Turborepo Deployment Best Practices

[bestickley]

Summary

I have a monorepo managed by turborepo. In this monorepo are separate microservice and web app packages. I'm thinking about using turborepo to manage my deploy scripts which deploy microservices and web apps to AWS (via CDK). My current thinking is that I could achieve doing this in 1 pipeline. However, I'm also aware that many recommend using a dedicated pipeline per microservice or app. For those with experience here, what do you recommend?

I like the idea of turborepo managing deploy script caching and cache invalidation for relationships between packages. For example, if web app A depends on microservice B and there is a code update to microsevice B then turborepo will take care of running the deploy script for microservice B and then web app A. Deployments will be fast - only deploying what's changed and only using 1 pipeline also keeps things simple. My first question: is anyone doing this? Seems like turborepo is often used for dev tasks but I don't see discussions around deployment commands. Any tips/best practices for using turborepo for deployments?

Secondly how are you achieving automatic rollbacks? I know this may be company/team specific, but has anyone achieved this with turborepo task controlled deployments? Can I try/catch a script and then deploy the previous known version if it fails? Also, how do I rollback across packages? To use previous example if microservice B succeeds but then web app A fails how can I rollback microservice B? Thank you!

Additional information

No response

Example

No response