⚡ Lightning Talk: Back To Basics: How To Measure Etcd Performance And Not To Die Trying
Lightningat KubeCon + CloudNativeCon North America 2022 · on Tuesday, 2022-10-25· at 18:10· Grand Riverview Ballroom AB·
summary
etcd performance diagnosis uses the built-in etcd benchmark tool and disk tests like fio against a three-member etcd cluster created from the provided scripts to measure latency and throughput; with 1,000 clients and hundreds of connections HDDs show severe throughput loss and high latency while switching to SSD doubles requests per second; practical fixes include moving off HDD or using 15K RPM disks, ionice disk priority, and tuning election/heartbeat timeouts to avoid overload.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 0:10Overview
- 0:30What is Etcd
- 1:07Etcd logs
- 1:36throughput and latency
- 2:08hcd Benchmark tool
- 3:40Multiple clients and multiple connections
- 4:00What can you do
- 4:48HDD vs SSD
- 5:08Recap
- 5:30Outro
related talks
42mHow Should You Effectively Use etcd Metrics
04mLightning Talk - Benchmarking Persistent Data Volumes
23mThe Secret Recipe of etcd
20mLessons Learned from Operating ETCD
19mDr etcd; or; How I Learned to Stop Worrying and Love the Datastore - Nick Young, VMware
47mNot Omnipotent etcd, or Why It Doesn't Handle Heavy Loads
26mDon't Let Your Kubernetes Cluster Go Wild: Ensuring Etcd Reliability
33mCassandra Performance Tuning Like You've Been Doing It for Ten Years