Closed zroubalik closed 4 years ago
This is a duplicate of https://github.com/kubernetes-incubator/custom-metrics-apiserver/issues/3 I think.
I think #3 would have been a solution to the problem, but I like that this issue states the problem rather than the solution.
Have any other solutions been discussed before? Is there a reason, for example, that all custom metrics must share the same apigroup?
@ecordell i don't know if the question is still relevant, but just to give another potential solution is to solve it in the underlying monitoring system. Probably it is not applicable to all adapters out there, but in the Prometheus ecosystem this can easily be achieved using a Thanos based setup, see [1] and [2].
[1] https://github.com/DirectXMan12/k8s-prometheus-adapter/issues/267 [2] https://www.youtube.com/watch?v=qTxunwzYO0g
@s-urbaniak I'd like to kindly ask why is this issue closed? Prometheus adapter is not the only usecase, from my POV this problem is still relevant. Thanks!
Is there a way how to deploy multiple adapters in a cluster? Eg. there's already deployed custom metrics server Foo and I need to deploy my own implementation Bar which provides different set of metrics.
If I understand it correctly, to be able to provide custome/external metrics API, the adapter needs to deploy cluster wide
APIService
with a specific name. Deploying second adapter in the cluster will results in reconfiguring the originalAPIService
object.