Peter Štibraný
Peter Štibraný
There is similar discussion in https://github.com/grafana/mimir/pull/2295.
> So the effective limit will change if the scheduler is scaled out or in. We don't suggest scaling schedulers, and recommend using 2.
As someone not very familiar with Helm and only changing jsonnet most of the time, I'm not too happy about this change. I think it will slow us down, and...
This is a bug to be fixed. We see it sometimes happening in our production environments. Not frequently, but it happens.
> Have you considered the option to wait until on-going uploads are completed before shutting down the compactor? I don't think that's an option. Compactor should honor shutdown request as...
> > Have you considered the option to wait until on-going uploads are completed before shutting down the compactor? > > I don't think that's an option. Compactor should honor...
> > Have you considered the option to wait until on-going uploads are completed before shutting down the compactor? > > I don't think that's an option. Compactor should honor...
https://github.com/grafana/mimir/pull/2285 fixes this, when it's enabled.
There are two values stored in KV store that need to be [synchronized between distributors](https://github.com/grafana/mimir/blob/14e4b3eea2c67a44da1d8aa90c53d12299006398/pkg/distributor/ha_tracker.proto#L13) - Current elected replica (`replica` field) - `received_at` If distributors have different view on these...
> Given I agree we need to handle conflicts (and we need to be prepared for them in edge cases), the conflict here could be solved picking the most recent...