Hi there! I am trying to implement an API Gateway by using AWS Classic registry (pulumi-aws). I am having some headache with the API Deployment part. I started with this code:

self._deployment = aws.apigateway.Deployment(f"{logic_name}-deployment",
            rest_api=self._args.api.id,
            triggers= self.compute_deploy_triggers(),
            opts=ResourceOptions(
                parent=self,
            )
        )

        self._stage = aws.apigateway.Stage(f"{logic_name}-stage",
            rest_api=self._args.api.id,
            stage_name=self._args.stage_name,
            # deployment=self._deployment.id,
            opts=ResourceOptions(
                parent=self
            )
        )

This is inside a ComponentResource, so the RestAPI and stage name are saved as _args. The issue is that with this configuration, whenever the deployment gets replaced i have this error:

sdk-v2/provider2.go572 sdk.helper_schema: deleting API Gateway Deployment (xxxxx): operation error API Gateway: DeleteDeployment, https response error StatusCode: 400, RequestID: xxxx, BadRequestException: Active stages pointing to this deployment must be moved or deleted: provider=aws@6.83.0

I have tried to work with some pulumi Resource Options, by changing the stage as follows:

self._stage = aws.apigateway.Stage(f"{logic_name}-stage",
            rest_api=self._args.api.id,
            stage_name=self._args.stage_name,
            deployment=self._deployment.id,
            opts=ResourceOptions(
                parent=self,
                depends_on=[self._deployment],
                delete_before_replace=True,
                replace_on_changes=["deployment"],
            )
        )

Doing this solves this error, as both resources are deleted before being re-created. In doing so, I break another piece of code though, because there is a usage-plan that uses the stage and when I replace it the usage plan doesn't react and updates it, therefore the only way is to also update the usage plan like this:

self.usage_plan = aws.apigateway.UsagePlan("usage-plan",
            api_stages=[
                aws.apigateway.UsagePlanApiStageArgs(
                    api_id=api.id,
                    stage=self._stage.stage_name
                )
            ],
            #
            opts=ResourceOptions(
                parent=self,
                replace_on_changes=["api_stages"]
            )
        )

All this seems kinda clunky for what I honestly thought was a trivial thing to implement. I am used to work in Terraform where it is way easier to create the same stuff. Am I missing something obvious maybe? Do you guys have implemented the same architecture in a easier way? Thank you in advance 🙂

That all looks correct, except the UsagePlan should be automatically deleted by Pulumi before the Stage even without your opts, since the

stage_name

property is an output of the Stage resource. Unless you've got some code to change that? Is

_stage

the actual Stage resource, and

stage_name

is an Output?

Thank you for your help @little-cartoon-10569! I think so. The only code part I have actually simplified above is that the stage is inside a ComponentResource and the usage plan is outside of that. Therefore, I have a "get_stage()" method that returns that "self._stage" (which is a _pulumi_aws.apigateway.Stage_ resource) and then I do a:

stage=component.get_stage().stage_name

Inside the usage_plan. Could it be something wrong with that?

If you're returning an output, then it should be good. That's what holds the dependency info.

You mean that the "get_stage()" method should return an Output[T] or that the .stage_name should be ? Just to be sure.

The output that came back from when the resource was created should never be resolved. Keep passing resource output properties around as outputs, and all will be well.