wondering if there's a good pattern for conditionally ignoring updates to resources, for example, based on the presence of an archive. the concrete use case here is updating lambdas in application stack. i don't want to even bother trying to update the lambdas if the archive isn't present on the file system. currently, trying to bring the stack up just fails because it can't locate the zip file for my lambdas. i still want to be able to deploy this stack locally without having to build the lambdas every time. any suggestions?

pulumi works via you declaring your desired state and then pulumi provisioning to reach that desired state. So if you conditionally declare something and it isn't declared, you are just deleting that resource because it is no longer part of your desired state. I think you need to consider reasoning about this from a different direction. If your lambda is deployed under different conditions to the rest of your pulumi project, than perhaps your lambda shouldn't be part of that pulumi project?

i mean, breaking it out into another stack is definitely a solution just adds more noise. the issue here is that lambdas have an environmental dependency on the filesystem, so it's not purely declarative any more.

It is declarative still. You still want the lambdas to exist, and that is why they are in their code. What you are really asking is - "How do I

pulumi up

everything except my Lambdas?", and you may not like the answer but the simple answer is -

pulumi up

them separately

not disagreeing with you, my preferred solution would be to deploy the lambdas as docker images so that it's entirely separate from pulumi, and we could just deploy the same tag if nothing has changed. so i realize that this is a limitation in our choice of deployment model, just having to proliferate yet-another-stack just for these feels bad

I guess my thought is if you are doing this in your CI/CD it's just another couple calls to the CLI. Anyways, good luck with whatever you decide!

def not a problem from automation perspective, i just inherited an aws account with 0 automation so this is a bit of an x/y problem, since ideally i'd never be bringing things up outside of the pipeline at all!