...
- kubernetes_reset.sh:
Purpose: There were certain issues occurring when the setup.sh file was executed again on the same setup. it threw error saying ‘the file/port in already in use’ when setup.sh file was executed again.
Solution: A list was prepared for errors occurring on re-execution of setup.sh file. To fix them, some commands were executed manually to reset the changes. Once the list was finalized, these were consolidated in a .sh file. The file is in eliot/scripts/ folder.
What it does:
-- Resets kubeadm changes on both Manager and Edge Node machine.
-- Installs iptables, ipvsadm required to restart kubectl service on both Manager and Edge Node machine.
-- Restarts kubelet service on both Manager and Edge Node machine.
-- Releases the port 10250 on both Manager and Edge Node machine.
-- Deletes kubernetes file on Edge Node machine. - kubernetes_cleanup.sh
Purpose: There was no file to rollback the changes done by setup.sh.
Solution: A list of changes done by setup.sh file was prepared and then commands were written to rollback those installations and changes. The file is in eliot/scripts/ folder.
What it does:
-- Uninstalls Docker, Kubernetes, Kubectl, Kubeadm, Kubelet on both Manager and Edge Node machine.
-- Resets kubeadm changes on both Manager and Edge Node machine.
-- Installs iptables, ipvsadm required to restart kubectl service on both Manager and Edge Node machine.
-- Restarts kubelet service on both Manager and Edge Node machine.
-- Releases the port 10250 on both Manager and Edge Node machine.
Troubleshooting
While checking with kubernetes pods instability issues like Error / CrashloopBack off along with network plugin Calico / Flannel
i.e when running
| Note |
|---|
1) Core-dns pod instability issue - CrashloopBack off / Error / Restarts while executing $ kubectl get pods - |
...
n kube-system |
...
Cause: /etc/resolv.conf |
...
is not configured properly to establish the network with coredns pod. Solution: check /etc/resolv.conf → whether it is having the content as below nameserver 8.8.8.8 |
...
2) Calico pods ( CNI ) instability issue - restarting the calico pods with unstable network connectivity while executing $ kubectl get pods -n kube-system or getting BGP connection not estabilished in calico-node-<xyz> . Bird is not ready while executing $ kubectl describe pods -n kube-system calico-node-<xyz> Cause: Error occurred due to network interface of the vm / machine which calico cni is using can't be able to connect to the internet. Solution: We have to execute below command in ELIOT Master node $ kubectl set env daemonset/calico-node -n kube-system IP_AUTODETECTION_METHOD=can-reach=www.google.com ( or ) $ kubectl set env daemonset/calico-node -n kube-system IP_AUTODETECTION_METHOD=interface=<interface-name> <interface-name> must be the interface which the machine / VM can be able to connect with the internet |
Error Message Guide
N/A
Maintenance
...