Add architecture to the binaries and images in the cache

[afbjorklund]

When supporting remote clusters of a different architecture, we need to add the arch in addition to the OS:

~/.minikube/cache/linux/v1.19.3
├── kubeadm
├── kubectl
└── kubelet

Simplest would be to add it after the OS, but we could support both in order to preserve existing user caches ?

~/.minikube/cache/linux
├── amd64
│   └── v1.19.3
│       ├── kubeadm
│       ├── kubectl
│       └── kubelet
└── arm64
    └── v1.19.3
        ├── kubeadm
        ├── kubectl
        └── kubelet

For instance when running an amd64 host, towards a arm64 machine. Either local hardware or cloud instance.

• Raspberry Pi

• AWS Graviton

Same thing for the images, if not using the preload (the preloaded-tarball files already include the arch, though)

~/.minikube/cache/images
├── gcr.io
│   └── k8s-minikube
│       └── storage-provisioner_v3
├── k8s.gcr.io
│   ├── coredns_1.7.0
│   ├── etcd_3.4.13-0
│   ├── kube-apiserver_v1.19.2
│   ├── kube-controller-manager_v1.19.2
│   ├── kube-proxy_v1.19.2
│   ├── kube-scheduler_v1.19.2
│   └── pause_3.2
└── kubernetesui
    ├── dashboard_v2.0.3
    └── metrics-scraper_v1.0.4

~/.minikube/cache/images/amd64
~/.minikube/cache/images/arm64

[afbjorklund]

Also, this code needs to be parameterized:

                        src, err := download.Binary(name, cfg.KubernetesVersion, "linux", runtime.GOARCH)
                        if err != nil {
                                return errors.Wrapf(err, "downloading %s", name)
                        }

                       if _, err := download.Binary(bin, version, "linux", runtime.GOARCH); err != nil {
                                return errors.Wrapf(err, "caching binary %s", bin)
                        }

func archTag(hasTag bool) string {
        if runtime.GOARCH == "amd64" && !hasTag {
                return ":"
        }
        return "-" + runtime.GOARCH + ":"
}

var defaultPlatform = v1.Platform{
        Architecture: runtime.GOARCH,
        OS:           "linux",
}

Basically split host arch and cluster (node?) arch.

Using runtime.GOARCH is a good default, but it needs to be possible to override.

It is still OK to hardcode the OS (for now...), since Kubernetes only runs on Linux.

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle stale

[sharifelgamal]

Ok, this has become more of an issue recently, since if someone accidentally runs an amd64 minikube binary on an arm machine, it corrupts their cache. Let's fix this one and for all.