Awesome! Have you tried kops and/or clusterapi? Those are imho great ways to run K8s on OS. Unfortunately to use clusterapi you'll need to build your own images if you need anything more modern than 1.22 I believe, but there is a packer template that works pretty well. Since it only uses kubeadm and cloudinit under the hood, you should have zero problems adding your Pi
I haven't tried either yet. But ClusterAPI is next on my agenda. I think I'll go back to a 3 node OKD cluster for things like Argo and monitoring. ClusterAPI can live there and deploy my k8s cluster for the Pi's. I believe OpenShift uses the ClusterAPI project within machine-api as well. So it would be good for me to work with directly and learn more about. Eg: github.com/openshift/machine-api-provider-openstack
@@triplewho1448 Yeah it does and the cool thing is that you can transfer the management to the newly created cluster, so you could basically even use kind locally for that
@@LampJustin So cluster-api is indeed really cool. One of the things I would like to improve about it is to allow it to leverage the OVN provider for Octavia. I'm not a huge fan of Amphora, particularly in a homelab environment. If I can save running a VM and just use OVN logic flows, that would be ideal. I did end up deploying k8s with cluster-api though, and I definitely can see the value. This is pretty cool: ua-cam.com/video/KzYV-fJ_wH0/v-deo.html
@@BrendanShephard I don't really like Amphora either... Especially annoying for me is that I sometimes have to failover the damn amphorae twice if I reboot/shutdown my server. The VIP just isn't accessible otherwise and I often shut them down as I just don't need it to run all the time. For 24/7 I have a small Proxmox host. But don't forget that all the healthchecking doesn't work with the OVN provider. It's just basic TCP loadbalancing, but I guess that's fine for Kuberntes if the pod itself has good healthchecking.
Awesome! Have you tried kops and/or clusterapi? Those are imho great ways to run K8s on OS. Unfortunately to use clusterapi you'll need to build your own images if you need anything more modern than 1.22 I believe, but there is a packer template that works pretty well. Since it only uses kubeadm and cloudinit under the hood, you should have zero problems adding your Pi
I haven't tried either yet. But ClusterAPI is next on my agenda. I think I'll go back to a 3 node OKD cluster for things like Argo and monitoring. ClusterAPI can live there and deploy my k8s cluster for the Pi's.
I believe OpenShift uses the ClusterAPI project within machine-api as well. So it would be good for me to work with directly and learn more about.
Eg:
github.com/openshift/machine-api-provider-openstack
@@triplewho1448 Yeah it does and the cool thing is that you can transfer the management to the newly created cluster, so you could basically even use kind locally for that
@@LampJustin So cluster-api is indeed really cool. One of the things I would like to improve about it is to allow it to leverage the OVN provider for Octavia. I'm not a huge fan of Amphora, particularly in a homelab environment. If I can save running a VM and just use OVN logic flows, that would be ideal.
I did end up deploying k8s with cluster-api though, and I definitely can see the value. This is pretty cool: ua-cam.com/video/KzYV-fJ_wH0/v-deo.html
@@BrendanShephard I don't really like Amphora either... Especially annoying for me is that I sometimes have to failover the damn amphorae twice if I reboot/shutdown my server. The VIP just isn't accessible otherwise and I often shut them down as I just don't need it to run all the time. For 24/7 I have a small Proxmox host.
But don't forget that all the healthchecking doesn't work with the OVN provider. It's just basic TCP loadbalancing, but I guess that's fine for Kuberntes if the pod itself has good healthchecking.