Unable to mount nfs without "nolock" option
dvassilyev-vi opened this issue · 2 comments
What happened:
I installed CSI driver for NFS volumes via generic Helm Chart. After installation it shows running pods as expected:
I created storage class and put it into K8s:
But PVC deployment stalls in "Pending" phase. This is what I get if try to describe it:
What you expected to happen:
I'd like to be able to mount these shared folders without "nolock" options. It can mount NFS points with "nolock" in Storage Class, but is it safe? So I logged in (kubectl exec -it ... /bin/sh) to controller pod and container "nfs", it has no separate "/run" mount point and "/" is mounted read-only (overlayfs). The same for node pod.
How to reproduce it:
Anything else we need to know?:
rpc.statd wants to create a lock file in /run, also rpcbind, and "/var/lib/nfs" might be needed to set rw?
Environment:
- CSI Driver version: 4.6.0
- Kubernetes version: 1.27.4 (EKS Anywhere 0.17.4)
- OS: Bottlerocket OS 1.14.3
- Kernel: 5.15.117
- Install tools: helm
- Others:
If anyone is interested, this happens when your filer has NFS4 disabled. NFS4 doesn't need all these RPC daemons to run. Enabled NFS4 - works as expected.
thanks for the info.