kamel install pulls image for incorrect architecture

[tdiesler]

On MacOS (M2) ...

> minikube start --addons registry
> kamel install --olm=false
Camel K installed in namespace default 

> CAMEL_K_POD=$(kubectl get pods --selector='name=camel-k-operator' --output=jsonpath='{.items[*].metadata.name}')
> kubectl logs $CAMEL_K_POD
exec /usr/local/bin/kamel: exec format error

This happens with 2.2.0 and the latest 2.3.0-nightly

[gansheer]

If this is an arm64 architecture processor you should follow the installation documentation. The Kamel CLI defaults to x86 image so you need to set the arm64 operator image explicitly, and the integration base image as well.

[tdiesler]

merci, that worked

[tdiesler]

Do we already have an issue about improving the getting started experience on ARM64?

Currently I do ...

> minikube start --addons registry

> CAMEL_K_VERSION=2.2.0
> kamel install --olm=false --operator-image apache/camel-k:${CAMEL_K_VERSION}-arm64 --base-image eclipse-temurin:17-jdk@sha256:a90cec83bb9b9ab19a63e377b20c3aa4e076b16c52d36e83c62c451b6d034e22

> kubectl logs -l camel.apache.org/component=operator --tail=400 -f

> kamel run helloworld.groovy
> kubectl logs -l camel.apache.org/integration=helloworld --tail=10 -f

[tdiesler]

PR: https://github.com/apache/camel-k/pull/5200

$ go run ./cmd/kamel install                                                                                                            
OLM is not available in the cluster. Fallback to regular installation.
Using arch specific operator image: docker.io/apache/camel-k:2.3.0-SNAPSHOT-arm64
Using arch specific base image: eclipse-temurin:17@sha256:a90cec83bb9b9ab19a63e377b20c3aa4e076b16c52d36e83c62c451b6d034e22
Camel K installed in namespace default 

[squakez]

As mentioned in the PR, we should publish manifests accordingly to solve this issue: https://github.com/apache/camel-k/pull/5200#pullrequestreview-1906815706

[tdiesler]

Is a dependency of https://github.com/apache/camel-k/issues/5238

[tdiesler]

Done