Hi team, I need an AWS Lambda to be able to connect to the Metaflow metadata endpoint, but it seems you aren't able to write the Metaflow config to

~/.metaflowconfig/config.json

in Lambdas. I also tried setting the config values as env variables using

os.environ

but that didn't seem to work either. Any advice?

I’m not the best person to answer, but you can probably write your config to some other writable location and set the METAFLOW_HOME environment variable to that new dir so that your config is picked up from the right place

I'll give that a try, thanks!

@acoustic-van-30942 I think you need to set

METAFLOW_HOME

to

/tmp/config.json

. I have a Docker image for my CI/CD tool that runs in PRs and I set mine to

ENV METAFLOW_HOME="/root/.metaflowconfig"

to get it to work.

Thanks so much Yudiesh. I will give it a try!

Are you using a AWS Lambda specific Docker image such as

public.ecr.aws/lambda/python:3.9

?

No, I'm using our own custom security hardened image based on

alpine

Maybe you could try

ENV METAFLOW_HOME="/root/.config"

instead?

Also tried that but AWS Lambda doesn't have permissions to write to

root

. I think only

tmp

is allowed.

Got it, since you are using a custom Docker image with Lambda have you setup the runtime interface client with it? https://docs.aws.amazon.com/lambda/latest/dg/python-image.html#python-image-clients

I have, yup

@acoustic-van-30942 you can try `METAFLOW_ARTIFACT_LOCALROOT`which was added in this PR.

That did the trick! Thanks Yudiesh. So close to finishing the CI/CD pipeline. Any workaround for the conda env creation in the Lambda? I tried setting

CONDA_PKGS_DIRS

and

CONDA_ENVS_PATH

to

/tmp

, but still getting this error. It's still creating the cache files to

/home/sbx_user

, which isn't writeable.

Creating Conda environment 'torch_model'...
CondaError: Error encountered while attempting to create cache directory.
Directory: /home/sbx_user1051/.cache/conda/notices
Exception: [Errno 30] Read-only file system: '/home/sbx_user1051'

I have the following Conda environment variables:

ENV CONDA_DIR /opt/conda
ENV PATH=$CONDA_DIR/bin:$PATH

Gotta make sure that it downloads it to the CONDA_DIR:

RUN apt-get update && \
  apt-get install -y \
  unzip \
  curl \
  jq \
  git \
  wget \
  && apt-get clean \
  && curl "<https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip>" -o "awscliv2.zip" \
  && unzip awscliv2.zip \
  && ./aws/install \
  && rm -rf awscliv2.zip \
  && rm -rf /var/lib/apt/lists/* \
  && wget --quiet <https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh> -O ~/miniconda.sh \
  && /bin/bash ~/miniconda.sh -b -p /opt/conda

Same for me too 🙂