Skip to content
pgbx

Postgres down?

When Postgres will not start, SQL is gone. pgbx still works.

Terminal window
pgbx diagnose
pgbx diagnose --log /var/log/postgresql/postgresql-16-main.log --pgdata /var/lib/postgresql/16/main --json
docker logs db 2>&1 > /tmp/pg.log && pgbx diagnose --log /tmp/pg.log

It reads (each optional): postmaster.pid and liveness, pg_ctl status, the newest server log, journalctl -u postgresql*, --log FILE, the kernel log (OOM kills), disk usage of the data dir, pg_wal, replication slots holding WAL, pgsql_tmp, the log dir, and the data dir owner/mode.

Probable causes: oom_kill, disk_full, wal_backlog, stale_pid, crash, corruption, permissions, config_error, too_many_connections, unknown.

{"ok": true, "command": "diagnose", "safety": "readonly",
 "postgres": "down", "probable_cause": "disk_full",
 "evidence": [{"source": "...", "line": "..."}],
 "steps": [{"tier": "destructive", "why": "...", "command": "...", "needs_human_approval": true}],
 "facts": {}}
  • pgbx never runs a step. Destructive steps are marked for human approval.
  • pg_wal, base, global and pg_xact are never offered for deletion.
  • pgbx doctor runs the diagnosis automatically when Postgres is unreachable (diagnosis in its JSON).
  • Lost the server for good? pgbx db-restore --from-s3 restores each database onto a new one; see Disaster recovery.