Propagate BeeGFS quota through NFS

[Joachim Meyer]

Hi,

we're currently setting up our BeeGFS to be quota'd.

On the "normal" nodes, this works okay, so far.

However, we're exporting BeeGFS through an NFS server to our submission nodes (as they're currently located in a different network, don't ask).

As far as I can tell, users on the submission nodes can just generate as much data as they'd like, even though their quota is exceeded.

Is there a way to convince NFSv4 to respect BeeGFS's quota?

We're exporting it like so:

/mnt/beegfs xxx.xxx.xxx.xxx/255.255.255.254(rw,async,fsid=2,crossmnt,no_subtree_check,no_root_squash)

Mounting:

134.96.59.10:/mnt/beegfs               /scratch      nfs4     tcp,hard,intr,rw,port=2049,relatime,nodiratime     0 0

I'll list a few potential issues with the setup, I could imagine, but I'd like confirmation if any of these might be the issue, how I could work around them:

- We're using IPA to define our groups and users. These can be resolved on the submission and compute nodes. However, the NFS server is not part of the IPA domain and thus does not resolve the user/group IDs. Newly created files appear the way they should on all systems (with the correct uid/gid).

- We currently don't have rquotad set up. However, my understanding is, that this should only be necessary to be able to query quota's as reported by quota for "normal" file systems, not for quota enforcement.

Does anyone have experience with this kind of setup and can give some pointers to where the issue might actually be/how we can find the issue/where I might have a misconception?

Thanks,

- Joachim Meyer