OD users 'disappearing'

[tenpa_gyatso]

We're running 10.5 on our servers and workstations, open computer labs authenticating against Open Directory and AFP for managing the sessions.

A number of our 600+ OD users have, for the last few days, been losing their login capability. When they attempt to login either from workstations on our network or into our email system or course management system, all of which auth against our OD server, the attempts fail. When looking at their accounts in Workgroup Manager all seems in place. And their user account on the Xserve where their storage lives is still there and all files/directories are intact.

When attempting to make changes in their OD account (password, group membership, etc.) using Workgroup Manager (WGM) when saving the changes I receive an error pane that includes the following: An unexpected error of type -14120 (eDSAuthFailed) occurred. All references I find of this point to failed authentication on the part of WGM. But I am able to authenticate as diradmin to the proper directory on my OD server at /LDAPv3/127.0.0.1.

When this started happening two days ago I was able to delete the user account in WGM, recreate the account to match the UID of the user's account on the storage Xserve and everyone was happy. As of last night, I also have to wipe the user's old files/directories from the storage Xserve after deleting the account. Once the WGM account is recreated, I see the user's account on the Xserve, new directories created properly and I put the backup of the user's files back in place.

When I think of 'what's changed recently?' all that comes to mind is that all four people using WGM to manage our user accounts upgraded to 10.6 on their laptops within the last month. But a great deal of work has been done managing accounts using WGM during that time and this problem didn't crop up until the beginning of this week.

Is there anything in particular I should be looking for in the OD and/or AFP log files? Does this sound more like an OS problem with the user accounts on the Xserve or only something broken in OD? And finally, can you think of anything to look for in user accounts who haven't reported this phenomenon that might point me in some direction? If you need any more specifics, please ask.

Thanks...

 

[tenpa_gyatso]

As a follow-up, the problem has ceased viz. additional users having this problem. After sifting through logs we found references to VM over-write which to me sounds like a 'simple' memory leak. We rebooted the Xserve and the problem has not surfaced for any additional users.

However, a few users who hadn't logged onto the network since this problem started at the beginning of this week did run into the same symptoms. Hopefully after today no additional user problems will crop up.

A memory leak would certainly explain the sort of 'hit' the OD database took and the sort of symptoms we saw. If anyone has anything to add, please do, either anecdotal or otherwise.

Thanks for reading...