Re: [HTCondor-users] Trying to understand DedicatedScheduler related problems

[Carsten Aulbert <carsten.aulbert@xxxxxxxxxx>]

Hi again,

we just had another job behave like this.

It was submitted (requesting 32 nodes which were free at that point),
one could watch

condor_status -const 'PartitionableSlot isnt true' -af ClientMachine
RemoteUser Cpus JobId

with a rising number of slots with an undefined JobId until it reached
30. At that point condor_q showed this job as running and within seconds
it re-appeared to be 'idle' and 12 nodes were left from condor_status'
view without a defined JobId.

Looking further through the logs, not much is seen, e.g.Schedlog:

10/02/20 14:43:44 (pid:1398) Starting add_shadow_birthdate(969.0)
10/02/20 14:43:44 (pid:1398) Started shadow for job 969.0 on
slot1@xxxxxxxxxxxxxxxxx
<10.10.82.1:9618?addrs=10.10.82.1-9618&noUDP&sock=2209_c4cc_3> for
DedicatedSchedule
r, (shadow pid = 1864058)
10/02/20 14:43:45 (pid:1398) Received a superuser command
10/02/20 14:43:45 (pid:1398) Number of Active Workers 0
10/02/20 14:43:46 (pid:1398) In DedicatedScheduler::reaper pid 1864058
has status 27648
10/02/20 14:43:46 (pid:1398) Shadow pid 1864058 exited with status 108
10/02/20 14:43:46 (pid:1398) Dedicated job abnormally ended, releasing claim
10/02/20 14:43:46 (pid:1398) Dedicated job abnormally ended, releasing claim
[..]

Thus, still being puzzled about it. Anyone with an idea, where to dig
out more information about what may have gotten wrong?

Cheers
Carsten

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature