How are people running migrations in these distributed SQL environments? Especially with something like Kubernetes that has canary / rollback functionality. D

How are people running migrations in these distributed SQL environments? Especially with something like Kubernetes that has canary / rollback functionality. Do people run their migrations as a manual step, is integrated into the CI/CD system? I can't find anybody talking about how they handle it in a scalable way. Not having a good DB migration story for dev/prod environments with automatic rollout / rollback makes using a SQL DB a non-starter for me. Most Docker / Kubernetes documentation and their ecosystem assumes the app is fully developed and development doesn't exist. Real world companies develop products and have to perform DB migrations on the production database on a regular basis with each release. What am I missing?