massive-website-banner

Common Issues

This section covers solutions to common issues

Error Messages

Issue

When attempting to use MASSIVE for GPU jobs I get the following error:

'Exiting due to CUDA error no CUDA-capable device is detected at line 104'

 

Solution

On MASSIVE GPU's are allocated at the request of the user and are unavailable otherwise. In the qsub command or pbs script you need to request the number of GPUs. Each node has 2 GPU's and 12 cores, so to use all cores and all GPUs on a node, for example, specifiy the following:

nodes=1:ppn=12:gpus=2

 

Issue

Using a program I get the following error:

'/usr/lib64/libstdc++.so.6: version `GLIBCXX_3.4.9' not found '

 

Solution

The program has been compiled against a newer version of GCC, than the one originally installed in the system, use modules to load the current version of GCC used on MASSIVE

module load gcc

 

If you are using a terminal session from the MASSIVE desktop, the issue is that your profile is not automatically loaded with KDE terminals (as they are not login shells). You can invoke your login profile with the following command:

source /etc/profile

 

Issue

While running a job I get an email stating the following:

PBS Job Id: 1000405.m1-m
Job Name: pbs_run_job
Exec host: m1042/0
job deleted
Job deleted at request of rootatm1-m
job 1000405 exceeded MEM usage hard limit (4003 > 1050)

(or "exceeded MEM usage soft limit")

 

Solution

The job has exceeded MASSIVE default memory limits. To use more memory you need to explicitly specify the amount of memory you require. See the memory section under Running PBS Script Jobs

 

Copyright © 2016 MASSIVE. All Rights Reserved.