[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Condor-users] Configuration problem with dedicated machines

Hello all.

We have installed Condor 6.7.8 in a cluster with 9 machines (Debian/Linux 3.1).
In this platform we want to run mpi, vanilla and standard jobs.

Three of these machines have been configured as dedicated resources. 

The "condor_status -run" command returns:

Name          OpSys       Arch   LoadAv RemoteUser           ClientMachine

sctr2.i3a.inf LINUX       INTEL  0.000  DedicatedScheduler@s sctr1.i3a.info
sctr3.i3a.inf LINUX       INTEL  0.000  DedicatedScheduler@s sctr1.i3a.info
sctr4.i3a.inf LINUX       INTEL  0.000  DedicatedScheduler@s sctr1.i3a.info

                     Machines         MIPS       KFLOPS   AvgLoadAvg

         INTEL/LINUX        3         6766      2415514   0.000

               Total        3         6766      2415514   0.000

When no mpi jobs are running, we want to run vanilla or standard jobs. However
these machines remain in Claimed/Idle state.    

According to the documentation, in the "condor_config.local" file of the central
manager we have added the parameter "UNUSED_CLAIM_TIMEOUT = 60" in order to
release the dedicated resources for other uses. Besides, the
"condor_config.local" file of the dedicated machines is: 

DedicatedScheduler = "DedicatedScheduler@xxxxxxxxxxxxxx"
## 2) Always run jobs, but prefer dedicated ones
START           = True
CONTINUE        = True
KILL            = False
WANT_VACATE     = False
RANK            = Scheduler =?= $(DedicatedScheduler)


STARTD_EXPRS = $(STARTD_EXPRS), DedicatedScheduler

How can we configure the dedicated machines to release the resources in order to
be able of running vanilla and standard jobs?

Thanks in advance.