---
name: storage-backup-disaster-recovery
description: Design, inspect, verify, and operate S3-compatible storage, MinIO, local volumes, file platforms, Restic, Borg, Proxmox Backup Server, database dumps, snapshots, retention, replication, restore testing, and disaster recovery. Use for cloud-drive, MariaDBAutobackup, Docker volumes, customer files, application backups, storage incidents, retention planning, and recovery exercises.
compatibility: Inventory and verification should use read-only access. Snapshot creation, retention changes, restore, deletion, bucket-policy changes, replication changes, and data migration require explicit authorization.
license: Apache-2.0
---

# Storage, backup and disaster recovery

A successful backup job is not proof of recoverability. Verify independent restore points, integrity and an executable recovery procedure.

## Inventory

Record datasets, owners, sensitivity, size, growth, source path or bucket, database consistency needs, backup method, encryption, retention, replication target, recovery-point objective, recovery-time objective and restore owner.

## Backup review

1. Confirm every production dataset is mapped to at least one backup policy.
2. Distinguish snapshots, replicas, synchronization and actual backups.
3. Verify recent job status, object count, bytes, checksums and retention pruning.
4. Check that backup credentials cannot modify the source and source credentials cannot delete all backups.
5. Confirm encryption keys and recovery instructions are independently protected.
6. Detect orphaned volumes, stale buckets, incomplete multipart uploads and unprotected databases.
7. Compare policy with observed restore points.

## Restore test

1. Select a representative restore point and isolated target.
2. Record expected files, database schema, application version and checksums.
3. Restore without overwriting production.
4. Validate integrity, permissions, ownership, application startup and domain-specific consistency.
5. Measure actual recovery duration.
6. Delete the isolated test only after evidence is retained.
7. Update the runbook with observed gaps.

## Retention and deletion

Generate a preview showing every object, snapshot or restore point that would be removed, its age, policy rule and whether another independent copy exists. Legal holds, customer retention and active incident evidence override normal pruning.

## Security rules

- Never return storage access keys, encryption keys, signed URLs or private object contents.
- Do not mount production credentials into restore tests.
- Do not call replication a backup when deletion or corruption propagates.
- Do not purge orphaned-looking data until ownership and references are checked.
- Preserve immutable or offline copies for critical datasets.

## Output

Return dataset coverage, last successful and last verified restore points, policy drift, unprotected data, retention preview, restore-test evidence, measured RPO/RTO, credential-boundary findings, actions performed and remaining recovery risk.