This message was deleted.

Upgrade of Kubernetes should be done 1 minor version at a time. If you are using longhorn maybe it has a lot of volumes. Another disk consumer might be images of containers or some application that holds too many data. Have you checked what is taking the space?

for what it's worth i'd be super cautious running on hardware that doesn't meet the etcd minimums

resolution note: seems like the additional upgrade images, longhorn, rke2 upgrade images, prometheus simply ate up enough of the disk space to cause the disk pressure issue (and the eviction of rke2-nginx-ingress) - fun. fix was simply to increase the size of the host nodes. ETCd was only a few hundred MiBs on each node and not a driving factor here. I think the situation is vague because its somewhat dependent on the cluster configuration: I am rke2 (no rancher) - HA, nginx, with longhorn and prom-kube-stack. This is mostly conforms to the ranchergovernment guidance. I rather should have noticed this a bit earlier but I was still working out some configuration bits on monitoring. And my question was really to provide backup for my documentation going forward for playbook generation.

I'm in a very similar setup as yourself