Łukasz Mierzwa

https://github.com/prometheus/prometheus/pull/10709#issuecomment-1129330804

`TestTombstoneCleanRetentionLimitsRace` does fail in CI, which is great because I couldn't get it to fail at all. That's helpful. But that only confirms that `TestTombstoneCleanRetentionLimitsRace` does work in CI. It's...

I've deployed this patch on 4th of October to some of our production instances where we've previously observed high latency on query API calls just after compaction runs. Results can...

I think I've updated all places that read `db.blocks` to have a read lock.

> Are we ok just removing the test? I read through the comments and it sounds like you hoped to find a way to test the race. Im not sure...

> Hello from the bug-scrub! It looks like this was approved and ready to go, but now it has a small merge conflict. > > I'll also link #14610 which...

rebased

> I would prefer to see the functions involved in this PR all document whether they need certain lock(s) on entry. Like is already done in a few places saying...

Test failure seems unrelated: ``` === NAME TestDelayedCompaction compact_test.go:2015: Error Trace: D:/a/prometheus/prometheus/tsdb/compact_test.go:1949 D:/a/prometheus/prometheus/tsdb/compact_test.go:2015 Error: "2.9755205s" is not less than "2s" Test: TestDelayedCompaction ```

> I want to note that this only works if the different sets of series don't overlap in time. I don't think I understand what you're trying to say here.