When I include a function in my configuration, the layer does not get included

Serverless Framework
1

I’m attempting to push a lambda layer stack.

My code, more-or-less following the official tutorial here:

My code looks like this:

service: ffmpeg
frameworkVersion: '2'
provider:
  name: aws
  runtime: nodejs14.x
  region: us-east-1
  stage: dev
  lambdaHashingVersion: 20201221

functions:
  mkgif:
    handler: handler.mkgif
    events:
      - s3: 
          bucket: my-bucket-name-here-foo-foo
          event: s3:ObjectCreated:*
          existing: true
          rules:
            - suffix: .mp4
            - prefix: test/
    layers:
      - {Ref: FfmpegLambdaLayer}

layers:
  ffmpeg:
    path: layer/ffmpeg

My file structure looks like this:

ffmpeg -
  -- layer
    -- ffmpeg
      -- *all the binaries and other files for ffmpeg
  -- handler.js
  -- .npmignore
  -- serverless.yml

Of note, when I deploy this, the uploaded zip package is ~2kb. When I go to the layer in the AWS console, the layer zip is also small and only contains the handler.js file and the .npmignore file.

The weird part is… if I comment out the functions area in my serverless.yml file, so my code looks like this:

service: ffmpeg
frameworkVersion: '2'
provider:
  name: aws
  runtime: nodejs14.x
  region: us-east-1
  stage: dev
  lambdaHashingVersion: 20201221

layers:
  ffmpeg:
    path: layer/ffmpeg

… THEN deploy… I get a 65mb zip file and a full-sized lambda layer in the AWS console, with the expected files.

Uh… why? How can I deploy a function AND the layer?

2

maybe a hint from myside, my layer reference in the function section of the serverless is different.
i use a reference output of the cloudformation layer-stack, it is the <layer-output-arn-from-cfstack), for example. My deployment is sepearated into 2 stacks, a sls deployment for the layer after i deploy my sls-function stack.

My reference looks like.
sls-xyz-mock-layer-dev-XyzDashmockDashdependenciesLambdaLayerQualifiedArn