Define a way to roll out updates/db migrations without downtimes

[Pitasi]

We need to figure it out a way of rolling updates into prod without downtimes for the users.

The most difficult upgrades are the ones involving a change in the database schema. We currently are aware that we have some microservices anti-patterns in place (i.e. sharing databases: tracelistener writing to a db that it's read by api-server, etc.).

I'm adding this ticket to the backlog so we can brainstorm some solutions and best practices that will improve our workflow.

[Pitasi]

I found this HN discussion interesting: https://news.ycombinator.com/item?id=30899362