From 4371e572123e432748e8309011be9adeb85b4080 Mon Sep 17 00:00:00 2001
From: wangbo <wangbolhr@163.com>
Date: Wed, 29 Jul 2026 18:51:26 +0800
Subject: [PATCH] =?UTF-8?q?fix(cluster):=20=E8=BF=9B=E4=B8=80=E6=AD=A5?=
 =?UTF-8?q?=E6=94=B6=E7=B4=A7=E5=AA=92=E4=BD=93=20Worker=20=E5=B9=B6?=
 =?UTF-8?q?=E5=8F=91=E4=B8=8A=E9=99=90?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

全局 48 槽在持续观察中仍导致宁波 Worker OOM。按实测内存占用将生产全局异步执行上限降至 16，双节点各 8 槽，单节点接管最多 16 槽；高并发请求继续由持久队列吸收。\n\n未开启 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 553f582..3fd57c6 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: "48"
+  AI_GATEWAY_ASYNC_WORKER_HARD_LIMIT: "16"
   AI_GATEWAY_ASYNC_WORKER_REFRESH_INTERVAL_SECONDS: "5"