Vladislav Zolotarov
Vladislav Zolotarov
> @vladzcloudius At the time of making that patch, I tried getting your feedback (asking in comment: [#10440 (comment)](https://github.com/scylladb/scylla/issues/10440#issuecomment-1112111452); adding you as a reviewer: #10456). My apologies but I must...
Unrelated but following the proposed solutions mentioned on #10440: you didn't need to revert the "2-sections" patch to restore the original 1-section behavior. Setting `SectionHitThreshold=0` (which not by accident is...
@slivne Please, note that the offending patch has been backported to 5.0. Therefore the fix needs to be backported to 5.0 too.
Related to https://github.com/scylladb/scylla/issues/7845 since cleanup are effectively compactions. So, we may want to separate these semantically but the machinery is going to be the same.
@eliransin @slivne @tarzanek @tzach FYI @slivne This enhancement is a precondition for further improvements in our compactions related tooling
@slivne @amnonh FYI
> what else is required? in in what priority? @bhalevy We need user initiated compactions become non-blocking a year (or more?) ago - this is blocking a SM major compaction...
> If anything comes into mind, please open a particular issue for it To make it very simple: every REST API that triggers a task should be asynchronous, not just...
> Modifying the whole API as you suggested is unrealistic, as this will require changing jmx to satisfy the synchronous nodetool semantics. You don't have to change `nodetool`, `jmx` or...