Documentation or Feature for Environment Use on Remote about pyiron HOT 3 CLOSED

Hi @pstaerk,

I am not 100% sure I got your setup correctly. You did install pyiron on an HPC and want to run there? Then you would not need the so called remote setup of the queuing system, but only the queueing itself (remote is for working at your laptop and sending stuff to another HPC - which also could be your setup...)

In either case, it is sadly not possible right now to specify a the environment in which the remote pyiron should be executed. The python process determining this is always started first (and thus would be in the standard login environment). (In addition to conda, there could also be several different other environmental variables be changing, like e.g. module load intel/2019 or the like to be considered). If it would not harm otherwise, you may alter your login script (e.g. .bashrc) to load the correct environment for pyiron from the beginning. In that case, all features should work like expected.

If you 'only' want to send specific jobs with external executables to the (HPC local) queuing system, you can alter the run scripts in the pyiron resources to load a specific environment before starting the actual calculation.

from pyiron.

I have a work-machine from which i submit jobs over ssh to a HPC slurm queue. And I installed pyiron via the usual conda install -c conda_forge pyiron command, however not in the base environment but into a conda environment. I will try to change the bashrc as you have suggested, sounds like a good idea!

Edit: Changing the .bashrc has worked very well and the environment gets loaded perfectly.

from pyiron.

Just for completeness: another point of injecting this would be via the queuing script defined for the queue adapter. You could do this already now by changing the script that is in the queues folder of your pyiron resources. Midterm, it might be nice to add this to pysqa, similar as requested here.

from pyiron.