From ba01cbef53d6b4861c063e0d1f77351d23fe7bee Mon Sep 17 00:00:00 2001
From: wangbo <wangbolhr@163.com>
Date: Wed, 29 Jul 2026 17:45:23 +0800
Subject: [PATCH] =?UTF-8?q?fix(cluster):=20=E9=99=90=E5=88=B6=E7=94=9F?=
 =?UTF-8?q?=E4=BA=A7=E5=BC=82=E6=AD=A5=20Worker=20=E6=89=A7=E8=A1=8C?=
 =?UTF-8?q?=E5=AE=B9=E9=87=8F?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

切换后集中恢复积压时，单 Worker 155 槽导致数据库连接池 24/24 饱和并触发香港 Pod OOM。将集群异步执行硬上限暂定为 48，使双 Worker 正常时各分配 24 槽，与每 Pod 连接池上限对齐。\n\n未开启任何 queue_size 生产策略。验证：kubectl kustomize、服务端 dry-run。
---
 deploy/kubernetes/production/application-config.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/deploy/kubernetes/production/application-config.yaml b/deploy/kubernetes/production/application-config.yaml
index 7c3aaa5..553f582 100644
--- a/deploy/kubernetes/production/application-config.yaml
+++ b/deploy/kubernetes/production/application-config.yaml
@@ -35,5 +35,5 @@ data:
   AI_GATEWAY_LOCAL_RESULT_MAX_BYTES: "268435456"
   AI_GATEWAY_LOCAL_RESULT_MAX_TASK_BYTES: "536870912"
   AI_GATEWAY_ASYNC_QUEUE_WORKER_ENABLED: "true"
-  AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT: "2048"
+  AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT: "48"
   AI_GATEWAY_ASYNC_WORKER_REFRESH_INTERVAL_SECONDS: "5"