Nick Pillitteri

Another option is to keep the query limit in place change the behavior of the query when the limit is hit. Instead of waiting for a slot to run, the...

One potential solution to this is to sort the tenant IDs (the parsing logic might already do this) and hash them and use the hash in the cache key instead...

> Do we have a fix coming in, or something to be done from our end? We apply a hash to keys to make them a consistent length elsewhere, we'll...

We inherit the `Bucket.Iter` method from the [Thanos objstore](https://github.com/thanos-io/objstore/blob/11ffbc490204bedf6d422e06a5fc1bfa0d861258/objstore.go#L71-L74) project. This seems like a pretty worthwhile fix with the potential to save a lot of work. We should discuss with...

> Design expected usage of Redis as cache (sentinel? cluster? master->replica setup?) For this, I'm going to go with the simplest Redis setup possible (single instances for each cache). My...

~I'm going to include https://github.com/grafana/dskit/issues/253 as part of this work. The multiple different cache abstractions in dskit are confusing. We (Mimir) are the only consumers of `RemoteCacheClient`, and only for...

Closing this issue since as discussed offline: * Minimal Jsonnet support for Redis config has been added * Helm support will _not_ be added because we don't run Redis and...

At some point in the future I'd like to add read-path request rate limits to the query-frontend so this would help with that :+1:

> #### Describe the solution you'd like > For the query-frontend to react to SIGTERM, by first marking itself as not ready and then proceeding with its shutdown procedure (flushing...

The CHANGELOG has just been cut to prepare for the next Mimir release. Please rebase `main` and eventually move the CHANGELOG entry added / updated in this PR to the...