Multi-Replica Backup

Multi-copy backup refers to the same backup task while remaining on different media or different failure domains, such as local disk + object storage/network disk. Multiple copies are not a substitute for recovery verification: each copy should be sampled regularly and confirmed for recovery.

Referral Process

1. First use backup run to generate a backup locally and complete the dry-run check.
2. Make sure the local archive is readable, then use backup upload-artifact to write to the second destination.
3. Perform a backup list-latest check on the remote object for key, size, and modification time.
4. Regularly download a remote copy to perform decryption, unpacking, or recovery drills in a temporary directory.
5. Set retention periods for local and remote to avoid the same failure and delete all replicas at the same time.

notice

• Do not place two copies on the same disk, on the same host, or in the same fault domain.
• The bucket, prefix, endpoint, and region of the object store must be recorded and periodically checked.
• backup run --upload does not automatically clean up remote history files; remote cleanup must be performed separately.

Related: Upload archive, backup download-artifact, cleanup object-storage.