As a community, we run a number of services. This ranges from developer-oriented stuff like CI and dashboards to user-facing stuff like k8s.io and the various shortcut names.
This document aims to explain how you can run something, too.
At the time of this writing we have one cluster, called "aaa" running in GKE. We will expand that as and when needed.
Whatever you run has to be legitimately community-property and has to be "reasonable". These are not free resources, but we're also not trying to stifle good work. We typically think about this in terms of web-services, but that is not necessarily all people will run. Over time we expect the diversity of workloads to increase.
Your escape hatch at every step is to reach out to the kubernetes community cluster-admins. You can email [email protected] or find many of us in the #wg-k8s-infra Slack channel.
You can include all of these steps in a single PR.
You will get one or more namespaces to run your thing. Decide what names you want and make sure they are available. Check the list in the namespaces config, and add your names there. Keep them short but descriptive.
Add a block to groups.yaml for each namespace you want. Each group
should be named k8s-infra-rbac-<namespace>.
Add the approriate set of users to these groups. Keep it as small as possible, but make sure there are enough to people to keep it running. If you have multiple namespaces that should be kept in sync, say so in the description.
Add these new groups to the group called "gke-security-groups", which will enable the RBAC linkage.
Once reviewed and merged, one of the group admins will create the google-group(s) you requested. Then a cluster admin will make your namespace(s) and install the RBAC rules for you.
At this point, any members of the "rbac" group(s) should be able to access the cluster.
To access our clusters, you will need to use Google's Cloud Shell. If you need to inspect more of the infrastructure, you can use Google's Cloud Console.
The first time you use this, you must get credentials for the cluster. From your CloudShell prompt, run:
gcloud container clusters get-credentials aaa --project kubernetes-public --region=us-central1You might want to abbreviate the context:
kubectl config rename-context gke_kubernetes-public_us-central1_aaa prod-aaaOnce that is done, you should be able to run kubectl against the cluster:
$ kubectl --context=prod-aaa get ns
NAME STATUS AGE
cert-manager Active 121d
default Active 155d
You can now run Deployments and expose Services and so on.
You are allowed to create Ingresses and Services with public IPs. Before you are done configuring your thing, you probably want your IPs to be earmarked as "static". This ensures that you can keep the same IP if you ever need to recreate your Service or Ingress. To do that, ask one of the cluster-admins to do it manually. Then you can configure your Service or Ingress to reference it explicitly.
If you are exposing a public IP, you probably want DNS for it. You can send a PR against the DNS config to add or update your name.
If you are exposing a public IP, you must do so securely. For HTTP, this
means TLS. We run cert-manager in the cluster(s), so you can
simply create a Certificate object. See the examples in this repo.
Your RBAC access does not include the ability to create arbitrary Secrets.
If you need to do that, talk to cluster-admins.
Coming soon. This is not yet fully decided.
It is your responsibility to keep your workloads running and secure, and to use resources responsibly. The cluster-admins team will provide some basic oversight (which will get better over time), but you are the front-line for your app.