polarismesh / polaris

Service Discovery and Governance Platform for Microservice and Distributed Architecture
https://polarismesh.cn
Other
2.41k stars 399 forks source link

【求助】北极星部分实例内存在短时间内飙升, 然后进程被OOM kill掉, 这种要怎么排查 #1385

Closed stonebirdjx closed 2 months ago

stonebirdjx commented 2 months ago

oom

stonebirdjx commented 2 months ago

单实例规格 8C16G,服务数量级20w左右,平时运行稳定,突然内存飙升,然后oom被kill。没有内存飙升的现场pprof,有没有有经验的懂哥告知一下排查方案

stonebirdjx commented 2 months ago

数据库lastFetchTime慢查询情况下返回0,导致数据库全量拉取 if maxWait != 0 && timePass > maxWait { log.Infof("[Store][database] query now spend %s, exceed %s, skip", timePass, maxWait) return 0, nil }