[Condor-users] Machine not releasing CurrentRank

["Doak Bane" <doak.bane@xxxxxxxxxx>]

I'm running Condor 6.7.3 and have configured a machine for "special" jobs - 
boosting the RANK. When the user sets the appropriate attribute, in his 
submit file, the special job (cluster) runs as expected, preempting (using 
MaxJobretirementTime) running jobs as needed. (My reference for this 
structure is the Bologna Batch paper.)

Machine:
SpecialServer = True
IsSpecialJob = ( Target.SpecialJob =?= True && $(IsVanillaJob)
RANK = $(IsSpecialJob)

Negotiator:
PRIORITY_HALFLIFE = 900

User:
+SpecialJob = True

User A submits a special job. Other user jobs are queued (not special). User 
A then submits jobs that are *not* special. Once the user A special jobs are 
done the not-special user A jobs and do not allow any jobs of users B or C 
to run. The priorities are such that they should run. And if a similar job 
mix - without any special jobs - is run then the priorities correctly 
schedule jobs.

When I check the machine classad when the special jobs are running, 
CurrentRank = 1.000. The machine does not "release" that rank so when I 
check while the not-special jobs are running I see the same value.

What is happening? And is there something I can do to make the machine 
release the rank correctly?

Thanks,
Doak Bane