Use PD MinIO S3’s Retention controls only when permanently removing a terminated service’s storage is intended. Purge can remove every associated bucket, object version and delete marker; it is not routine connection or quota repair.

Resolve recovery needs first

Identify the exact service, all associated buckets, retention deadline and approval state. Complete required recovery and verify the copy before allowing deletion. A retained module inventory is not a backup of deleted files.

Approve due work or hold it

1. Open PD MinIO S3 → Retention.
2. Review the intended terminated service and its countdown.
3. Choose Approve purge and review its confirmation when deletion is intended.
4. Inspect the saved approval state.
5. Choose Hold before further queue work if recovery or review is still needed.

Approval permits the scheduled queue to act when due. Hold stops future queue work, including later runs of a partial purge; it cannot roll back removed objects.

Understand Purge now and completion

For deliberate immediate deletion, expand the service row, type the exact requested primary bucket name and review Purge now and its final confirmation. This can advance eligibility and start or queue work; large storage can take several runs.

The purge processes historical versions and unfinished uploads, but a permission refusal or provider protection can stop completion after earlier deletion succeeded. Inspect Activity and final state rather than treating a button response as proof every bucket is gone. A refused bucket check is not proof the bucket is absent.

Never retry an uncertain destructive action blindly. Retain the service, original action time and redacted error while your hosting team checks the remaining state; do not include credentials or private object contents in diagnostics.

Check the deletion state

Your hosting team should compare the exact terminated service, complete associated bucket set, saved approval/Hold, due time and latest Activity result with the intended deletion decision. A large purge can span runs, and a refusal can occur after some objects have already gone. Your storage administrator should inspect the actual remaining buckets, versions, multipart uploads and permission or protection refusal before further deletion is authorised. Hold affects later queue work; do not assume it interrupts a deletion request already executing.

See the PD MinIO S3 Blesta guide for related tasks, or explore the module.