Scrape metrics from kubernetes nodes and container runtime

[arturo-seijas]

Enhancement Proposal

Given the COS is a kubernetes bundle, it think I'd be a really nice addition to discover and scrape all the kubernetes components without the need of manual configuration and also nodes and pods, although the latter would probably require another charm to be deployed along the current charms in the bundle.

Also, adding grafana dashboards for these metrics could come in handy too.

Thank you!

[simskij]

I agree with you. We've previously discussed this and it is really a two-part proposal.

1. We need to be able to collect telemetry for CK8s. This is likely to be solved using the planned grafana-agent machine charm, which among other things include node_exporter.
2. We need to be able to collect telemetry for MicroK8s. This is likely to be solved using a proxy charm and a telemetry plugin that will deploy node_exporter, kubestate-metrics`, and the kubernetes metrics server.

What we don't want to do is to open the flood gate that is kubernetes_sd_config as that would invite less ideal usage models that don't leverage everything that the juju topology has to offer.

[simskij]

This will be addressed by the K8s team. Closing this issue as it is a no-op on our side.