Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

Diagnosing Slow Elasticsearch Search with STAROps

AI summary

A Juejin tutorial demonstrates how engineers can use the STAROps Elasticsearch 智能助手 to inspect per-node CPU, query load, and shard distribution during a search slowdown. Its worked example shows how a Green cluster and healthy average CPU can conceal sustained pressure on one node, while emphasizing that slow logs and thread-pool rejection deltas are still needed to establish the root cause.

In-depth insight

The article provides a concrete AI-assisted observability workflow for Elasticsearch incidents and shows why cluster-level health indicators alone may miss uneven node pressure.

本文作者:但丁(孔祥超)、杜尘(王明)、逸陵(阳其凯)

某电商平台的商品搜索服务使用一台 6 个数据节点的 Elasticsearch 实例。周五北京时间 20:00 起,业务陆续反馈搜索变慢。业务监控中的搜索接口 P99 从平时约 200 ms 升到 2s 以上,部分请求超时。值班工程师打开控制台:集群状态为 Green,没有未分配分片,集群平均 CPU 约 40%,乍看不像资源已经吃紧。

按节点展开后,问题才显露出来:节点 A 从 20:05 起 CPU 持续高于 95%,其余 5 个节点在 20%~40%。线程池里能看到 search 累计拒绝上万次,却不知道今晚新增了多少;等抓取热线程时,峰值已经回落,堆栈也没能解释刚才为什么慢。

业务还在等恢复进展。工程师需要弄清楚:今晚请求更多了,某类查询变重了,还是压力集中到了节点 A?这个节点承载了哪些分片,慢查询又落在哪里?这些问题需要把同一时段的证据放在一起看。

#STAROps#Elasticsearch 智能助手#Elasticsearch

Original · 8 min readRead the original
Group chat

No comments yet. Start the conversation.