fix(worker): 拆分准入事务并收敛验收失败状态

将批量准入改为每任务独立事务，避免 P24 压测时 48 条任务共享长事务造成 transactionid 锁潮与 Worker 槽位空转。失败验收在保持 validation 的同时清理未提交任务，并记录精确压力门禁原因。统一生产 0+2 Worker 拓扑、运行时 ConfigMap 与发布配置，补充镜像预检诊断和回归测试。\n\n验证：Go 全量测试、go vet、真实 PostgreSQL 48 任务集成测试、迁移安全检查、发布脚本测试、bash -n、ShellCheck、Compose 配置均通过。

deploy/kubernetes/easyai-ai-gateway-cluster-release

@@ -15,8 +15,8 @@ source "$config_file"
 : "${PUBLIC_BASE_URL:=https://ai.51easyai.com}"
 : "${K3S_BIN:=/usr/local/bin/k3s}"
 : "${DESIRED_STATE_DIR:=/usr/local/share/easyai-ai-gateway-release/production}"
-: "${AI_GATEWAY_WORKER_REPLICAS_NINGBO:=1}"
-: "${AI_GATEWAY_WORKER_REPLICAS_HONGKONG:=1}"
+: "${AI_GATEWAY_WORKER_REPLICAS_NINGBO:=0}"
+: "${AI_GATEWAY_WORKER_REPLICAS_HONGKONG:=2}"
 : "${AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT:=24}"
 : "${AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT:=48}"
 : "${AI_GATEWAY_ASYNC_WORKER_GLOBAL_HARD_LIMIT:=$AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT}"
@@ -32,10 +32,10 @@ source "$config_file"
 : "${AI_GATEWAY_MEDIA_MATERIALIZATION_CONCURRENCY:=24}"
 : "${AI_GATEWAY_MEDIA_REQUEST_CONCURRENCY:=24}"
 : "${AI_GATEWAY_WORKER_AUTOSCALING_ENABLED:=false}"
-: "${AI_GATEWAY_WORKER_MIN_REPLICAS_NINGBO:=1}"
+: "${AI_GATEWAY_WORKER_MIN_REPLICAS_NINGBO:=0}"
 : "${AI_GATEWAY_WORKER_MIN_REPLICAS_HONGKONG:=1}"
-: "${AI_GATEWAY_WORKER_MAX_REPLICAS_NINGBO:=1}"
-: "${AI_GATEWAY_WORKER_MAX_REPLICAS_HONGKONG:=1}"
+: "${AI_GATEWAY_WORKER_MAX_REPLICAS_NINGBO:=0}"
+: "${AI_GATEWAY_WORKER_MAX_REPLICAS_HONGKONG:=2}"
 : "${AI_GATEWAY_WORKER_TARGET_OUTSTANDING_PER_REPLICA:=$((AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT * 2))}"
 : "${AI_GATEWAY_WORKER_SCALE_UP_WINDOW_SECONDS:=20}"
 : "${AI_GATEWAY_WORKER_SCALE_DOWN_STABILIZATION_SECONDS:=600}"
@@ -199,8 +199,55 @@ capacity_config_hash() {
 }
 
 record_capacity_config() {
-  local config_hash
-  config_hash=$(capacity_config_hash)
+	local config_hash
+	local runtime_patch
+	runtime_patch=$(jq -nc \
+		--arg replicasNingbo "$AI_GATEWAY_WORKER_REPLICAS_NINGBO" \
+		--arg replicasHongkong "$AI_GATEWAY_WORKER_REPLICAS_HONGKONG" \
+		--arg minNingbo "$AI_GATEWAY_WORKER_MIN_REPLICAS_NINGBO" \
+		--arg minHongkong "$AI_GATEWAY_WORKER_MIN_REPLICAS_HONGKONG" \
+		--arg maxNingbo "$AI_GATEWAY_WORKER_MAX_REPLICAS_NINGBO" \
+		--arg maxHongkong "$AI_GATEWAY_WORKER_MAX_REPLICAS_HONGKONG" \
+		--arg autoscaling "$AI_GATEWAY_WORKER_AUTOSCALING_ENABLED" \
+		--arg instanceLimit "$AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT" \
+		--arg globalLimit "$AI_GATEWAY_ASYNC_WORKER_GLOBAL_HARD_LIMIT" \
+		--arg workerPool "$AI_GATEWAY_DATABASE_MAX_CONNS" \
+		--arg materialization "$AI_GATEWAY_MEDIA_MATERIALIZATION_CONCURRENCY" \
+		--arg mediaRequest "$AI_GATEWAY_MEDIA_REQUEST_CONCURRENCY" \
+		--arg targetOutstanding "$AI_GATEWAY_WORKER_TARGET_OUTSTANDING_PER_REPLICA" \
+		--arg scaleUp "$AI_GATEWAY_WORKER_SCALE_UP_WINDOW_SECONDS" \
+		--arg scaleDown "$AI_GATEWAY_WORKER_SCALE_DOWN_STABILIZATION_SECONDS" \
+		--arg drainTimeout "$AI_GATEWAY_WORKER_DRAIN_TIMEOUT_SECONDS" \
+		--arg memoryTarget "$AI_GATEWAY_NODE_MEMORY_TARGET_PERCENT" \
+		--arg memoryHard "$AI_GATEWAY_NODE_MEMORY_HARD_PERCENT" \
+		--arg cpuTarget "$AI_GATEWAY_NODE_CPU_TARGET_PERCENT" \
+		--arg postgresBudget "$AI_GATEWAY_POSTGRES_CONNECTION_BUDGET" \
+		'{data:{
+			AI_GATEWAY_WORKER_REPLICAS_NINGBO:$replicasNingbo,
+			AI_GATEWAY_WORKER_REPLICAS_HONGKONG:$replicasHongkong,
+			AI_GATEWAY_WORKER_MIN_REPLICAS_NINGBO:$minNingbo,
+			AI_GATEWAY_WORKER_MIN_REPLICAS_HONGKONG:$minHongkong,
+			AI_GATEWAY_WORKER_MAX_REPLICAS_NINGBO:$maxNingbo,
+			AI_GATEWAY_WORKER_MAX_REPLICAS_HONGKONG:$maxHongkong,
+			AI_GATEWAY_WORKER_AUTOSCALING_ENABLED:$autoscaling,
+			AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT:$instanceLimit,
+			AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT:$globalLimit,
+			AI_GATEWAY_ASYNC_WORKER_GLOBAL_HARD_LIMIT:$globalLimit,
+			AI_GATEWAY_WORKER_DATABASE_MAX_CONNS:$workerPool,
+			AI_GATEWAY_MEDIA_MATERIALIZATION_CONCURRENCY:$materialization,
+			AI_GATEWAY_MEDIA_REQUEST_CONCURRENCY:$mediaRequest,
+			AI_GATEWAY_WORKER_TARGET_OUTSTANDING_PER_REPLICA:$targetOutstanding,
+			AI_GATEWAY_WORKER_SCALE_UP_WINDOW_SECONDS:$scaleUp,
+			AI_GATEWAY_WORKER_SCALE_DOWN_STABILIZATION_SECONDS:$scaleDown,
+			AI_GATEWAY_WORKER_DRAIN_TIMEOUT_SECONDS:$drainTimeout,
+			AI_GATEWAY_NODE_MEMORY_TARGET_PERCENT:$memoryTarget,
+			AI_GATEWAY_NODE_MEMORY_HARD_PERCENT:$memoryHard,
+			AI_GATEWAY_NODE_CPU_TARGET_PERCENT:$cpuTarget,
+			AI_GATEWAY_POSTGRES_CONNECTION_BUDGET:$postgresBudget
+		}}')
+	"${kubectl[@]}" patch configmap easyai-ai-gateway-config -n "$NAMESPACE" \
+		--type merge -p "$runtime_patch" >/dev/null
+	config_hash=$(capacity_config_hash)
   "${kubectl[@]}" patch configmap easyai-gateway-release -n "$NAMESPACE" \
     --type merge -p "{\"data\":{\"configHash\":\"$config_hash\"}}" >/dev/null
   "${kubectl[@]}" annotate namespace "$NAMESPACE" \
@@ -512,9 +559,18 @@ spec:
       emptyDir:
         sizeLimit: 64Mi
 EOF
-  if ! "${kubectl[@]}" wait --for=condition=Ready "pod/$pod" \
-    -n "$NAMESPACE" --timeout=120s; then
-    "${kubectl[@]}" delete pod "$pod" -n "$NAMESPACE" \
+	if ! "${kubectl[@]}" wait --for=condition=Ready "pod/$pod" \
+		-n "$NAMESPACE" --timeout=300s; then
+		"${kubectl[@]}" get pod "$pod" -n "$NAMESPACE" -o json 2>/dev/null | jq -c '
+			{
+				phase:(.status.phase // "unknown"),
+				node:(.spec.nodeName // "unscheduled"),
+				ready:([.status.containerStatuses[]?.ready] | all),
+				restarts:([.status.containerStatuses[]?.restartCount] | add // 0),
+				waitingReasons:[.status.containerStatuses[]?.state.waiting.reason // empty],
+				terminatedReasons:[.status.containerStatuses[]?.state.terminated.reason // empty]
+			}' >&2 || true
+		"${kubectl[@]}" delete pod "$pod" -n "$NAMESPACE" \
       --ignore-not-found=true --wait=true >/dev/null || true
     echo 'capacity controller exact-image preflight failed' >&2
     return 1
@@ -866,7 +922,7 @@ activate_manifest() {
       -n "$NAMESPACE" --replicas=0
   fi
 
-  if { [[ $api_changed != true ]] ||
+	if { [[ $api_changed != true ]] ||
        { rollout_worker_site hongkong "$api_image" &&
          rollout_worker_site ningbo "$api_image" &&
          { [[ $action == rollback ]] ||
@@ -875,16 +931,17 @@ activate_manifest() {
      rollout_site ningbo "$api_image" "$web_image" "$api_changed" "$web_changed" &&
      wait_for_url 'public readiness' "$PUBLIC_BASE_URL/api/v1/readyz" '"ok":true'; then
     :
-  else
+	else
     if [[ -n $previous_api && -n $previous_web ]]; then
       echo '[cluster-release] restoring exact pre-release Deployment snapshot' >&2
       restore_application_deployments "$deployment_snapshot" || true
     fi
     rm -f -- "$deployment_snapshot" "$current_file"
-    return 1
-  fi
+		return 1
+	fi
 
-  record_current "$manifest" "$action" "$started_at"
+	record_capacity_config
+	record_current "$manifest" "$action" "$started_at"
   rm -f -- "$deployment_snapshot" "$current_file"
   echo "production_release=PASS action=$action release=$source_sha"
 }

deploy/kubernetes/easyai-ai-gateway-cluster-release.conf.example

@@ -6,8 +6,8 @@ DESIRED_STATE_DIR=/usr/local/share/easyai-ai-gateway-release/production
 
 # Worker 容量只需修改本发布配置并执行获授权的 `capacity`，不再修改 Go 代码。
 # 副本数属于 Deployment 配置，不能由容器内部环境变量改变；发布工具读取这两个变量并写入 replicas。
-AI_GATEWAY_WORKER_REPLICAS_NINGBO=1
-AI_GATEWAY_WORKER_REPLICAS_HONGKONG=1
+AI_GATEWAY_WORKER_REPLICAS_NINGBO=0
+AI_GATEWAY_WORKER_REPLICAS_HONGKONG=2
 AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT=24
 AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT=48
 AI_GATEWAY_ASYNC_WORKER_GLOBAL_HARD_LIMIT=48
@@ -23,10 +23,10 @@ AI_GATEWAY_DATABASE_MAX_CONN_IDLE_SECONDS=30
 AI_GATEWAY_MEDIA_MATERIALIZATION_CONCURRENCY=24
 AI_GATEWAY_MEDIA_REQUEST_CONCURRENCY=24
 AI_GATEWAY_WORKER_AUTOSCALING_ENABLED=false
-AI_GATEWAY_WORKER_MIN_REPLICAS_NINGBO=1
+AI_GATEWAY_WORKER_MIN_REPLICAS_NINGBO=0
 AI_GATEWAY_WORKER_MIN_REPLICAS_HONGKONG=1
-AI_GATEWAY_WORKER_MAX_REPLICAS_NINGBO=1
-AI_GATEWAY_WORKER_MAX_REPLICAS_HONGKONG=1
+AI_GATEWAY_WORKER_MAX_REPLICAS_NINGBO=0
+AI_GATEWAY_WORKER_MAX_REPLICAS_HONGKONG=2
 AI_GATEWAY_WORKER_TARGET_OUTSTANDING_PER_REPLICA=48
 AI_GATEWAY_WORKER_SCALE_UP_WINDOW_SECONDS=20
 AI_GATEWAY_WORKER_SCALE_DOWN_STABILIZATION_SECONDS=600

deploy/kubernetes/production/application-config.yaml

@@ -46,12 +46,12 @@ data:
   AI_GATEWAY_ASYNC_WORKER_GLOBAL_HARD_LIMIT: "48"
   AI_GATEWAY_ASYNC_WORKER_REFRESH_INTERVAL_SECONDS: "5"
   AI_GATEWAY_WORKER_AUTOSCALING_ENABLED: "false"
-  AI_GATEWAY_WORKER_REPLICAS_NINGBO: "1"
-  AI_GATEWAY_WORKER_REPLICAS_HONGKONG: "1"
-  AI_GATEWAY_WORKER_MIN_REPLICAS_NINGBO: "1"
+  AI_GATEWAY_WORKER_REPLICAS_NINGBO: "0"
+  AI_GATEWAY_WORKER_REPLICAS_HONGKONG: "2"
+  AI_GATEWAY_WORKER_MIN_REPLICAS_NINGBO: "0"
   AI_GATEWAY_WORKER_MIN_REPLICAS_HONGKONG: "1"
-  AI_GATEWAY_WORKER_MAX_REPLICAS_NINGBO: "1"
-  AI_GATEWAY_WORKER_MAX_REPLICAS_HONGKONG: "1"
+  AI_GATEWAY_WORKER_MAX_REPLICAS_NINGBO: "0"
+  AI_GATEWAY_WORKER_MAX_REPLICAS_HONGKONG: "2"
   AI_GATEWAY_WORKER_TARGET_OUTSTANDING_PER_REPLICA: "48"
   AI_GATEWAY_WORKER_SCALE_UP_WINDOW_SECONDS: "20"
   AI_GATEWAY_WORKER_SCALE_DOWN_STABILIZATION_SECONDS: "600"

deploy/kubernetes/production/application.yaml

@@ -434,7 +434,7 @@ metadata:
     app.kubernetes.io/component: worker
     app.kubernetes.io/part-of: easyai-ai-gateway
 spec:
-  replicas: 1
+  replicas: 0
   strategy:
     type: RollingUpdate
     rollingUpdate:
@@ -607,7 +607,7 @@ metadata:
     app.kubernetes.io/component: worker
     app.kubernetes.io/part-of: easyai-ai-gateway
 spec:
-  replicas: 1
+  replicas: 2
   strategy:
     type: RollingUpdate
     rollingUpdate: