Centrally managed neurocommands

[marcelzwiers]

Installing neurocommands on our HPC works as expected, following the instructions from the documentation. This installs modules and containers in the user directory. What I would like to do is:

• Install it for all HPC users, e.g. in /opt/neurocommand/. Can I ask our sys admin to do just that or is there something that needs to be adapted (e.g. the build script)?
• Install it without changing the Python base environment. I know that can be done using venv (or conda or so), but would that require users to activate that environment before running neurocommands? Or is the Python venv not needed to run neurocommands and can I simply run the build script in the venv, fetch containers, deactivate it, and then load the module and run neurocommands in the base environment?

[stebo85]

Dear @marcelzwiers

yes, you can install neurocommand in a global directory for all users on an HPC. /opt/ might not be the best location and I would suggest a network-mount that's available on all compute nodes.

The python environment is only needed for installing it, not for using it. So you can do exactly what you said: Install the containers in the python environment for neurocommand and then change back to the base environment for using them. The users of the HPC would only need to add the modules directory to their lmod modulepath (or run module use).

Let me know how you go with this and if the documentation needs more details on this

Thank you
Steffen

[marcelzwiers]

Thanks! Our /opt is mounted on all nodes, and yes, it would be nice if the installation docs mention that the requirements are not needed for using the containers, and that one best installs the python libs in a venv. Thanks again, I will try to make things available centre wide, and will let you know if we encounter issues. Cheers from Nijmegen :)

…

_____________________ Sent from my phone Op ma 29 jan 2024 23:40 schreef Steffen Bollmann ***@***.***>:

Dear @marcelzwiers <https://github.com/marcelzwiers> yes, you can install neurocommand in a global directory for all users on an HPC. /opt/ might not be the best location and I would suggest a network-mount that's available on all compute nodes. The python environment is only needed for installing it, not for using it. So you can do exactly what you said: Install the containers in the python environment for neurocommand and then change back to the base environment for using them. The users of the HPC would only need to add the modules directory to their lmod modulepath (or run module use). Let me know how you go with this and if the documentation needs more details on this Thank you Steffen — Reply to this email directly, view it on GitHub <#261 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADTUGL6R4S4CXSEY2EOXXC3YRAQMHAVCNFSM6AAAAABCPSNM6OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMJVGY4TSOBWGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[stebo85]

Great :) I added more details to the documentation.