Tips and tricks for Kubernetes storage
summary
Walks through Kubernetes storage mechanics and concrete fixes: PVCs/StorageClasses expose protocol, replicas and mount options; ReadWriteOnce vs ReadWriteMany means RWO is per node not per pod; use wait-for-first-consumer to place volumes in the pod’s zone; reclaim policies can silently leak backend volumes so OpenEBS MayaStore adds garbage-collect orphan volumes; compares node-local vs replicated OpenEBS drivers, warns on multi-attach errors after node failure, and recommends tools for migration (Velero, Volsync, PV-migrate) and testing graceful vs forced node shutdown.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:07Kubernetes basics
- 9:23Stateful vs stateless
- 10:35Common pitfalls
- 12:52Reclaim policy dangers
- 15:24Wait for first consumer
- 16:28Access modes confusion
- 17:36Multi-node volume access
- 18:42Storage anti-affinity
- 19:24Volume monitoring
- 20:31Snapshot consistency
- 21:40Local vs replicated storage
- 24:14Topology-aware scheduling
- 25:22Node failure handling
- 26:27Pod shutdown modes
- 28:13CPU manager policies
- 30:20Data migration techniques
- 31:23Summary and Q&A
related talks
37mStorage on Kubernetes - Learning From Failures
49mDevOps: Kubernetes Storage - Myths, Facts and Tips
37mKubernetes Storage 101
35mKubernetes Storage Lingo 101
30mKubernetes Storage Cheat Sheet for VM Administrators
19mKeynote: Debunking the Myth: Kubernetes Storage is Hard
28mPersistent Storage with Kubernetes in Production - Which Solution and Why?
28mKubernetes Storage Evolution: Enabling High Performance Distributed Datastores