standard_init_linux.go:219: exec user process caused: exec format error ECR + Quarkus

Ok, I already know that there are many errors on SO, however I cannot find any that suits with my case.

PREMISE:

My environments is MacOS with M1.

I have an EKS cluster with linux (amd64) as OS architecture on AWS.

When trying to build and deploy my micro service with Quarkus, I set on application.properties:

quarkus.container-image.image=<my ECR repository>
quarkus.container-image.build=true
quarkus.container-image.push=true
quarkus.jib.platforms=linux/amd64

I'm using jib extension for building image.

Once the image is succesfully deployed on my ECR aws registry, I launch:

 kubectl apply -f target/kubernetes/kubernetes.yaml

then

 kubectl get po

I see

NAME                              READY   STATUS             RESTARTS   AGE
kube-cm-example-6ffd58655-rzzqh   0/1     CrashLoopBackOff   8          16m

Then

kubectl logs kube-cm-example-6ffd58655-rzzqh

return the following output

standard_init_linux.go:219: exec user process caused: exec format error

I read on many sources it's a platform image issue, however using jib I set linux/amd64 which is the same arch of the node.

I also tried to manually build a docker image specifying platform=linux/amd64 and then deploying to ECR, but nothing changes.

Solution

Based on the observations, it may be that your Kubernetes cluster is not pulling newly updated images from ECR when starting a pod. Unless some conditions are met, the default pull policy of k8s skips pulling images if they already exist. During development, you may consider making it always pull images. Take a look at this Stack Overflow question.