execbuilder: fix a rare flake #82642
Conversation
This commit fixes a rare flake in the `vectorize_local` test. I don't quite understand why the test flakes tbh, but deduplicating the messages from the trace makes it non-flaky while still ensuring that the test exercises what it is intended to (that two spans are scanned in parallel). Release note: None
|
This change is |
There was a problem hiding this comment.
If you think this could be more than a testing artifact, maybe we should open an issue to look into it. I guess it's hard to know, since it seems this is not easily reproducible. Could something be added to tracing to help debug it in the future?
Reviewed 1 of 1 files at r1, all commit messages.
Reviewable status:complete! 1 of 0 LGTMs obtained (waiting on @cucaroach)
|
To be honest, I'm very puzzled by this failure. It is reproducible, but quite rare - like once in several hundreds of runs, sometimes in a thousand. I thought it was related to a range split, so I added a My guess is still that it has something to do with the testing setup around the zone configs, range boundaries, etc, since the first failure is relatively recent, and we had some changes in that area, but I don't think it's worth investigating this issue further - I spent like an hour and a half on it already, and I highly doubt it's a correctness issue. TFTR! bors r+ |
|
Build succeeded: |
|
I did bisect of this failure and got to 77adcaf (which was my hypothesis too). So it looks like we could include the issue into the list of #75639. I still don't think it's worth spending more time on this. cc @irfansharif for visibility. |
This commit fixes a rare flake in the
vectorize_localtest. I don'tquite understand why the test flakes tbh, but deduplicating the messages
from the trace makes it non-flaky while still ensuring that the test
exercises what it is intended to (that two spans are scanned in
parallel).
Fixes: #81636.
Release note: None