Add metric label to identify namespace resources impacted

[sebglon]

Hi,
We have a unique CAPI that manage multiple clusters isolated by namespace.
We need a way a way to split alert by namespaces at least for the CAPI controller.

Can we add the resource_namespace or target_namespace label to each metrics?
I am aware to create a PR but i prefer to validate the label name to use before.

[sbueringer]

Which metrics are you referring to?

What do you mean with CAPI, ClusterAPI?

[sebglon]

Hi,
We need this to monitor reconciliation error, and reconcileTotal, ReconcilePanic on all resources managed by cluster-api-controller (CAPI) and all of our providers.
We use the AWS, Outscale and in e few months GCP; and all use the controller-runtime lib.

[sbueringer]

I think adding the request namespace to this metric would lead to a pretty high cardinality metric if you have a lot of namespaces.

I think we should not do this in controller-runtime, as it will affect everyone using controller-runtime.

That being said, for the same reason I wouldn't want to make the same change by adding an additional metric in Cluster API.

We have a unique CAPI that manage multiple clusters isolated by namespace.

If that means you have separate CAPI controllers reconciling specific namespaces you should be able to add a namespace label when you scrape the metrics from your controllers.