phillipedwards
commented
1 year ago I took a brief look at what exists today in this repo and what it would take to add additional providers such as AWS, GCP, etc. I don't think it's too big of a lift, but there are a few questions that come to mind.
For starters, identifying and lifting non-provider specific code into a common module(s) is probably a good first step. The unknown in my mind is how each provider specific task would interface with the the provider and that is mainly concerned with how values from the ADO environment are made available to each provider.
Generally speaking there are two answers to the above question and they are:
- environment variables
- pulumi stack configuration
I think both have their downsides and upsides but we also have to consider how default/explicit providers might impact that decision.
Looking at the aws-toolkit-azure-devops there is a clear direction we can follow as to how support different providers (ec2, cfn, etc) although we do not need to be as complex as the linked repo.
The pipeline task supports azure service connections but does not support AWS or other providers.
Ask: the pipeline task should support Service Connections for other providers - e.g. AWS.