Homelab / A primary, a streaming replica, and runbooks that record what was actually run.
Data and recovery
One guest holds the PostgreSQL primary and Redis, another keeps a streaming standby. Each recovery procedure records whether it was ever run against this cluster: the replica promotion was tested through timeline 2 and re-seeded, and the etcd restore was promoted, verified, and reverted. A procedure nobody has run is a hypothesis.
One guest holds the PostgreSQL primary and Redis; another keeps a streaming standby.
Each recovery procedure lives as a document next to the scripts it runs.
That document records whether the procedure was ever run against this cluster.
How I built it
The replica promotion was tested through timeline 2 and then re-seeded; the etcd restore was promoted, verified, and reverted. The cases that have not been run are marked as such in the same table, because a procedure nobody has run is a hypothesis, not a backup.