Skip to content

Latest commit

 

History

History
75 lines (55 loc) · 2.77 KB

Jupyterlab.md

File metadata and controls

75 lines (55 loc) · 2.77 KB

Jupyter lab setting up

Jupyter lab can run remotely. You can set as follows.

Configurate on the server

Inside the sever, do

  • Create configuration file: jupyter notebook --generate-config

  • Enter into the configuration folder: cd ~/.jupyter/

  • Optional: generating a password to login instead of using token each time: jupyter notebook password

  • Create a certificate

    openssl req -x509 -nodes -days 365 -newkey rsa:2048 -keyout mykey.key -out mycert.pem

  • Edit the file jupyter_notebook_config.py (it is in your current folder) to look like

    c.ServerApp.certfile = u'$HOME/.jupyter/mycert.pem'
c.ServerApp.keyfile = u'$HOME/.jupyter/mykey.key'
c.ServerApp.ip = '*'
c.ServerApp.open_browser = False
c.ServerApp.port = PORT

    Replace PORT by any integer number between 8001 and 9999.

    Also, replace $HOME to the aboslute path of your own HOME directory: /home/YOUR_USER_NAME

Lauch and Connect

  • On server, from any folder Now Launch : jupyter lab in conda base evironment

  • On local machine:

    1. On your local terminal run the following: 
      ssh -N -L PORT:localhost:PORT user@hpc02.sbms.hku.hk
    2. Go to your local browser and open https://localhost:PORT

    Note, lauch HKU VPN first if you are not within HKU network. Please use https instead of http otherwise you won't be able to use it. Also, Chrome does not work well with this setup. Use Firefox, Safari, or others instead.

  • Alternatively On your local machine, open https://hpc02.sbms.hku.hk:PORT/lab/; however, this may requires the PORT is open from firewall setting, which may requires root permission. You can check this discussion.

Using Jupyter on computing node of a cluster

Comparing to school server (a plain system), running interactive job on cluster can take some extra efforts to set up, but generally doable.

Here, using the Portable Batch System (PBS) as an example, as used in CPOS cluster, we can also run jupyter lab.

  • step 1: submit a job to run your jupyter lab with qsub. You can follow the example cluster-jupyter.sh. Note, you may need to specify your absolute path of the jupyter binary file.

  • step 2: check which computing node it uses via the qstat -n command. For example, it may use compute-0-1.local

  • step 3: connect to the computing node from your local computer. This time you need to know the computing node it uses.

    ssh -N -L PORT:COMP_NODE:PORT USER@xomics.cpos.hku.hk

    Please change $PORT, $COMP_NODE and $USER to your own setting. Then go to your local browser and open https://localhost:PORT