Running Kubernetes Locally

Working on the Kubernetes source on your own local machine can be rewarding:

spinning up a “cluster” is very fast
debugging is simple
low-level tools, like netfilter logging and bpf, are close at hand

Of course, this is no substitute for real end-to-end tests, but when you want to work quickly, I’ve found it invaluable.

hack/run-locally

There is an existing Kubernetes script that starts a bare-bones cluster. However, it doesn’t work out-of-the-box on Fedora. Here’s what you have to do to enable it:

Get CRI-O and crun

sudo dnf module enable cri-o:1.19
sudo dnf install cri-o crun
sudo systemctl enable cri-o

Configure cri-o to use crun

This is needed (for now) to support cgroups v2. Edit the file /etc/crio/crio.conf

Add the following section:

[crio.runtime.runtimes.crun]
runtime_path = "/usr/bin/crun"
runtime_root = "/run/crun"

Edit the default_runtime directive:

default_runtime = "crun"

Restart cri-o:

systemctl restart cri-o

Start kubernetes locally

In your kubernetes checkout, run

$ hack/install-etcd.sh
$ CGROUP_DRIVER=systemd CONTAINER_RUNTIME=remote \
    CONTAINER_RUNTIME_ENDPOINT='unix:///var/run/crio/crio.sock' \
    ./hack/local-up-cluster.sh

It will spit out a KUBECONFIG for you to use - have fun!

Debugging

You can find more information in the logs:

journalctl -u crio
less /tmp/kubelet.log