JupyterNotebook
This is not a complete article: This is a draft, a work in progress that is intended to be published into an article, which may or may not be ready for inclusion in the main wiki. It should not necessarily be considered factual or authoritative.
Introduction
Jupyter notebook comes in one Python model on Graham. You can get it working on the login node (not recommended), and the compute nodes (highly recommended). Note that Login nodes impose various user- and process-based limits, so notebooks running there may be killed if they consume significant cpu-time or memory. You will have to submit a job requesting the # of CPU (or even GPU), amount of memory and runtime. Here, we give the instructions to submit a Jupyter job.
Connecting to a manually spawned Jupyter Notebook
Create a Tunnel
To access the notebook running on a compute node from your web browser, you will need to create a tunnel between the cluster and your computer since the compute nodes are not directly accessible from the Internet. To create that tunnel, we recommend the usage of the Python package sshuttle.
On your computer, open a new terminal window, and run the following sshuttle command to create the tunnel
[name@my_computer $] sshuttle --dns -Nr userid@machine_name
Load the module
Log onto graham.sharcnet.ca (or graham.computecanada.ca) or cedar.computecanada.ca and load the python module
ssh user@{graham|cedar}.computecanada.ca
module load python35-scipy-stack
Install Python modules
Refer to the Python documentation.
Submit the job
Create a bash script for submitting a Jupyter job on the slurm scheduler, i.e., slurm_jupyter.sh and add the following. Note do not forget to change the lines with <> since you need to choose your email (line 10) and your account (line 11).
{{File
|name=slurm_jupyter.sh |lang="sh" |lines=yes |contents=
- !/bin/bash
- SBATCH --gres=gpu:1 #number of GPUs if needed
- SBATCH --time=0-01:00 #time in dd-hr:mm
- SBATCH --nodes 1 #nodes
- SBATCH --ntasks-per-node 2 #cores per node
- SBATCH --mem-per-cpu 4000 #mem in MB
- SBATCH --job-name tunnel #name of the job
- SBATCH --output jupyter-log-%J.txt #output file
- SBATCH --mail-type=BEGIN #send email when job begins
- SBATCH --mail-user=<email_to_notify> #to this email address
- SBATCH --account=<account> #account
- load cuda (remove if you don't need GPUs) and python modules
module load cuda module load python35-scipy-stack
- get tunneling info
XDG_RUNTIME_DIR=""
- choose a random port
ipnport=$(shuf -i8000-9999 -n1)
- get hostname's IP and remove whitespaces
ipnip=$(hostname -i | xargs)
- print tunneling instructions to jupyter-log-{jobid}.txt
echo -e "
Copy/Paste this in your local terminal to ssh tunnel with remote
-----------------------------------------------------------------
sshuttle -r $USER@${CC_CLUSTER}.computecanada.ca -v $ipnip/24
-----------------------------------------------------------------
Then open a browser on your local machine to the following address ------------------------------------------------------------------ http://$ipnip:$ipnport (prefix w/ https:// if using password) ------------------------------------------------------------------ "
- start an ipcluster instance and launch jupyter server
jupyter-notebook --no-browser --port=$ipnport --ip=$ipnip }}
Submit the previous script
sbatch slurm_jupyter.sh
Example on Graham
Once you submit the job, and this job has started, you can check the instructions in the log file jupyter-log-*.txt. For example, let's say that you submit a job with a JOBID 131620, then you will have a file jupyter-log-131620.txt which will have something like the following
Copy/Paste this in your local terminal to ssh tunnel with remote
-----------------------------------------------------------------
sshuttle -r jnandez@graham.computecanada.ca -v 10.29.76.27/24
-----------------------------------------------------------------
Then open a browser on your local machine to the following address
------------------------------------------------------------------
http://10.29.76.27:9799 (prefix w/ https:// if using password)
------------------------------------------------------------------
[I 13:58:14.393 NotebookApp] Serving notebooks from local directory: /home/jnandez/jupyter_notebook_HPC
[I 13:58:14.394 NotebookApp] 0 active kernels
[I 13:58:14.394 NotebookApp] The Jupyter Notebook is running at: http://10.29.76.27:9799/
Opening in a web browser
Open your local browser and type,
http://10.29.76.27:9799
Example on Cedar
Once you submit the job, and this job has started, you can check the instructions in the log file jupyter-log-*.txt. For example, let's say that you submit a job with a JOBID 1418055, then you will have a file jupyter-log-1418055.txt which will have something like the following
Copy/Paste this in your local terminal to ssh tunnel with remote
-----------------------------------------------------------------
sshuttle -r jnandez@cedar.computecanada.ca -v 172.16.136.103/24
-----------------------------------------------------------------
Then open a browser on your local machine to the following address
------------------------------------------------------------------
http://172.16.136.103:8975 (prefix w/ https:// if using password)
------------------------------------------------------------------
[I 09:31:44.233 NotebookApp] Serving notebooks from local directory: /home/jnandez/jupyter_notebook_HPC
[I 09:31:44.233 NotebookApp] 0 active kernels
[I 09:31:44.233 NotebookApp] The Jupyter Notebook is running at: http://172.16.136.103:8975/
[I 09:31:44.233 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).
Making the tunneling
Open a new terminal window, and run the sshuttle (only on Linux and MacOS X) command to port-forward the Jupyter port, e.g.
sshuttle -r jnandez@cedar.computecanada.ca -v 172.16.136.103/24
Opening in a web browser
Open your local browser and type,
http://172.16.136.103:8975