nfs, mounted loop devices

Question

I'm usign a Debian server (Jessie) as an NFS server. All exported NFS-shares are mounted filesystems using loop devices. They are mounted in the fstab without any special options & the mountpoints are managed within a common root directory:

• /nfsshares *(rw,fsid=0,async)
• /nfsshares/mp1 *(rw,sync,no_root_squash,no_subtree_check,nohide)
• /nfsshares/mp2 *(rw,async,no_root_squash,no_subtree_check,nohide)

mp1,mp2 are mountpoints for the loop devices!

Starting the nfs server causes an entry in the log file: /nfsshares/mp1 und /nfsshares/mp2 have same filehandle for * using first.

All NFS-clients always mount the first share - regardless of the used mount points. I think that this has something to do with the used loop-device for the mountpoints.

Isn't possible to use mounted images for NFS shares?

ThanX /franzR

Answer 1

Depending on the locations of the directories being mounted, as well as client and server configurations it may be valuable to ensure that /etc/exports is setup correctly on the server. This answer on setting up NFS with mount-points as sub-directories might be helpful.

Answer 2

This answer is probably a bit late to the party, but in the hopes of helping others with the same problem:

The problem is likely an identical fs UUID (at least it was for me). Specifying a fsid in the export solved the issue. (Do note that fsid=0 is used for the 'full export', see https://linux.die.net/man/5/exports).

So adding fsids != 0 for the 2 loop devices exports should fix this issue. Result should look like:

`/nfsshares        *(rw,fsid=0,async)`
`/nfsshares/mp1    *(rw,fsid=1,sync,no_root_squash,no_subtree_check,nohide)`
`/nfsshares/mp2    *(rw,fsid=2,async,no_root_squash,no_subtree_check,nohide)`

Got the inspiration from: here