WARNING: This driver is beta and should not be used in performance critical applications
DISCLAIMER: This is not an officially supported Google product
The GCP Compute Persistent Disk CSI Driver is a CSI Specification compliant driver used by Container Orchestrators to manage the lifecycle of Google Compute Engine Persistent Disks.
Status: Beta
Latest stable image: gcr.io/gke-release/gcp-compute-persistent-disk-csi-driver:v0.5.1-gke.0
| Driver Version | Kubernetes Version | Test Status |
|---|---|---|
| HEAD Latest | HEAD | |
| 0.5.x Stable | HEAD | |
| 0.4.x Stable | HEAD | |
| 0.4.x Stable | 1.13.5 | |
| 0.3.x Stable | HEAD | |
| HEAD Latest | HEAD (Migration ON) | |
| HEAD Stable | HEAD (Migration ON) |
This plugin is compatible with CSI versions v1.0.0
| GCE PD CSI Driver\Kubernetes Version | 1.10.5 - 1.11 | 1.12 | 1.13 | 1.14 | 1.15+ |
|---|---|---|---|---|---|
| v0.1.x.alpha | yes | no | no | no | no |
| v0.2.x (alpha) | no | yes | no | no | no |
| v0.3.x (beta) | no | no | yes | yes | yes |
| v0.4.x (beta) | no | no | yes | yes | yes |
| v0.5.x (beta) | no | no | no | yes | yes |
| dev | no | no | no | no | yes |
See Github Issues
| Parameter | Values | Default | Description |
|---|---|---|---|
| type | pd-ssd OR pd-standard |
pd-standard |
Type allows you to choose between standard Persistent Disks or Solid State Drive Persistent Disks |
| replication-type | none OR regional-pd |
none |
Replication type allows you to choose between Zonal Persistent Disks or Regional Persistent Disks |
This driver supports only one topology key:
topology.gke.io/zone
that represents availability by zone.
| Feature | Stage | Min Kubernetes Master Version | Min Kubernetes Nodes Version | Min Driver Version | Deployment Overlay |
|---|---|---|---|---|---|
| Topology | Beta | 1.14 | 1.14 | v0.5.0 | Stable |
| Snapshots | Alpha | 1.13 | Any | v0.3.0 | Alpha |
| Resize (Expand) | Alpha | 1.14 | 1.14 | dev | Alpha |
See Github Issues
As part of the deployment process, the driver is deployed in a newly created namespace by default. The namespace will be deleted as part of the cleanup process.
Controller-level and node-level deployments will both have priorityClassName set, and the corresponding priority value is close to the maximum possible for user-created PriorityClasses.