This is a guide on getting set up with the OCF's Kubernetes cluster. It will be updated as we develop a staging environment and more accessible workflows for staff.

Accessing the Cluster:

1. Connect to an OCF machine (desktop or login server)
2. Run kubectl auth whoami
3. :D

By default this targets the staging cluster, but you can pass --context=dna to target production if you have access.

(Currently only ocfroot can interact with the cluster, but we're working on a staging environment)

If your command hangs, you likely need a Kerberos ticket (i.e. run kinit). Alternatively, you could copy the contents of the $KUBECONFIG file to your ~/.kube/config and add the --no-browser flag. If you want to connect from your own device, just copy our configuration and it should work as is. For more information on configuring auth, please see the kubelogin documentation.

kubectl

kubectl is the primary tool for a user to call the Kubernetes API. It is fairly extensive, but here are some basic commands that may be useful:

• kubectl get nodes - List all nodes in the cluster, useful to see if any nodes are NotReady or SchedulingDisabled (cordoned).
• kubectl get ns - List namespaces. Almost all resources should be in a namespace, which needs to be specified with -n [namespace] when appropriate.
• kubectl get [resource] - List all of the given resource in a namespace. Some common resource types are deploy, pod, svc, ing, cm.
• kubectl api-resources - List all resource types. This will be very overwhelming due to the number of CRDs, and probably not helpful.
• kubectl get [resource]/[name] - Get information on the named resource. Specify output type with -o, such as -o yaml for the object's manifest.
• kubectl describe [resource]/[name] - Get details on the current state of the named resource.
• kubectl logs [pod|deployment]/[name] - Get logs for the container running in a pod. You can specify a container with -c and follow the logs with -f.

There are a lot of other commands and flags, which you can learn about in the docs. Some useful flags are -A to list all namespaces, and --watch to watch for changes in output. Flux should manage almost all resources during normal cluster operation, so directly mutating state with kubectl should be reserved for when it is absolutely needed for debugging purposes.

k9s

If you'd prefer a tui wrapper with vim-like keybinds, k9s is installed on our machines:

• :q to quit.
• :ns to view available namespaces.
• Scroll with j and k down to the ocfweb namespace (you can also scroll with your mouse). Press enter.
• You are now viewing the pods for ocfweb; press enter on one to see its containers. Press esc (maybe twice) to go back to the previous page.
• Press l to see the logs for that pod. Use j and k to scroll through them.
• Press d on the pod to get additional information about it.
• See more keybinds: k9scli.io/topics/commands.