NFS Disconnected

Question

Level 1

0 points

NFS Disconnected

We have a Solaris 10 server that shares by NFS a directory.
On the Mac side, we mount (nosuid, nodev, rdirplus) this directory as home directory for some users.
In the kernel.log file we have the following messages :
Jan 19 15:57:51 Mac-Pro-de-xxxx kernel[0]: nfs server 10.10.7.2:/DATA/PROFIL/home_dir: lockd not responding
Jan 19 15:58:07 Mac-Pro-de-xxxx kernel[0]: nfs server 10.10.7.2:/DATA/PROFIL/home_dir: lockd alive again
Jan 19 15:58:21 Mac-Pro-de-xxxx kernel[0]: nfs server 10.10.7.2:/DATA/PROFIL/home_dir: lockd not responding
Jan 19 15:58:32 Mac-Pro-de-xxxx kernel[0]: nfs server 10.10.7.2:/DATA/PROFIL/home_dir: lockd alive again
Jan 19 15:58:46 Mac-Pro-de-xxxx kernel[0]: nfs server 10.10.7.2:/DATA/PROFIL/home_dir: lockd not responding

The mac displays a box with a message that says 'Connexion interrompue' (lost connexion). But when we did a 'ls -al' in a terminal all looks fine.

On the solaris server side there's no messages even in debug mode.

Any ideas ?

MAC PRO, Mac OS X (10.6.6)

Posted on Jan 25, 2011 6:23 AM

Reply

Answer 1

hig3

Level 1

4 points

Aug 25, 2017 7:47 AM in response to taz31

If you are absolutely sure that you do not need locks for that directory, you can give "locallocks" mount option to mount_nfs command in order to disable communication with server lockd on macOS 10.12 client.

mount -t nfs -o vers=3,locallocks,... host:directory directory

Reply

Answer 2

Jan 25, 2011 9:18 AM in response to taz31

I had this problem with a lab of 25 eMacs and iMacs which would go into "disconnect storms" where every single client would disconnect and become unusable. There were log messages on the clients saying the the nfs server was not responding. And it was very strange -- it would start at wildly random intervals, go on for sometimes minutes, sometimes hours, sometimes days. Finally we realized that spotlight seemed to be indexing on some of the affected computers. (My users' home directories are on the server.) There is a command that you can execute on the files on the server which will mark them to exclude from spotlight indexing. We found that we needed to mark every file on every partition on every disk on the server -- somehow spotlight was able to traverse up and down directory trees. Perhaps that was the problem -- spotlight on files that the client did not have privileges for.

One thing -- do not press the "disconnect" button when it is offered to you. Instead, unplug the ethernet cable and plug it back in, and/or turn off the airport and turn it back on. This will reset the network and it often comes back when you do. Otherwise you have to reboot to get the network back.

Reply

Answer 3

taz31 Author

Level 1

0 points

Jan 26, 2011 5:36 AM in response to cathy fasano

Thanks for the indication. We disable indexing by sudo mdutil -a -i off (snow loepard). The command sudo mdutil -a -s shows us the indexing is disabled on all volumes (including NFS mounts). The problem is still alive !!!

Reply

Answer 4

dajak

Level 1

0 points

Feb 15, 2011 4:35 AM in response to taz31

Got the same problem. It happens whenever I write a file to the NFS disk and Mac OS creates a "._*"-File.
I do a

find . -name ._* | xargs rm

on the NFS share and reconnect the network to get rid of the "beachball" in the Applications working with the share.

THIS IS REALLY ANNOYING! I have to work with a lot of filese from the server!

Reply

Answer 5

jonesy16

Level 1

4 points

Mar 1, 2011 8:03 AM in response to taz31

We have a similar problem. I have an entire small business (about 11 users) connected from OS X 10.6 clients via ethernet to a CentOS NFS server. Each box is connecting to about 6 different NFS shares on the linux box. Every once in a while (about 1 share / user / day) will pop up saying it has lost connection. You can pull up a Finder window and browse to the share via /net/server/share and it works fine, but the original NFS mount has to be disconnected and reconnected to restore usability. This is happening only on OS X clients, none of our linux clients have any issues. What gives?

Reply

Answer 6

Mar 16, 2011 4:10 PM in response to taz31

Add a "me too" here. This has been happening since January on my MacBook Pro. Otherwise all other systems in our fleet that read off our Mandriva NFS server have been doing fine fine.

But with my MacOS 10.6.6, I get the "lockd not responding" error. (no lockd alive again) and I also get the popup asking if I want to disconnect.

This error is new. And happens no matter where it is plugged in.

I connect on the macbook using Automator since it's a portable unit, I just call it from the command line via an alias when I doc with our network.

Reply

Answer 7

etresoft

Level 9

51,022 points

Mar 16, 2011 4:59 PM in response to Bill Capehart

It seems to be an old problem.

http://lists.apple.com/archives/macosx-interop/2006/Apr/msg00014.html
http://discussions.apple.com/message.jspa?messageID=6597979
http://discussions.apple.com/thread.jspa?messageID=12318712&#12318712
http://opensolaris.org/jive/thread.jspa?messageID=477052&tstart=0

You have to make sure your server is up to date, make sure all of your firewalls are fully functional or disabled, and debug the low-level networking. In other words, standard NFS.

Reply

Answer 8

Feb 24, 2012 4:20 AM in response to taz31

I am also experiencing this problem. Running the newer Oracle Solaris 11 Express and Snow Leopard as clients.

Doing a rsync of a home directory looks like this...

Desktop/Personal/Internet Usage/

Desktop/Personal/Internet Usage/1395696624.png

Desktop/Personal/Internet Usage/._1395696624.png

Desktop/Personal/Internet Usage/1395699028.png