kubernetes / kubernetes/perf-tests
Pod startup time phases are inaccurate in longer tests.
Open
area/clusterloader
good first issue
help wanted
- Dominant language
- Go
- Stars
- 989
- Forks
- 654
- Avg merge
- 3d 12h
- Merged PRs (30d)
- 56
Description
gatherScheduleTimes:
https://github.com/kubernetes/perf-tests/blob/master/clusterloader2/pkg/measurement/common/slos/pod_startup_latency.go#L268
is currently listing events at the end of the test.
Given that events by default have 1h TTL, for measurements across longer periods it just returns incomplete results. Given that we don't 100% accuracy, we should switch to a mechanism that is similar to the one that slo-monitor is using (added in #1477 ).
Contributor guide
Assessment
This issue has not been assessed yet.