Thank you so much for getting back to me; I'll try some of your suggestions tonight and I'll let you know how I get on!
All the best,
On 3 Nov 2011, at 13:58, The Admin wrote:
Thanks for the email. It sounds very similar to the issue I was experiencing. The connection isn't being dropped, the finder is screwing up. If you traverse that directory with terminal you should still find it connected.
I made some changes to the server which seemed to fix it. Firstly I turned off all folder redirection as after checking user accounts I found it wasn't working anyway.
I then went through All user accounts looking for the old folder redirector symlinks and deleted them. I also looked for all other symlinks and deleted those too. For good measure i removed all ACLs (chmod -RN) and then used server admin to set them again. Then i rebooted all the clients and the server.
What i think is happening is when you make a symlink from a remote directory to a local directory or vis versa some older API somewhere in OS X creates a new filesharing connection. As you cant connect to an AFP server with more than one set of credentials at a time i believe this caused a conflict and then the finder then tries to resolve the alias to one of the connections. Or something like that anyway. I actually don't know for certain i just know the issue has gone away now and I didn't have to rebuild the server!
On 2 Nov 2011, at 13:52, Mr X wrote:Hi there,
Hope you don't me e-mailing you, please feel free to ignore me if this is an imposition. I've been having a problem with network account users on and randomly finding all desktop icons are gone, after much googling I happened upon your blog and your post about the Carbon API Kerfuffle. The issue is seemingly fixed by a restart so I can't see any need to run a combo update every night (and I don't have a task server set up at the moment anyway), and although I have very recently turned on Folder Redirection the issue predates that.
After much wailing and gnashing of teeth and poking around it appears the AFP automount is seemingly losing it's connection and it cannot be re-established; if I browse to /Network/Serveand drill down to connect to the share where the user's folders are I get a Delete/Fix Alias prompt, but it's fine after a restart. Did you manage to gain any more insight into why this might be happening?
Looks like it's going to be a weekend job of hosing then rebuilding the server, but I'm loathed to do that because it feels like I never find out the cause of the problem (not to mention I have to spend my weekend in a cold, dusty server room).
Here's a post I threw up on the Mac OSX Server Admin mailing list with a bit more info:
I have found the advice on this list invaluable over the last few years, I'm hoping you guys can shed some light on an issue I'm currently having.
I have a Mac Pro server on 10.6.8, 1TB boot partition, 2 x 3TB drives on RAID 1 with SoftRAID 4.0.7 for primary file storage, basic services (DNS, OD master handing out MCX prefs, AFP, SMB, SUS). I have about 15 client machines authenticated-bound to the OD, and about 50 users who hot desk between some of the client machines using pure network homes (I have a nice speedy network, so I thought they could live without the long log-ins of PHDs).
I'm getting an intermittent problem that I can't nail down, seems to happen pretty much every day to at least one client machine - when a user logs in, their desktop icons disappear; none of their files appear, none of the MCX-login mounted shares appear, not even the Macintosh HD icon, and you can't save or drag or create any files on the desktop - so it looks like a permissions error. However if they browse to their Home folder (e.g. from the sidebar), or browse to it through the automounted homefolders share all their files are there with the correct permissions. Restarting the client machine solves this problem.
The clients are a hotch potch of different generation iMacs and a couple of recent MBPs.
DNS is set up correctly and resolving properly on all machines (nslookup OK from client end)
OD and Kerberos seem to be OK - sudo changeip -checkhostname reports all clear, kinit xxxxx working from client machine
Clients bind to the OD without issue, and users with laptops and PHDs have no problems
So far I have tried:
- Deleting the space in the home folders automount - "Home Folders" to "homefolders"; I read that might cause an issue
- Enabled Guest access to the automount, I know this is standard practice but I really don't like having guest access on anything
- Trashing prefs to do with a wide variety of things on an affected client machine, one at a time - Kerberos, DirectoryService, Finder, MCX etc.
- Removed quotas from all users (was previously set to 20 GB, which should've been plenty)
- Trawled the logs in Console for anything suspicious, but can't find anything that I think is particularly relevant after some googling. Three that are fairly common and could be problematic are:
10/31/11 4:49:17 PM com.apple.launchd (com.apple.SystemStarter) Failed to count the number of files in "/System/Library/StartupItems": No such file or directory
10/31/11 4:49:18 PM com.apple.launchd.peruser.1080 (com.apple.Kerberos.renew.plist) Exited with exit code: 1
10/31/11 4:49:18 PM ServerScanner Not scanning because node /LDAPv3/server.site.company is in searchPath
I've only seen one other case of what I'm experiencing, but it doesn't look like there was a resolution (and the deleting files bit doesn't tally up here, my issue appears to be random):
I'm totally stumped, and can't find a pattern to it - if I'm being honest I'm not even 100% sure whether it's a server or a client problem. Thoughts I've had as to what it might be are:
- The home folder automount is on a SoftRAID volume - perhaps this is causing the issue, though I have tried moving the home folders to a different, standard Apple volume
- The Mac Pro has both ethernet ports aggregated/bonded - could this cause problems?
- You're going to hate me for this one - as I could only find a 10.6.3 server disk (which wouldn't boot on the Mac Pro), I created my own server installer using InstaDMG and the 10.6.8 Combo Update. It installed perfectly and I've noticed no other issues with it (and I've set up a similar network at another site with no such issues using the same installer), but I guess this is what it could be too.Many thanks in advance for any input you might have.
With best wishes,