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

[HTCondor-users] missing resource request with grid_resource = nordugrid

Hi all,

we are trying to get a local research group to push jobs from their HTCondor to our Tier 1 ARC-CE+HTCondor cluster. However, none of the REQUEST_* parameters of their jobs are passed on to us, e.g. an 8-core job gets only a 1-core slot. We only see the information explicitly added via the nordugrid_rsl parameter.
Needless to say, they are at the top of our job kill highscore now.

We have confirmed this with minimal (non-)working JDL [1], which does not generate the proper request for 8 cpus [2].

I am quite sure the error is not with the CE, since there are a lot of VOs properly submitting jobs.
Do they need to configure anything specifically to get their HTCondor to pass on these parameters to us?


[1] Local HTCondor: submit.jdl
executable  = test.sh
output      = job.$(Cluster).$(Process).stdout
error       = job.$(Cluster).$(Process).stderr
log         = job.$(Cluster).log

request_cpus = 8
universe    = grid
nordugrid_rsl = (jobname=\"Ekp HTC to Gridka ARC\")

should_transfer_files = YES
when_to_transfer_output = ON_EXIT
x509userproxy          = /tmp/x509up_u12089

grid_resource = nordugrid arc-1-kit.gridka.de
queue 1

[2] ARC-CE: job.pvLLDm2NDerntzUVznaTWOeoABFKDmABFKDmzDTWDmABFKDmJ8ehrn.description
&("savestate" = "yes" )("action" = "request" )("hostname" = "ekpbms2.ekp.kit.edu" )("executable" = "test.sh" )("executables" = "test.sh" )("inputfiles" = (".gahp_complete" "" ) ("test.sh" "" ) )("stdout" = "_condor_stdout.ekpbms2.ekp.kit.edu_2667890.0_1511945855" )("stderr" = "_condor_stderr.ekpbms2.ekp.kit.edu_2667890.0_1511945855" )("outputfiles" = ("_condor_stdout.ekpbms2.ekp.kit.edu_2667890.0_1511945855" "" ) ("_condor_stderr.ekpbms2.ekp.kit.edu_2667890.0_1511945855" "" ) )("jobname" = "Ekp HTC to Gridka ARC" )