In correlation with the end of life (EOL) for MKE 3.6.x, maintenance of this documentation set was discontinued as of 2024-OCT-13. Click here for the latest MKE 3.x version documentation.

Run hybrid workloads in Kubernetes

To run Windows workloads in a hybrid Windows Kubernetes cluster, you must target your workloads to nodes that are running the correct Windows version. Failure to correctly target your workloads may result in an error when Kubernetes schedules the Pod on an incompatible node:

Error response from daemon: hcsshim::CreateComputeSystem win2019-deployment-no-nodeselect: The container operating system does not match the host operating system.

1. Note the Windows version associated with each of the nodes in your cluster:

   kubectl get node
   

   Example output:

   NAME                         STATUS   ROLES    AGE   VERSION
   manager-node                 Ready    master   51m   v1.23.4-mirantis-1
   win2019-node                 Ready    <none>   44m   v1.23.4
   win2022-node                 Ready    <none>   38m   v1.23.4
   

2. Create a deployment with the appropriate node selectors. Use 10.0.17763 for Windows Server 2019 workloads and 10.0.20348 for Windows Server 2022 workloads.

   For example purposes, paste the following content into a file called win2019-deployment.yaml:

   apiVersion: apps/v1
   kind: Deployment
   metadata:
     labels:
       app: win2019-deployment
     name: win2019-deployment
   spec:
     replicas: 5
     selector:
       matchLabels:
         app: win2019-deployment
     template:
       metadata:
         labels:
           app: win2019-deployment
         name: win2019-deployment
       spec:
         containers:
         - name: win2019-deployment
           image: mcr.microsoft.com/windows/nanoserver:1809
           command: ["cmd", "/c", "ping -t localhost"]
           ports:
           - containerPort: 80
         nodeSelector:
           kubernetes.io/os: windows
           node.kubernetes.io/windows-build: 10.0.17763
   

3. Apply the deployment:

   kubectl apply -f win2019-deployment.yaml
   

4. Verify that the Pods are scheduled on the required node:

   kubectl get pods -o wide
   

   Example output:

   NAME                                                READY   STATUS             RESTARTS      AGE     IP              NODE                      NOMINATED NODE   READINESS GATES
   win2019-deployment-57d75f6f9f-ldsqf                 1/1     Running            0             6m39s   192.168.50.76   win2019-node              <none>           <none>
   win2019-deployment-57d75f6f9f-n5b25                 1/1     Running            0             6m39s   192.168.50.79   win2019-node              <none>           <none>
   win2019-deployment-57d75f6f9f-r5mz6                 1/1     Running            0             6m39s   192.168.50.78   win2019-node              <none>           <none>
   win2019-deployment-57d75f6f9f-xggmt                 1/1     Running            0             6m39s   192.168.50.77   win2019-node              <none>           <none>
   win2019-deployment-57d75f6f9f-zltk2                 1/1     Running            0             7m7s    192.168.50.73   win2019-node              <none>           <none>