What It Captures
The slow query log writes an entry for every statement whose execution exceeds a configured threshold, often with the duration, the full text, and the actual parameters used. Where aggregate stats show you a normalized query, the slow log shows a real offending instance you can replay.
Setting the Threshold
- Too low a threshold floods the log and adds logging overhead to every request.
- Too high a threshold misses the moderately slow queries that add up.
- Start with a threshold above normal latency, then lower it as you fix the worst cases.
Using It Well
- Pair the slow log with statistics views for ranking; use the log to grab a concrete example with parameters.
- Reproduce a logged statement under EXPLAIN ANALYZE to see why it was slow.
- Watch for spikes in slow log volume as an early warning of regression or a load shift.
A Caution
The threshold is on a single statement's time, so it can miss a fast query run a million times. Combine it with cumulative stats so neither blind spot is left uncovered.
Key idea
The slow query log records real statements that cross a time threshold, giving concrete examples to replay, but pair it with cumulative stats to catch fast queries run very often.