AWS Amplify Console & builidng dotnet core lambda functions

I added my first dotnet core lambda function to an Amplify app, and on my first Amplify console build, I got the error: Unable to find dotnet version 3.1 on the path.

The problem

The cause of the error message was that dotnet tools are not installed in the default Amplify Console build image. I was looking into installing the dotnet tools as steps in the amplify.yml when I found this the Github issue #685 reported by @CoreyPritchard.

Lucky for me @alexkates had already described a better fix in a comment to the github issue, with creating a custom Docker image to build with instead.

The solution

Since I love CI/CD pipelines, I decided to automate the steps provided by @alexkates in an AWS CodeBuild project.

This Github repo

The files that are doing the job here are:

• buildspec.yml Automate the steps from cloning the AWS Amplify console latest image repo, login into Docker Hub, building the images and pushing the new build image to Docker Hub.
• Dockerfile Describes the additions needed in the docker image to make dotnet core builds work. See comments in it for more info.

AWS Codebuild project

I followed the this Docker sample for CodeBuild when setting up the project, but connected it to this repository so this buildspec.yml and Dockerfile are used.

Environment variables that muyst be set up for the project:

• IMAGE_REPO_NAME The Docker Hub image repository name including the user/organisation. Example your docker hub user or organisation/repository name .
• IMAGE_TAG Normally "latest".
• DOCKER_HUB_USER A Docker Hub username
• DOCKER_HUB_TOKEN An access token for the username yoo provided

Make your Amplify app use the new build image

After the first succesfull build of the new docker image, you must wire up your custom image in the Amplify Console Build Settings. See Configure Amplify to use the custom docker image on how to do it.

Possible improvements

• The provision of the build image from Docker Hub seems to be slow. A possible improvement is to push it to Amazon Elastic Container Registry instead of Docker Hub
• The access token is printed in the log Use the -password-stdin instead of -password argument when login to docker hub. See docs.
• Customize the AWS Dockerfile and remove everything that is not necesarry to improve load time.