When setting up a cluster with kubeadm in an on-premises environment to build a monitoring server (prometheus, zabbix), I ran into several errors. Here I’ve collected them.

1.

Unable to connect to the server: x509:
certificate signed by unknown authority
(possibly because of "crypto/rsa: verification error" while trying to verify candidate authority certificate "kubernetes")

This error can occur when running kubeadm init or kubeadm join. It usually happens when using an old or incorrectly configured config.

rm -rf ~/.kube
mkdir ~/.kube
cp -i /etc/kubernetes/admin.conf ~/.kube/config
chown $(id -u):$(id -g) ~/.kube/config

2.

When kube-flannel and kube-proxy are in CrashLoopBackOff

This error may occur when installing kube-flannel for communication within the cluster.

Use the kubectl describe po -n kube-flannel ${kube-flannel pod-name} command to check which node it is running on, then connect to that node server and run
containerd config default | /etc/containerd/config.toml
Then, in the generated /etc/containerd/config.toml, change
[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc.options]
SystemdCgroup = false
to
SystemdCgroup = true
Then restart containerd with systemctl restart containerd and check whether the status has changed.

3.

Taint error

I encountered this error when I created only a master node and tried to launch a Pod as a test. Pod creation is not allowed on the Control Plane, but I removed the Taint temporarily to test it. Taint is configured on nodes where you don’t want Pods to run. Therefore, when a Taint is set, Pods are not scheduled on the node with that Taint.

• Check Taint kubectl describe node ${node-name} | grep -i "taint"
• Remove Taint kubectl taint nodes -all node-role.kubernetes.io.master-

Thank you.