Investigate and optimize PostReceive worker performance bottlenecks based on instrumentation data

Part of Improve PostReceive Worker Performance to Meet ... (&19159)

Summary

Using data from enhanced instrumentation (see Add granular instrumentation to PostReceive wor... (#566545)), identify slowest components of the PostReceive worker to achieve the 10‑second performance target.

Kibana: https://log.gprd.gitlab.net/app/r/s/Gd0Pq+

Problem

Based on initial analysis and the parent issue investigation, several performance bottlenecks have been identified:

1. Redis Cache Operations

   • 35,566 repository cache calls that took 9.7 seconds
   • Related to the Epic gitlab-org&17190 for cache optimization

2. Lock Contention

   • Multiple PostReceive jobs competing for the same locks
   • Particularly problematic with pull mirroring scenarios

3. Database Query Performance

   • Transaction overhead and query optimization opportunities
   • Potential N+1 query patterns

4. CPU-Intensive Operations

   • Repository expiration logic
   • Event processing overhead

Approach

• Analyze the results of instrumentation from Add granular instrumentation to PostReceive wor... (#566545)
• Analyze patterns and identify top bottlenecks
• Create issues to optimize each bottleneck

Acceptance Criteria

• Performance analysis report based on instrumentation data
• Create issues for identified bottlenecks (ideally, with suggestions how to fix them)

Related Issues

• Related to: #553426