Hi all. Could anyone tell me how to spread

REGION: stack.region,
USER_POOL_ID: auth.userPoolId,
IDENTITY_POOL_ID: auth.cognitoIdentityPoolId,
USER_POOL_CLIENT_ID: auth.userPoolClientId,

among lambda functions in SST? My stacks are being called in the following way:

app
  .stack(ApiStack)
  .stack(StorageStack)
  .stack(AuthStack)
  .stack(SocketStack)
  .stack(FrontendStack)

This is a new SST API so I had to use

API Stack

in

Auth Stack

via

sst.use

What is interesting is

API Stack

had being imported in

Auth Stack

to attach permissions. BUT when I am trying to do smth like this:

const lambda = api.getFunction('POST /routeName')

lambda?.addEnvironment(
 'USER_POOL_ID',
 auth.userPoolId
)

I am getting the following error in the terminal

Error: 'dev-ApiStack' depends on 'dev-sst-AuthStack' (dev-sst-ApiStack -> dev-sst-AuthStack/Auth/UserPool/Resource.Ref, dev-sst-ApiStack -> dev-\sst-AuthStack/Auth/IdentityPool.Ref, dev-sst-ApiStack -> dev-sst-AuthStack/Auth/UserPoolClient/Resource.Ref). Adding this dependency (dev-sst-AuthStack -> dev-sst-ApiStack/API/Api/Resource.Ref) would create a cyclic reference.

I gave up on this one.

I haven't found the solution except of deploying both stages to DEV & PROD and preserve outputs in

.env

file. But this is a bad solution because I need to save DEV_variables & PROD_variables + using

env.IS_LOCAL

to distinguish the environment etc.. I remember in SST versions

0.60x

it could be easily done. Hope someone can help me to find a better way

@Ivan Roskoshnyi can you show me what the AuthStack looks like?

The cyclic reference happens b/c: • When u do

auth.attachPermissionsForAuthUsers([api])

, this makes the AuthStack depends on the ApiStack, b/c AuthStack is creating an IAM role whose

resources

field references the Api’s id. • And when u do

lambda?.addEnvironment(
 'USER_POOL_ID',
 auth.userPoolId
)

this makes the ApiStack depends on the AuthStack, b/c ApiStack is creating a Lambda function whose

environment

field references the User Pool’s id. Hence the cyclic dependency. Both stacks need the other stack to be deployed first 😞

Glad to hear Option 2 is a Frank approved solution.

An upside of Option 2 is that when testing the function code, the value for the User Pool ID will similarly be loaded from SSM. As opposed to needing to manually figure out and set the environment variables (ie.

USER_POOL_ID

).

@Frank Thanks for the help! Is there any guide of using SSM with SST?

Would another alternative be to: • move the

.stack(AuthStack)

to the top of the list in index.ts •

sst.use(AuthStack)

in ApiStack & StorageStack and attach the permissions in those stacks instead of using them in the AuthStack. That would make the AuthStack generic and allow you to define how and what resources you apply it to in their respective stacks.