Niclas Schad

@danfromtitan > they have to have different names, which the chart doesn't support I don't really understand that point. When you use a different helm release name the name also...

> > When you use a different helm release name the name also changes also with Kubernetes namespaces I don't see any Issue > > You'd want a single helm...

> Yeah, memberlist is another one, I use etcd so I don't have that problem. > > The idea is to create a dynamic config file in Cortex and add...

> mutating admission controller ~Okay interesting so we basically watch for resource "node"~ (didn't know about that binding thing, even better) and then call our webhook which will then read...

Do you wanna give that a go? @danfromtitan Would be great if we could just use something that already exists

> I'm looking into it. Do you want the webhook config/deployment added to the chart with a zone_aware_enabled flag around it ? :+1: sounds good. > There ain't any that...

> @kd7lxl - there is one fundamental issue with the idea of using an admission controller: at the time the API request makes it through the controller, there is no...

> > The pod/binding event looked promising. Have you tried that? > > Couldn't find any "pod/binding" resource in the API reference. The [implementation of binding a pod to machine](https://github.com/kubernetes/kubernetes/blob/master/pkg/registry/core/pod/storage/storage.go#L162)...

I had a peak at some admission webhook examples and the whole setup can be a bit tricky, especially since you need a proper certificate for your webhook server. Which...

yes that was intentional as CPU scaling for ingesters in my opinion is not really relevant.