Backup and restore

Operations Center can back up its state into a gzip compressed tar archive and restore it again.

Content of a backup

A backup holds:

  • the configuration (config.yml)

  • the server and client certificates and keys

  • a consistent snapshot of the database

  • the cluster artifacts

  • the images served by Operations Center, only if requested

  • the cached update files, only if requested

The first entry of the archive is the manifest backup.yaml. It records the format version of the backup, the version of Operations Center, which created it, the time of creation and if the images and the cached update files are included.

The inventory is not part of a backup. It is synced again from the clusters after a restore.

Note

A backup contains private keys and credentials. Store it safely.

Create a backup

operations-center system backup operations-center-backup.tar.gz

The images and the cached update files are only included on request:

  • --with-images includes the images

  • --with-updates includes the cached update files

  • --complete includes both

Without the cached update files, a restore keeps the cached update files, which are known to the restored database, and the missing updates are fetched again from the update source.

Without the images, the backup holds the image sources but no images. A restore removes the images present on the system, they are fetched again from the image sources. Uploaded images have to be uploaded again.

Restore a backup

operations-center system restore operations-center-backup.tar.gz

The backup is validated first. A backup without manifest, with a format version not supported by this version of Operations Center or created by a newer version of Operations Center is rejected. A backup of an older version is upgraded on restore.

Operations Center then restarts and replaces its state with the one from the backup. If Operations Center does not start successfully with the restored state, it puts back the state from before the restore on the next start.

A restore is refused, while a server is being deployed, updated, evacuated or restored or while a cluster update is in progress. Operations, which have been in progress, when the backup has been created, are aborted after the restore:

  • Cluster updates are aborted.

  • Server deployments are canceled, the servers are left untouched.

Backup and restore require the admin role on the server object.