q - I ran into some, let's call it confusion, w.r.t. cordon/drain while using a strict-local longhorn volume when upgrading rke2 using its upgrade job. It seems in this scenario you can't really re-schedule any pods when a node like that needs to go down. In this case would I want cordon=true and drain=false (?) or simply set them both to false and let the scheduler be confused while the node reboots? I am sort of a fan of the latter. (for reference we have mmap'd databases that would not play well in a replicated environment, so single-volume set to localhost - so I get mgmt, logging, metrics - but in almost a localPath/hostPath setup). I wanted to use the auto-upgrader images from docker.hub since it seemed more structured than me doing it manually, however I am now rethinking that maybe I should have just done things manually.