Callum Styan

@QuentinBisson am I understanding correctly, this should supersede your other PR https://github.com/grafana/loki/pull/13023 ?

@QuentinBisson need to merge in again after some of the other PRs have been merged

hey @harry671003, can you share more details about what you're seeing? I see samples being sent correctly when testing locally with both sets of CLI arguments you've provided.

It looks to me like this is fixed already, at least looking at our internal deployment. I see a template which has an activity bump config of 3 hours, and...

@mafredri requested a review as you opened the original issue, please lmk if I've misunderstood `scaletest` and any parallelism in play there, which is the reason for using atomic updates...

The only other suggestion I would have would be to make the metric a package global, especially when using promauto. Alternatively we could replace promauto usage with regular metrics registration...

@bwplotka when you're back can you take a quick look here.

I can rebase this PR and run the benchmarks again sometime this week 👍

Benchmark diff is minimal still: ``` benchstat old new goos: linux goarch: amd64 pkg: github.com/prometheus/prometheus/storage/remote cpu: AMD Ryzen 7 9800X3D 8-Core Processor │ old │ new │ │ sec/op │...

@arejula27 there's a few more, `sendv2Samples` for example. We left comments on some functions that should be cleaned up. [Here's](https://github.com/prometheus/prometheus/pull/14348) an attempt @bwplotka started at DRYing the code using generics,...