[Condor-users] kbbd issues

[Eric Abel <Eric.Abel@xxxxxxxxxx>]

Fellow Condor users,

 

I recently have encountered an issue with kbbd.  I have a windows pool, with a mixture of windows7 and XP workstations.  Recently, one of the windows7 machines ceased to kill jobs when the user logged on.  Looking deeper into the problem, I discovered that the issue surfaced when a user would log in, then leave the computer for a period of time.  The conditions for job acceptance were met (per UWCS_START), and jobs started.  Upon trying to resume his session (note he never logged out), the jobs continued to run, utilizing 100% of his CPU.  I looked into it, and saw that there was no kbbd associated with his username (although a SYSTEM kbbd was running).  I then logged in, using the “switch user” feature, and this spawned a kbbd under my username.  This kbbd sensed my presence and took appropriate action with the running jobs.  We then logged out of my session, returned to the other user’s session…still no kbbd associated with his name.  We then logged him out, then back in again, and now there was a kbbd associated with his username.  So it seems that kbbd is terminating for some reason, and not being restarted upon session resume, only for the start of a new session.  I should point out that both user accounts are administrators on this machine.  I should also point out that the system has been working fine for about 6 months, and this only surface after the last round of windows updates.  I checked the kbbd log and found nothing out of the ordinary.   Has anybody encountered this before?  Any suggestions?

 

Thanks in advance,

 

Eric