Hello!

Occasionally on my Nobara client I will be ripping a bluray, using mkvtoolnix or handbrake and my nfs connection drops causing any application connected to freeze up. This happens semi consistently. I've tried increasing fuse file descriptors and setting the number of threads to maximum. I've switched to NFSv3 only on Unraid and I've tried multiple mount options. Currently my mount options for nobara are (nfs defaults 0 0). I've experienced the issue through several versions of Unraid, including the latest 7.2.4, 7.2.3 and 7.1.12. I don't know if this is an issue with the OS (nobara/fedora 43) Kernel (6.19.5-200) NFS client (nfs-utils-2.8.4-0.fc43.x86_64) or Unraid.

I was finally able to capture this with rpcdebug. I unmounted the directory with umount -fl and then attempted to remount with mount -a -vvvv during the test in these logs. Unfortunately I lost the verbose output from the mount command.

Any help or suggestions greatly appreciated.

NFS Clients:

10.0.5.7 (Rocky 7.9)

10.0.6.2 (Nobara 43)

Unraid:

10.0.5.1 (Unraid 7.2.4)

Client rpcinfo:

rpcinfo
  program version netid     address                service    owner
   100000    4    tcp6      ::.0.111               portmapper superuser
   100000    3    tcp6      ::.0.111               portmapper superuser
   100000    4    udp6      ::.0.111               portmapper superuser
   100000    3    udp6      ::.0.111               portmapper superuser
   100000    4    tcp       0.0.0.0.0.111          portmapper superuser
   100000    3    tcp       0.0.0.0.0.111          portmapper superuser
   100000    2    tcp       0.0.0.0.0.111          portmapper superuser
   100000    4    udp       0.0.0.0.0.111          portmapper superuser
   100000    3    udp       0.0.0.0.0.111          portmapper superuser
   100000    2    udp       0.0.0.0.0.111          portmapper superuser
   100000    4    local     @/run/rpcbind.sock     portmapper superuser
   100000    3    local     @/run/rpcbind.sock     portmapper superuser
   100024    1    udp       0.0.0.0.182.182        status     29
   100024    1    tcp       0.0.0.0.210.11         status     29
   100024    1    udp6      ::.185.55              status     29
   100024    1    tcp6      ::.138.73              status     29
   100021    1    udp       0.0.0.0.189.142        nlockmgr   superuser
   100021    3    udp       0.0.0.0.189.142        nlockmgr   superuser
   100021    4    udp       0.0.0.0.189.142        nlockmgr   superuser
   100021    1    tcp       0.0.0.0.139.239        nlockmgr   superuser
   100021    3    tcp       0.0.0.0.139.239        nlockmgr   superuser
   100021    4    tcp       0.0.0.0.139.239        nlockmgr   superuser
   100021    1    udp6      ::.147.243             nlockmgr   superuser
   100021    3    udp6      ::.147.243             nlockmgr   superuser
   100021    4    udp6      ::.147.243             nlockmgr   superuser
   100021    1    tcp6      ::.129.159             nlockmgr   superuser
   100021    3    tcp6      ::.129.159             nlockmgr   superuser
   100021    4    tcp6      ::.129.159             nlockmgr   superuser

Unraid nfsstat:

nfsstat
Server rpc stats:
calls      badcalls   badfmt     badauth    badclnt
386503     32         29         3          0        

Server nfs v3:
null             getattr          setattr          lookup           access            
10        0%     837       0%     7         0%     430       0%     373       0%      
readlink         read             write            create           mkdir             
0         0%     269113   69%     108284   28%     12        0%     2         0%      
symlink          mknod            remove           rmdir            rename            
0         0%     0         0%     1         0%     0         0%     0         0%      
link             readdir          readdirplus      fsstat           fsinfo            
0         0%     0         0%     158       0%     2448      0%     26        0%      
pathconf         commit            
4         0%     4806      1%     


client_rpc_debug.log unraid_nfsd_debug.log turing-diagnostics-20260306-1102.zip

Edited March 6Mar 6 by timjtech

RPC: xs_error_report ... error=104

nfs: server 10.0.5.1 not responding, still trying

error=104 = TCP connection was reset mid-transfer.

The server message confirms it:

nfsd attempted to send ~131 KB but only part of it was accepted before the socket closed.

That means the TCP session broke while the NFS server was writing data.

This pattern is typically one of:

1. Network driver / NIC issue

2. TCP segmentation offload (TSO/GSO/GRO) bugs

3. MTU / jumbo frame mismatch

If you are sharing a user, it could also be FUSE-related; NFS doesn't always play nice with it, so it may be worth retesting with a disk or exclusive share.

Thanks, I wouldn't be surprised if it's the NIC, it has problems with power management that I've had to script a fix to. I might just buy a new pcie adapter. I confirmed the MTU shouldn't be an issue. I have an old dell switch that might be doing this too so I'll do a more thorough network test.

What do you mean by sharing a user? I have squash on so everything is nobody/99 users/100. I hate that I have to use FUSE but my shares are too big for a single volume (30TB+). I could probably figure out some symlinks and/or workarounds but that kind of sucks for usability.

11 hours ago, timjtech said:

What do you mean by sharing a user?

Sorry, I meant using a user share that is going through FUSE (/mnt/user)