fix: 隔离既有发布的数据库容器生命周期（09-05-existing-deploy-db-lifecycle-isolation）

deploy/trialrun/deploy.sh

@@ -248,7 +248,7 @@ cd "$REMOTE_DIR"
 
 if [[ "$DEPLOY_DATA_MODE" == "initialize" ]]; then
   mkdir -p "$POSTGRES_DATA_DIR" "$STORAGE_HOST_DIR"
-else
+elif [[ "$DEPLOY_DATA_MODE" == "existing" ]]; then
   if [[ ! -d "$POSTGRES_DATA_DIR" || ! -f "$POSTGRES_DATA_DIR/PG_VERSION" ]]; then
     echo "[ERROR] existing 模式要求 POSTGRES_DATA_DIR 已存在且包含 PG_VERSION" >&2
     exit 1
@@ -257,6 +257,9 @@ else
     echo "[ERROR] existing 模式要求 STORAGE_HOST_DIR 已存在" >&2
     exit 1
   fi
+else
+  echo "[ERROR] DEPLOY_DATA_MODE 必须显式设为 existing 或 initialize" >&2
+  exit 1
 fi
 if [[ "$ENABLE_MINIO" == "1" ]]; then
   mkdir -p deploy/trialrun/data/minio
@@ -312,23 +315,96 @@ run_logged_deploy_stage() {
 
 echo ">>> 4.1 校验远端 compose 配置"
 "${compose_cmd[@]}" "${profile_args[@]}" config >/dev/null
+if ! command -v timeout >/dev/null 2>&1; then
+  echo "[ERROR] db_health_timeout_unavailable：缺少有界就绪检查工具，停止发布" >&2
+  exit 1
+fi
 
-echo ">>> 4.2 启动基础服务"
-"${compose_cmd[@]}" up -d db
+wait_for_database_health() {
+  if timeout --signal=KILL 120 bash -s -- "${compose_cmd[@]}" <<'DB_HEALTH'
+set -euo pipefail
+while true; do
+  if ! database_id="$("$@" ps -a -q db 2>/dev/null)"; then
+    echo "[ERROR] db_health_query_failed：无法查询目标数据库容器，停止发布"
+    exit 1
+  fi
+  if [[ ! "$database_id" =~ ^[a-f0-9]{12,64}$ ]]; then
+    echo "[ERROR] db_container_missing_or_ambiguous：目标数据库容器不存在或不唯一，停止发布"
+    exit 1
+  fi
+  if ! database_health="$(docker inspect --type container --format '{{.State.Status}}|{{if .State.Health}}{{.State.Health.Status}}{{else}}missing{{end}}' "$database_id" 2>/dev/null)"; then
+    echo "[ERROR] db_health_query_failed：无法读取目标数据库健康状态，停止发布"
+    exit 1
+  fi
+  case "$database_health" in
+    'running|healthy') exit 0 ;;
+    'running|starting') sleep 2 ;;
+    *)
+      echo "[ERROR] db_not_healthy：数据库未运行、不健康或缺少健康检查，停止发布；不会自动启动或修复"
+      exit 1
+      ;;
+  esac
+done
+DB_HEALTH
+  then
+    echo "[PASS] 目标数据库健康检查通过；本次等待上限 120 秒"
+  else
+    local health_status=$?
+    if [[ "$health_status" == "124" || "$health_status" == "137" ]]; then
+      echo "[ERROR] db_health_timeout：数据库就绪检查超过 120 秒，停止发布；不执行后续 init 或应用切换" >&2
+    fi
+    return 1
+  fi
+}
+
+echo ">>> 4.2 按部署模式检查基础服务"
+if [[ "$DEPLOY_DATA_MODE" == "initialize" ]]; then
+  "${compose_cmd[@]}" up -d db
+fi
+wait_for_database_health
 if [[ "$ENABLE_MINIO" == "1" ]]; then
-  "${compose_cmd[@]}" "${profile_args[@]}" up -d minio minio_init
+  "${compose_cmd[@]}" "${profile_args[@]}" up -d --no-deps minio minio_init
 fi
 
 echo ">>> 4.3 构建最新 init 镜像"
 run_logged_deploy_stage "build-init" "${compose_cmd[@]}" build init
 
 echo ">>> 4.4 预检新配置并执行 init"
-run_logged_deploy_stage "run-init" "${compose_cmd[@]}" run --rm -T --entrypoint sh init -lc 'cd /app/server && node -e '\''require("./src/config").getConfig()'\'' && npm run prisma:migrate:deploy && npm run seed:config && npm run seed:trialrun'
+wait_for_database_health
+run_logged_deploy_stage "run-init" "${compose_cmd[@]}" run --rm --no-deps -T --entrypoint sh init -lc 'cd /app/server && node -e '\''require("./src/config").getConfig()'\'' && npm run prisma:migrate:deploy && npm run seed:config && npm run seed:trialrun'
 
 echo ">>> 4.5 构建并切换应用服务"
-run_logged_deploy_stage "build-switch-server" "${compose_cmd[@]}" "${profile_args[@]}" up -d --build server
+wait_for_database_health
+run_logged_deploy_stage "build-server" "${compose_cmd[@]}" "${profile_args[@]}" build server
 if [[ "$ENABLE_WORKER" == "1" ]]; then
-  "${compose_cmd[@]}" "${profile_args[@]}" up -d worker
+  wait_for_database_health
+  if ! worker_container_id="$("${compose_cmd[@]}" "${profile_args[@]}" ps -a -q worker)"; then
+    echo "[ERROR] worker_image_query_failed：无法检查 worker 容器，停止发布" >&2
+    exit 1
+  fi
+  worker_image_ready=0
+  if [[ -n "$worker_container_id" ]]; then
+    if [[ ! "$worker_container_id" =~ ^[a-f0-9]{12,64}$ ]] || ! worker_image_reference="$(docker inspect --type container --format '{{.Config.Image}}' "$worker_container_id" 2>/dev/null)"; then
+      echo "[ERROR] worker_image_query_failed：worker 容器不唯一或镜像引用不可读，停止发布" >&2
+      exit 1
+    fi
+    if [[ ! "$worker_image_reference" =~ ^[A-Za-z0-9][A-Za-z0-9._/:@-]*$ ]]; then
+      echo "[ERROR] worker_image_query_failed：worker 镜像引用无效，停止发布" >&2
+      exit 1
+    fi
+    if docker image inspect "$worker_image_reference" >/dev/null 2>&1; then
+      worker_image_ready=1
+    fi
+  fi
+  if [[ "$worker_image_ready" != "1" ]]; then
+    run_logged_deploy_stage "build-worker" "${compose_cmd[@]}" "${profile_args[@]}" build worker
+  fi
+fi
+wait_for_database_health
+run_logged_deploy_stage "build-switch-server" "${compose_cmd[@]}" "${profile_args[@]}" up -d --no-build --no-deps server
+if [[ "$ENABLE_WORKER" == "1" ]]; then
+  wait_for_database_health
+  "${compose_cmd[@]}" "${profile_args[@]}" up -d --no-build --no-deps worker
 fi
 
 echo ">>> 4.6 校验 Prisma migration status"