Björn Rabenstein
Björn Rabenstein
Yeah, sure, if we end up with tweaking the number of samples per chunk, it could be a hidden flag for the time being. But maybe going down the path...
I would say, let's put it into a (hidden) flag here in this PR (similar to `--storage.tsdb.min-block-duration` and friends, e.g. `--storage.tsdb.max-samples-per-chunk`). With the flags, users can experiment, and based on...
Thanks, merging now…
Very good point about storing the setting per memSeries. That seems unneeded indeed. Even if it were not fully responsible for the memory increase, it is still confusing conceptually (because...
Updated the PR title to not cause confusion in the future.
I think #12055 is an easy thing to do and should totally get merged if the more sophisticated approach here will take its time. (However, as commented in #12055, I'm...
Do you want more review comments now or are you still working on this? (Asking because the PR is marked as draft.)
I don't have a strong opinion here. All my ideas expressed so far are based on educated guessing. For the real decision, we should look at benchmarks. If there is...
OK, so let me reiterate what I have understood: * This PR introduces a size limit for histogram chunks with the same extrapolation logic as previously applied for the sample...
Yes, please let's move the bug fix for the snapshotting to another PR. For one, we should let @marctc and @codesome review it separately. Also, the fix should probably go...