Connection Parameters conn_params

App developers can use conn_params in the submit.yml.erb to pass runtime data generated in the before.sh.erb back to the view.html.erb.

This is helpful for:

  • Data that is only known after the job submits and starts running.

  • Data that needs to be used to connect to the application.

This technique will generate a file in the jobs working directory called connection.yml when the app launches which will contain the defined variables and their associated values.

Jupyter Notebook Example

Here's an example using a Jupyter application which needs needs to know the exact API to connect to. We can either connect to JupyterLab at /lab or Jupyter Notebook at /tree, but this information is not known until the job has been submitted.

So once the job is submitted, we need to export the jupyter_api environment variable that can then be written to connection.yml which OnDemand will consume and use in the view.html.erb.

Warning

The environment variables in before.sh.erb must be lowercase and exported through the export function.

# within template/before.sh.erb

JUPYTER_API="<%= context.jupyterlab_switch == "1" ? "lab" : "tree" %>"

export jupyter_api="$JUPYTER_API"

Now with that variable exported, you need to add it to conn_params in submit.html.erb to ensure that OnDemand makes use of it.

In the view.html.erb, which renders after the submission in the interactive apps page, you can access the value of this variable with:

<%-
  ...
  next_url  = "#{base_url}/#{jupyter_api}"

  full_url="#{login_url}?next=#{CGI.escape(next_url)}"
  ...
%>

...

<form id="<%= form_id %>" action="<%= full_url %>" method="post" target="_blank" onsubmit="changeTarget()" >

This configuration uses the value from jupyter_api to populate the action attribute in the rendered page in the interactive sessions.