aws-lambda: DockerImageCode.fromImageAsset not caching as-expected

[mccauleyp]

Describe the bug

I have a simple Dockerfile:

FROM public.ecr.aws/lambda/python:3.11

# Copy requirements.txt separately and run `pip install` first to facilitate build caching
COPY requirements.txt .
RUN pip install --upgrade pip && \
    pip install -r ./requirements.txt

COPY handlers ${LAMBDA_TASK_ROOT}/handlers

If I build this outside of CDK, the caching works as-expected in that the RUN pip install command only runs if requirements.txt changes, otherwise the cache is used.

When I build it inside CDK, e.g.:

        self.my_function = aws_lambda.DockerImageFunction(
            scope,
            "MyLambda",
            code=aws_lambda.DockerImageCode.from_image_asset(
                platform=aws_ecr_assets.Platform.LINUX_AMD64,
                directory="/path/to/dockerfile/dir/,
                cmd=["handlers.entrypoint.handler"],
            ),
        )

The pip install re-runs any time the code changes, event if requirements.txt is unchanged. I'm guessing that's related to the Dockerfile, requirements, and app code all being part of the same new asset, e.g.:

- cdk.out/
- - asset.___/
- - - handlers/
- - - Dockerfile
- - - requirements.txt

Is it possible to refactor the CDK code to allow caching? Is there a way for me to merge two assets in a single build, so the Dockerfile and requirements.txt don't get duplicated in a new hash?

Expected Behavior

CDK docker build caches in the same way as outside of CDK.

Current Behavior

CDK build requires rerunning lengthy steps.

Reproduction Steps

See above.

Possible Solution

No response

Additional Information/Context

No response

CDK CLI Version

2.128.0

Framework Version

No response

Node.js Version

v20.11.0

OS

OSX

Language

Python

Language Version

No response

Other information

No response

[pahud]

The pip install re-runs any time the code changes

What is the code you mentioned? Is it the python code in your docker directory or the CDK code?

I just tested this

    new lambda.DockerImageFunction(this, 'MyLamsbda', {
      code: lambda.DockerImageCode.fromImageAsset(
        path.join(__dirname, '../docker.d/python.d'),
      ),
      memorySize: 256,
    })

And my Dockerfile like

FROM public.ecr.aws/lambda/python:3.11

# Copy requirements.txt separately and run `pip install` first to facilitate build caching
COPY requirements.txt .
RUN pip install --upgrade pip && \
    pip install -r ./requirements.txt

# COPY handlers ${LAMBDA_TASK_ROOT}/handlers
COPY index.py ${LAMBDA_TASK_ROOT}/index.py

I modified the index.py and I saw the pip install is actually cached.

#7 [3/4] RUN pip install --upgrade pip &&     pip install -r ./requirements.txt
#7 CACHED

[mccauleyp]

@pahud Thanks for looking into this! I was talking about the Python code, and you tested it in the way I intended. I'm still seeing my CDK builds not use the cache while a regular build does, but your test suggests that it's probably something specific to my environment, so I'll close this and try to sort that out. Thanks!

[github-actions]

⚠️COMMENT VISIBILITY WARNING⚠️

Comments on closed issues are hard for our team to see.
If you need more assistance, please either tag a team member or open a new issue that references this one.
If you wish to keep having a conversation with other community members under this issue feel free to do so.