Investigating PostgreSQL performance issues
Problem to solve
Sometimes Support gets tickets and emergencies for GitLab environments where performance of the database is identified as the root cause.
Purpose of this issue is to stash queries, learning points, context etc. from these investigations.
Contents
- ANALYZE
- collect data on ACTIVE workload
- environment information
- PgBouncer
- config changes
- idle connections
- server load and metrics
- potential vacuum problems
- investigate specific slow queries (EXPLAIN)
- index bloat
- monitor sessions
- investigating sessions that are blocked by locked rows
- table / index sizes
- Make sure statement timeout is set
- Finding slow queries
- auto_explain
deep dives
Further details
Proposal
Who can address the issue
Other links/references
Edited by Ben Prescott_