Performing your first rollback
Trigger a rollback and verify the cluster returns to a known-good state.
A rollback restores an application to one of its captured snapshots. The in-cluster controller re-applies the manifests that the snapshot holds, in dependency order, and returns the rollback record once the apply phase finishes.
Pick a snapshot
List the snapshots for an application:
GET /v1/applications/{name}/snapshotsEach snapshot has an id, the change class that produced it, a
captured-at timestamp, and the list of manifest kinds covered.
Trigger the rollback
POST /v1/applications/{name}/rollback
Content-Type: application/json
{
"snapshot": "snap_2026-05-17T08:42:11Z",
"reason": "config drift on payments"
}The API returns immediately with a rollback id and a pending
status. Poll the rollback resource until it reaches restored or
failed.
Verify
When the rollback reports restored:
- The application generation is incremented.
- A change cycle of class
rollbackis recorded with severityinformational. - A fresh snapshot is captured of the restored state, so the next rollback target is the post-restore baseline.
If the controller fails to apply any manifest, the rollback record
stops at failed and the partial state is left untouched.