Using the Boston University cluster#

These instructions are for Boston University students using the openplaces project on Boston University’s shared computing cluster (SCC).

Connect#

You need to be at Boston University or use the BU VPN.
Open Terminal.
Log onto BU’s cluster via the geo server (Earth & Environment):
```
ssh your_username@geo.bu.edu
```
You will be prompted for your password and two-factor authentication.

Note

When accessing restricted data, you need to log onto scc4

ssh your_username@scc4.bu.edu

Create your copy of the code#

Navigate to the openplaces user code directory:
```
cd /project/openplaces/code/user
```
Create a new folder with your BU user name and change into it:
```
mkdir your_username
cd your_username
```

Clone the openplaces Github repository:

git clone https://github.com/chrnolte/openplaces.git

Change into the repository directory:
```
cd openplaces
```
No need to run the setup script (Install) unless you want to manage your own Anaconda environment.
Create a symbolic link to your repository. Call it openplaces and leave it in your user directory (~):
```
ln -s /project/openplaces/code/user/your_username ~/openplaces
```
Next time you log onto the server, you can type cd openplaces to change to your repository directory.

Start Jupyter notebook#

Using OnDemand#

Log onto SCC OnDemand: https://scc-ondemand.bu.edu.
Select Interactive Apps > Jupyter Notebook.
In the form that opens, edit these fields:

List of modules to load (space separated)
miniconda/25.3.1 (remove python if present)

Pre-Launch Command (optional):
conda activate /projectnb/openplaces/.conda/envs/openplaces

Interface
notebook

Working Directory
/project/openplaces/code/user/your_username/notebooks

Number of GPUs
0 (unless you need GPUs)

GPU compute capability
0 (unless you need GPUs)

Project
openplaces

Check Save settings, so you don’t have to enter these again.
Launch the request.

Wait until your interactive session has been created.

After a while, a new button will appear: Connect to Jupyter. Click it!
Continue with Set up your data folders.

Using the Terminal#

Activate the openplaces environment:

module load miniconda
conda activate openplaces

Go to the notebooks folder:
```
cd notebooks
```
Start a session of Jupyter without the browser but with a port:
```
jupyter notebook --no-browser --port=7049
```
Change the port number if somebody else is using it.
In a second, new Terminal / Prompt window, tunnel to the same port you used in the previous command:
```
ssh your_username@scc4.bu.edu -L 7049:localhost:7049 -X
```
- The -L argument creates the tunnel from the internal port of the SCC to the port on your machine. If you use the same port number on both sides, you can copy & paste the displayed Jupyter link in your browser.
- The -X argument is optional. Use it if you want to use graphical user interface (GUI) applications on the cluster (e.g. interactive backends for matplotlib, QGIS). They will be more responsive. It’s not needed for static matplotlib plots.
Now you can copy & paste the link from the first Terminal into the address bar of your browser.

It will start with http://localhost:7046/tree?token=....

This should open the familiar Jupyter landing page, generated by Jupyter on the computing cluster.

You can now open a Jupyter notebook and start coding.
Continue with Set up your data folders.

Caution

Keep both Terminal windows open while you work on the cluster. If you close the tunnel, you can’t interact with the Jupyter notebook anymore.

Shortcut#

You can edit ~/.bashrc and give this series of commands an alias:

openplaces() {
  conda activate openplaces
  cd /project/openplaces/code/user/your_username
  jupyter notebook --no-browser --port=7049
}

This way, after logging onto the cluster, you just type openplaces and get an active Jupyter session.

Set up your data folders#

Upon the first call of import openplaces, a configuration script is launched.

See configuring your installation for the full explanation.

On the cluster, you will have to pick this configuration:

Root directory for data, models, and reports
/projectnb/openplaces/earth

Choose your configuration mode
b (Multi-user)

Directory paths
a (Choose defaults)

Transfer files#

Using OnDemand#

Log onto SCC OnDemand: https://scc-ondemand.bu.edu.
Browse the directory structure in Files.

Using an FTP client#

To move files between your computer and the cluster, use an FTP client (like a Finder / Explorer for managing remote files).

FileZilla is a great free cross-platform option and a good pick for Mac users.

WinSCP is a good option on Windows.

Connection settings#

File protocol: SFTP
Host name: scc4.bu.edu
Port number: 22
User name: your_username
Password: Leave this empty and use your Kerberos password only when prompted (safer)
Local folder: path/to/your/code
Remote folder: /project/openplaces/code/user/your_username