Performance Diagnostics with Open Distro for Elasticsearch
summary
Open Distro ships a Performance Analyzer plugin that instruments every HTTP and shard RPC, writes events to shared memory/tempfs and a separate agent reads and correlates /proc per-thread CPU samples to attribute CPU and latency by index, shard and operation. Metrics land in SQLite snapshots every 5s, the Node.js PerfTop client fetches them over SSH, and the stack handles workloads up to 100k docs/sec. Built diagnostics expose CPU, JVM, disk service rate and hot-shard causes and supports encrypted transport.
AI-written summary · about
chapters (24)
- 0:00Introduction
- 1:19Overview
- 2:45Why do we care
- 3:59Open Distro
- 4:31Security
- 6:24Performance Analyzer
- 6:45Performance Analyzer Overview
- 8:30Performance Analyzer Use Cases
- 10:03Performance Analyzer Components
- 13:34Data Flow
- 17:18Metrics
- 18:33What I miss
- 19:03Key innovation
- 19:17CPU utilization
- 20:11Query results
- 21:12Query operations
- 23:05Database creation
- 23:29Matrix processor
- 24:11Web service
- 24:38Disk utilization
- 25:50Root cause analysis
- 26:54Typical issues
- 28:58Perftop
- 29:29Network Memory
related talks
26mApplication performance management w/ open...
54mUnderstanding Performance w/DTrace (While The Customer Yells)...
29mThe osquery Agent. Performance Tuning @Scale
24mDevOps. Monitoring Databases. On-Premise and Cloud environments
26mObserving Node.js: Using Metrics to Improve your Application Performance
48mFantastic Performance and Where to Find It
29mApache Spark Performance Tuning Session with Delight
39mLinux Server Deep Dives