Why do I get MPI_Abort errors when trying to submit a parallel job?

3 visualizaciones (últimos 30 días)
Paul Zhang
Paul Zhang el 23 de Mayo de 2014
Comentada: Edric Ellis el 23 de Mayo de 2014
The core of my job submission code is below:
jopt.email_notif = 0;
jopt.toggleleft = left_list(j);
jopt.toggleCausalDir = dir_list(k);
jopt.toggleChoice = choice(l);
jopt.od_number = od_list(i);
jopt.connectivity = 1;
sched = findResource('scheduler', 'configuration', 'NeuroEcon.local')
set(sched,'SubmitArguments', '-l walltime=0:20:00')
pjob = createParallelJob(sched);
set(pjob, 'FileDependencies', {'multiDCMset1.m'})
set(pjob, 'MaximumNumberOfWorkers', 1)
set(pjob, 'MinimumNumberOfWorkers', 1)
t = createTask(pjob, @multiDCMset1, 1, {jopt})
t_all{1,jj}=t; jj=jj+1;
submit(pjob);
---------------------------------------
The following is the error message I get in the job submission log, after the job finishes running. I don't understand the error or what could cause it. I do know that the same script runs fine on another person's computer. Do I need some specific settings to submit parallel jobs?
------------------
Node file: /opt/torque/aux//2075983.neuroecon.caltech.edu
Starting SMPD on compute-1-30 ...
ssh compute-1-30 "/opt/matlab//bin/mw_smpd" -s -phrase MATLAB -port 25983
All SMPDs launched
"/opt/matlab//bin/mw_mpiexec" -phrase MATLAB -port 25983 -l -n 1
-machinefile /opt/torque/aux//2075983.neuroecon.caltech.edu -genvlist
MDCE_DECODE_FUNCTION,MDCE_STORAGE_LOCATION,MDCE_STORAGE _CONSTRUCTOR,MDCE_JOB_LOCATION,MDCE_DEBUG
"/opt/matlab/bin/worker" -parallel
[0]which: no shopt in
(/opt/matlab/bin:/usr/kerberos/bin:/usr/java/latest/bin:/opt /intel/itac/7.1/bin:/opt/intel/fce/10.1.018/bin:/opt/intel /idbe/10.1.018/bin:/opt/intel/cce/10.1.018/bin:/usr/local /bin:/bin:/usr/bin:/opt/ganglia/bin:/opt/ganglia/sbin:/opt /openmpi/bin/:/opt/maui/bin:/opt/torque/bin:/opt/torque/sbin: /opt/rocks/bin:/opt/rocks/sbin)
[0] < M A T L A B (R) >
[0] Copyright 1984-2009 The MathWorks, Inc.
[0] Version 7.8.0.347 (R2009a) 64-bit (glnxa64)
[0] February 12, 2009
[0]
[0] To get started, type one of these: helpwin, helpdesk, or demo.
[0] For product information, visit www.mathworks.com.
[0]
job aborted:
rank: node: exit code[: error message]
0: compute-1-30: -2: application called MPI_Abort(MPI_COMM_WORLD, 42) -
process 0
Stopping SMPD on compute-1-30 ...
ssh compute-1-30 "/opt/matlab//bin/mw_smpd" -shutdown -phrase MATLAB -port
25983
Exiting with code: 42
  1 comentario
Edric Ellis
Edric Ellis el 23 de Mayo de 2014
Is there any error in the task of the job? Check using:
pjob.Tasks(1).Error
or even
getReport(pjob.Tasks(1).Error)

Iniciar sesión para comentar.

Respuestas (0)

Categorías

Más información sobre Cluster Configuration en Help Center y File Exchange.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by