Ganesh Vernekar
Ganesh Vernekar
Looks like this should have happened pre 2.40.x as well (the error existed before). Testing out what changed in between. We should not be erroring out here ideally.
Good to know that this existed before as well, and we don't need to error here. Earlier it didn't give a problem because of a bug, which 2.40.x fixed. See...
This was fixed by https://github.com/prometheus/prometheus/pull/11623
Looping in @cstyan and @csmarchbanks since things here are closely tied to remote write.
/prombench main
/prombench cancel Some noticeable reduction in allocation. It may be a bigger difference in certain setups.
> I want to know why not create a new field SubQueryStep. It not only passes down the interval of the SubQuery, but differs subquery from other scenes. Because for...
> What happens when Prometheus itself writes OOO? TSDB is agnostic to this. If a sample is OOO and within the out_of_order_time_window, it will be added to the TSDB. Regarding...
Yes TSDB could have some interface to allow appending samples without OOO logic. Could you please open an issue for that one so that we can track its progress?
Sounds good. I would prefer `UnsortedLabelValues` instead of passing a bool to avoid a lot of code changes (as `UnsortedLabelValues` would not be used very often)