Skip to content

FAQ

How to run Jupyter notebooks remotely on EDI?

Assuming jupyter is running on edi07 (port 8888) and that the connection is established with sshuttle (see Getting started section for more details) it is possible setup the tunnel as follows: 

ssh -NL 8888:localhost:8888 your_username@edi07

Afterwards you should be able to see the running Jupyter instance via browser at the URL: http://localhost:8888

How to install python packages on EDI?

python3 (3.6.9) and python2 (2.7.17) along with the recent pip and virtualenv are installed system-wide on each EDI.

You can simply install packages either with pip install --user or use virtualenv (this will allow to handle multiple projects with possibly conflicting dependencies).

Finally, if you need newer versions of python or want to handle complicated dependencies you can install a local version of anaconda in your $HOME directory.

I want to use program XXX on EDI, can I install it on my own?

You can install any software you like in your $HOME directory (as long as you have a valid license to use it).

If you need support in setup of some program or want it to be installed system-wide, please contact the administrators.

What are the guidelines for /home/nfs/ distributed filesystem use?

  • Store only the necessary data in your /home/nfs/your_username directory. The space in this filesystem is limited.
    • For example: storing your conda distribution, scripts, inputs for jobs run on the cluster and small outputs that you'll move to personal space after completion of the calculations is OK.
  • Avoid prolonged and heavy I/O operations (like reading and writing large files).
    • For example: Writing output from the MD simulations in amber (successive, small write operations) is OK. Reading a 100 GB file from nfs directory is not OK.