https://pulumi.com logo
Powered by
p

polite-kite-18322

08/11/2023, 1:09 AM
I'm really struggling with how to create an aws Lambda Archive that does NOT include the folders between the "workdir" and the files i'm trying to include in the archive. Consider this folder structure:
Copy code
pulumi.yaml
pretokengeneration/dist/{bunch of files to archive}
What I need is to create an archive that ONLY includes the files in 
pretokengeneration/dist/
without the parent folder structure. Hoisting the files to the root of the zip file. I keep getting this structure that includes all paths from "work dir" to where the files are and it blows up in aws lambda. I've scoured the internets... and pulumi community. For some reason this is either so obvious that no one has ever asked this question, or I'm doing something really stupid.
I've tried many combinations of prefixes and values for 
dir
to see if I could generate the desired zip file but to no avail:
Copy code
const lambdaBuild = local.runOutput({
		dir: "pretokengeneration",
		command: `cd src && npm install && tsc`,
		archivePaths: ["dist/**", "!dist/**/test/**"]
	});

...
     // then I provide it to pulumi.aws.Function like this
     code: lambdaBuild.archive,
@aloof-bear-23820/ @proud-winter-66045 looks like no traction 😞
b

brave-planet-10645

08/11/2023, 4:02 PM
Hi Bruce, is there any reason you can’t do this:
Copy code
const fn = new aws.lambda.Function("fn", {
    code: new pulumi.asset.AssetArchive({
        ".": new pulumi.asset.FileArchive("./dist")
    },
    ... rest of lambda
});
I think that should work
p

polite-kite-18322

08/11/2023, 4:02 PM
I saw that usage elsewhere and tried it. Pretty sure it didn't work as I'd hoped. I can try again in a bit...
and thank you for responding!
b

brave-planet-10645

08/11/2023, 4:04 PM
That’s how I usually do this
p

polite-kite-18322

08/11/2023, 5:15 PM
Still showing as nested in lambda. Is there a way to inspect the generated archive locally without having to upload to aws during debugging?
b

brave-planet-10645

08/11/2023, 5:25 PM
You’re using what looks like JavaScript/typescript to define your resources. Use those tools to read the directory structure and you can log them to the screen
So I think it’s the archive paths input in the command resource that’s doing this. Can you not use that and use the assets.archive bit in the function resource to zip it up?
p

polite-kite-18322

08/11/2023, 5:59 PM
I am using Typescript. In the debugger I see this. Not sure what "use those tools to read the directory structure" means?
The pulumi approach of using but not using Promises is quite confusing.
pulumi.log.info() is demanding a "Resource" but these operations provide 
FileArchive
so not sure how to understand what they actually resolve to.
@brave-planet-10645 I can find "archivePaths" in my Typescript, but you are losing me with the rest of this statement:
So I think it’s the archive paths input in the command resource that’s doing this. Can you not use that and use the assets.archive bit in the function resource to zip it up?
What is an "assets.archive" bit? What is a "function resource"? Is that the Lambda defn?
I'm trying exactly what you described - ignoring the 
local.runOutput
approach.
Copy code
const fn = new aws.lambda.Function("fn", {
    code: new pulumi.asset.AssetArchive({
        ".": new pulumi.asset.FileArchive("./dist")
    },
    ... rest of lambda
});
The resulting archive from this seems to include the 
./dist
folder structure in it:
Copy code
new pulumi.asset.AssetArchive({
        ".": new pulumi.asset.FileArchive("./dist")
    },
should I just use bash and zip tooling that I understand? This is really frustrating.
q

quaint-army-32339

08/11/2023, 8:19 PM
Bruce, have you tried the built in file handling fom Typescript/Javascript for this? As an example: https://stackoverflow.com/questions/62031171/file-handling-in-typescript
p

polite-kite-18322

08/11/2023, 8:20 PM
I know how to work with Typescript. It's the pulumi libraries that are opaque to me.
So if bypassing Pulumi for this is the recommendation, then I'll do that. It just seems like this should be "pretty easy" since this is a common work flow and I see examples of other people using this online. I guess they all just dump the typescript artifacts into the folder that contains 
pulumi.yaml
b

brave-planet-10645

08/11/2023, 8:24 PM
This is my usual approach to deploying lambdas with api gateway https://github.com/pierskarsenbarg/useful-pulumi/tree/main/api-gateway-lambda lines 15-16 of index.ts is how I’ve done the file handling and the files uploaded are in the 
app
folder.
In this case I’m just uploading a JavaScript lambda with no preprocessing (so no typescript-JavaScript compiling)
p

polite-kite-18322

08/11/2023, 8:29 PM
Thank you for sharing the example. That should be easily adapted to include the Typescript=>Javascript preprocessing step without issue. So the resulting archive contains just 
index.js
and 
package.json
at the root? Then my issue is somewhere else like: • i'm not looking at the archive I think I'm looking at • linux is doing something different than other platforms • some other simple failure on my end.
I do appreciate your time here. Just getting a bit frustrated but that is not your problem.
b

brave-planet-10645

08/11/2023, 8:31 PM
It doesn’t have to contain the index.js and the package.json, that’s just my example for a super simple lambda
p

polite-kite-18322

08/11/2023, 8:32 PM
i get that - just clarifying that the contents of the 
./app
folder are appearing in the root of the archive and not in an 
app
folder in the archive.
b

brave-planet-10645

08/11/2023, 8:39 PM
That’s correct
p

polite-kite-18322

08/14/2023, 4:19 PM
@brave-planet-10645 so looks like the UI in 
aws
is misrepresenting what is in the package. When I downloaded the archive (export function) it looks correct!!!
Now I just need to ensure that my 
tsc
step is triggered reliably before the archive is created. Trying to to figure out how a 
Output<runResult>
can be used as a DependsOn (it only accepts 
Input<Resource>
types).
b

brave-planet-10645

08/14/2023, 4:43 PM
Personally I would have the build as a step in your CI/CD pipeline pre-Pulumi. Then you can store the output as an artifact that can be accessed at a later time if necessary. I’ve got a demo of how I use Pulumi and webpack to build a mono-repo: https://github.com/pierskarsenbarg/useful-pulumi/tree/main/serverless-webpack
p

polite-kite-18322

08/14/2023, 5:20 PM
Sounds like there is no way to depend on the output of runResult. Not sure what value it provides then. I'll try your approach.
b

brave-planet-10645

08/15/2023, 9:36 AM
What’s the 
runResult
coming from?
Powered by