Question

Your team uses Google Cloud Deploy to release a Java microservice to GKE across dev, staging, and prod targets. Before each new version's pods receive traffic, a database schema migration script must run exactly once against the target environment's Cloud SQL instance, and the rollout must halt if the migration fails. You want to implement this within the delivery pipeline using native Cloud Deploy capabilities. What should you do?

Accepted Answer

Add a pre-deploy hook to the Cloud Deploy pipeline that runs the migration in a container image; if the hook fails, Cloud Deploy marks the rollout as failed.

Answer

Configure a Cloud Build trigger on the same branch that runs the migration in parallel with the Cloud Deploy release so both complete around the same time.

Answer

Embed the migration command in the application container's entrypoint so it runs when the first pod starts up in each environment.

Answer

Create a separate Cloud Scheduler job that invokes the migration script a few minutes before each scheduled promotion to the next target.