I need to use NFS with Kubernetes in GitHub codespaces for reasons. k3d¹ already handled my throwaway clusters nicely, so the obvious move was to add an NFS-backed² ReadWriteMany volume via nfs-subdir-external-provisioner. I expected it to “just work.”

It did not. Pods sat in ContainerCreating until they timed out with mount.nfs: Operation not permitted, the same failure called out in k3d issue #1109. The root cause is boring but important: the stock k3s image ships from scratch³, so it contains none of the NFS client tools (nfs-utils, rpcbind⁴, or even the OpenRC⁵ metadata needed to start services). GitHub Codespaces run on cgroup v2⁶, and k3d tries to be helpful in that environment by swapping in its own entrypoint script, which meant every hand-rolled fix I wrote was silently discarded.