Update prometheus and ingress-nginx dependencies, multiple major versions behind

[consideRatio]

With #2180 we now have automation to do regular dependency bumps of our charts. But, at this point, we are several major versions behind for prometheus and ingress-nginx.

I've outlined steps for how to upgrade these:

• ingress-nginx (research notes here):

  • [x] upgrade the chart to 3.36.0 (appVersion: 0.49.0) (merge #2214)
  • [x] upgrade our Ingress resources from v1beta1 to v1 (merge #2224)
  • [x] upgrade the chart to the latest version (merge #2203) prometheus: https://github.com/jupyterhub/mybinder.org-deploy/pull/2201#issuecomment-1140235678

• prometheus:

  • [ ] Delete deployment resources that can't be helm upgraded, without deleting its owned replicasets and pods, for all deployments.

    # this command will delete the deployment resource, but not the replica-sets/pods it owns
    # instead, these pods will be orphaned, and can later be managed by the new deployment
    # to be created
    kubectl delete deploy -n <namespace here> -l app.kubernetes.io/instance=prometheus,app.kubernetes.io/name=kube-state-metrics --cascade=orphan

  • [ ] GKE Staging
  • [ ] GKE Prod
  • [ ] Turing
  • [ ] OVH
  • [ ] upgrade the chart to the latest version (merge #2201)

[minrk]

upgrade our Ingress resources from v1beta1 to v1

I believe all of our ingresses are already v1. Is that not the case?

$ kubectl get ing -o yaml | grep apiVersion
apiVersion: v1
- apiVersion: networking.k8s.io/v1
- apiVersion: networking.k8s.io/v1
    - apiVersion: acme.cert-manager.io/v1alpha2
- apiVersion: networking.k8s.io/v1
- apiVersion: networking.k8s.io/v1
- apiVersion: networking.k8s.io/v1
- apiVersion: networking.k8s.io/v1
- apiVersion: networking.k8s.io/v1

[consideRatio]

Those directly part of the mybinder helm chart's templates are as of #2224 v1, but I bet there can be old resources generated by the use of old helm charts like prometheus.

Ah, you inspected all ingress resources in some namespace, so, if that namespace included prometheus of old version, we should be good. Otherwise, I was thinking we should also bump prometheus before dropping support for v1beta1.

[minrk]

Yup, that's all our ingress resources from this whole chart. And #2203 succeeded on staging, so going ahead with that one