CI/CD of Satellite UI: Automated Production Deployments

[ferristocrat]

Description:

In Scope:

Out of Scope:

What is the problem/pain point?

What is the impact?

Why now?

Links:

Acceptance Criteria:

Milestones:

• https://github.com/storj/storj/milestone/42

[mobyvb]

The bulk of this will probably be setting up "manual frontend deployments", and automating it will be easy, and just dependent on #60 to switch from manual to automatic
This doc outlines the notes we took about building up to CI/CD, for reference
This infra PR is a proof-of-concept for what infrastructure change we would need to separate out a container for the UI from the main backend deployment. It can probably be used as a basis for a lot of this work.

Steps:

1. Do something like in the infra PR above to separate the UI from the backend infrastructurally
   • AC: both backend and UI containers still behind the same domain/load balancer (e.g. us1.storj.io); however, UI is served from one container, and a different container handles backend requests, e.g. metainfo/uplink, calls to /api endpoints from UI
2. Discussion to determine other requirements:
   • discussion with QA what a "manual UI deploy" outside of a scheduled backend deploy could/should look like, including who needs to approve, who is allowed to conduct the deploy, and what level of manual testing is necessary for UI-only deployments (this discussion should also help us with some pending tasks for CI/CD of Satellite UI: Automated UI Tests #60)
   • discussion with Console about what habits need to be formed, what automated enforcement we can implement, etc... with the goal being to prevent any issues due to a more-up-to-date frontend being run against a more-out-of-date backend
     • what are all the possible issues that could arise due to a newer version of the frontend? (e.g. calls to endpoints that don't exist yet on the backend). Also consider a more up-to-date backend and a more out-of-date frontend. There is a possibility that if we make backend vs. frontend deployments completely independent, that both situations could occur.
     • what does the team need to account for in code changes and reviews in order to prevent the issues from the above point?
     • can we build any automatic checkers/linters etc... to prevent backwards incompatibility issues?
3. Separate frontend and backend deployments - details of this step are largely dependent on the discussions above, so I won't try to guess what it will look like here
4. When CI/CD of Satellite UI: Automated UI Tests #60 is finished we should be able to turn on continuous deployment. However, we should start on the above tasks asap, regardless of the status of CI/CD of Satellite UI: Automated UI Tests #60

cc @ferristocrat @NiaStorj @jewharton

[NiaStorj]

Manual deployments - 2- 4 sprints
Automated deployment testing will be blocked by #60 and manual deployment steps. Once finished automated deployment would be able to start. ROM 2-3 after being unblocked my #60 and manual deployments.