Can't setup HA Cluster with 1.24.6 about kubernetes

[justmeandopensource]

@michaelschem Thanks for raising this issue. I will test the process with Ubuntu 20.04 and K8s version 1.24.6 and let you know if I see the same problem. Cheers.

[ak2766]

Which OS is this on?

[michaelschem]

ubuntu 20.04

[ak2766]

I'll leave it to Venkat to sort out as Vagrant and I are not acquaintances. I was looking to rule out the OS as the culprit.

I've built several clusters on Ubuntu 20.04, the latest being 1.23.9, and haven't had issues though I build my clusters using home made bash scripts. I'll make them publicly available once I find time to remove all the hard coded bits.

[michaelschem]

I'd be happy to pull out the hard coded bits if it meant getting a script faster!

[michaelschem]

If I remove the --upload-certs option it comes up but coredns never becomes stable

[michaelschem]

I'm running this on proxmox. Not sure if that makes any difference.

[justmeandopensource]

Ok. I have tested this on Ubuntu 20.04 with K8s 1.24.6 and is working fine. No issues.
Just followed the documentation (except changing 1.22.0-00 to 1.24.6-00 in three places in one line).
Host machine is a Linux machine. All kubernetes and haproxy nodes brought up using vagrant with libvirt provider.
Same as in this video.
https://www.youtube.com/watch?v=SueeqeioyKY

[michaelschem]

Ok, well I just realized I'm actually running 22.04. Is that not supported?

[justmeandopensource]

Pretty sure it will work on 22.04. I will test and report.

[justmeandopensource]

Hmm. I get all sorts of error with Ubuntu 22.04. On first attempt at kubeadm init command I get this.

[kubelet-finalize] Updating "/etc/kubernetes/kubelet.conf" to point to a rotatable kubelet client certificate and key
error execution phase addon/coredns: couldn't retrieve DNS addon deployments: Get "https://172.16.16.100:6443/apis/apps/v1/namespaces/kube-system/deployments?labelSelector
=k8s-app%3Dkube-dns": context deadline exceeded

I will have to work on this. But as mentioned in the documentation (and in the YouTube video), it works and meant for Ubuntu 20.04. Never been tested on 22.04 until today. I will get this working on 22.04 and add documentation and video for it (if I get it working). Cheers.

[michaelschem]

@justmeandopensource I've got it working on 22.04 now. I'll create a PR shortly.

[michaelschem]

@justmeandopensource PR created #119 I don't have vagrant setup so I haven't tested it yet, but that's more or less what I did to get it to work.

[justmeandopensource]

The documentation you originally referred to in this issue is this.
https://github.com/justmeandopensource/kubernetes/tree/master/kubeadm-ha-keepalived-haproxy/external-keepalived-haproxy

Your PR is for this one
https://github.com/justmeandopensource/kubernetes/blob/master/kubeadm-ha-multi-master/README.md

The documentation you have created PR definitely needs to be updated to use containerd instead of docker. But in the first link, I used containerd and not docker. But I didn't do systemdcgroup thing. As this documentation originally covers Ubuntu 20.04, I need to test this cgroup thing with 20.04 before pushing changes.

In my vagrant setup though, I have already got that in place.
https://github.com/justmeandopensource/kubernetes/blob/master/vagrant-provisioning/bootstrap.sh

Anyways I am glad that you sorted your issue. Well done.
Leave these documentation update with me. I will do my best. All these documentations are referred in my YouTube videos. SO I can't just change Ubuntu 20.04 to 22.04. I need to update videos etc. :)

Thanks for your help.

[michaelschem]

Ah, my bad on the PR.

Yea, I understand the need for consistency with your youtube videos. Glad I could help with the upgrade.

The systemdcgroup is definitely needed otherwise kubeadm crash loops.

Thanks for the vids and tutorials!