Why should you care?

PostgresVerse payment primary AZ failure triggers failover so Swiggy-scale order capture resumes in minutes.

See it live — copy this example

Run in pgAdmin or psql.

-- After old primary is gone, on standby:
SELECT pg_is_in_recovery();  -- true before promote

-- promote (exact command depends on install)
-- pg_ctl promote -D /var/lib/postgresql/data

SELECT pg_is_in_recovery();  -- false after promote

What happened?

• Standby starts as recovery=true.
• Promote ends recovery and opens writes.
• Applications must use virtual IP or service discovery — connection string change.

Practice next

1. Lab: stop primary container/process.
2. Run promote on standby per your lab doc.
3. INSERT test row on new primary.
4. Document timeline: detection, promote, DNS TTL, app pool refresh.
5. Use pg_rewind to rejoin old primary as standby after split.

Remember

Failover = promote standby + reroute clients. Measure RTO/RPO in drills not disasters. Rewind or rebuild old primary as new standby after.

PostgresVerse AZ outage

Patroni promotes standby in 45s; K8s updates service endpoint.

Outcome: Orders lost only uncommitted sub-second window — within SLA.