Kubernetes 1.37 and storage: test SELinux, CSI, and PVC before the first node

The signal and the facts that change the decision

Kubernetes 1.37 brings relevant changes to volume operations: evolution of SELinuxMount, storage-capacity scheduling, awareness of CSI limits, and PVC unused state. Impact depends on drivers, labels, volume sharing, and feature gates in each distribution.

The release announcement describes capacity and CSI functions with their own feature gates and defaults; some remain opt-in. PVC 'unused since' state reflects controller observation and neither orders deletion nor proves the storage backend unmounted at that instant.

SELinux and volume sharing need tests with actual labels and drivers; theoretical cluster compatibility does not cover CSI implementation. A 'Running' Pod does not prove integrity or recoverability. Label problems may appear only on remount, CSI limits during scale-from-zero, and PVC conditions can be mistaken for automatic deletion authority.

Decision, limits, and stop condition

Create a storage test track separate from the general smoke test. Reproduce drivers, StorageClasses, access modes, SELinux labels, expansion, snapshots, attach limits, autoscaling, failover, and recovery; only then include nodes with stateful workloads in the canary.

A 'Running' Pod does not prove integrity or recoverability. Label problems may appear only on remount, CSI limits during scale-from-zero, and PVC conditions can be mistaken for automatic deletion authority.

Execution and proof of completion

Execution should start by inventorying csi drivers, versions, storageclasses, modes, snapshots, expansion, topology, selinux, and workloads sharing volumes. Then, restoring test data, mounting, writing, restarting, moving pods, scaling from zero, and forcing failure and remount scenarios. Closure requires comparing hashes, latency, events, csi errors, and pvc conditions and keeping automated cleanup off until governed by its own policy.

The storage gate closes when drivers declare support, data survive the scenarios, attach and autoscaling respect limits, labels do not block remount, and snapshots restore within stated RPO and RTO.

• inventorying CSI drivers, versions, StorageClasses, modes, snapshots, expansion, topology, SELinux, and workloads sharing volumes.
• restoring test data, mounting, writing, restarting, moving Pods, scaling from zero, and forcing failure and remount scenarios.
• comparing hashes, latency, events, CSI errors, and PVC conditions and keeping automated cleanup off until governed by its own policy.
• The storage gate closes when drivers declare support, data survive the scenarios, attach and autoscaling respect limits, labels do not block remount, and snapshots restore within stated RPO and RTO.