Here's my main function: queue.onEvent('send-to-stackdriver', new aws.lambda.CallbackFunction("stackdriver-writer", { memorySize: 512, // this is an excessive amount of RAM, but its running very slow with less. callback: (e) => { logger.logger(e, 'client-events') }, environment: { variables: { gcpKey: keyFile, projectId: projectId } }, policies: [ 'arnawsiam:awspolicy/service-role/AWSLambdaSQSQueueExecutionRole' ], runtime: "nodejs10.x" // required to use NodeJS 8.13+ runtime for GCP SDK } ))

and the logging function it calls: import * as Logging from "@google-cloud/logging"; export async function logger(event: any, logName = "client-events") { const keyFile = JSON.parse(

${process.env.gcpKey}

) const projectId =

${process.env.projectId}

const logging = new Logging.Logging({ projectId: projectId, credentials: { client_email: keyFile.client_email, private_key: keyFile.private_key }}); const log = logging.log(logName); const metadata = { resource: { type: "global" } }; event.Records.forEach(async (record: { body: {} | undefined; }) => { const entry = log.entry(metadata, record.body); await log.write(entry); console.log(

Logged: ${record}

); }); } just one dependency

so we have a way to help out here.

Here's the way Pulumi works (and we desperately must documentthis somewhere).

1. when you serialize a JS function (like you have with

callback: (e) => { logger.logger(e, 'client-events') },

) by default we include all your node_modules brought in by your package.json.

This is clearly not optimal in many scenarios (but it makes things super simple for common cases).

To improve this, we allow you to do the following:

1. in your package.json file add a new

"pulumi": { ... }

section.

in there, add a

"runtimeDependencies": { ... }

where you list only the node_modules that you want to be sucked up into the cloud.

note: this will still need to be in your normal

"dependencies"

section so that tools like npm/yarn know about it and pull it down.

so, based on your description, it should be sufficient to just pull in

@google-cloud/logging

.

Note: this will impact all lambdas you create. If you want to exercise fine-grained control over your individual lambdas, you can do so when calling

new aws.lambda.CallbackFunction("stackdriver-writer", {

There is a

.codePathOptions

property that can be used to flexibly control things here.

@swift-painter-31084 Let me know if this helps you out!

did it not work though? what's in the final uploaded zip?

oh. i see the problem

we respect this sectoin in referenced packages, but not your own.

for now, it's likely you'll have to setup the

codePathOptions: {     extraExcludePackages... }

yourself

😞

sorry

we should make this work for your package as well, i'll file bug

Hard to say. I don't have experience with lerna myself

it should be able to give intellisense.

i believe you should be able to do

codePathOptions: { extraExcludePackages: ["@..."] }

if that doesn't work, i'll have to investigate

(where you fill in the part in the quotes btw)

i believe we literally check the names of the properties in your 'dependencies' section, i.e.

so you'd put in

"@google-cloud/language"

ideally we'll make the former method work as well

so you don'thave to have this big goop for all your functions