MATLAB: Do jobs hang when using the ‘-nodesktop’ and ‘-nojvm’ flag and LSF Scheduler

clusterhangingnodisplayparallelParallel Computing Toolboxserver

I am using '-nojvm' and '-nodesktop' flags when running an application on LSF scheduler. I have noticed about 10% of all my jobs hang with the following log output:
Exited with exit code 140.
Resource usage summary:
CPU time : 23018.00 sec.
Max Memory : 237.48 MB
Average Memory : 231.58 MB
Total Requested Memory : 24000.00 MB
Delta Memory : 23762.52 MB
Max Swap : 3501 MB
Max Processes : 4
Max Threads : 50
Run time : 23043 sec.
Turnaround time : 23073 sec.
The output (if any) follows:
------------------------------------------
Setting up environment variables
---
LD_LIBRARY_PATH is .:/home/tools/matlab/mcr_r2019a/v96//runtime/glnxa64:/home/tools/matlab/mcr_r2019a/v96//bin/glnxa64:/home/tools/matlab/mcr_r2019a/v96//sys/os/glnxa64:/home/tools/matlab/mcr_r2019a/v96//sys/opengl/lib/glnxa64
Maximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reached
Maximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reachedMaximum number of clients reache
dMaximum number of clients reachedMaximum number of clients !
Why does this happen and is there any workaround for this?
 

Best Answer

  • Adding the flag "-nodisplay" seems to resolve the issue. This is because the "-nodisplay" startup option suppresses allocation of an X-client.