Spike: Infrastructure + DevOps of SSR Migration #14

Choose a provider

• Cloud Run looking like the best choice but investigate if a constant server would be better / worth it. (For caching a consistent server would be better)
• Look into Vercel (people say it's the best for NextJS apps). Is it worth the extra setup there? How would it be better?
• How powerful a provider instance do we need?

Session management:

• Client session should propagate to the server-side process

CICD

• Create a branching strategy keeping SSR changes isolated from the current production UI source code
• Deploy application to a "new" domain keeping the main domain "clean" until code and CI pipeline are ready for production
• Assure that the application can: build, lint, test (unit and e2e), set ENV variables from Github Actions
• The procedure to generate 'preview-links'. The current idea is to create a Cloud Run instance for the pr, then on PR close -> delete the instance.

Migration

• What would be the smoothest way to transfer mobilitydatabase.org to the new provider?
• How can we have a mechanism to roll back to the previous hosting if things do not go well?

Observability

• When a provider is chosen, how could we see if things are running normally? Ex: If cloud run, see if something is poorly configured and would cost a lot.