Please add __meta_kubernetes_pod_status so 'completed' pods can be dropped from scrape

[bboreham]

Proposal

Use case. Why is this important?

Pods that are created by Kubernetes Job objects will go to Completed state if they finish without error.
Without a way to stop scraping, the Prometheus up state goes to 0 and (in my current environment) this triggers an alert.

The Kubernetes SD inserts various labels such as __meta_kubernetes_pod_annotation... which allow relabel config to drop some targets. I would like to extend this to a meta label which would hold the Completed status, so I can drop those pods from the targets.

[brian-brazil]

We'll generally accept any metadata that's available, so feel free to send a PR.

[bboreham]

Looking more closely, I find "Completed" comes from the (multiple) container.State.Terminated.Reason field, and requires gymnastics to obtain.
There is a simpler Phase field which will say Running, Succeeded, etc., and I propose to implement that.