At boot time, two of my Linux (RedHat 8.0 and RedHat 9.0) machines mount an nfs share exported by a Mac OS X (10.2.5) machine; the mount information is specified in /etc/fstab. The mount succeeds, but it takes about four minutes to happen. Once it's done, the mount is fine; it's just really slowing down my boot process. The same delay happens if I umount the share and redo it with a mount command from the shell. On the RedHat side, the location of the share is specified with an IP address, so it's not a DNS problem. Now:

*If I try to mount the Mac OS X share on another Mac OS X machine, it succeeds immediately.

* If I export a share on one Linux machine and mount it on the other, it succeeds immediately.

* I'm having trouble mounting the Linux share on a Mac OS X machine, but I've seen this before, and Linux-Linux shares didn't have this problem then.

Arggh. Any advice?

Thanks,
Jim

Hi James Miller,

Seeing as you seem to be having timeout delays on your Macintosh machine when you ssh from there to your Linux box, it would appear that the problem is a Macintosh machine related problem and not a Linux one.

Therefore your question would be better suited to a Macintosh forum where knowledgable Macintosh experts maybe able to solve your Macintosh networking problems.

I'm getting this problem as well on my home system trying to NFS mount a filesystem from my Linux Alpha server.

The server is running 2.4.21 on an alpha. The client is 2.4.20 / 2.4.21 on an x86.

The delay is on the order of 15+ minutes.

Using tcpdump and ethereal I see that the mount RPC succeeds, followed by successful GETATTR, FSSTAT, and FSINFO calls. Towards the end of the delay before the client mount is successful the FSSTAT and FSINFO RPC calls are repeated again. (This is using NFSv3 UDP. telling the client to use v2 doesn't help)

We use v2 and v3 NFS between linux systems all the time at work without problems. I don't understand what is going wrong here.

Problem solved: The portmapper service wasn't running on my NFS client.

Make sure /sbin/portmap is running (most distros have a portmap init script to do it for you) on the client.