Snapshots Aren’t Backups: The Blast Radius Problem in Kubernetes DR
summary
CSI volume snapshots are fast and cheap but CSI snapshots live in the same blast radius, so they often fail as disaster recovery. The panel argues snapshot ≠ backup: reliable DR requires application‑aware quiescing, offsite immutable copies, and isolated sandbox restores; expect nontrivial RTO/RPO tradeoffs and cross‑team runbooks. Practical steps include local snapshot + copy to protected object store, periodic end‑to‑end restore drills, and test restores regularly to validate the plan.
AI-written summary · about
chapters (14)
- 0:00Introduction and panelists
- 1:53Why workload protection matters
- 4:47Runbooks and emergency planning
- 10:55Snapshots vs backups in Kubernetes
- 13:06Complexity of Kubernetes backups
- 17:49Ransomware and backup security
- 20:20Cost versus RPO and RTO
- 22:51Blast radius and segmentation
- 26:44Incident response and troubleshooting
- 31:41Defining disaster scenarios
- 34:30Where to store snapshots
- 36:38Database snapshot best practices
- 38:46Testing recovery plans
- 40:10Actionable takeaways
related talks
35mBeyond Storage Management
30mIs Your Kubernetes Disaster Recovery Actually Ready?
58mProtecting Stateful Workloads With CSI Snapshot
36mPolicy-Based Volume Snapshots Management in Kubernetes
25mChaos To Control: Kubernetes Resiliency at the Edge
25mIntro & Deep Dive: Kubernetes Data Protection WG
15mWhen Clusters Collide: The Misadventures of Disaster Recovery in K8s
30mWhy a Layered Storage Architecture is Critical for Cyber Resilience