Make DB metric expectations resilient to zero durations
-
Please check this box if this contribution uses AI-generated content (including content generated by GitLab Duo features) as outlined in the GitLab DCO & CLA. As a benefit of being a GitLab Community Contributor, you receive complimentary access to GitLab Duo.
What does this MR do and why?
In our MergeRequests::Mergeability::Logger specs we were asserting that both mergeability.expensive_operation.db_main_count.values and mergeability.expensive_operation.db_main_duration_s.values
must always be present. In reality, whenever an operation makes zero SQL calls (or the total duration rounds to 0), our implementation deliberately skips emitting the <.db_main_duration_s> key:
# observe_sql_counters in app/services/merge_requests/mergeability/logger.rb
result = value - start_db_counters.fetch(key, 0)
next if result == 0
observe("mergeability.#{name}.#{key}", result)This guard makes the spec flaky, because sometimes db_main_duration_s.values is genuinely omitted when its value is zero.
Consideration for Logger Behavior (Future Enhancement):
While removing next if result == 0 entirely in the logger would ensure all metric keys are always present, investigation showed this could add approximately 35 extra zero-valued keys per mergeability check, which might be considered excessive log noise.
A potential future enhancement to the logger could involve an allowlist of specific metric keys (like durations corresponding to non-zero counts) that should always be emitted even if their value is zero, to balance consistency with conciseness. For now, the test-side fix is preferred to maintain the current logging behavior.
References
How to set up and validate locally
- Check out this branch.
- Run the affected spec:
bundle exec rspec spec/services/merge_requests/mergeability/logger_spec.rb - The tests, particularly those under "when its a query", should now pass consistently.
MR acceptance checklist
Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.