sooooo... I was trying to model my gke cluster like I would model an on-prem cluster by tainting nodes as noexecute and dedicating them to longhorn but that seems to be causing more problems that what it's worth... as evidenced by my constraints for deploying minio and performing an RKE2 upgrade (the system upgrade plan kinda wants to execute on nodes... who would have thought! LOL). Ah... live and learn.