Bartlomiej Plotka

### AC * We have tests ensuring we update caBundle on self-deploy * We have tests ensuring we don't try updating caBundle if those are updated by GKE on managed...

Following recommendations those should be moved out of GH actions. Also we could use ADC instead flow https://cloud.google.com/docs/authentication/application-default-credentials#GAC (common envvar and file)

Improved scrape configuration debuggability.

1

One common misconfiguration on GMP is a mismatched port/port name on PodMonitoring. Unfortunately this issue surfaces nowhere other than your metric is not scraped and not on GCM. It causes...

See https://github.com/googleapis/google-cloud-go/blob/main/migration.md

Related to https://github.com/GoogleCloudPlatform/prometheus-engine/issues/536

This is due to https://cloud.google.com/kubernetes-engine/docs/concepts/autopilot-resource-requests#defaults which causes our example pods to not work in the typical GKE autopilot cluster as unrealistic requests is assigned (2GiB mem). 

Occasionally customers see failed export on histograms: > export.go:827 level=error component=gcm_exporter msg="send batch" size=106 err="rpc error: code = InvalidArgument desc = One or more TimeSeries could not be written: Field...

`make lint` fails for multiple reasons (old golang-lint version, errors we want to ignore, but should do this explicitly or ignore in lint config). Let's either remove make lint or...

Currently histogram matching has to match all three metrics (_bucket, _sum, _count). Summary is similar (_sum _count). This means customers have to do something like `_.+` regex or all 3...

Ref: https://github.com/prometheus/prometheus/blob/main/.github/stale.yml This is to ensure PRs are not open forever. It's common people throw PRs and never come back. If they come back we can always reopen (: