Is it better to use k3s with multipass or docker (via k3d) on a Mac M1? Using multipass VMs seems a bit closer to how k8s is deployed in production as at least you can interact with the individual nodes. Just curious if anyone has noticed any advantages/disadvantages with either approach. Thanks!

I guess that's personal preference on ease-of-use and resource usage. With k3d (docker) you can also interact with the individual nodes (e.g.

docker exec -it k3d-k3s-default-server-0 sh

to get a shell in the first server node). However, some things may work better in a VM where OS dependencies are present (e.g. deploying Longhorn that relies on iscsi-drivers of the host).