Cache Layer according to "Pipfile.lock" or "poetry.lock".

[dnp1]

Considering that layer contains only the dependencies, we could just cache e avoid redeploy according to "Pipfile.lock", it is possible?

[estahn]

I'm not sure if my case is the same or needs a new ticket.

We're using dockerizePip: true. Running serverless package multiple times locally will result in about 19s packaging time. The CI process takes about 5min in total incl deployment, but a large chunk is packaging. I assume this is due to the docker build process, as "docker" is not being executed locally in consequent runs.

Log from CI:

...
Generating requirements.txt from "pyproject.toml"
Parsed requirements.txt from pyproject.toml in /home/runner/work/<proj>/<proj>/.serverless/requirements.txt
Installing requirements from "/home/runner/.cache/serverless-python-requirements/df293b6184a94fbfd982a4f95fea1698fe858cb805f178dd2112d29905e6a12e_x86_64_slspyc/requirements.txt"
Docker Image: sls-py-reqs-custom
Using download cache directory /home/runner/.cache/serverless-python-requirements/downloadCacheslspyc
Running docker run --rm -v /home/runner/.cache/serverless-python-requirements/df293b6184a94fbfd982a4f95fea[16]...
Python executable not found for "runtime": python3.9
...

Config:

  pythonRequirements:
    slim: true
    # If changes do not reflect, use `serverless requirements cleanCache`
    # See See https://github.com/serverless/serverless-python-requirements/issues/437
    slimPatterns:
      - "**/tests/**"

      # Removal of boto3 due to Lambda size limitation.
      # This should ideally be packaged with the app.
      - "**/botocore/**"
      - "**/boto3/**"

      # Django files we don't need for the API.
      - "**/locale/!(en|en_*)/**"
    strip: false
    dockerizePip: true
    dockerFile: ./deploy/Dockerfile-requirements
    #dockerImage: public.ecr.aws/sam/build-python3.9
    zip: false
    # Set to `true` to cleanup cache in case of requirements issues.
    invalidateCaches: false
    useDownloadCache: true
    useStaticCache: true
    layer: true

1. Is there a current solution to speed up or circumvent the docker build if requirements have not changed?
2. Are there ideas about a future solution to reduce the runtime if requirements have not changed?

---

UPDATE:

After further investigation I found the following line and realised the issue is with the caching in GitHub actions:

Using static cache of requirements found at /Users/me/Library/Caches/serverless-python-requirements/64ee303d22f5bbbd46285b7b97c678c83eee6084133162ae1866640723712dd5_x86_64_slspyc

The correct approach needs to include the hash of the poetry.lock.

      - uses: actions/cache@v3
        with:
          key: serverless-python-requirements-${{ hashFiles('poetry.lock') }}
          path: |
            /home/runner/.cache/serverless-python-requirements

Hope this helps someone having the same issue.