View snapshots
vCluster Platform can take snapshots of a cluster at scheduled intervals. For configuration and examples, see Configure auto snapshots.
View snapshots​
After enabling automatic snapshots, you can view the list of snapshots for each cluster in the details of the cluster. This menu is only available to project admins and platform admins.
Snapshot name​
Snapshots are identified by a generated name formatted as <Virtual_Cluster_Name>-<Snapshot_Timestamp>.tar.gz.
Snapshot schedule lifecycle​
Snapshots move through statuses as they run. A snapshot starts as Scheduled. At its scheduled time it moves to
In Progress while the cluster writes it to storage, then to Completed once the archive is stored. If the snapshot
can't be taken, it moves to Failed. If a snapshot is removed from the storage backend but is still listed in
vCluster Platform, it appears as Not Found.
| Snapshot Status | Description |
|---|---|
Scheduled | A snapshot that is scheduled to be stored in the storage backend service. Only one snapshot can have the Scheduled status at a time. |
In Progress | A snapshot that the cluster is writing to the storage backend service. It also stays in this status while the platform retries after a connection error. |
Completed | A snapshot that has been successfully stored in the storage backend service. |
Failed | A snapshot that failed, or whose retry window ran out. The snapshot details show the reason. |
Not Found | A snapshot that appears in the list but has been removed from the storage backend service. |
If the cluster is sleeping at the scheduled time, the snapshot fails with the reason vCluster is sleeping.
Retries after connection errors​
In vCluster Platform v5.0 and later, a connection error doesn't fail a snapshot. The platform starts each snapshot by
creating a snapshot request in the cluster, then checks that request until the cluster reports a result. The cluster
keeps writing the snapshot even while the platform can't reach it. So an error that says nothing about the snapshot
itself, such as a dropped network peer or a timed-out API call, leaves the snapshot In Progress. The platform
retries about every 10 seconds within a retry window.
| Stage | Retry window | Measured from |
|---|---|---|
| Creating the snapshot request | 5 minutes | The scheduled time |
| Checking an existing snapshot request | 30 minutes | When the request was created |
When the window runs out, the snapshot moves to Failed, and its reason starts with gave up after, followed by the
window and the last error. The windows limit how long the platform retries connection errors, not how long a
snapshot can take. A snapshot that the cluster reports as still running stays In Progress without a time limit.
If the platform created the request but lost the response, it finds that request on its next attempt and follows it, rather than starting a second snapshot to the same location.
While the platform retries, the UI keeps showing In Progress. The platform log records each attempt with the message
could not check in-progress snapshot, retrying and the error.
These errors fail a snapshot at once, without retrying:
- The cluster reports that the snapshot failed or partially failed.
- The snapshot request is missing from the cluster, or its data can't be read.
If you turn off automatic snapshots, the platform removes scheduled snapshots. It keeps checking an in-progress snapshot whose request exists, with the same retry windows, until the snapshot completes or fails. An in-progress snapshot whose request was never created is removed.
Retention​
snapshots.auto.retention sets how many days to keep snapshots (period) and how many to keep (maxSnapshots).
The platform applies both settings to two groups of snapshots, and counts each group separately:
- Completed snapshots. The platform prunes these only right after a new snapshot completes, so a run of failures never deletes the snapshots you would restore from. Pruning a completed snapshot also deletes its archive from storage.
- Failed and Not Found snapshots. These share their own count and are pruned even while recent snapshots keep failing. In vCluster Platform v5.0 and later, if storage holds an archive for a pruned failed snapshot, the platform deletes that archive too. That happens when a snapshot partially failed, or when the cluster finished writing it after the retry window ran out.
Scheduled and In Progress snapshots are never pruned.