Repository navigation
Headlamp plugin
From v2.4.0 the Headlamp dashboard has a k4all section. It shows the state of the cluster, runs upgrades and changes the common settings, without editing YAML.
The full user guide is docs/headlamp-plugin.md. Developer notes: DESIGN.md and DISTRIBUTION.md.
Open Headlamp on NodePort 32323 (https://<node-ip>:32323/) and log in with the admin token (as root):
kubectl get secret admin-user -n headlamp -o jsonpath='{.data.token}' | base64 -dThe k4all entry is in the sidebar, with four pages. Headlamp's Node page also gets a k4all section: role, release, upgrade phase, network interfaces.
| Page | What it's for |
|---|---|
| Overview | The k4all release, whether an upgrade is available or running, health, components, features, ingress, and each node with its release and upgrade phase. |
| Upgrades | Move every node to a newer k4all release and follow it. See Upgrading k4all (v2.x). |
| Settings | Features (Virtualization, Argo CD, OVS secondary networks), ingress (dedicated IP, default IngressClass, Ingresses still on the nginx class), Cilium options. |
| Components | Every component of the release: state, versions, Helm value overrides that survive upgrades, management (Managed, Unmanaged, Removed), re-apply. |
The Ingresses still on the nginx class have a Switch to traefik button: see Moving from ingress-nginx to Traefik (v2.4.0).
Every change opens a review first. The API server checks it (dry run), and the review lists what changes and what that does to the cluster, with the YAML diff. Nothing is saved until you apply it. If someone else changed the same thing in the meantime, the review tells you.
The Overview and Upgrades pages keep the last known state while the cluster doesn't answer (for example while the API server or Headlamp restarts during an upgrade), and retry by themselves.
The admin token can do everything. For other users, bind one of the ClusterRoles in headlamp-plugin/distribution/rbac.yaml:
| Role | Can |
|---|---|
k4all-viewer |
see every page |
k4all-editor |
also change settings, overrides, management, re-apply, and switch Ingresses to traefik |
k4all-upgrader |
also start, pause, resume, stop and retry upgrades |
- Switching the CNI (Calico, Cilium) and turning the OVS bridge on or off. These are read-only for now, marked "Coming with k4all-operator v0.6.0": k4all v2.4.0 pins k4all-operator v0.5.0, which can't do them yet. The same rows become editable once the cluster runs that future operator.
- Install-time settings (pod and service networks, API endpoint, HA, firewalld, proxy). They're shown read-only, with the reason.
-
Running a hop on a node directly. The plugin sets the ClusterUpgrade target; the k4all-operator and
k4all-upgradeon each node do the rest. When a node gets stuck, the fix is on the node: the page shows the commands to run. - Editing component versions or values in the ReleaseManifest. The next release sync would overwrite them. Use the Helm value overrides in Components: they survive upgrades.
- Clusters before v2.4.0 have no plugin. Upgrade them from the command line first.
The plugin is part of the k4all release, like every component. A release that brings a new version updates Headlamp during the release sync. Reload the page afterwards.