add epp servicemonitor

[sallyom]

Adds EPP servicemonitor

[kalantar]

Is the intent of this to add monitoring to the epp node itself?

[kalantar]

I notice the creation of a ClusterRole and a ClusterRoleBinding. I wonder if this is specific to monitoring the epp pod or if this is a general issue. I notice that in the upstream GAIE inferencepool template, they split the permissions between a Role and a ClusterRole.

[kalantar]

With a push to move function to the upstream GAIE, I wonder if this should be there instead?

[sallyom]

@kalantar servicemonitor with rbac/bearertoken is required to scrape metrics from EPP - the equivalent changes are proposed in the GAIE charts (kubernetes-sigs/gateway-api-inference-extension#1425) - yes if charts are moved upstream, this won't be required. This change will enable metrics collection from EPP pod until the charts are consolidated into 1 upstream chart. Clusterrole & Clusterrolebinding are required to scrape metrics from EPP. Same as in GAIE charts.

[jgchn]

Tracking the upstream GAIE PR. It sounds like once that is merged, this PR against ms won't be necessary.

[sallyom]

@kalantar @jgchn upstream PR merged - the wide-ep guide deploys epp with modelservice charts, so we should add the servicemonitor to modelservice charts as well since these will remain through the next release.

PR to add servicemonitor to guides

/cc @Gregory-Pereira

[sallyom]

closing this, will look at new format for kustomize