nginxinc/nginx-prometheus-exporter

Support different Container Registry

torbendury opened this issue · 11 comments

Is your feature request related to a problem? Please describe.
DockerHub has rate limiting for non-authenticated IPs, which is a big blocker for our Google Kubernetes Engine clusters. We have several applications in our clusters which run nginx-prometheus-exporter as sidecar containers and therefore need to pull the image. I know that Google caches popular images and doesn't tunnel requests directly to DockerHub, but with nginx-prometheus-exporter this does not seem to be the case. As a result, our Pods are stuck with ImagePullBackOff when we have bigger scaling events (the image is not the only one that pulls from DockerHub, unfortunately..).

Describe the solution you'd like
Also push to other container registries as quay.io or even the GitHub Container Registry.

Describe alternatives you've considered
Subjecting ourselves under the new Docker license+pricing model is not a considerable alternative, imho.

Additional context
none

another argument is the new DockerHub licensing model
+1

I see there is already a PR #230
so maybe there's just docu missing

Ah, I see that you already push to AWS ECR and GHCR - maybe document that somewhere? :-)

Hi @torbendury @philipp-hb

yes we started testing ECR and GHCR with edge, but we haven’t had a release yet, so the stable packages are not available there.

I guess it makes sense to update the docs with this information 🙂

Would you like for us to push to more registries?

Right now GitHub Container Registry does not have any SLA, neither does DockerHub. AWS registry has 99,9% SLA, but using it may conflict with some.."policital" problems regarding Amazon. Maybe something more neutral, e.g. quay.io could be a possiblity?

@torbendury Are you looking for a registry that you can use in your pipeline that has high reliability and won't throttle?
I get the impression that is really what you are looking for.

@brianehlert Less in Pipelines, more in bigger clusters with big scaleups. Yes, availability (and reliability) and also throttling are points to consider, imho.

This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 10 days.

For now, ECR and GHCR are a sufficient solution for us.

https://quay.io/repository/nginx/nginx-prometheus-exporter was added.

We now support DockerHub, ECR, GHCR and Quay.io. If any other is needed, please open a new issue for that specific one.