Stress testing of the configurable-http-proxy and of JupyterHub.

Build the container:

invoke build_worker

You'll also need to have the jupyter/configurable-http-proxy image downloaded.

Start proxy listening:

invoke proxy

To stop the server and clean up, run:

invoke cleanup_proxy

Build containers:

invoke build_fds

Run fd exhaustion test:

invoke stress_fds <hostname> <n> --proxy=<ip>

where <hostname> is an SSH-friendly hostname (i.e., it can be an alias), <n> is the number of workers to start, and <ip> is the IP address of the proxy (probably the IP address of <hostname>).

The first time you run the server processes, you'll need to run:

invoke setup_server

After that, you shouldn't need to run that again. Next, build the necessary JupyterHub image:

invoke build_jupyterhub

Now, run the restuser service:

invoke restuser

Finally, run JupyterHub:

invoke jupyterhub

To clean up the JupyterHub docker images and any users that were created, run:

invoke cleanup_jupyterhub

Build containers:

invoke build_hub

Run stress test:

invoke stress_hub <url> --port=<port> -N=<num_users>

where <url> is the url of the host where JupyterHub is running and <port> is the port number that JupyterHub is running on. The <num_users> argument is for the number of users to test simultaneously (maximum 100). You can also pass a --hubid argument if you want to run multiple stress processes at once.