Thursday, May 12, 2005

The Never Ending Connection

We had a situation a few weeks ago when Analysis Services was not responding to any user connections. Thinking that this was perhaps a return of some of the historical problems that we have had in terms of locking, we promptly logged a call.

This particular situation was slightly unusual in that it appeared to come and go with no warning which made tracking it down very difficult.

During our investigations with MS it became apparent that something else was going on here, not only was AS affected, SQL Server was also demonstrating the same symptoms. Connections appeared to either take in excess of 50-60 seconds or else they did not work at all.

Many memory dumps later and after plenty of diagnostic work with MS it would appear that the user connection was being help up inside LSASS.EXE when talking to the primary domain controller.

After a bit of internal searching we eventually found that one user was logging onto the server with a NULL domain name, which was causing the old NT 4 PDC to broadcast to all the other domain controllers and wait for a response to see if they could validate the user. When this failed control was eventually returned to LSASS.EXE and some logons could be serviced before the user tried to connect again, and the whole vicious cycle started all over again.

I have never heard of a problem like this before, so I thought it worthwhile to share. If you need any further details on this please contact me.