i have a project called

infrastructure

, which i successfully spun up. in this project is a Kubernetes cluster. in the script i used to create it, I exported a whole bunch of outputs which i will need to reference in other projects. this code is from the next project i want to spin up, where i try to reference the values of the first project. my teammate - who has since left the company 😒ad-panda: - created this handy function which passes your logic into an `apply`:

export const useClusterAttributes = (infraStackName: string, func: (outputs: ClusterAttributes) => any) => {
    const stackRef = new pulumi.StackReference(infraStackName)
    stackRef.getOutput('clusterName').apply(clusterName => { console.log(`The clusterName should be a string here, right?: ${clusterName}`) })
    return pulumi.all([
        stackRef.getOutput('vpcId'),
        stackRef.getOutput('publicSubnetIds'),
        stackRef.getOutput('privateSubnetIds'),
        stackRef.getOutput('clusterName'),
        stackRef.getOutput('kubeconfig'),
        stackRef.getOutput('clusterOidcProviderArn'),
        stackRef.getOutput('clusterOidcProviderUrl'),
    ])
        .apply(([
            vpcId,
            publicSubnetIds,
            privateSubnetIds,
            clusterName,
            kubeconfig,
            clusterOidcProviderArn,
            clusterOidcProviderUrl,
        ]) => {
            console.log(`At least should be a string here: ${clusterName}`);
            func({
                vpcId,
                publicSubnetIds,
                privateSubnetIds,
                clusterName,
                kubeconfig,
                clusterOidcProviderArn,
                clusterOidcProviderUrl,
            })
        })
}

i added those

console.log

statements, we'll get to them in a sec.

this is how you use this function:

useClusterAttributes(infraStackName, async ({
    vpcId,
    clusterName,
    kubeconfig,
    clusterOidcProviderArn,
    clusterOidcProviderUrl,
}) => {
    const k8sProvider = ensureProvider(clusterName, kubeconfig)
   ... declare more resources down here ...
})

this is an abstraction away from a simpler

apply

syntax. the goal of

userClusterAttributes

is to prevent us from rewriting

pulumi.all

and

apply

commands for the same outputs again and again. please let me know if anyone needs clarification on this function and i'd be happy to explain more!

the real problem is that all of my Outputs -

clusterName

,

vpcId

, etc. - all evaluate to

undefined

. so when i try to create that

k8sProvider

, i get this error:

Error: Missing resource name argument (for URN creation)

and my

console.log

statements from earlier?

hello orgname/infrastructure/stackname
    OutputImpl {
      __pulumiOutput: true,
      resources: [Function (anonymous)],
      allResources: [Function (anonymous)],
      isKnown: Promise { <pending> },
      isSecret: Promise { <pending> },
      promise: [Function (anonymous)],
      toString: [Function (anonymous)],
      toJSON: [Function (anonymous)]
    }
    The clusterName should be a string here, right?: undefined
    At least should be a string here: undefined

can anyone tell me why all my

stackRef.getOutput

Outputs are evaluating to

undefined

?

btw, i did confirm that the outputs are actually defined when they are exported from my first project, `infrastructure`:

you don’t need to do an

.apply

on the stack ref

take these:

stackRef.getOutput('vpcId'),
        stackRef.getOutput('publicSubnetIds'),
        stackRef.getOutput('privateSubnetIds'),
        stackRef.getOutput('clusterName'),
        stackRef.getOutput('kubeconfig'),
        stackRef.getOutput('clusterOidcProviderArn'),
        stackRef.getOutput('clusterOidcProviderUrl'),

outside the apply

ah no sorry, I misread

so previously the exports were done with the name

process

in front of them, e.g.

processClusterName

,

processVpcId

, etc. I rewrote them all to not have

process

, and I am searching my codebase and there is no instance of anything that says

processClusterName

anymore. also, i'm pasting the screenshot i showed above from my terminal when i ran

pulumi up

on my stack. it wrote that the outputs were just

clusterName

, no

process

.

BUT, i just saw in my web browser, when looking at this stack, that there are still outputs called "process"!

i changed my code to search for

processClusterName

and it actually found something. no more

undefined

🎉 so now my question is: why don't my terminal's outputs match what appears to actually be running?

should i tear down this whole stack and rebuild it? is that necessary whenever you change your outputs?

(this will take a while, the Kubernetes clusters are slow to spin up)

OMG you know what. i think i know what happened. i feel absolutely silly 🤦‍♂️ if i'm right, it's all a stack mixup. total user error 😅 we will confirm once this

pulumi up

is done

ok, it's not a stack mixup, but i see what happened. i missed an error when deploying the

infrastructure

project before. i thought i had wiped out all those old outputs, but i missed that the deployment failed. as always, the issue lied between the chair and the keyboard 🤦‍♂️ thank you for your help @billowy-army-68599!! you walking me through it led to finding this, so you definitely contributed here. i should be able to straighten this out now that i have a handle on things.