Install GitHub Desktop.

• Install ( https://desktop.github.com/ )
• Login to GitHub with the app
• Quit
• Restart
• This should configure your ssh key for cloning of private repos.

Setup passwordless ssh to nodes on the cluster (required for job submission).

• Connect to the cluster ssh login1.int.janelia.org.
• Create a key and setup passwordless ssh for launching jobs by running this ssh-keygen -t rsa && cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys and hitting enter all the way through.
• Copy everything returned by this cat ~/.ssh/id_rsa.pub.

Configure the cluster

• Connect to the cluster ssh login1.int.janelia.org.
• Run bsub -Is -tty /bin/bash.
• Run our configuration file curl https://raw.githubusercontent.com/nanshe-org/nanshe_workflow_deploy_janelia/master/1_preinstall_nanshe_workflow.sh | bash.
• Alternatively, this can be run in the background echo 'curl https://raw.githubusercontent.com/nanshe-org/nanshe_workflow_deploy_janelia/master/1_preinstall_nanshe_workflow.sh | bash' | bsub. Check to make sure it is done by running bjobs and verifying STDIN is no longer in the job list returned.
• Logout of the cluster. All changes will take effect on the next login.

Installing/Updating

• Connect to the cluster ssh login1.int.janelia.org.
• Run this singularity build ~/nanshe_workflow.simg docker://nanshe/nanshe_workflow.
• Alternatively, this can be run in the background with bsub singularity build ~/nanshe_workflow.simg docker://nanshe/nanshe_workflow. Check to make sure it is done by running bjobs and verifying the job is no longer in the list returned.
• If this fails for any reason, try opening GitHub Desktop. Once it is fully loaded close it. Then, open a new Terminal or Git Bash (GitHub). Finally, rerun these steps.

Running

• Connect to the cluster ssh login1.int.janelia.org.
• Switch to a directory with data using cd.
• Run bsub -Is -tty /bin/bash and then run singularity run --bind /misc --bind /scratch ~/nanshe_workflow.simg.
• Alternatively run the bsub singularity run --bind /misc --bind /scratch ~/nanshe_workflow.simg and then wait a little and run bpeek.
• It will take a little bit to start then it will print a bunch of stuff to the screen and hang (this is intentional).
• Copy the URL from the output (removing (/) and or 127.0.0.1 from it) and enter this into your browser.
• Select nanshe_ipython.ipynb.

Shutting down

• Save and close your notebook.
• Use Ctrl+c back to the login node.
• Run bjobs and terminate any remaining jobs with bkill.

Security

• Everything I just told you can still leave you vulnerable!
• An improvement on this procedure is to create a password to access the notebook.
• To do this run the following python -c "from IPython.lib import passwd; print passwd()".
• Enter a password ideally different from your existing one, twice.
• This will return something like this sha1:HHHHHHHHHHHH:HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH where H is a hex number.
• Go into the file ~/.ipython/profile_default/ipython_notebook_config.py on the cluster and find the line with c.NotebookApp.password.
• Uncomment it and place the full string in there. It will look like this c.NotebookApp.password = u'sha1:HHHHHHHHHHHH:HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH'
• Now if someone tries to get into your notebook, they will have to enter a password.
• However, this password is going over the wire unencrypted (in an SSH tunnel). If this is a concern for you, create a certificate.
• Run the following, openssl req -x509 -nodes -days 365 -newkey rsa:1024 -keyout ipython.pem -out ipython.pem.
• This will give you a certificate that will last for roughly a year. It contains the key and certificate data.
• To use this certificate, go into the file ~/.ipython/profile_default/ipython_notebook_config.py on the cluster look for the line with c.NotebookApp.certfile.
• Uncomment replace that line with this c.NotebookApp.certfile = u'ipython.pem'.
• This certificate will be unsigned, which will make your browser very upset, but it can be coerced into acceptance.