Runbook: Auto-Update Rollback

Alert: OpenClawAutoUpdateRollback

Meaning

An automatic version update was applied but the updated pod failed health checks, triggering an automatic rollback to the previous version.

Impact

The instance is running on the previous version. The failed version is recorded in status.autoUpdate.failedVersion and will be skipped in future checks until a newer version is released.

Diagnosis

# Check auto-update status
kubectl get openclawinstance <name> -n <namespace> -o jsonpath='{.status.autoUpdate}' | jq .

# Check rollback count
kubectl get openclawinstance <name> -n <namespace> -o jsonpath='{.status.autoUpdate.rollbackCount}'

# Check the failed version
kubectl get openclawinstance <name> -n <namespace> -o jsonpath='{.status.autoUpdate.failedVersion}'

# Check events for details
kubectl describe openclawinstance <name> -n <namespace> | grep -A5 AutoUpdate

# Check pod logs from the failed version (if still available)
kubectl logs <name>-0 -n <namespace> -c openclaw --previous --tail=100

Mitigation

1. Wait for fix - The failed version is automatically skipped; the next release will be tried
2. Investigate failure - Check pod logs to understand why the new version failed health checks
3. Disable auto-update - Set spec.autoUpdate.enabled: false to stop automatic updates
4. Increase health check timeout - If the new version needs more startup time, increase spec.autoUpdate.healthCheckTimeout
5. Reset rollback count - After 3 consecutive rollbacks, auto-update pauses. Fix the issue, then manually update spec.image.tag to reset