/pg-backup — backup & PITR review

Cluster/stanza: $ARGUMENTS

Invoke postgres-ha-ops. Remember: replication is NOT backup — it propagates a DROP TABLE, corruption or ransomware to every replica. Only pgBackRest + continuous WAL archiving gives PITR.

Walk this checklist and report what is present vs missing (read-only):

• pgBackRest stanza exists; pgbackrest check validates the archive path end-to-end
• archive_mode=on + archive_command on every node (the new primary must archive after failover)
• archive_timeout set so idle-WAL loss is bounded (RPO ceiling)
• Schedule: full + differential + incremental with a retention policy
• Off-host + off-site + immutable (Object Lock) + encrypted repo — 3-2-1-1-0
• Backups taken from a standby (offload the primary)
• pgbackrest verify scheduled; WAL archive gap count = 0; last-backup age alerted
• Restore tested — monthly restore + quarterly timed DR drill with recorded RPO/RTO
• A one-off pg_dump on the node is a supplementary export, not the strategy

If asked to take a backup now, do it from the leader/standby and verify with pg_restore -l before reporting success. Depth: reference/PostgreSQL-HA-Production-Install-Guide.md §10.