XFS project quota overrun => NFS client and server unresponsive

[Jan Schulze]

Hi all,

we are using XFS with project quota enabled on our NFS server. Server and clients are running Scientific Linux 5.2, NFS version is 3.

When a user overruns the project quota while copying data from an NFS client, the server and client become very unresponsive. On the server, one NFS thread per core is stuck in uninterruptible (D) state with a load of 1. The only solution to decrease the load, is to reboot the client. Any other method I tried (restarting NFS service, restarting network on client and/or server) did not help. I tried to vary mount options for XFS, but that did not solve the problem. It does not seem to be a hardware issue, as the problem could be easily reproduced on other server/client hardware.

Are there any ideas, how to debug this? Anyone having the same issue? Any help is highly appreciated.


Best Regards,
Jan