Add karpenter_provisioners_usage_pct metric back about karpenter-provider-aws HOT 2 OPEN

We are actively using this metric to monitor and alert on the provisioner usage

You should be able to construct this metric by taking the usage metric and dividing it by the limits in a PromQL query. We opted for this path over directly generating the metric ourselves, since this effectively tracks redundant information. Is there a limitation around constructing the query this way?

from karpenter-provider-aws.

@jonathan-innis which metrics in particular? I am trying to do this and I am struggling to do so. The metric karpenter_nodepool_usage seems to report correctly the provisioned resources by nodepool but metrics like karpenter_nodes_total_pod_requests don't seem to have the nodepool label and I am not sure what they return as they don't seem to match the resources requested by node, or at least they don't match what kube-capacity returns for cpu/memory.

from karpenter-provider-aws.