Reputation: 1259
Kubernetes - Can you build a cluster with master node on x86 and worker node on arm?
I have an Intel NUC (I5) and Raspberry-Pi Model-B . I tried to create a kubernetes cluster by making the Intel-NUC as master node and Raspberry-Pi as worker node.When I try the above set up, I see that the worker node crashing all the time . Here's the output . This happens only with the above set up. If I try creating a cluster with two Raspberry-Pis (One master and one worker node), it works fine.
What am I doing wrong?
sudo kubectl get pods --all-namespaces
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system etcd-ubuntu 1/1 Running 0 13h
kube-system kube-apiserver-ubuntu 1/1 Running 0 13h
kube-system kube-controller-manager-ubuntu 1/1 Running 0 13h
kube-system kube-dns-6f4fd4bdf-fqmmt 3/3 Running 0 13h
kube-system kube-proxy-46ddk 0/1 CrashLoopBackOff 5 3m
kube-system kube-proxy-j48fc 1/1 Running 0 13h
kube-system kube-scheduler-ubuntu 1/1 Running 0 13h
kube-system kubernetes-dashboard-5bd6f767c7-nh6hz 1/1 Running 0 13h
kube-system weave-net-2bnzq 2/2 Running 0 13h
kube-system weave-net-7hr54 1/2 CrashLoopBackOff 3 3m
I examined the logs for kube-proxy and found the following entry
Logs from kube-proxy
standard_init_linux.go:178: exec user process caused "exec format error"
This seem to stem from the issue that the image picked up is arm arch as oppose x86 arch. Here's the yaml file
{
"kind": "Pod",
"apiVersion": "v1",
"metadata": {
"name": "kube-proxy-5xc9c",
"generateName": "kube-proxy-",
"namespace": "kube-system",
"selfLink": "/api/v1/namespaces/kube-system/pods/kube-proxy-5xc9c",
"uid": "a227b43b-27ef-11e8-8cf2-b827eb03776e",
"resourceVersion": "22798",
"creationTimestamp": "2018-03-15T01:24:40Z",
"labels": {
"controller-revision-hash": "3203044440",
"k8s-app": "kube-proxy",
"pod-template-generation": "1"
},
"ownerReferences": [
{
"apiVersion": "extensions/v1beta1",
"kind": "DaemonSet",
"name": "kube-proxy",
"uid": "361aca09-27c9-11e8-a102-b827eb03776e",
"controller": true,
"blockOwnerDeletion": true
}
]
},
"spec": {
"volumes": [
{
"name": "kube-proxy",
"configMap": {
"name": "kube-proxy",
"defaultMode": 420
}
},
{
"name": "xtables-lock",
"hostPath": {
"path": "/run/xtables.lock",
"type": "FileOrCreate"
}
},
{
"name": "lib-modules",
"hostPath": {
"path": "/lib/modules",
"type": ""
}
},
{
"name": "kube-proxy-token-kzt5h",
"secret": {
"secretName": "kube-proxy-token-kzt5h",
"defaultMode": 420
}
}
],
"containers": [
{
"name": "kube-proxy",
"image": "gcr.io/google_containers/kube-proxy-arm:v1.9.4",
"command": [
"/usr/local/bin/kube-proxy",
"--config=/var/lib/kube-proxy/config.conf"
],
"resources": {},
"volumeMounts": [
{
"name": "kube-proxy",
"mountPath": "/var/lib/kube-proxy"
},
{
"name": "xtables-lock",
"mountPath": "/run/xtables.lock"
},
{
"name": "lib-modules",
"readOnly": true,
"mountPath": "/lib/modules"
},
{
"name": "kube-proxy-token-kzt5h",
"readOnly": true,
"mountPath": "/var/run/secrets/kubernetes.io/serviceaccount"
}
],
"terminationMessagePath": "/dev/termination-log",
"terminationMessagePolicy": "File",
"imagePullPolicy": "IfNotPresent",
"securityContext": {
"privileged": true
}
}
],
"restartPolicy": "Always",
"terminationGracePeriodSeconds": 30,
"dnsPolicy": "ClusterFirst",
"serviceAccountName": "kube-proxy",
"serviceAccount": "kube-proxy",
"nodeName": "udubuntu",
"hostNetwork": true,
"securityContext": {},
"schedulerName": "default-scheduler",
"tolerations": [
{
"key": "node-role.kubernetes.io/master",
"effect": "NoSchedule"
},
{
"key": "node.cloudprovider.kubernetes.io/uninitialized",
"value": "true",
"effect": "NoSchedule"
},
{
"key": "node.kubernetes.io/not-ready",
"operator": "Exists",
"effect": "NoExecute"
},
{
"key": "node.kubernetes.io/unreachable",
"operator": "Exists",
"effect": "NoExecute"
},
{
"key": "node.kubernetes.io/disk-pressure",
"operator": "Exists",
"effect": "NoSchedule"
},
{
"key": "node.kubernetes.io/memory-pressure",
"operator": "Exists",
"effect": "NoSchedule"
}
]
},
"status": {
"phase": "Running",
"conditions": [
{
"type": "Initialized",
"status": "True",
"lastProbeTime": null,
"lastTransitionTime": "2018-03-15T01:24:45Z"
},
{
"type": "Ready",
"status": "False",
"lastProbeTime": null,
"lastTransitionTime": "2018-03-15T01:35:41Z",
"reason": "ContainersNotReady",
"message": "containers with unready status: [kube-proxy]"
},
{
"type": "PodScheduled",
"status": "True",
"lastProbeTime": null,
"lastTransitionTime": "2018-03-15T01:24:46Z"
}
],
"hostIP": "192.168.178.24",
"podIP": "192.168.178.24",
"startTime": "2018-03-15T01:24:45Z",
"containerStatuses": [
{
"name": "kube-proxy",
"state": {
"waiting": {
"reason": "CrashLoopBackOff",
"message": "Back-off 5m0s restarting failed container=kube-proxy pod=kube-proxy-5xc9c_kube-system(a227b43b-27ef-11e8-8cf2-b827eb03776e)"
}
},
"lastState": {
"terminated": {
"exitCode": 1,
"reason": "Error",
"startedAt": "2018-03-15T01:40:51Z",
"finishedAt": "2018-03-15T01:40:51Z",
"containerID": "docker://866dd8e7175bd71557b9dcfc84716a0f3abd634d5d78c94441f971b8bf24cd0d"
}
},
"ready": false,
"restartCount": 8,
"image": "gcr.io/google_containers/kube-proxy-arm:v1.9.4",
"imageID": "docker-pullable://gcr.io/google_containers/kube-proxy-arm@sha256:c6fa0de67fb6dbbb0009b2e6562860d1f6da96574d23617726e862f35f9344e7",
"containerID": "docker://866dd8e7175bd71557b9dcfc84716a0f3abd634d5d78c94441f971b8bf24cd0d"
}
],
"qosClass": "BestEffort"
}
}
Upvotes: 3
Views: 2685
Answers (2)
Reputation: 501
Yes, it's possible and i've just done that for one of my customers.
Basically there is an issue that the KubeProxy DaemonSet deployed automatically on the master is compiled to x64 - because you wanted that the master will be x64 and the nodes will be ARM.
When you add ARM nodes to the cluster, the DaemonSet tries to deploy the x64 image on them, and fails.
You'll need to edit the default DaemonSet after installation to select only x64 nodes, and deploy another DaemonSet for ARM nodes. This gist will walk you through: Multiplatform (amd64 and arm) Kubernetes cluster setup
Hope this helps, Ofir.
Upvotes: 6
Reputation: 22874
it is perfectly possible to do that, but you need to ensure that you launch pods for given cpu architecture on appropriate machines. Use nodeSelector or Node. Otherwise, you get what you experience - amd64 binaries failing to run on arm, or vice versa.
Hint: use beta.kubernetes.io/arch label to distinct between cpu architectures.
Upvotes: 1
Related Questions
- Is a Kubernetes Arm Master and x86 nodes wise/possible?
- Do Kubernetes worker nodes have to have exactly the same hardware and OS?
- Can I use multiple docker containers on a computer as Kubernetes worker nodes?
- Can worker nodes be different OS and version in a k8s cluster?
- Installing kubernetes cluster on master node
- Will (can) Kubernetes run Docker containers on the master node(s)?
- Feasibility of using multi master Kubernetes cluster architecture
- Kubernetes Cluster with different CPU configuration
- How can I run a Kubernetes cluster with a single machine as a master and node?