Allow the NFS service to be configured as a load balancer

[tibbe]

Is this a bug report or feature request?

• Feature Request

What should the feature do:
Allow for the NFS service, generated implicitly by the NFS operator from the NFS configs described in Network Filesystem (NFS), to be configured as a load balancer with an IP set through spec.loadBalancerIP. As a consequence of allowing an external IP, the generated NFS-Ganesha config might also need to change to allow connections from outside the cluster.

I believe the latter is needed as I tried adding a second (load balancer) service to expose the pod created by the Rook NFS operator, which resulted in hosts on the host network (e.g. 192.168.1.0/24) to be able to connect to the load balancer service while hosts (e.g. 172.19.27.38) outside that network couldn't. This is the config I used:

---
apiVersion: v1
kind: Service
metadata:
  labels:
    app: rook-nfs
  name: rook-nfs-external
  namespace: rook-nfs
spec:
  type: LoadBalancer
  loadBalancerIP: 192.168.1.11
  ports:
  - name: nfs
    port: 2049
    protocol: TCP
    targetPort: 2049
  - name: rpc
    port: 111
    protocol: TCP
    targetPort: 111
  selector:
    app: rook-nfs

(This is the same as rook/rook#7674 but for the regular, non-Ceph-specific, Rook NFS support.)

What is use case behind this feature:

Expose NFS outside the cluster through an IP.

Environment:

[tibbe]

Does this feature request make sense?

[travisn]

If Rook just needs to create the loadbalancer service when the loadBalancerIP is set, I could see this being straight forward to implement. The question then becomes what else needs to be configured in ganesha as you mentioned to get it working end to end.

@tibbe Are you interested in implementing the feature? We are looking for more contributors to the nfs operator.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in a week if no further activity occurs. Thank you for your contributions.

[lknite]

running into this issue, after a power outage, my nfs nodeport or maybe it was a loadbalancer ip didn't come back up ... this is because it wasn't all done via gitops, i had to rig a solution and now i've got to figure out again how i did that, if it were defined in the helm chart then it'd be taken care of, so if this still hasn't been implemented i upvote it