A service gets volumes with a mount path, size and storage class. Clusterward creates the claims, schedules snapshots with retention and deletes data only after a confirmed snapshot.
A volume in Clusterward is a persistent disk belonging to a service: a PersistentVolumeClaim on Scaleway Block Storage with a name, mount path, size and storage class, created before the deployment and mounted into the container. Snapshots are Scaleway Block Storage snapshots of the disk behind the claim, taken on a schedule with retention in days, or once before deletion. Undeploy leaves volumes untouched, a service with volumes cannot be deleted, and the only way to get rid of data leads through a confirmed snapshot.
Name, mount path, size, storage class and, optionally, the snapshot schedule.
During a deployment, the claim is created before the Deployment and mounted.
An hourly run creates due snapshots of deployed services.
Expired snapshots are only deleted from Clusterward’s own list.
First a snapshot, which must be ready, then the claim, then the volume.
The rules you learn the hard way with block storage on Kubernetes are built in here.
Each volume becomes its own claim, named after the release and volume, created before the deployment. Mount paths must not overlap, and names are unique.
A Deployment with volumes uses the Recreate strategy, because a rolling update gets stuck on a single-attach disk. Scaling is pinned to one instance.
One schedule and one retention period in days per volume. Snapshots are created as Scaleway Block Storage snapshots, appear in the console and can be restored with a click in the cockpit.
A volume’s size can increase, never decrease, and the storage class stays fixed. Removing a volume from the list is refused.
Undeploy doesn’t touch any claim. A service with volumes cannot be deleted until its data has been explicitly removed.
The only way: delete volume data, confirmed by typing, while the service is not deployed. A snapshot must be ready before the claim goes.
The claims a chart creates itself – including those of a database in a StatefulSet – appear on the Volumes card of the Helm service, marked “from the chart”, with mount path and size. Snapshots immediately or on a schedule.
When restoring, Clusterward stops every Deployment and StatefulSet that uses the disk, backs up the current state and recreates the claim with Helm’s metadata. The next upgrade still recognizes it as its own.
A disk the chart leaves behind on uninstall stays visible and can be deleted – with a final snapshot first.
Disks and snapshots via the Scaleway API, per zone.
The disk is resolved via PVC, PV and CSI handle.
Deleting the claim leaves the disk to the storage class.
Optional per volume, so non-root containers can write.
Each rule corresponds to a mistake that would otherwise cost data on block storage.
| Rule | Behavior |
|---|---|
| Scaling | Fixed, exactly one instance, no autoscaler |
| Strategy | Recreate instead of RollingUpdate |
| Size | Increase only |
| Storage class | Immutable after creation |
| Removing a volume | Only via “Delete volume data” with a snapshot |
| Undeploy | Claims remain |
| Snapshot deletion | Only entries from Clusterward’s list |
| Helm chart disks | Snapshots and restore as with your own volumes; the chart determines the size |
| Snapshot status | Shown as ready about a minute after completion at Scaleway |
Volumes belong to the service and are created or enlarged with every deployment. Snapshots require the Block Storage permission set in the cluster’s provisioning profile; see Cluster provisioning.
Environment building blocks of a pipeline carry volumes too; their rollback backs up the data before tearing anything down, as described under Tenant pipelines.
Which disks a Helm chart creates and what helm uninstall does with them is explained in the article Helm charts and their disks.
Git, container image or Helm chart – rolled out with a health check.
Go to DeploymentsPostgres and MySQL connected privately, one database per service.
Go to Managed databasesCustomer onboarding from building blocks, live in under a minute.
Go to Tenant pipelinesCan’t find your question? Write to us – we usually reply on the same business day.
Ask a questionWe attach a volume to a service, schedule snapshots and show the protected deletion path.
