Port variable: Single point of truth

[ckunki]

In the initial implementation of the AI-Lab (copied from SLC Sandbox) the Jupyter port 8888 was mentioned literally and redunant in different files: Ansible, Python, Tests, CloudformationStack, Dockerfile, Documentation, ...

Ticket #75 requested to change the port of the Jupyter server to 49494.

AI-Lab Docker Edition

For AI-Lab Docker Edition this could be done the usual way by

• Defining the final value in the Ansible scripts,
• Forwarding it to Ansible facts, and
• Finally using it for port forwardings, welcome message, etc.

AMI and VM Edition

AMI and VM Editions cannot use this approach:

Function run_setup_ec2_and_install_dependencies in file lib/setup_ec2/run_setup_ec2_and_install_dependencies.py

1. Calls run_lifecycle_for_ec2()
2. Before calling run_install_dependencies()
3. Writes a log message about how to connect to the Jupyter server

Ansible is run in step (2) and Ansible facts are available only after step (2).

Step (1) in contrast already needs the port information to create an instance of CloudformationStack wrapped into CloudformationStackContextManager which in method enter() calls the CF stack's method upload_cloudformation_stack().

This finally loads the jinja template templates/ec2_cloudformation.jinja.yaml and could inject the value for the port to avoid redundancies.

Acceptance Criteria

The current ticket requests to

• Define the port in python
• In run_setup_ec2_and_install_dependencies
  • Replace the literal values by the value of the port variable in the log message
  • forward the variable to calls for rendering the template
• For Ansible
  • Add it to ConfigObject
  • Pass it as extra_vars to Ansible in method run_install_dependencies()
  • Use the value in Ansible scripts, e.g. like version: "{{slc_version}}"