Simple K8s controller to clean-up orphaned local PVs (using nvme's).
We often use projects like TopoLVM and OpenEBS LocalPV to provision PersistentVolumes (PVs) backed by ephemeral local-instance storage, such as NVMe disks on AWS EC2 instances. Even when the retention policy is set to Retain, the underlying storage is lost when the instance shuts down. This results in orphaned PVs, leading to errors when Kubernetes attempts to reattach them especially if a new node reuses the same IP address. Most of the CSI drivers won't delete the PVs in these scenarios.
To address this, we need a solution that continuously monitors these scenarios and automatically deletes orphaned PVs.
- Automatic orphaned PV cleanup: Identifies and deletes PVs that are not bound to any existing node.
- Dry-run mode: Allows testing without performing actual deletions.
- Configurable Volume Node Affinity labels: Supports custom node selector labels for determining volume node affinity. Since, CSI drivers define their own topology label.
- StorageClass Filters: Allows filter the volumes based on multiple storage classes.
To deploy the Local PV Cleanup Controller in your Kubernetes cluster using Kustomize plugin in Kubectl:
kubectl apply -k ./config/default/The controller supports the following additional flags than the default flags:
| Flag | Default | Description |
|---|---|---|
--dry-run |
false |
Run in dry-run mode without making actual changes. |
--node-selector-keys |
topology.topolvm.io/node |
Comma-separated list of labels used in PV node affinity to determine the node name. |
--storage-class-names |
topolvm |
Comma-separated list of StorageClass Names used to filter the PVs. |
--requeue-duration |
15m |
Duration for PV reconciler requeue if the node exists (e.g., 5m, 10m, 1h). |
Feel free to open issues or submit PRs if you have any improvements or bug fixes.