Initially we had the same type of problem, machine objects were being created but not User objects. We were pushing the client with SMS/SCCM to 10 machines at a time and the problem was happening on 4 out of 5 machines.
Eventually we increased the number of admin logins allowed from 1 to 20 and the timeout in the Autodomain script from 20 secs. to 45. It seems to be working much better and is probably around 98% successful.
you don't happen to be out of user licenses do you?
No, I'm way below my license count.
EE is prone to connection problems, especially when they are multiple simultaneous requests coming from slower clients to a large database. Therefore you need to implement your own recovery process (multiple, timed retries).
Thanks for the response. I think the issue may be with the DB. I just ran two seperate install tests and received longer than normal wait times (4 minutes) for the command sbadmclDumpMachineUsers to complete. The client is on the same subnet so I'm pretty sure the latency isn't becuase of the network. Here's what I see in the log:
1/29/2010 10:42:00.55 AM: Running sbadmclIsMachineActive for ENDPOINTWS01...
1/29/2010 10:42:00.55 AM: Existing Connection
1/29/2010 10:42:05.55 AM: 0x00000000 | The operation completed successfully.
1/29/2010 10:42:05.55 AM: sbadmclIsMachineActive returned results for ENDPOINTWS01
1/29/2010 10:42:05.55 AM: Getting the current list of users for machine "ENDPOINTWS01"
1/29/2010 10:42:05.55 AM: Running sbadmclDumpMachineUsers (ENDPOINTWS01)...
1/29/2010 10:42:05.55 AM: Existing Connection
1/29/2010 10:46:24.02 AM: 0x00000000 | The operation completed successfully.
Am I on the right track about this being a database issue?
In these tests the user actually did get created.
Do you have database cache (index) turned on?
Is there something in AutoD (besides what's already been suggested) I can set to ensure the script completes? Last week I deployed 300 new machines at once and 86 of them failed user assignment. I've deployed this many at once before with about 6% of the users not getting assigned. Now that I have 1800 users and machines in the DB I'm realizing (painfully) that the number needs to be much lower. I still have about 2000 machines left to encrypt and management is pushing hard to get these done ASAP. Also, this week I implemented the dbcfg.ini file in the hopes of improving DB performance. I don't have the toastcache.bat file though. Will caching still occur without having run that?
Cache refresh should be done as maintenance (with all database services off) at least once a week.
For AutoDomain support you need to reach Simon, who created it.
There are ways to resolve this problem, but you might need to create your own queue mechanism to serialize EE requests.