evidence-focused query evidence
Query audit
Postgres slow query audit
Database Optimization Tool reviews slow query evidence before anyone changes production. The report ranks expensive statements, separates repeat workload from one-off latency, and points to the next safe diagnostic step.
pg_stat_statements prioritization
Temp write and read pressure
EXPLAIN before changes
What counts as a slow query finding
A slow query candidate should show measurable cost: high total time, high mean time, excessive rows, shared block reads, temp writes, or repeated execution under normal workload.
How queries are ranked
Database Optimization Tool prefers aggregate evidence over anecdotes. Statements with repeated cost, large IO pressure, or spill behavior are promoted ahead of isolated outliers.
Where pg_stat_statements helps
When pg_stat_statements is available, the audit can group normalized statements and compare calls, timing, rows, block reads, and temporary block activity.
Safe follow-up
The report does not rewrite SQL or apply changes automatically. It recommends EXPLAIN review, parameter checks, staging validation, and human-owned rollout decisions.
FAQ
Frequently asked questions
These answers describe the product focus: careful database evidence, clear findings, and team-approved next steps.
Does Database Optimization Tool monitor slow queries in real time?
No. It produces evidence-focused audit findings from collected PostgreSQL evidence. It is not a live monitoring or alerting system.
Do I need pg_stat_statements enabled?
It is strongly recommended. Without pg_stat_statements, slow query analysis has less query-level detail and depends more on surrounding table and index signals.
Will the report automatically fix slow SQL?
No. The report identifies likely causes and safe diagnostics. SQL changes, index changes, and rollout decisions stay with your team.
Why require EXPLAIN before changing production?
Slow query symptoms can have multiple causes. EXPLAIN helps verify plan shape, row estimates, scan type, joins, and buffer usage before a fix is treated as safe.