etcd V3.6.0 and etc...
summary
etcd v3.6.0 ships with downgrade support (manual validate → enable → rolling binary swap) and drops v2 store migration toward a v3 bootstrap path; upgrading from 3.5 requires patches because a leader-promotion bug affects 3.5.1–3.5.19 so must upgrade to 3.5.20+ before moving to 3.6. Performance improves—memory drops after changing snapshot_count 100,000→10,000 and compaction tweaks—RC3 exists due to the upgrade bug. etcd-operator v0.1.0 can create and customize clusters in Kubernetes but is not recommended for production; roadmap adds TLS, auto cert providers, backups and DR.
AI-written summary · about
chapters (14)
- 0:00etcd 3.6.0 and operator 0.1
- 2:13New features in etcd 3.6.0
- 3:39Migration to WAL3 store
- 4:43Full downgrade support
- 6:32Feature gates transition
- 8:41New health check endpoints
- 10:06Cluster bootstrap discovery
- 11:11Critical 3.5 to 3.6 upgrade
- 12:13Performance improvements
- 16:56Release process automation
- 18:00etcd operator working group
- 20:25Operator 0.1.0 milestone
- 21:34Future operator roadmap
- 23:36Q&A on multi-cluster sync
related talks
32mSIG etcd Maintainer Track Talk
38metcd in 2026: Enhancements, V3.7 Roadmap...
36mWhat You Should Know about Etcd v3
05mLet’s Deploy etcd Operator
19mDr etcd; or; How I Learned to Stop Worrying and Love the Datastore - Nick Young, VMware
43mSelf-hosted, Scale, and Federation with Kubernetes v1.4 and Beyond
36mKoping with Change: What kops Learned Adopting etcd3, cluster-api and CRDs
35mSafely Upgrading Kubernetes Clusters