Marco Pracucci

> Limiting series Currently series are not limited. The limit is always initialized to 0 (disabled), but @duricanikolic is working on it (see https://github.com/grafana/mimir/issues/3382), so we should try to improve...

> I'm tempted to also have some tolerance for incorrect estimations. Maybe if we're within the same order of magnitude as the limit, we can still attempt to complete the...

👍 Another approach could be having rulers discovering alertmanager via the ring, like proposed in https://github.com/grafana/mimir/issues/3900.

👋 Hi! Thanks for reporting it. Looks like an error not correctly handled. To better understand it, could you give me the query run by the "Error rate / component"...

I'm glad I've asked for the metric :) The metric is tracked in another project, specifically here: https://github.com/thanos-io/objstore/blob/main/objstore.go However, the failures metric is increased only if context wasn't canceled. For...

> i think @colega did some work to _not_ classify cancelled contexts as failed queries in #3837, that's still not in any released version though Right, but since the failure...

Hi! The short answer is that replication factor = 2 is not really supported in Mimir. The configuration accepts any value, but replication factor = 2 has never been battle...

> The default value of ingester.ring.unregister-on-shutdown is true in the document. However, in the helmchart, the value is set to false. What is the impact of this configuration? How to...

> I missed a key config: shuffle_sharding_ingesters_enabled = false. According to my own understanding, when shuffle_sharding_ingesters_enabled = false, Querier will query all Ingesters and return the results, and finally the...

> trying after while ... are there some news in mimir to be able to properly handle RF=2 RF=2 is still not properly supported and there's currently not a strong...