Lambda functions bundling deploy inside container

[pmdp]

We use SBT from Gitlab-CI to implement CI/CD, so we need to be able to deploy (via cdk cli) from a gitlab runner job, i.e. from a container with the docker in docker service active (shared docker socket). Gitlab-CI docker in docker

The problem comes when PythonFunctions that SBT implements, by default, uses the BundlingOptions with BundlingFileAccess.BIND_MOUNT value. This causes the following error: Error: Bundling did not produce any output. Check that content is written to /asset-output. when deploying.

As far as we have been able to investigate, it could be fixed by changing from BIND_MOUNT to VOLUME_COPY, but we did not find the possibility to do this when using the SBT (ControlPlane) contructs.

Has this problem been solved in the past or how could be solve it?

[suhussai]

Hi @pmdp , I haven't come across this exact problem, but based on my research, I think there a few ways we can try and fix this:

1. look for an environment variable that indicates we are running in GitLab runner job (or GitHub action, or something similar). For example, GitLab sets the CI variable as per https://docs.gitlab.com/ci/variables/predefined_variables/#predefined-variables. So, we could update SBT to read that env var and set the BundlingOptions argument appropriately.

2. Check for a custom env var (such as DIND_MODE) which, if set, configures the BundlingOptions argument appropriately.

3. Update the ControlPlane construct to include another argument that sets the bundling options for all PythonFunctions created as part of that construct.

I'm leaning towards option 2 as that seems to be the more broadly applicable one. What are your thoughts?

[pmdp]

Hi @suhussai, We seem to have fixed it, by correctly configuring the Gitlab-CI docker environment (config.toml) with the repository directory mounted as a volume in the container running the deploy cdk.

In principle it is not necessary to have to modify the default BundlingOptions operation of the lambda functions.

Thank you very much for your reply.