INDEX expansion_add must be greater than 0".to_string(),
+                }
+                .fail();
+            }
+            result.expansion_add = value;
+        }
+
+        if let Some(s) = options_map.get("expansion_search") {
+            let value = s.parse::<u32>().map_err(|_| {
+                InvalidSqlSnafu {
+                    msg: format!(
+                        "invalid VECTOR INDEX expansion_search: {s}, expected positive integer"
+                    ),
+                }
+                .build()
+            })?;
+            if value == 0 {
+                return InvalidSqlSnafu {
+                    msg: "VECTOR INDEX expansion_search must be greater than 0".to_string(),
+                }
+                .fail();
+            }
+            result.expansion_search = value;
+        }
+
+        Ok(Some(result))
+    }
+
    pub fn build_json_structure_settings(&self) -> Result<Option<JsonStructureSettings>> {
        let Some(options) = self.json_datatype_options.as_ref() else {
            return Ok(None);
@@ -893,4 +990,92 @@ AS SELECT number FROM numbers_input where number > 10"#,
            _ => unreachable!(),
        }
    }
+
+    #[test]
+    fn test_vector_index_options_validation() {
+        use super::{ColumnExtensions, OptionMap};
+
+        // Test zero connectivity should fail
+        let extensions = ColumnExtensions {
+            fulltext_index_options: None,
+            vector_options: None,
+            skipping_index_options: None,
+            inverted_index_options: None,
+            json_datatype_options: None,
+            vector_index_options: Some(OptionMap::from([(
+                "connectivity".to_string(),
+                "0".to_string(),
+            )])),
+        };
+        let result = extensions.build_vector_index_options();
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("connectivity must be in the range [2, 2048]")
+        );
+
+        // Test zero expansion_add should fail
+        let extensions = ColumnExtensions {
+            fulltext_index_options: None,
+            vector_options: None,
+            skipping_index_options: None,
+            inverted_index_options: None,
+            json_datatype_options: None,
+            vector_index_options: Some(OptionMap::from([(
+                "expansion_add".to_string(),
+                "0".to_string(),
+            )])),
+        };
+        let result = extensions.build_vector_index_options();
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("expansion_add must be greater than 0")
+        );
+
+        // Test zero expansion_search should fail
+        let extensions = ColumnExtensions {
+            fulltext_index_options: None,
+            vector_options: None,
+            skipping_index_options: None,
+            inverted_index_options: None,
+            json_datatype_options: None,
+            vector_index_options: Some(OptionMap::from([(
+                "expansion_search".to_string(),
+                "0".to_string(),
+            )])),
+        };
+        let result = extensions.build_vector_index_options();
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("expansion_search must be greater than 0")
+        );
+
+        // Test valid values should succeed
+        let extensions = ColumnExtensions {
+            fulltext_index_options: None,
+            vector_options: None,
+            skipping_index_options: None,
+            inverted_index_options: None,
+            json_datatype_options: None,
+            vector_index_options: Some(OptionMap::from([
+                ("connectivity".to_string(), "32".to_string()),
+                ("expansion_add".to_string(), "200".to_string()),
+                ("expansion_search".to_string(), "100".to_string()),
+            ])),
+        };
+        let result = extensions.build_vector_index_options();
+        assert!(result.is_ok());
+        let options = result.unwrap().unwrap();
+        assert_eq!(options.connectivity, 32);
+        assert_eq!(options.expansion_add, 200);
+        assert_eq!(options.expansion_search, 100);
+    }
 }
--- a/src/store-api/src/storage.rs
+++ b/src/store-api/src/storage.rs
@@ -27,5 +27,8 @@ pub use datatypes::schema::{

 pub use self::descriptors::*;
 pub use self::file::{FileId, FileRef, FileRefsManifest, GcReport, IndexVersion, ParseIdError};
-pub use self::requests::{ScanRequest, TimeSeriesDistribution, TimeSeriesRowSelector};
+pub use self::requests::{
+    ScanRequest, TimeSeriesDistribution, TimeSeriesRowSelector, VectorDistanceMetric,
+    VectorIndexEngine, VectorIndexEngineType, VectorSearchMatches, VectorSearchRequest,
+};
 pub use self::types::{SequenceNumber, SequenceRange};
--- a/src/store-api/src/storage/requests.rs
+++ b/src/store-api/src/storage/requests.rs
@@ -14,11 +14,66 @@

 use std::fmt::{Display, Formatter};

+use common_error::ext::BoxedError;
 use common_recordbatch::OrderOption;
 use datafusion_expr::expr::Expr;
+// Re-export vector types from datatypes to avoid duplication
+pub use datatypes::schema::{VectorDistanceMetric, VectorIndexEngineType};
 use strum::Display;

-use crate::storage::SequenceNumber;
+use crate::storage::{ColumnId, SequenceNumber};
+
+/// A hint for KNN vector search.
+#[derive(Debug, Clone, PartialEq)]
+pub struct VectorSearchRequest {
+    /// Column ID of the vector column to search.
+    pub column_id: ColumnId,
+    /// The query vector to search for.
+    pub query_vector: Vec<f32>,
+    /// Number of nearest neighbors to return.
+    pub k: usize,
+    /// Distance metric to use (matches the index metric).
+    pub metric: VectorDistanceMetric,
+}
+
+/// Search results from vector index.
+#[derive(Debug, Clone, PartialEq)]
+pub struct VectorSearchMatches {
+    /// Keys (row offsets in the index).
+    pub keys: Vec<u64>,
+    /// Distances from the query vector.
+    pub distances: Vec<f32>,
+}
+
+/// Trait for vector index engines (HNSW implementations).
+///
+/// This trait defines the interface for pluggable vector index engines.
+/// Implementations (e.g., UsearchEngine) are provided by storage engines like mito2.
+pub trait VectorIndexEngine: Send + Sync {
+    /// Adds a vector with the given key.
+    fn add(&mut self, key: u64, vector: &[f32]) -> Result<(), BoxedError>;
+
+    /// Searches for k nearest neighbors.
+    fn search(&self, query: &[f32], k: usize) -> Result<VectorSearchMatches, BoxedError>;
+
+    /// Returns the serialized length.
+    fn serialized_length(&self) -> usize;
+
+    /// Serializes the index to a buffer.
+    fn save_to_buffer(&self, buffer: &mut [u8]) -> Result<(), BoxedError>;
+
+    /// Reserves capacity for vectors.
+    fn reserve(&mut self, capacity: usize) -> Result<(), BoxedError>;
+
+    /// Returns current size (number of vectors).
+    fn size(&self) -> usize;
+
+    /// Returns current capacity.
+    fn capacity(&self) -> usize;
+
+    /// Returns memory usage in bytes.
+    fn memory_usage(&self) -> usize;
+}

 /// A hint on how to select rows from a time-series.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Display)]
@@ -38,7 +93,7 @@ pub enum TimeSeriesDistribution {
    PerSeries,
 }

-#[derive(Default, Clone, Debug, PartialEq, Eq)]
+#[derive(Default, Clone, Debug, PartialEq)]
 pub struct ScanRequest {
    /// Indices of columns to read, `None` to read all columns. This indices is
    /// based on table schema.
@@ -66,6 +121,9 @@ pub struct ScanRequest {
    pub sst_min_sequence: Option<SequenceNumber>,
    /// Optional hint for the distribution of time-series data.
    pub distribution: Option<TimeSeriesDistribution>,
+    /// Optional hint for KNN vector search. When set, the scan should use
+    /// vector index to find the k nearest neighbors.
+    pub vector_search: Option<VectorSearchRequest>,
 }

 impl Display for ScanRequest {
@@ -138,6 +196,16 @@ impl Display for ScanRequest {
        if let Some(distribution) = &self.distribution {
            write!(f, "{}distribution: {}", delimiter.as_str(), distribution)?;
        }
+        if let Some(vector_search) = &self.vector_search {
+            write!(
+                f,
+                "{}vector_search: column_id={}, k={}, metric={}",
+                delimiter.as_str(),
+                vector_search.column_id,
+                vector_search.k,
+                vector_search.metric
+            )?;
+        }
        write!(f, " }}")
    }
 }
--- a/tests-fuzz/targets/failover/fuzz_failover_metric_regions.rs
+++ b/tests-fuzz/targets/failover/fuzz_failover_metric_regions.rs
@@ -19,7 +19,7 @@ use std::sync::Arc;
 use std::time::Duration;

 use arbitrary::{Arbitrary, Unstructured};
-use common_meta::distributed_time_constants;
+use common_meta::distributed_time_constants::default_distributed_time_constants;
 use common_telemetry::info;
 use libfuzzer_sys::fuzz_target;
 use rand::{Rng, SeedableRng};
@@ -254,10 +254,7 @@ async fn execute_failover(ctx: FuzzContext, input: FuzzInput) -> Result<()> {
    recover_pod_failure(ctx.kube.clone(), &ctx.namespace, &chaos_name).await?;
    wait_for_all_datanode_online(ctx.greptime.clone(), Duration::from_secs(60)).await;

-    tokio::time::sleep(Duration::from_secs(
-        distributed_time_constants::REGION_LEASE_SECS,
-    ))
-    .await;
+    tokio::time::sleep(default_distributed_time_constants().region_lease).await;
    // Validates value rows
    info!("Validates num of rows");

--- a/tests-fuzz/targets/failover/fuzz_failover_mito_regions.rs
+++ b/tests-fuzz/targets/failover/fuzz_failover_mito_regions.rs
@@ -19,7 +19,7 @@ use std::sync::Arc;
 use std::time::Duration;

 use arbitrary::{Arbitrary, Unstructured};
-use common_meta::distributed_time_constants;
+use common_meta::distributed_time_constants::default_distributed_time_constants;
 use common_telemetry::info;
 use common_time::util::current_time_millis;
 use futures::future::try_join_all;
@@ -322,10 +322,7 @@ async fn execute_failover(ctx: FuzzContext, input: FuzzInput) -> Result<()> {
    recover_pod_failure(ctx.kube.clone(), &ctx.namespace, &chaos_name).await?;
    wait_for_all_datanode_online(ctx.greptime.clone(), Duration::from_secs(60)).await;

-    tokio::time::sleep(Duration::from_secs(
-        distributed_time_constants::REGION_LEASE_SECS,
-    ))
-    .await;
+    tokio::time::sleep(default_distributed_time_constants().region_lease).await;
    // Validates value rows
    info!("Validates num of rows");
    for (table_ctx, expected_rows) in table_ctxs.iter().zip(affected_rows) {
--- a/tests-fuzz/targets/migration/fuzz_migrate_metric_regions.rs
+++ b/tests-fuzz/targets/migration/fuzz_migrate_metric_regions.rs
@@ -19,7 +19,7 @@ use std::sync::Arc;
 use std::time::Duration;

 use arbitrary::{Arbitrary, Unstructured};
-use common_meta::distributed_time_constants;
+use common_meta::distributed_time_constants::default_distributed_time_constants;
 use common_telemetry::info;
 use libfuzzer_sys::fuzz_target;
 use rand::{Rng, SeedableRng};
@@ -275,10 +275,7 @@ async fn migrate_regions(ctx: &FuzzContext, migrations: &[Migration]) -> Result<
        wait_for_migration(ctx, migration, &procedure_id).await;
    }

-    tokio::time::sleep(Duration::from_secs(
-        distributed_time_constants::REGION_LEASE_SECS,
-    ))
-    .await;
+    tokio::time::sleep(default_distributed_time_constants().region_lease).await;

    Ok(())
 }
--- a/tests-fuzz/targets/migration/fuzz_migrate_mito_regions.rs
+++ b/tests-fuzz/targets/migration/fuzz_migrate_mito_regions.rs
@@ -19,7 +19,7 @@ use std::sync::Arc;
 use std::time::Duration;

 use arbitrary::{Arbitrary, Unstructured};
-use common_meta::distributed_time_constants;
+use common_meta::distributed_time_constants::default_distributed_time_constants;
 use common_telemetry::info;
 use libfuzzer_sys::fuzz_target;
 use rand::{Rng, SeedableRng};
@@ -274,10 +274,7 @@ async fn migrate_regions(ctx: &FuzzContext, migrations: &[Migration]) -> Result<
        .await;
    }

-    tokio::time::sleep(Duration::from_secs(
-        distributed_time_constants::REGION_LEASE_SECS,
-    ))
-    .await;
+    tokio::time::sleep(default_distributed_time_constants().region_lease).await;

    Ok(())
 }
--- a/tests-integration/src/standalone.rs
+++ b/tests-integration/src/standalone.rs
@@ -259,9 +259,8 @@ impl GreptimeDbStandaloneBuilder {
        let grpc_handler = instance.clone() as Arc<dyn GrpcQueryHandlerWithBoxedError>;
        let weak_grpc_handler = Arc::downgrade(&grpc_handler);
        frontend_instance_handler
-            .lock()
-            .unwrap()
-            .replace(weak_grpc_handler);
+            .set_handler(weak_grpc_handler)
+            .await;

        let flow_streaming_engine = flownode.flow_engine().streaming_engine();
        let invoker = flow::FrontendInvoker::build_from(
--- a/tests-integration/tests/http.rs
+++ b/tests-integration/tests/http.rs
@@ -1397,6 +1397,8 @@ max_recv_message_size = "512MiB"
 max_send_message_size = "512MiB"
 flight_compression = "arrow_ipc"
 runtime_size = 8
+http2_keep_alive_interval = "10s"
+http2_keep_alive_timeout = "3s"

 [grpc.tls]
 mode = "disable"
@@ -1586,6 +1588,7 @@ fn drop_lines_with_inconsistent_results(input: String) -> String {
        "endpoint =",
        "region =",
        "enable_virtual_host_style =",
+        "disable_ec2_metadata =",
        "cache_path =",
        "cache_capacity =",
        "memory_pool_size =",
Author	SHA1	Message	Date
discord9	1d3cfdc0e5	clippy Signed-off-by: discord9 <discord9@163.com>	2025-12-22 21:07:24 +08:00
discord9	088401c3e9	c Signed-off-by: discord9 <discord9@163.com>	2025-12-22 21:05:23 +08:00
discord9	4419e0254f	refactor: add list test Signed-off-by: discord9 <discord9@163.com>	2025-12-22 21:05:13 +08:00
discord9	709ccd3e31	c Signed-off-by: discord9 <discord9@163.com>	2025-12-22 18:39:14 +08:00
discord9	5b50b4824d	wt Signed-off-by: discord9 <discord9@163.com>	2025-12-22 18:35:19 +08:00
discord9	1ef5c2e024	chore Signed-off-by: discord9 <discord9@163.com>	2025-12-22 18:22:03 +08:00
discord9	d20727f335	test: better fuzz Signed-off-by: discord9 <discord9@163.com>	2025-12-22 18:15:02 +08:00
discord9	2391ab1941	even more test Signed-off-by: discord9 <discord9@163.com>	2025-12-22 16:21:27 +08:00
discord9	ec77a5d53a	sanity check Signed-off-by: discord9 <discord9@163.com>	2025-12-22 13:54:19 +08:00
discord9	dbad96eb80	more test Signed-off-by: discord9 <discord9@163.com>	2025-12-22 13:44:00 +08:00
discord9	c0652f6dd5	chore: release push check against Cargo.toml (#7426 ) Signed-off-by: discord9 <discord9@163.com>	2025-12-19 13:16:15 +00:00
Yingwen	fed6cb0806	fix: flat format use correct encoding in indexer for tags (#7440 ) * test: add inverted and skipping test Signed-off-by: evenyag <realevenyag@gmail.com> * test: Add tests for fulltext index Signed-off-by: evenyag <realevenyag@gmail.com> * fix: index dictionary type in correct encoding in flat format Signed-off-by: evenyag <realevenyag@gmail.com> * refactor: use encode_data_type() in SortField Signed-off-by: evenyag <realevenyag@gmail.com> * refactor: refine imports Signed-off-by: evenyag <realevenyag@gmail.com> * test: add tests for sparse encoding Signed-off-by: evenyag <realevenyag@gmail.com> * chore: remove logs Signed-off-by: evenyag <realevenyag@gmail.com> * test: update list test Signed-off-by: evenyag <realevenyag@gmail.com> * test: simplify tests Signed-off-by: evenyag <realevenyag@gmail.com> --------- Signed-off-by: evenyag <realevenyag@gmail.com>	2025-12-19 07:36:44 +00:00
discord9	69659211f6	chore: fix bincode version (#7445 ) Signed-off-by: discord9 <discord9@163.com>	2025-12-19 07:36:28 +00:00
LFC	6332d91884	test: reduce execution time of test `test_suspend_frontend` (#7444 ) Signed-off-by: luofucong <luofc@foxmail.com>	2025-12-19 07:25:36 +00:00
Weny Xu	4d66bd96b8	feat: make distributed time constants and client timeouts configurable (#7433 ) Signed-off-by: WenyXu <wenymedia@gmail.com>	2025-12-19 02:23:20 +00:00
Ning Sun	2f4a15ec40	ci: ensure commits from main branch for whitelisted git dependencies (#7434 ) * chore: update proto to include native histogram * ci: add a CI check to ensure whitelisted dependencies are using their main branch * chore: add changes to Cargo.toml to trigger CI * chore: update proto * test: update test to include histogram	2025-12-18 14:10:33 +00:00
Lanqing Yang	658332fe68	chore(mito): nit remove extra hashset in gc workers (#7399 ) chore(mito): remove extra hashset in gc workers Signed-off-by: lyang24 <lanqingy93@gmail.com>	2025-12-18 13:09:32 +00:00
shuiyisong	c088d361a4	chore: expose `disable_ec2_metadata` option (#7439 ) chore: add option for disable ec2 metadata Signed-off-by: shuiyisong <xixing.sys@gmail.com>	2025-12-18 11:55:08 +00:00
shuiyisong	a85864067e	chore: remove `canonicalize` (#7430 ) * chore: remove canonicalize Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: add match file name option Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: update field name Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: modify tls option Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: update config file Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: update config md Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: update option to `enable_filename_match` Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: address CR issues Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: remove option Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: remove unused test Signed-off-by: shuiyisong <xixing.sys@gmail.com> --------- Signed-off-by: shuiyisong <xixing.sys@gmail.com>	2025-12-18 09:39:10 +00:00
LFC	0df69c95aa	chore: use official etcd-client (#7432 ) Signed-off-by: luofucong <luofc@foxmail.com>	2025-12-18 06:25:48 +00:00
McKnight22	72eede8b38	refactor(cli): unify storage configuration for export command (#7280 ) * refactor(cli): unify storage configuration for export command - Utilize ObjectStoreConfig to unify storage configuration for export command - Support export command for Fs, S3, OSS, GCS and Azblob - Fix the Display implementation for SecretString always returned the string "SecretString([REDACTED])" even when the internal secret was empty. Signed-off-by: McKnight22 <tao.wang.22@outlook.com> * refactor(cli): unify storage configuration for export command - Change the encapsulation permissions of each configuration options for every storage backend to public access. Signed-off-by: McKnight22 <tao.wang.22@outlook.com> Co-authored-by: WenyXu <wenymedia@gmail.com> * refactor(cli): unify storage configuration for export command - Update the implementation of ObjectStoreConfig::build_xxx() using macro solutions Signed-off-by: McKnight22 <tao.wang.22@outlook.com> Co-authored-by: WenyXu <wenymedia@gmail.com> * refactor(cli): unify storage configuration for export command - Introduce config validation for each storage type Signed-off-by: McKnight22 <tao.wang.22@outlook.com> * refactor(cli): unify storage configuration for export command - Enable trait-based polymorphism for storage type handling (from inherent impl to trait impl) - Extract helper functions to reduce code duplication Signed-off-by: McKnight22 <tao.wang.22@outlook.com> * refactor(cli): unify storage configuration for export command - Improve SecretString handling and validation (Distinguishing between "not provided" and "empty string") - Add validation when using filesystem storage Signed-off-by: McKnight22 <tao.wang.22@outlook.com> * refactor(cli): unify storage configuration for export command - Refactor storage field validation with macro Signed-off-by: McKnight22 <tao.wang.22@outlook.com> * refactor(cli): unify storage configuration for export command - support GCS Application Default Credentials (like GKE, Cloud Run, or local development with ) in export (Enabling ADC without validating or to be present) (Making optional in GCS validation (defaults to https://storage.googleapis.com)) Signed-off-by: McKnight22 <tao.wang.22@outlook.com> * refactor(cli): unify storage configuration for export command This commit refactors the validation logic for object store configurations in the CLI to leverage clap features and reduce boilerplate. Key changes: - Update wrap_with_clap_prefix macro to use clap's requires attribute. This ensures that storage-specific options (e.g., --s3-bucket) are only accepted when the corresponding backend is enabled (e.g., --s3). - Simplify FieldValidator trait by removing the is_provided method, as dependency checks are now handled by clap. - Introduce validate_backend! macro to standardize the validation of required fields for enabled backends. - Refactor ExportCommand to remove explicit validation calls (validate_s3, etc.) and rely on the validation within backend constructors. - Add integration tests for ExportCommand to verify build success with S3, OSS, GCS, and Azblob configurations. Signed-off-by: McKnight22 <tao.wang.22@outlook.com> * refactor(cli): unify storage configuration for export command - Use macros to simplify storage export implementation Signed-off-by: McKnight22 <tao.wang.22@outlook.com> Co-authored-by: WenyXu <wenymedia@gmail.com> * refactor(cli): unify storage configuration for export command - Rollback StorageExport trait implementation to not using macro for better code clarity and maintainability - Introduce format_uri helper function to unify URI formatting logic - Fix OSS URI path bug inherited from legacy code Signed-off-by: McKnight22 <tao.wang.22@outlook.com> Co-authored-by: WenyXu <wenymedia@gmail.com> * refactor(cli): unify storage configuration for export command - Remove unnecessary async_trait Signed-off-by: McKnight22 <tao.wang.22@outlook.com> Co-authored-by: jeremyhi <jiachun_feng@proton.me> --------- Signed-off-by: McKnight22 <tao.wang.22@outlook.com> Co-authored-by: WenyXu <wenymedia@gmail.com> Co-authored-by: jeremyhi <jiachun_feng@proton.me>	2025-12-18 03:16:53 +00:00
jeremyhi	95eccd6cde	feat: introduce granularity for memory manager (#7416 ) * feat: introduce granularity for memory manager Signed-off-by: jeremyhi <fengjiachun@gmail.com> * chore: add unit test Signed-off-by: jeremyhi <fengjiachun@gmail.com> * chore: remove granularity getter for mamanger Signed-off-by: jeremyhi <fengjiachun@gmail.com> * Update src/common/memory-manager/src/manager.rs Co-authored-by: Lei, HUANG <6406592+v0y4g3r@users.noreply.github.com> * feat: acquire_with_policy for manager Signed-off-by: jeremyhi <fengjiachun@gmail.com> --------- Signed-off-by: jeremyhi <fengjiachun@gmail.com> Co-authored-by: Lei, HUANG <6406592+v0y4g3r@users.noreply.github.com>	2025-12-17 11:08:51 +00:00
fys	0bc5a305be	chore: add wait_initialized method for frontend client (#7414 ) * chore: add wait_initialized method for frontend client * fix: some * fix: cargo fmt * add comment * add unit test * rename * fix: cargo check * fix: cr by copilot	2025-12-17 08:13:36 +00:00
discord9	1afcddd5a9	chore: feature gate vector_index (#7428 ) Signed-off-by: discord9 <discord9@163.com>	2025-12-17 07:14:25 +00:00
shuiyisong	62808b887b	fix: using anonymous s3 access when ak and sk is not provided (#7425 ) * chore: allow s3 anon Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: disable ec2 metadata Signed-off-by: shuiyisong <xixing.sys@gmail.com> --------- Signed-off-by: shuiyisong <xixing.sys@gmail.com>	2025-12-17 06:34:29 +00:00
discord9	04ddd40e00	chore: bump version to beta.3 (#7423 ) chore: bump to beta.3 Signed-off-by: discord9 <discord9@163.com>	2025-12-17 04:18:23 +00:00
liyang	b4f028be5f	chore: change etcd endpoints to array in the test scripts (#7419 ) chore: change etcd endpoint Signed-off-by: liyang <daviderli614@gmail.com>	2025-12-17 03:14:35 +00:00
Lei, HUANG	da964880f5	chore: expose symbols (#7417 ) * refactor/expose-symbols: ## Refactor `bulk/part.rs` to Simplify Mutation Handling - Removed the `mutations_to_record_batch` function and its associated helper functions, including `ArraysSorter`, `timestamp_array_to_iter`, and `binary_array_to_dictionary`, to simplify the mutation handling logic in `bulk/part.rs`. - Deleted related test functions `check_binary_array_to_dictionary` and `check_mutations_to_record_batches` from the test module, along with their associated test cases. Signed-off-by: Lei, HUANG <mrsatangel@gmail.com> * refactor/expose-symbols: ### Commit Message Refactor and Enhance Deduplication Logic - `flush.rs`: Refactored `maybe_dedup_one` function to accept `append_mode` and `merge_mode` as parameters instead of `RegionOptions`. This change enhances flexibility in deduplication logic. - `memtable/bulk.rs`: Made `BulkRangeIterBuilder` struct and its fields public to allow external access and modification, improving extensibility. - `sst.rs`: Corrected a typo in the schema documentation, changing `__prmary_key` to `__primary_key` for clarity and accuracy. Signed-off-by: Lei, HUANG <mrsatangel@gmail.com> --------- Signed-off-by: Lei, HUANG <mrsatangel@gmail.com>	2025-12-17 01:29:36 +00:00
dennis zhuang	a35a39f726	feat(vector_index): adds the foundational types and SQL parsing support for vector index (#7366 ) * feat: adds the foundational types and SQL parsing support for vector index Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * refactor: by suggestions Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * fix: ensure index option values must be greater than zero Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * chore: validate connectivity strictly Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * fix: compile error Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * feat: disable SIMD for ci Signed-off-by: Dennis Zhuang <killme2008@gmail.com> --------- Signed-off-by: Dennis Zhuang <killme2008@gmail.com>	2025-12-16 22:45:36 +00:00