fix(integration): stage PR #367 on big-main

Signed-off-by: XiaoSeS <87064762+XiaoSeS@users.noreply.github.com>

deploy/k8s/README.md

@@ -236,6 +236,12 @@ skillhub-storage-s3-secret-key: your-secret-key
 | postgres-data-0 | 10Gi | PostgreSQL 数据（with-infra only） |
 | redis-data-0 | 5Gi | Redis 数据（with-infra only） |
 
+#### PostgreSQL 数据目录兼容性
+
+`with-infra` 会在启动时检查 PostgreSQL PVC 根目录：如果已存在 `PG_VERSION`，继续使用根目录中的旧集群；否则在 `pgdata/` 子目录初始化新集群，避免新 ext4 卷中的 `lost+found` 阻止 `initdb`。升级现有部署不需要移动数据库文件。
+
+回滚到不包含该检测逻辑的旧清单时，如果集群位于 `pgdata/`，必须保留当前启动命令，或显式设置 `PGDATA=/var/lib/postgresql/data/pgdata`。不要把正在运行的数据库目录手动移动到 PVC 根目录。
+
 ## 镜像说明
 
 | 组件 | 镜像 |

deploy/k8s/overlays/with-infra/postgres-statefulset.yaml

@@ -18,6 +18,19 @@ spec:
       containers:
         - name: postgres
           image: postgres:16-alpine
+          # Legacy installations store PostgreSQL directly at the PVC root.
+          # Fresh ext4 volumes may contain lost+found, so initialize new clusters
+          # in a child directory without hiding an existing root-level cluster.
+          command:
+            - sh
+            - -ec
+            - |
+              if [ -f /var/lib/postgresql/data/PG_VERSION ]; then
+                export PGDATA=/var/lib/postgresql/data
+              else
+                export PGDATA=/var/lib/postgresql/data/pgdata
+              fi
+              exec docker-entrypoint.sh postgres
           ports:
             - containerPort: 5432
               name: postgres

docs/skillhub/en/guide/kubernetes.md

@@ -214,6 +214,12 @@ skillhub-storage-s3-secret-key: your-secret-key
   value: us-east-1
 ```
 
+### PostgreSQL data-directory compatibility
+
+The `with-infra` overlay checks the PostgreSQL PVC root at startup. If `PG_VERSION` already exists there, the legacy root-level cluster remains in use. Otherwise, a new cluster is initialized under `pgdata/`, avoiding the `lost+found` directory that can block `initdb` on a fresh ext4 volume. Existing installations do not need to move their database files during upgrade.
+
+When rolling back to an older manifest without this detection, keep the current startup command or set `PGDATA=/var/lib/postgresql/data/pgdata` explicitly if the cluster lives under `pgdata/`. Do not move an active database directory back to the PVC root by hand.
+
 ### Image Reference
 
 | Component | Image |

docs/skillhub/guide/kubernetes.md

@@ -214,6 +214,12 @@ skillhub-storage-s3-secret-key: your-secret-key
   value: cn-shanghai
 ```
 
+### PostgreSQL 数据目录兼容性
+
+`with-infra` 会在启动时检查 PostgreSQL PVC 根目录：如果已存在 `PG_VERSION`，继续使用根目录中的旧集群；否则在 `pgdata/` 子目录初始化新集群，避免新 ext4 卷中的 `lost+found` 阻止 `initdb`。升级现有部署不需要移动数据库文件。
+
+回滚到不包含该检测逻辑的旧清单时，如果集群位于 `pgdata/`，必须保留当前启动命令，或显式设置 `PGDATA=/var/lib/postgresql/data/pgdata`。不要把正在运行的数据库目录手动移动到 PVC 根目录。
+
 ### 镜像说明
 
 | 组件 | 镜像 |

scripts/tests/k8s-postgres-storage-test.sh

@@ -0,0 +1,122 @@
+#!/usr/bin/env bash
+
+set -euo pipefail
+
+REPO_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
+MANIFEST="$REPO_ROOT/deploy/k8s/overlays/with-infra/postgres-statefulset.yaml"
+POSTGRES_IMAGE="${POSTGRES_IMAGE:-postgres:16-alpine}"
+RUN_ID="skillhub-k8s-pgdata-$$"
+FRESH_VOLUME="${RUN_ID}-fresh"
+LEGACY_VOLUME="${RUN_ID}-legacy"
+CONTAINERS=()
+VOLUMES=("$FRESH_VOLUME" "$LEGACY_VOLUME")
+
+cleanup() {
+  for container in "${CONTAINERS[@]}"; do
+    docker stop "$container" >/dev/null 2>&1 || true
+  done
+  for volume in "${VOLUMES[@]}"; do
+    docker volume rm "$volume" >/dev/null 2>&1 || true
+  done
+}
+trap cleanup EXIT
+
+for command in docker sed; do
+  if ! command -v "$command" >/dev/null 2>&1; then
+    echo "required command not found: $command" >&2
+    exit 1
+  fi
+done
+
+POSTGRES_START_COMMAND="$(
+  sed -n '/^            - |$/,/^          ports:$/p' "$MANIFEST" \
+    | sed '1d;$d;s/^              //'
+)"
+if [[ -z "$POSTGRES_START_COMMAND" ]]; then
+  echo "could not extract the PostgreSQL startup command from $MANIFEST" >&2
+  exit 1
+fi
+
+wait_for_postgres() {
+  local container="$1"
+  for _ in $(seq 1 30); do
+    if docker exec "$container" pg_isready -U skillhub -d skillhub >/dev/null 2>&1; then
+      return 0
+    fi
+    sleep 1
+  done
+  docker logs "$container" >&2
+  return 1
+}
+
+start_postgres() {
+  local container="$1"
+  local volume="$2"
+  local mode="${3:-default}"
+  CONTAINERS+=("$container")
+  if [[ "$mode" == "patched" ]]; then
+    docker run --detach --rm \
+      --name "$container" \
+      --env POSTGRES_DB=skillhub \
+      --env POSTGRES_USER=skillhub \
+      --env POSTGRES_PASSWORD=storage-test \
+      --volume "$volume:/var/lib/postgresql/data" \
+      --entrypoint sh \
+      "$POSTGRES_IMAGE" \
+      -ec "$POSTGRES_START_COMMAND" >/dev/null
+  else
+    docker run --detach --rm \
+      --name "$container" \
+      --env POSTGRES_DB=skillhub \
+      --env POSTGRES_USER=skillhub \
+      --env POSTGRES_PASSWORD=storage-test \
+      --volume "$volume:/var/lib/postgresql/data" \
+      "$POSTGRES_IMAGE" >/dev/null
+  fi
+  wait_for_postgres "$container"
+}
+
+stop_postgres() {
+  docker stop "$1" >/dev/null
+}
+
+assert_marker() {
+  local container="$1"
+  local expected="$2"
+  local actual
+  actual="$(docker exec "$container" psql -U skillhub -d skillhub -Atqc \
+    'SELECT value FROM storage_upgrade_verification WHERE id = 1')"
+  [[ "$actual" == "$expected" ]]
+}
+
+docker volume create "$FRESH_VOLUME" >/dev/null
+docker run --rm \
+  --volume "$FRESH_VOLUME:/data" \
+  --entrypoint sh \
+  "$POSTGRES_IMAGE" \
+  -ec 'mkdir -p /data/lost+found'
+
+start_postgres "${RUN_ID}-fresh-1" "$FRESH_VOLUME" patched
+docker exec "${RUN_ID}-fresh-1" test -f /var/lib/postgresql/data/pgdata/PG_VERSION
+docker exec "${RUN_ID}-fresh-1" psql -U skillhub -d skillhub -v ON_ERROR_STOP=1 -qc \
+  "CREATE TABLE storage_upgrade_verification (id integer PRIMARY KEY, value text NOT NULL); \
+   INSERT INTO storage_upgrade_verification VALUES (1, 'fresh-volume');"
+stop_postgres "${RUN_ID}-fresh-1"
+
+start_postgres "${RUN_ID}-fresh-2" "$FRESH_VOLUME" patched
+assert_marker "${RUN_ID}-fresh-2" "fresh-volume"
+stop_postgres "${RUN_ID}-fresh-2"
+echo "PASS: fresh volume with lost+found initializes in pgdata and survives restart"
+
+docker volume create "$LEGACY_VOLUME" >/dev/null
+start_postgres "${RUN_ID}-legacy-1" "$LEGACY_VOLUME"
+docker exec "${RUN_ID}-legacy-1" test -f /var/lib/postgresql/data/PG_VERSION
+docker exec "${RUN_ID}-legacy-1" psql -U skillhub -d skillhub -v ON_ERROR_STOP=1 -qc \
+  "CREATE TABLE storage_upgrade_verification (id integer PRIMARY KEY, value text NOT NULL); \
+   INSERT INTO storage_upgrade_verification VALUES (1, 'legacy-root');"
+stop_postgres "${RUN_ID}-legacy-1"
+
+start_postgres "${RUN_ID}-legacy-2" "$LEGACY_VOLUME" patched
+assert_marker "${RUN_ID}-legacy-2" "legacy-root"
+stop_postgres "${RUN_ID}-legacy-2"
+echo "PASS: legacy root-level cluster remains visible after the manifest upgrade"