FieldManual

Operator field manual: playbooks, runbooks, checklists, and postmortems.

Postgres Failover

Severity: critical Last verified: 2026-03-01
databasereplication write-failuresconnection-reset promote-replicadns-flip

Procedure

  1. Confirm primary is unavailable for 60+ seconds.
  2. Promote healthiest replica and freeze schema migrations.
  3. Update connection routing and recycle API pools.

Post-checks

  • Write throughput restored
  • Replica lag below 3 seconds