Skip to content

Latest commit

 

History

History
120 lines (98 loc) · 4.24 KB

windows.md

File metadata and controls

120 lines (98 loc) · 4.24 KB
Raw

Windows Container Support

The Argo server and the workflow controller currently only run on Linux. The workflow executor however also runs on Windows nodes, meaning you can use Windows containers inside your workflows! Here are the steps to get started.

Requirements

  • Kubernetes 1.14 or later, supporting Windows nodes
  • Hybrid cluster containing Linux and Windows nodes like described in the Kubernetes docs
  • Argo configured and running like described here

Schedule workflows with Windows containers

If you're running workflows in your hybrid Kubernetes cluster, always make sure to include a nodeSelector to run the steps on the correct host OS:

apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: hello-windows-
spec:
  entrypoint: hello-win
  templates:
    - name: hello-win
      nodeSelector:
        kubernetes.io/os: windows    # specify the OS your step should run on
      container:
        image: mcr.microsoft.com/windows/nanoserver:1809
        command: ["cmd", "/c"]
        args: ["echo", "Hello from Windows Container!"]

You can run this example and get the logs:

$ argo submit --watch https://raw.githubusercontent.com/argoproj/argo/master/examples/hello-windows.yaml
$ argo logs hello-windows-s9kk5
hello-windows-s9kk5: "Hello from Windows Container!"

Schedule hybrid workflows

You can also run different steps on different host OSs. This can for example be very helpful when you need to compile your application on Windows and Linux.

An example workflow can look like the following:

apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: hello-hybrid-
spec:
  entrypoint: mytemplate
  templates:
    - name: mytemplate
      steps:
        - - name: step1
            template: hello-win
        - - name: step2
            template: hello-linux

    - name: hello-win
      nodeSelector:
        kubernetes.io/os: windows
      container:
        image: mcr.microsoft.com/windows/nanoserver:1809
        command: ["cmd", "/c"]
        args: ["echo", "Hello from Windows Container!"]
    - name: hello-linux
      nodeSelector:
        kubernetes.io/os: linux
      container:
        image: alpine
        command: [echo]
        args: ["Hello from Linux Container!"]

Again, you can run this example and get the logs:

$ argo submit --watch https://raw.githubusercontent.com/argoproj/argo/master/examples/hello-hybrid.yaml
$ argo logs hello-hybrid-plqpp
hello-hybrid-plqpp-1977432187: "Hello from Windows Container!"
hello-hybrid-plqpp-764774907: Hello from Linux Container!

Artifact mount path

Artifacts work mostly the same way as on Linux. All paths get automatically mapped to the C: drive. For example:

 # ...
    - name: print-message
      inputs:
        artifacts:
          # unpack the message input artifact
          # and put it at C:\message
          - name: message
            path: "/message" # gets mapped to C:\message
      nodeSelector:
        kubernetes.io/os: windows
      container:
        image: mcr.microsoft.com/windows/nanoserver:1809
        command: ["cmd", "/c"]
        args: ["dir C:\\message"]   # List the C:\message directory

Remember that volume mounts on Windows can only target a directory in the container, and not an individual file.

Limitations

  • Sharing process namespaces doesn't work on Windows so you can't use the Process Namespace Sharing (pns) workflow executor.

Building the workflow executor image for Windows

To build the workflow executor image for Windows you need a Windows machine running Windows Server 2019 with Docker installed like described in the docs.

You then clone the project and run the Docker build with the Dockerfile for Windows and argoexec as a target:

git clone https://github.com/argoproj/argo.git
cd argo
docker build -t myargoexec -f .\Dockerfile.windows --target argoexec .