Sharing Lambda layers across stack fails with error: Export EXPORT_NAME cannot be updated as it is in use by STACK_1, STACK_N

[pabacham]

1. Stack 1 has reached the limit of 500 resources per stack, so I decided to create some additional stacks and share the API between them.

export default class Stack2 extends sst.Stack {
  constructor(scope: sst.App, id: string, props: StackProps) {
    super(scope, id, props);

    props.api.addRoutes(this, {
      ...testRoutes,
      ...otherTestRoutes,
    });
  }
}

2. Stack1 has a common PrismaLayer

layers: [
      new lambda.LayerVersion(stack, 'PrismaLayer', {
        code: lambda.Code.fromAsset('./lambda-layers-prisma-client'),
      }),
    ],

3. When I try to deploy the updates, I have got an error
   Stack1 failed: Export Stack1:ExportsOutputRefPrismaLayer958161C6A54DA9DC cannot be updated as it is in use by Stack2 and Stack3

Currently, I can remove Stack2 and Stack3 manually and then update the App, is there another way how to fix the error?

[fwang]

Hey @pabacham, do you get this error when you try to deploy after changing the layer? And can you share the stack export values from stack 1?

[pabacham]

@fwang

1. Yes. Before every deployment, I run "prisma generate" command and then put it as a layer. The idea was taken from here https://dev.to/eddeee888/how-to-deploy-prisma-in-aws-lambda-with-serverless-1m76
2. The only exported value is API. this.api = api; https://monosnap.com/file/Hlh5uTIKpAE4vs7YcIQt6CwCD2kh1t

а) props.api.addLayers method could potentially help me
b) another way of updating and deploying Prisma

[fwang]

@pabacham, when you are sharing a layer across stack, I think the layer's ARN is automatically added as a stack export. If you go to your CFN console, find stack 1, and look at its export values, you should see it there.

Let's just confirm that is the issue.

[pabacham]

@fwang Correct. it is there https://monosnap.com/file/ZpghS66KJhRf94xBGtA0tcx9U3WZxF

[fwang]

Thanks @pabacham, does the output value for the Prisma layer end with a version number?

[pabacham]

@fwang

Stack 1
Status: deployed
Outputs:
ExportsOutputRefPrismaLayer958161C6A54DA9DC: arn:aws:lambda:us-east-1:660857977215:layer:PrismaLayer958161C6:53

Let me know if you mean smth other than that

[fwang]

Hey @pabacham, here's what I found.

What is the issue?

The Lambda layer is created in stack 1 and shared in stack 2. When you share a layer across stacks, CDK creates a stack export with the layer's ARN in stack 1. And everytime the layer changes (ie. when you run codegen), the layer's ARN changes (ie. the ARN always ends with a number). However, CloudFormation does not allow updating exported values. You can read more about this CDK issue here - aws/aws-cdk#1972

How are we going to fix it?

Like mentioned in the above thread, when SST detects a Lambda layer is shared across stacks, it will create an SSM parameter in stack 1 with the layer's ARN. And stack 2 will reference the SSM value, instead of the ARN directly.

I'm working on a fix for this rn.

[fwang]

@pabacham Fixed in v0.35.1. Give it a try!

[Morriphi]

@fwang Does this fix only use SSM when adding layers to each function individually? I was using addDefaultFunctionLayers and noticed it defaulted back to using stack exports. It would be great if we could use addDefaultFunctionLayers and not have the dependency issue.