Solaris 10 NFS Share Issue

Hello all,

I am having an issue with an NFS share I have created between two Solaris 10 boxes. I want the share to have read/write permissions, but for some reason it is coming up as read-only on the client side mount despite "rw" being specified in the mount options.

Here is what I have in /etc/vfstab on the client:

server:/share - /share nfs - yes rw

Here is what I have in /etc/dfs/dfstab on the server:

share -F nfs -o rw,root=clientservername /share

The problem appears to be with the client-side mount, because if I perform the following command it becomes read-write:

mount -o remount /share

Any ideas why the share started out as read-only despite being specified otherwise? Any help is greatly appreciated!

Regards,

Chris

Please post the /share directory permissions on the server.

Who is the owner?

(Could try setting to 777 and see if the problem goes away)

The owner is root and the permissions are currently 755, but I currently have read/write after performing the remount option. I am troubleshooting this on a production server, so I have not had a chance to reboot the server to see if it reverts to read-only (I suspect it will since it started out that way), but when I do I will give 777 a shot.

Thanks,

Chris

Unfortunately I tried setting the permissions to 777, but it still comes up as read-only initially. So everytime the system reboots or the NFS client service is restarted...it comes up as read-only and I have to use the remount command to fix it. I also tried forcing NFSv4 just to see if that changes anything, but no change.

I just find it odd that the remount command is supposed to mount an already mounted share adding in the "rw" option even though I already mounted it with the rw option in the first place.

Regards,

Chris

I agree that this is weird, especially since I do this all the time between Solaris 10 boxes with no issues. I appreciate that doesn't help you.

Having looked at how I configure mine, I notice one major difference:

In vfstab on clients I always use:

 
server:/share - /share nfs - yes rw,soft
 

for soft not hard (default) mount

Other than that I can't spot any differences.

Perhaps cut and post the actual vfstab and dfstab contents in case I can see a syntax error or the like.

I confgure dfstab on the server side as you are doing it.

I admit that I am confused.

Weird. Try mounting with "-o vers=3,proto=tcp" and see what happens.

And don't use soft mounts. Soft mounts do not support mmap'd files because once a file is mmap'd, the system assumes any pages needed can be read/written as necessary and there is no way to tell a running application otherwise. This can lead to processes crashing with SEGV/SIGBUS and/or data corruption. Also, writes cached on the client can be lost if the IO attempt times out after the application received a successful return from its write() call, which can also lead to data corruption. Never mind the fact that you're assuming all applications will properly handle possible intermittent IO errors any reads/writes. If they don't - data corruption.

IMO, if you know you're never going to have mmap'd files (and executables - applications and shared objects - are all mmap'd when they are run...) and ALL the apps running on your system are known to properly handle intermittent IO errors soft mounts are OK.

1 Like

Another thought on this issue..........

On your server can you

ping <clientname>

just to check that the client name can be resolved. Does it translate to the correct ip address of the client?

If ping cannot resolve the name then neither will the share command and the server will not recognise the inbound nfs mount request as a node it should give rw,root access to. A read-only access might result (I guess).