Radoslaw Zarzynski

@markhpc: testing the improved `ceph_test_rados` alone (without the EC partial read changes) has shown there are related failures. Most of them happened in cache tiering scenarios, away from EC. To...

@ljflores: yeah, the problem with this run is it's on the code from 24th of March (`/a/yuriw-2024-03-24_22:19:24-rados-wip-yuri10-testing-2024-03-24-1159-distro-default-smithi/7620593`), so it doesn't contain the recent fix (https://github.com/ceph/ceph/pull/55196/commits/888963a10e40de8fd38d5f4011b88a17d493862d) found with the changed `ceph_test_rados`.

This was on the agenda of the yesterday's RADOS Team Meeting. In short: * going to make a group review, * this is a deep change, requiring long baking in...

@pdvian: is there any progress in QA? Maybe just osdmaptool could help...

Let's see whether we have any coverage for excessive shuffling in teuthology but do not merge till disproving the hypothesis.

> If we change the crush rule, will we have this excessive data reshuffling going on? I guess so. This is being actually tested in downstream right now. > Update:...

Thank you, @pdvian!

I had a discussion with Prashant. We need at least a release note as for the multi-take (single take is fine!) rules a data shuffling can actually occur. Relabelling for...

> This is called *every* time somebody calls `.then()` on a `future` or every time somebody does anything with Seastar. I disagree, this piece of code is called when the...

> You don't really disagree. I do. _Every_ means _every_. If you use the great quantifier, please honor the consequences. > We do agree that Seastar has these per-call allocations....