Keep Persistent Volumes Healthy for Stateful Workloads
Talkat KubeCon + CloudNativeCon + Open Source Summit China 2021 · on Friday, 2021-12-10· at 13:15· Kubecon + CloudNativeCon 演讲厅, 虚拟会议 (Virtual)·
summary
Kubernetes now exposes a volume health monitoring capability where a CSI external health-monitor controller and kubelet via nodeGetVolumeStats query drivers and storage backends to emit PVC and Pod events when volumes are missing, degraded, corrupt, or out of capacity. The current implementation (alpha since 1.19, expanded in 1.21 and targeting 1.23) only reports issues; a local demo with the hostpath CSI driver v3.0.0 successfully produces abnormal events. Roadmap items include automatic reactions (auto-resize, delete PVC/Pods, push alerts).
AI-written summary · about
chapters (8)
- 0:00Start
- 0:45Agenda
- 1:21What Happened to my Volume?
- 6:23CSI Deployment Example
- 8:52Volume Health Monitoring (Controller)
- 10:47Volume Health Monitoring (Node)
- 12:09Volume Condition in CSI Spec
- 13:02Demo
related talks
48mCSI in Kubernetes
37mKubernetes Storage 101
28mKubernetes Storage Evolution: Enabling High Performance Distributed Datastores
29mStateful Workloads in Kubernetes: A Deep Dive
27mKubernetes SIG Storage: Intro & Deep Dive
24mCapacity-aware Dynamic Volume Provisioning For LVM-based Local Storage
58mProtecting Stateful Workloads With CSI Snapshot
36mPolicy-Based Volume Snapshots Management in Kubernetes