AWS App Runner issue due to mismatching arch when building docker image in Apple silicon

[mauricioschneider]

When running yarn ops deploy -a all, the App Runner service for the API fails with the following logs:

App Runner event logs

 [AppRunner] Deployment with ID : x started. Triggering event : SERVICE_CREATE
 [AppRunner] Deployment Artifact: [Repo Type: ECR], [Image URL: x.dkr.ecr.us-west-2.amazonaws.com/letsgo-main-api], [Image Tag: 60f617e3ad194880]
 [AppRunner] Pulling image x.dkr.ecr.us-west-2.amazonaws.com/letsgo-main-api from ECR repository.
 [AppRunner] Successfully pulled your application image from ECR.
 [AppRunner] Provisioning instances and deploying image for publicly accessible service.
 [AppRunner] Performing health check on protocol `HTTP` [Path: '/v1/health'], [Port: '8080'].
 [AppRunner] Your application stopped or failed to start. See logs for more information.  Container exit code: 1
 [AppRunner] Deployment with ID : x failed.

Application logs

exec /usr/local/bin/docker-entrypoint.sh: exec format error

I'm running letsgo on an Macbook Pro M3 Max, and according to this SO post, it's related to mismatching architecture used to build the docker image vs the one where the docker image will run. Here's a more in-depth explanation.

[mauricioschneider]

Besides changing the FROM in the Dockerfile, it's possible to use a environment variable, export DOCKER_DEFAULT_PLATFORM=linux/amd64.

[mauricioschneider]

After changing the first line of api/Dockerfile to FROM --platform=linux/amd64 node:18-alpine AS base, the deployment of the image to App Runner is successful.

[mauricioschneider]

The line above should be used in all Dockerfile copies in api, web, and worker.

[tjanczuk]

Good catch, fix coming.