I was given some very raw instructions to set up a kubernetes cluster, including some setup scripts to run, and no leisure time to read all the docs. Also I had some guest VMs setup for me in ESX. Everyone I work with is new to kubernetes, so present ability to answer questions is limited, and we are feeling our way to establish best practices. After some initial success running a deployment, getting pods to run, software worked as expected, my configuration is now broken. I had started to alter my network configuration, and also tweak my .yaml deployment files. All the kubectl commands presently fail with a message like this:
Unable to connect to the server: dial tcp 10.x.y.z:6443: getsockopt: no route to host
The IP address now longer resolves. I believe now the issue is I had switched from DHCP to static on my guest VMs at a later time. I had not realized the VMs were originally booted with DHCP, as the intent had always been to use static addresses. So when I ran these kubernetes setup scripts, I promulgated this now-bogus DHCP address into the cluster, and now need to fix it. So for example:
root@k8-node-01:~# grep server $HOME/.kube/config server: https://10.x.y.z:6443 root@k8-node-01:~#
- Can I address the problem simply by writing the correct new address into the above file and others, and then rebooting my kubernetes nodes?
- What other config files might I need to change, besides my deployment yaml files?
- I see the configuration file includes some certificate info, and I am thinking it may be tied to my original IP address. If so, then merely changing to the IP might not work, so what would I do then?
- Originally I was thinking to just "reinstall" the cluster, but since kubectl doesn't work right now, best steps aren't clear. Can I just apt-get purge kubernetes?
- Seems like so much of the kubernetes instructions I have seen have hard-coded IPs instead of DNS addresses. Why is that? Are there any actual limitations?