DevOps: Kubernetes Storage - Myths, Facts and Tips
summary
Covers concrete storage patterns: use PersistentVolume/PersistentVolumeClaim and StorageClass for dynamic provisioning instead of embedding volume IDs, use StatefulSet (not Deployment) so each replica gets its own PVC, and prefer local persistent volumes for low-latency workloads (author cites ~5% difference versus bare metal). Explains EBS/GCE limits (single-AZ attach), shows how resize and snapshots work via Container Storage Interface (CSI), and warns that sharing one PVC across replicas causes data corruption.
AI-written summary · about
chapters (16)
- 0:00Introduction to Kubernetes storage
- 1:59Speaker introduction and agenda
- 4:26Challenges with stateful containers
- 10:30Volume plugin types overview
- 13:41Ephemeral storage explained
- 18:14Remote storage with direct references
- 20:44Persistent volumes and claims
- 27:48Dynamic provisioning
- 32:07Storage class configuration
- 34:54Stateful applications and deployments
- 37:43Stateful sets for stateful apps
- 38:46Local persistent volumes
- 40:52Container Storage Interface
- 43:41CSI features: resize and snapshots
- 45:06Mythbusters: storage myths debunked
- 47:16Key takeaways and summary
related talks
37mKubernetes Storage 101
36mTips and tricks for Kubernetes storage
30mKubernetes Storage Cheat Sheet for VM Administrators
28mPersistent Storage with Kubernetes in Production - Which Solution and Why?
19mKeynote: Debunking the Myth: Kubernetes Storage is Hard
35mKubernetes Storage Lingo 101
28mKubernetes Storage Evolution: Enabling High Performance Distributed Datastores
10m⚡ Lightning Talk: Navigating the Complexities of Stateful Workloads in Kubernetes