This directory contains the Ansible playbook to set up a virtual machine (VM) with a JupyterHub, for use within the EUCP. The VM tested has been a Ubuntu 18.04 (LTS) system. The current setup uses one running in the SURF HPC cloud.
Run as follows:
ansible-playbook hpccloud-playbook.yml
Note that the ansible.cfg template file in this repository directory
points to the Ansible inventory, or specifically, the hosts file (a
template file is given in this directory), which sets the host(s) to
run the installation on. Update hosts as necessary for your
situation; you can use host aliases defined in your $HOME/.ssh/config
file.
Running the playbook performs the following steps:
-
Installation of some standard (Ubuntu) packages
-
Creation and mounting of data storage virtual disks
-
Downloading and installation of MiniConda
-
Installation of JupyterHub, JupyterLab and a few extensions (for Python version 3.6)
-
-
Setup running for JupyterHub under a separate account
-
Sudo access for JupyterHub group to spawn a JupyterLab for each account
-
Systemd setup for JupyterHub
-
-
Creation of user accounts with random passwords
The creation and mounting of data storage uses an Ansible variable
file, mount_vars.yml, to match devices and mount points. Every
device is xfs formatted and mounted under /mnt/data/<dir>/. Create
the mount_vars.yml file from the mount_vars_template.yml template.
If you don’t need any data disks mounted, provide an empty list:
mounts: []
The Ansible playbook can automatically create users with a random
password and with the correct groups. The user name and corresponding
main group are listed in the user_vars.yml file as a list of
mappings. Copy the user_vars_template.yml to user_vars.yml and
edit it accordingly. If you don’t need any users right now, make the
users variable an empty list:
users: []
Note: when adding users yourself, make sure that their (secondary)
groups are users and jupyterhub, otherwise new users will not be
able to start a JupyterLab.