Test clusters¶

The repo ships a script that spins up two kind clusters loaded with sample workloads — perfect for exploring or developing against without touching anything real.

What it sets up¶

hack/dev-clusters.sh creates:

kind-kubus-a — the full demo:
- metrics-server (patched for kind's self-signed kubelet certs), so the metrics and overview charts light up;
- a podinfo Helm release with 3 replicas — great for aggregated logs and Helm;
- intentionally broken workloads (crash-loops, bad images, a failing Deployment, a CronJob) so the overview dashboard has something to flag.
kind-kubus-b — an empty cluster, so you can try the multi-cluster merged views and the cluster column.

Prerequisites¶

kind
kubectl
helm
A working container runtime (Docker)

Run it¶

./hack/dev-clusters.sh

Then launch Kubus and select the kind-kubus-a and kind-kubus-b contexts in the cluster switcher.

kind + inotify limits

On Linux, low inotify limits make kind's system pods crash-loop with "too many open files". If you hit that, raise them:

sudo sysctl fs.inotify.max_user_instances=512 fs.inotify.max_user_watches=524288

Tear down¶

kind delete cluster --name kubus-a
kind delete cluster --name kubus-b