Run the services that make up the JupyterHub backend. It includes multiple instances of JupyterHub, load-balanced by HAProxy. All services are run in docker containers. Notebook servers are launched on the Rancher Docker platform. Users are pre-authenticated on the frontend Apache server through web-SSO (shibboleth).

• Set the environment variable JHUB_SETUP_DIR and point it to this directory.
• Install docker in the latest version. See doc/README_docker for tips about docker
• If you have not yet generated a hub api token call bin/generate_hub_api_token.sh and save the string printed on the screen in the variable 'api_token' in section 'JUPYTERHUB' in etc/config.ini
• Create HAProxy configuration file by calling bin/generate_hub_api_token.sh You need to do that only once.
• Adjust configuration file etc/config.ini as needed. In particular the following parameters need to be adjusted: RANCHER section: env_id, rest_base_url, access_key, secret_key JUPYTERHUB section: hub_ip, api_token
• Make sure the local firewall allows connections from the frontend server and from the docker VMs (cows) as described in doc/README_iptables

Launch services by e.g. calling

bin/manage_service.sh start nbmon hub lb

Stop services by e.g. calling

bin/manage_service.sh stop lb hub nbmon

(Find out options by calling bin/manage_service.sh without any options)

• The jupyterhub frontend web server must be able to connect to the backend on port 80
• All docker VMs (Rancher lingo: cows) must be able to connect to all hub ports and hub api ports (not the proxy api ports though)

See doc/README_iptables for potentially useful testing commands before puppetising the rules