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

Re: [Condor-users] partitionable slots are not returned?



On 03/28/2011 03:02 AM, Carsten Aulbert wrote:
hi all,

we just started to venture into partitioning slots, the set-up on one machine
is pretty simple right now:

SLOT_TYPE_1 = cpu=4, ram=7800, swap=30%
NUM_SLOTS_TYPE_1 = 1
SLOT_TYPE_1_PARTITIONABLE = True


This slot is then advertised correctly. However, running a couple of jobs and
we see that the "partitioned slots" are not returned to the pool:

gpu016:~# condor_status -direct testhost

Name               OpSys      Arch   State     Activity LoadAv Mem
ActvtyTime

slot1@xxxxxxxxxxxx LINUX      X86_64 Owner     Idle     0.000  6775
0+00:44:47
slot1_1@xxxxxxxxxx LINUX      X86_64 Claimed   Busy     0.000  1024
0+00:16:16
slot1_2@xxxxxxxxxx LINUX      X86_64 Owner     Idle     0.000     1
0+00:21:50
                      Total Owner Claimed Unclaimed Matched Preempting Backfill

         X86_64/LINUX     3     2       1         0       0          0        0

                Total     3     2       1         0       0          0        0

slot 1_2 was not used for about 30 minutes now, slots1_1 was idle before and a
new job took that slot again - however, this is pretty bad, as the new job
only requested a single core and slot1_1 had 3 cores reserved.

I cannot find anything about configuring the return into Condor, thus I assume
it should *just* happen, any idea what's wrong here?

Cheers

Carsten

I would guess that your START is preventing the slot from handling any jobs or going back to the Unclaimed state. When a dynamic slot hits Unclaimed it gets folded back into the partitionable slot.

You can identify the slot types with PartitionableSlot, DynamicSlot attributes.

Best,


matt