Why should you care?

PostgresVerse SRE needs alerts when p95 checkout latency correlates with new deploy query plan regression.

See it live — copy this example

Run in pgAdmin or psql.

-- Enable auto explain in session lab
LOAD 'auto_explain';
SET auto_explain.log_min_duration = '200ms';
SET auto_explain.log_analyze = on;

SELECT o.* FROM orders o
JOIN customers c ON c.customer_id = o.customer_id
WHERE c.email LIKE '%@flipkart.com';

What happened?

• auto_explain logs EXPLAIN ANALYZE for queries exceeding 200ms to server log.
• Combined with pg_stat_statements you get aggregate plus per-slow occurrence plans.

Practice next

1. Set auto_explain in postgresql.conf or session.
2. Run intentionally heavy JOIN.
3. Tail log for EXPLAIN output.
4. Export pg_stat_statements to Prometheus postgres_exporter.
5. Set statement_timeout on role analyst to 5min cap.

Remember

Layer stats view, logs, and APM. auto_explain captures plans of slow queries automatically. Baseline before sale events.

PostgresVerse Grafana board

Dashboard plots top query total time and active sessions during sale.

Outcome: Rollback decision made in 10 min when new deploy spikes mean_ms.