Lambda/API Gateway integration not working

[Rafael17]

🐛 Bug Report

After configuring the server using Lambda/API Gateway setup, I'm running into issues where all the caches are a MISS. Similar to: #28

When there is a cache miss, the server properly adds the caches to S3. However, on a sub sequential turbo run with no changes to source code, the server responds with a 200, saying that we have a cache hit on the GET /artifacts/:id call. But then there is no POST /artifacts/events, so it thinks it was a cache miss and makes a sub sequential PUT /artifacts/:id, replacing already existing cache.

Running this locally works without problems

here is what my configs for openapi/REST API Gateway look like

paths:
  '/v8/{proxy+}':
    x-amazon-apigateway-any-method:
      parameters:
        - name: proxy
          in: path
          required: true
          schema:
            type: string
      responses: {}
      x-amazon-apigateway-integration:
        # ref: https://docs.aws.amazon.com/apigateway/api-reference/resource/integration/
        uri: --REDACTED--
        httpMethod: POST
        passthroughBehavior: when_no_match
        type: aws_proxy
        credentials: --REDACTED--
     

• server: turborepo-remote-cache: "1.14.1"
• client: turbo: "1.8.8"

[Rafael17]

Here is a little more info on what I'm seeing on the API Gateway logs request/responses in sequential order:

Known MISS after code changes:

# first run
npm run lint

1 GET - Response - 404, Artifact not found

2a - POST - Request - /v8/artifacts/events?slug=team_myteam

Method request body before transformations: [
{
    "duration": 0,
    "event": "MISS",
    "hash": "6f6bb0a759b0a7fe",
    "sessionId": "9a841c25-c0db-44df-a701-573262bc0207",
    "source": "REMOTE"
}
]

2b - POST - Response - 200
{}

3a - PUT - Request
[Binary Data]

3b - PUT - Response - 200

{"urls":["team_myteam/6f6bb0a759b0a7fe"]}

Should be cache HIT but getting a MISS on the client

# 2nd run
npm run lint

1 - GET - Response - 200 (it found the cache!!!)

{
    "statusCode": 200,
    "body": "KLUv/QBYzQYAksskHXBFdbRZlsI2sREC3JiBV6+axS2ECFpk+Sr934sMxixNcY4i/ddcdDWYf83pokUMcM5hQP9XNIKoBlOQSA7JW81efTSrYzDYtdmr11qLN5VqhCJsdgmK0K2QgjPaM0U4dxOyj0oumnwwH5vZandzkTj0r7mTFkXoCeYBCxNIyQEqj5I/c4hI/tri+UB0p8AUGw0WHCAQg6otN5H5gTMOHwy7gNeTOI1NBmCygBEH7ZIkQ6QImM13DtqnEwYIDsuACtU40HZtnjpAchjY6oAIduMF1YRfBg==",
    "headers": {
        "content-type": "application/octet-stream",
        "date": "Fri, 07 Apr 2023 23:19:43 GMT",
        "connection": "keep-alive",
        "transfer-encoding": "chunked"
    },
    "isBase64Encoded": true
}

no POST request on the API Gateway

2a - PUT - Request
[Binary Data]

2b - PUT - Response - 200

{"urls":["team_myteam/6f6bb0a759b0a7fe"]}

So the POST /v8/artifacts/events?slug=team_myteam request appears to be missing causing a cache MISS on the client

[Rafael17]

Basically I'm looking for help to get the Lambda server working with AWS API REST Gateway, the setup in the docs works only for AWS API HTTP Gateway

[Elliot-Alexander]

I'm seeing a similar issue with my AWS deployment. Can see the caches in S3 and can confirm that the requests for the cache objects are returning 200s however it still causes a cache miss :( I'm seeing this issue when running in GitHub Actions.

[Elliot-Alexander]

As a follow up, I can provide a CDK files or Cloudformation templates if that would assist with triage?

[WANZARGEN]

There is a body size limit for lambda...
So I used EC2 instance, and it works perfectly. 😄

[Rafael17]

Thanks for the response @WANZARGEN

The limit in this case is not the issue, since I'm testing lint and the output is a tiny file. The issue has to do with the lambda server and AWS REST Gateway. Did you happen to front your EC2 with AWS REST Gateway? I'm assuming no.

[Elliot-Alexander]

Does using the cache from local work for you @Rafael17? As I only see this issue when running turbo in GitHub Actions. I'm wondering if it's an issue with the turbo CLI when running in CI rather than the cache server itself

[Rafael17]

@Elliot-Alexander For me everything works correctly when using AWS HTTP Gateway. The problem only appears when I use AWS REST Gateway

[fox1t]

Can you reiterate the issue you are facing, so we can add this information to the documentation?

[EloB]

I think lambda only can handle files up to 6mb.
https://docs.aws.amazon.com/lambda/latest/dg/gettingstarted-limits.html

[duveborg]

Yeah, did anyone get around the fact that the max payload is 6MB?

[EloB]

@duveborg Hello mate ;) Long time no see. I think this repo needs to implement presigned urls to workaround that limitation.
https://docs.aws.amazon.com/AmazonS3/latest/userguide/using-presigned-url.html

[EloB]

@duveborg I sit in the same boat as you.

My workaround right now is to deploy a small docker container using ec2 and run the server below until any lambda turbo repo implementation could handle above.
https://github.com/salamaashoush/turbo-remote-cache-rs

[EloB]

I'm not sure if this helps anyone who want to implement this but I think that TURBO_PREFLIGHT or --preflight would be of good use to do this.

vercel/turborepo#956
vercel/turborepo#1052

[EloB]

I've created another repo that just solves s3 and lambda with presigned urls that allows bigger files.
https://github.com/EloB/turborepo-remote-cache-lambda

git clone [email protected]:EloB/turborepo-remote-cache-lambda.git
cd turborepo-remote-cache-lambda/
npm --prefix=function install
sam build
zip -j app.zip .aws-sam/build/TurboRepoRemoteCacheFunction/*
# Create an lambda function. Manually create the lambda function as explained in this repo.

Take a look at this file. The actual lambda function is only 130 lines of code.
https://github.com/EloB/turborepo-remote-cache-lambda/blob/main/function/app.ts

I haven't had time to write documentation but it's working for me.

You need to include teamId in the JWT token. Can be created at https://jwt.io/ see image. Enter some team and secret (see image below). Don't forgot to create .turbo/config.json in your Turbo repo so you enable remote caching.

.turbo/config.json

{
  "teamid": "team_yourteamhere",
  "apiurl": "https://YOUR-LAMBDA-URL.lambda-url.eu-central-1.on.aws"
}

You also need to set environment variables JWT_SECRET (most be same as the secret from jwt.io if you created it there) and S3_BUCKET for the lambda.

Also important to use turbo run yourtask --preflight because it enables turbo to run OPTIONS request where you can change the location where to download/upload that makes s3 presigned urls to function. It's not documented yet in their openapi but I learnt it from their issues.
https://turbo.build/repo/docs/core-concepts/remote-caching#self-hosting
https://turbo.build/api/remote-cache-spec
vercel/turborepo#956 (read more about preflight).