23 Feb
2012
23 Feb
'12
2:59 a.m.
I can use a yt build that lives in my home directory to process data that lives on scratch from the compute nodes using an interactive qsub -I job. On an non-interactice job I do need to take care that the matplotlib backend is set to something that doesn't require X. I don't know how what type of filesystem $HOME is in terms of the mounting, but the situation would be substantially improved if the builds used by "module load yt/dev" lived on a filesystem that is more like /nics/d/home and less like /lustre/scratch.
It is possible to install the yt on NFS. But you can not access it from compute nodes. You can only use it on login nodes.
Pragnesh