Incremental backups in Kubernetes: Where we stand and what's missing
summary
Kubernetes lacks a native changed-block-tracking facility so incremental backups copy too much data; a CNCF study shows stateful workloads are ~40% so this matters. Proposal adds change block tracking via a CSI extension that returns block-diff metadata through a two‑hop flow (a session CR and a secure gRPC proxy) to avoid bloating the API server. An internal prototype exists and the design is approved by the Data Protection WG but still needs CSI and Kubernetes architecture approval.
AI-written summary · about
chapters (13)
- 0:00Intro
- 0:39Agenda
- 1:43Stateful applications
- 3:57Data protection in Kubernetes
- 5:19Data production working group
- 6:43Why we need new data protection approach
- 9:26How to achieve data protection
- 10:05Whats missing
- 15:21Workflow
- 16:07Change block tracking
- 17:47Incremental backup workflow
- 20:41Client workflow
- 22:47Conclusion
related talks
30mBringing Incremental Backups to Kubernetes with CSI Snapshot Metadata Service
30mKubernetes Data Protection WG Intro & Deep Dive
26mKubernetes Data Protection WG Intro and Deep Dive
31mKubernetes Data Protection WG Intro & Deep Dive
35mBeyond Storage Management
28mKubernetes Data Protection WG Deep Dive
25mIntro & Deep Dive: Kubernetes Data Protection WG
30mKubernetes Data Protection WG Deep Dive