Inconsistent UX for multi-stack cdktf commands

[daniel-laszlo]

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave "+1" or other comments that do not add relevant new information or questions, they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

cdktf & Language Versions

cdktf 0.13.2

Affected Resource(s)

cdktf synth
cdktf diff
cdktf deploy

Debug Output

Expected Behavior

Can we standardize options for handling multi-stack workflows? I need to be able to run all subcommands for 1) a specific stack, 2) for all stacks.

Actual Behavior

• cdktf synth runs for all stacks without supplying any further options, but I cannot specify to run for only one stack
• cdktf diff requires specifying a stack, but there is no option to run for all stacks
• cdktf deploy requires specifying a stack or '*' to run for all stacks

Steps to Reproduce

Important Factoids

References

[DanielMSchmidt]

Yeah this part is a bit confusing and we don't really call the reason it explicitly in our documentation, so we should definitely add a section.

Synth synthesizes all stacks because synthesizing means running the app command. The command is arbitrary, but most commonly we run ts /python/ etc to execute your cdktf program. We can't run half a program that easily, so that's why we run it entirely.

When we built the multi-stack feature we wanted to allow to plan multiple stacks as well, but that's not really feasible right now. With cross stack references you might use a value from one stack in another. If we didn't apply yet or the changes would change that referenced value we can not make a (correct) plan. This is why we limit the plan to one stack.

[daniel-laszlo]

Thank you for the explanation

[daniel-laszlo]

@DanielMSchmidt followup question. Would I lose the ability to run plan for a stack which references a value from another stack? And then I would be restricted to only deploy in this case?

[DanielMSchmidt]

No, you can plan the stack housing the value, apply it and the plan the stack using the value. If you plan the stack using the value first your plan might be incorrect / erroring

[LyleDavis]

@DanielMSchmidt in the scenario of running a plan on a stack where it depends on another stack that hasn't yet had the relevant changes applied, is it guaranteed that it will error or will it sometimes show an incorrect plan without erroring?

I'm new to CDKTF and terraform in general, but in CI I was hoping to be able to run a diff on all stacks (if not together then one by one) and then deploy * on merge. Obviously that would be dangerous if there is the potential for the diff to be wrong due to stack dependencies.

[fathom-parth]

@DanielMSchmidt a tiny bump if you have time to answer @LyleDavis's question! I'm setting up cdktf for my organization and trying to make sure our CI/CD setup is optimal.

I think given this issue, the issue at #2157, and the cdktf best practices telling us to use multiple stacks over a single stack per-environment/project (we use gcp), there's no real way to run cdktf diff during CI?