Upgraded to btrfs and now esxi host is getting "bad file descriptor" for all files on NFS

Hi all,

My OMV is the latest release with all patches (1.7).

The other night, I decided to upgrade the filesystem on my RAID to BTRFS (still using mdadm). So, I unmounted the filesystem and ran the btrfs-convert tool on the volume.

Everything went OK, and short of having to adjust the shared folders to use the new uuid name and minor stuff like that, I'm up and running (CIFS/SSH/etc.).

Except for my ESXi host that was running from the NFS share...

Now, when I try to browse the NFS from the VIClient (VIC) it shows the two VM folders at the top level, but when I try to click into the folder, I see nothing. On the esxi cmdline, if I go into the folder and type 'ls' I see this:

/vmfs/volumes/ddfc7551-fa56ac4a # cd NetTools/
/vmfs/volumes/ddfc7551-fa56ac4a/NetTools # ls
ls: ./NetTools-flat.vmdk: Bad file descriptor
ls: ./NetTools.vmdk: Bad file descriptor
ls: ./NetTools.nvram: Bad file descriptor
ls: ./vmware.log: Bad file descriptor
ls: ./vmware-35.log: Bad file descriptor
ls: ./vmware-36.log: Bad file descriptor
ls: ./vmware-37.log: Bad file descriptor
ls: ./NetTools.vmx: Bad file descriptor
ls: ./NetTools.vmxf: Bad file descriptor
ls: ./NetTools.vmsd: Bad file descriptor
ls: ./vmware-40.log: Bad file descriptor
ls: ./vmware-38.log: Bad file descriptor
ls: ./vmware-39.log: Bad file descriptor

Looking in the vmkernel.log, I see this repeated many times every time I try to view one of the NFS dirs:

2014-12-26T21:21:29.803Z cpu0:41172)WARNING: NFS: 1359: File handle too big (60)
2014-12-26T21:21:29.803Z cpu0:41172)WARNING: NFS: 1359: File handle too big (60)
2014-12-26T21:21:29.804Z cpu0:41172)WARNING: NFS: 1359: File handle too big (60)
2014-12-26T21:21:29.805Z cpu0:41172)WARNING: NFS: 1359: File handle too big (60)
2014-12-26T21:21:29.805Z cpu0:41172)WARNING: NFS: 1359: File handle too big (60)

Note that this ESXi host was working just fine and running VMs from the NFS like this prior to the BTRFS upgrade. Is there some kind of oddity with BTRFS that is making its way through NFS to the client to upset it? The share ACL for 'unknown' is r/w/x and the share is setup in OMV as r/w.

I'm at a loss for what might be going wrong here, especially since it worked just fine for months up to the upgrade and I thought the point of NFS was to hide filesystem details from clients. Right now, I'm scp-ing the VM files from OMV onto the ESXi host's internal disk just fine, so all is not lost. But, I'd prefer to run the VMs from the NFS to take advantage of the RAID/etc.

Thanks,
Mike

I'll try that once the data is scp off the drive.

No luck. Same issue with filehandle too big (60)

Nope, doing the ls on the omv machine works, as does mounting the NFS share to itself. This is likely some deficiency with the ESXi NFS client, but what's odd to me is that it was working perfectly with my OMV install for probably over a year (since I built my OMV machine) until the BTRFS upgrade.

I erased the volume and formatted EXT4 and the ESXi NFS mount is now operational. Guess this can serve as a warning to anyone else considering upgrading to BTRFS who uses ESXi.