fix(cluster): 限制生产异步 Worker 执行容量
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。
This commit is contained in:
@@ -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"
|
||||
|
||||
Reference in New Issue
Block a user