AWS cost attribution (2c/4): convert Grafana dashboard into a .jsonnet template and version control it

[consideRatio]

The naming 2x/n is meant to indicate it can be worked in parallell amongside other 2x/n issues

The cost attribution dashboard central source of truth is currently in https://grafana.openscapes.2i2c.cloud where Grafana has it written to a small disk after being developed in the UI. But, we need to deploy it to more places and keep those places updated. So, how do we do that?

Technical context

The https://github.com/jupyterhub/grafana-dashboards project version controlls a few dashboards in .jsonnet format. They are rendered into the Grafana JSON format using the go-jsonnet's jsonnet CLI using the GoLang graffonet library.

Its nicer to work with a dashboard in a .jsonnet format than the big JSON blob it is by default, but its a bit of work to convert the Grafana JSON blob to its equivalent in .jsonnet.

Dashboard ID issue

Currently the documented procedure to deploy the dashboard involves updating a dashboard ID. I think maybe that can be avoided by introducing a dashboard variable that gets autopopulated, and then we reference the variable. I think this is how the jupyterhub/grafana-dashboard function currently with the prometheus datasource - they reference a datasource variable.

Definition of done

• The current openscapes dashboard JSON blob is converted to a .jsonnet dashboard file
• The converted .jsonnet dashboard is verified to look and behave alike the original by deploying it using the deploy.py script from jupyterhub/grafana-dashboards
  Use flags like --dashboards-dir=<a folder where we have the .jsonnet file>, --folder-name=<Some grafana folder name>, --folder-uid=aws-ce-grafana-backend
• The developed .jsonnet file is version controlled somewhere
  Its fine and perhaps suitable to initially do it in this repository.

[GeorgianaElena]

@consideRatio, I am correct to think that the intention is for the jssonet to live in juptyerhub/grafana-dashboards, right?

[consideRatio]

@GeorgianaElena no as that project is part of the jupyterhub org, and its inclusion there means also aws-ce-grafana-backend needs go be available and open sourced properly.

Let this be entirely 2i2c specific for now, but do make use of the jupyterhub/grafana-dashboard's deploy script to render and deploy the dashboard.

[consideRatio]

Should we go for a timeboxed 15 min chat to get started, with the goal to have an empty dummy dashboard to deploy?

[GeorgianaElena]

The definition of done was respected for the first 3 panels. The reaming depend on the existence of a variable that I could not figure out and I've opened to track this work #4944.