Backup and restore

A backup is proven only by restoring it. This tool is the database's own backup and restore, or the managed service's snapshots, found in the project's operations config rather than assumed.

Finding it

• The backup job: a scheduled job, a Kubernetes CronJob, or the managed database's backup settings.
• Where backups go: the bucket or the list of snapshots, and who may read them.
• The database's version, so the restore tool matches it: pg_restore cannot read a file from a newer pg_dump.

Using it

• Restore the latest backup into a scratch database, never over a live one: pg_restore --dbname=<scratch> <file>, or a new instance from a snapshot.
• Check what came back: the row counts of the main tables, and the newest row's time against the backup's.
• Time the restore: it bounds how long an outage lasts.
• On a schedule, and after any change to the backup job.

Examples: pg_dump and pg_restore (PostgreSQL), mysqldump (MySQL), mongodump and mongorestore (MongoDB), a managed database's snapshots.