Github action running cicd pipeline with already existing/provided credentials azure machine

[cbr7]

It would be good to have some way to execute cicd pipelines on an already existing/not spot instance azure machine by providing the IP and login credentials to the job and the job connecting to the machine instead of creating a spot instance. This would have several benefits, it would allow test debugging to be much faster, since we would no longer have to wait for spot instance availability during test debugging and also allow manual connection to the azure instance in order to observe/investigate issues locally on the machine.

[odockal]

Possible approach:

• Have a job without Create and destroy windows machine step that runs on workflow_dispatch and accepts host (password passed via inputs does not seems like secure option) via input param with other inputs (as usual: branch, fork, repo, etc.) and that uses a gh repo secret (see below) REMOTE_SSH and possibly REMOTE_PASS
• Create machine manually/locally (gets host, pass, etc.)
• Use the secrets produced and update the gh repo secrets
• Run the above workflow to debug on particular machine without a need to wait for machine to spin up
• Finish you debugging
• Destroy windows machine locally

[odockal]

The issue can be closed once we verify that the workflow works as expected. Workflow is at place, so we can actually test it.