Support for health and liveness checks when running on Kubernetes

[MrPickles]

Does the Stackdriver Sidecar expose any endpoints for liveness and readiness probes? Prometheus itself has such endpoints. If not, are there plans to support those?

We've been running into recent issues where the sidecar pods randomly seem to fail to fetch GCP credentials. (It seems to be a transient error, but I'm not completely certain.) However, the pod is still deemed healthy and will continue to fail at both fetching credentials and exporting metrics until the pod is killed off.

level=warn ts=2021-01-12T14:56:18.277Z caller=queue_manager.go:534 component=queue_manager msg="Unrecoverable error sending samples to remote storage" err="google: could not find default credentials. See https://developers.google.com/accounts/docs/application-default-credentials for more information."

Having some means of detecting transient errors in the sidecar would be really helpful with catching bad pods. Thanks!