Don't Panic — Database Backup and Recovery

A backup is a carefully preserved chance to have the database you used to have. Recovery is the longer, less photogenic job of turning that chance into a service people can use. The distinction is useful because a green backup job is only a message about writing something somewhere. It has not yet volunteered to start the database, find its keys, obey its permissions, or answer an application query. Machines are admirably literal about this.

The first two ideas are RPO, the point in time to which data must return, and RTO, the maximum acceptable outage. RPO decides how much recent change can vanish. RTO must pay for retrieval, restore, log replay, validation, and cutover, all of which insist on occupying time at once. A short backup window is therefore not a short recovery. It is one ingredient in a soup that is inconveniently served during an incident.

The third idea is the recovery chain. A base or full backup may need incremental backups, ordered transaction logs, a catalog, credentials, and a compatible database engine. For PostgreSQL, archived write-ahead logs let recovery stop at a selected point. For MySQL, later binary logs can bring a full backup forward. Lose a required link, and the desired point becomes a fine historical concept with nowhere to go.

The surprise is that a running database is not the finish line. Rows may return while roles, configuration, extensions, or application connections do not. A business action may also have changed something outside the database. That is why a clean, isolated target is so useful: it gives the recovery team space to check the engine, schema, critical data, permissions, and application behavior before the recovered state meets real traffic. Replication helps continuity, but it can copy a bad deletion with impressive punctuality.

Read the Intro for the full map from objective to controlled cutover. Use Slides when the backup forms and recovery chain need a quick visual comparison. Keep the Cheatsheet nearby for selection rules and validation gates. The Practice tab turns the theory into a bounded restore drill, and the Quiz checks the decisions that make a backup a recovery plan rather than an elaborate collection of optimism.