Nebari Distribution Implementation Details

[Adam-D-Lewis]

Nebari distributions were discussed. Several required features of distributions have been implemented including config sets and the ability to specify startup apps for jhub apps in the nebari config. This issue seeks to detail how users will install and deploy a distribution of Nebari.

Steps to deploy a Nebari distribution (e.g. gcp-ragna-distro)

1. A distribution is a python package. You install the package. It will have nebari and any needed nebari plugins defined as dependencies.

• e.g. pip install gcp-ragna-distro

2. They then run the usual nebari commands to initialize the config and deploy it

• nebari init gcp ...
• nebari deploy -c nebari-config.yaml

How to create a distribution

1. All of the distribution use cases listed in the Nebari Distribution discussion can be achieved with just specifying nebari plugins as dependencies as usual in the pyproject.toml file and by specifying nebari config set(s) automatically. Users should still be able to specify their own config set(s) when running cli commands e.g. nebari init --config-set my-config-set.yaml .... I think config sets specified via the CLI should take priority over the distribution config set(s) in the case of conflicts. I also think distributions should have a way to see the final config and throw an error if certain required conditions aren't met.

Code for defining the config set part of a distribution might look like

# gcp-ragna-distro/__init__.py
from nebari.hookspecs import hookimpl

@hookimpl
def config_set():
	# load or define config_set
	config_set = ...
	return config_set

[Adam-D-Lewis]

I think for now a distro should just be the python package way of adding config sets. This encompasses all we need. See PR https://github.com/orgs/nebari-dev/discussions/2934 for progress