These days I'm playing with Kubernetes and managed to install a single-node cluster (on a single computer). I see it offers many tools to add / modify / remove configuration parts (services, pods, deployments, ...) but I was wondering what could one do if a node doesn't start anymore - i.e. the machine is fine but the configuration is broken.
Are there tools that can help in that situation? I'm talking about services, deployments, etc.
Kubeadm seems to only provide node configuration, while kubectl requires a running node to retrieve informations.
kubectl talks to the cluster through the API and for this to work we need to have the kube-apiserver running.
$ kubectl get pods -A -l component=kube-apiserver
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system kube-apiserver-yaki-test-1 1/1 Running 0 3d18h
There are a few ways to access you cluster but all of them requires you to have you API server running.
The best approach would be to fix the what is causing the cluster to fail. Here you can read about ways to troubleshoot your cluster.