I'm interested in using Fargate to run a cluster with multiple containers written in different languages. I've got a basic setup working but it's all in one repo. I'd like to move each container into a different repo and still have Pulumi 'do the right thing'. What is the right way to do this?

Do you have everything in one stack as well?

Currently, yes. Should I break that up?

I use a monorepo and don't use pulumi to build my containers. Rather, pulumi takes as input the image + tag for containers I want to deploy. When I want to deploy new containers to an environment/stack, I just run

pulumi up

with new config values. Having pulumi build your containers makes things qutie a bit harder IMO. I'd probably have a pulumi repo which uses submodules to pull in each container repo. Pulumi really wants you to have one program to do your deployment - so spreading your pulumi program across all those repos might be really painful (or result in a lot of code duplication).

So separate the infrastructure from the containers and link them together using references to containers via a image repository. Do image repos have webhooks to trigger the 'up' call?

I'm not sure. I use a CodePipeline to deploy when code is merged to master, so the

pulumi up

process is just part of the CI/CD pipeline.

Good to know, I'll look into that