Every dashboard in a tenant is a ClusterIP service and none is exposed outside the cluster. You reach them with kubectl port-forward, which opens a tunnel through the API server for as long as the command runs.
#Before you begin
kubectl pointed at the tenant from the controller:
export KUBECONFIG=~/.bedrock/kubeconfig#Forward a port
Run the command for the dashboard you want and open the address in a browser. Stop the forward with Ctrl+C.
| Dashboard | Command | Address |
|---|---|---|
| Grafana | kubectl -n application port-forward svc/grafana 3000:3000 | http://localhost:3000 |
| Prometheus | kubectl -n application port-forward svc/prometheus-operated 9090:9090 | http://localhost:9090 |
| Alertmanager | kubectl -n application port-forward svc/alertmanager-operated 9093:9093 | http://localhost:9093 |
| OpenCost | kubectl -n application port-forward svc/opencost-ui 9091:9090 | http://localhost:9091 |
| Hubble | kubectl -n kube-system port-forward svc/hubble-ui 12000:80 | http://localhost:12000 |
| Longhorn | kubectl -n longhorn-system port-forward svc/longhorn-frontend 8080:80 | http://localhost:8080 |
Grafana’s user is admin. The bootstrap makes a random password once, stores it in OpenBao at apps/grafana/admin and keeps a copy on the controller in ~/.bedrock/grafana-admin:
cat ~/.bedrock/grafana-adminWarning
The Longhorn and Hubble interfaces have no login of their own, and the Longhorn interface can delete volumes. Only open them when you need them, and close the forward afterwards.#Run several at once
Each forward holds the terminal. To run them in the background and stop them together:
kubectl -n application port-forward svc/grafana 3000:3000 &
kubectl -n application port-forward svc/prometheus-operated 9090:9090 &
waitCtrl+C stops both.