I'm also having this problem. However it's hanging on powershell as well.

Edit: typo

What host OS and Docker images are you using?

stuck_tech

Host OS:, Windows Server 2019 

Docker images tried:

• mcr.microsoft.com/windows/servercore:ltsc2019
• mcr.microsoft.com/powershell:preview
• One of my own based off of  mcr.microsoft.com/dotnet/framework/sdk:3.5

Same experience on every one. 

Currently my test pipeline looks like. this

pipeline {  
    agent any  
    stages {
        stage('hello world') {
            parallel {
                // this one doesn't work
                stage('Hello world on Windows') {
                    agent {
                      docker {
                          label 'Windows&&Docker&&aws'
                          image 'mcr.microsoft.com/windows/servercore:ltsc2019'
                      }
                     }
                    options {
                     timeout(1) // in case the pipeline hangs
                    }
                    steps {
                     // hangs here
                     powershell 'powershell "cmd /c echo test inside"'
                    }
                }
                // This one works just fine
                stage('Hello world on Linux') {
                    agent {
                      docker {
                          label 'linux&&docker&&aws'
                          image 'alpine:latest'
                      }
                    }
                    options {
                     timeout(1) // in case the pipeline hangs
                    }
                    steps {
                     sh 'echo "hello world"'
                    }
                }
      }
    }
  }
}

Interesting. So different Windows Server OS and different images than mine. Also interesting that your powershell commands hang while ours do not.  Do things work as expected when you manually start or attach to the containers?

Might be irrelevant but what method did you use to install Docker on the host node?  We used this.

Install-Module -Name DockerMsftProvider -Repository PSGallery -Force
Install-Package -Name docker -ProviderName DockerMsftProvider
Restart-Computer -Force

And also had to do these steps (install git-bash and update paths) to get around a nohup error. When running the Jenkins jobs.

https://stackoverflow.com/questions/45140614/jenkins-pipeline-sh-fail-with-cannot-run-program-nohup-on-windows/53395989#53395989

 I'm curious if the chosen workaround for nohup has something to do with it in our case. There is a PR to address the root of the nohup issue. Perhaps I will try building that PR and see if it solves the issue. But would still like to know more about your setup.

https://github.com/jenkinsci/pipeline-model-definition-plugin/pull/354

I'm taking a research day, so I'm working from home and haven't tried anything on the Windows server at work.

Anyways, I tried to get a Windows Docker container working in Jenkins on my home Windows 10 machine and it worked.

I'm thinking the server at work must have a configuration incorrectly set. I just wish I could figure out what the heck the problem is. I'm not even sure where to look.

I agree it may be server config related but hard to understand why or where to start down that path. I feel like the nohup thing is a logical place to start as it has to do with how the plugin interfaces with the container shells.

Did you run into the nohup initially and if so how did you solve it?

We haven't experienced this.

Our host OS is windows 2019.

nohup is never called during bat.

Our workaround for nohup when using `sh` is to install git using chocolatey with GitAndUnixToolsOnPath  parameter.

This is our Jenkins windows agents (our Windows base images only have Docker installed)

Today I would use the existing DockerImage created in Jenkins org: https://github.com/jenkinsci/docker-jnlp-slave/blob/master/Dockerfile-windows

FROM mcr.microsoft.com/windows/servercore:ltsc2019

SHELL ["powershell", "-Command", "$ErrorActionPreference = 'Stop'; $ProgressPreference = 'SilentlyContinue';"]

RUN Set-ExecutionPolicy Bypass -Scope Process -Force; iex ((New-Object System.Net.WebClient).DownloadString('https://chocolatey.org/install.ps1')); `
  cinst docker-cli docker-compose adoptopenjdk8jre powershell-preview -y --no-progress; `
  cinst git --params '/GitAndUnixToolsOnPath /SChannel' -y --no-progress

# temporary fix for powershell preview
RUN [Environment]::SetEnvironmentVariable("Path", $env:Path + ";C:\Program Files\PowerShell\7-preview", [EnvironmentVariableTarget]::Machine)

# Adding Jenkins Setup last to avoid rebuild
ADD JenkinsAgentSetup.ps1 .

ENTRYPOINT ["powershell.exe","-executionpolicy", "bypass", "./JenkinsAgentSetup.ps1"]

This is how we run the agent

docker run `
  --dns 10.0.0.1 `
  --dns 1.1.1.1 `
  --dns-search company.io `
  --name jenkins.agent `
  -v "$($WORKSPACE):$($JENKINS_WORKSPACE)" `
  -v \\.\pipe\docker_engine:\\.\pipe\docker_engine `
  -e "JENKINS_MASTER_URL=$JENKINS_MASTER_URL" `
  -e "JENKINS_AGENT_NAME=$JENKINS_AGENT_NAME" `
  -e "JENKINS_AGENT_PARAMETERS=$JENKINS_AGENT_PARAMETERS" `
  -e "JENKINS_AGENT_SECRET=$JENKINS_AGENT_SECRET" `
  artifactory.company.io/docker/jenkins.agent.windows:latest

I'd highly encourage using Windows Server 2019 cause it allows you to pipe your docker engine.

Our experience with Windows Server 2016 was a HUGE nope.

It started working on our end when we upgraded the version of Docker to the latest. 

It looks related to JENKINS-59903, which was fixed in Jenkins version 2.201.
Sorry, it was reported, not fixed.

If you pass `–isolation=hyperv` you can run images based on any windows kernel, regardless of what kernel the host is on.

solvingj Does not work for me. At least not when trying to pull a 1903 image on Server 2019 1809. Maybe I only works for backward comparability, not forward? 

>docker build --isolation="hyperv" -t "test_full" -f Dockerfile_1903 . Sending build context to Docker daemon 13.82kB Step 1/7 : FROM mcr.microsoft.com/powershell:7.0.0-preview.5-nanoserver-1903 7.0.0-preview.5-nanoserver-1903: Pulling from powershell a Windows version 10.0.18362-based image is incompatible with a 10.0.17763 host

Edit: If I try and older version like nanoserver-1803 I get "The container operating system does not match the host operating system." on a powershell step.without the --isolation flag. When adding the flag I get "The request is not supported." on the same step.

I don't use quotes, but i don't think that's the issue.  I think you either need to be on a newer version of docker, or you need to enable experimental features.  

We're on the latest version 19.03.4 on Server 2019 (1809) now. Just edited my previous comment with more info. I get different results from docker build if I add the --isolation flag. Fails either way but getting a different result makes me thing it's attempting to apply the isolation setting. Either way I don't think it will solve our issues.

This matrix is how I learned about it.  It might be helpful to you: https://docs.microsoft.com/en-us/virtualization/windowscontainers/deploy-containers/version-compatibility

We've now tried this on a Windows Server 2019 host (1809) and a vanilla{{ }}mcr.microsoft.com/windows/servercore:1809 image and the bat call still hangs.  Even worse when we try a nanoserver image such as mcr.microsoft.com/powershell:nanoserver-1809 it will hang on both powershell and bat and when the Jenkins job is manually cancelled the server will experience a critical error and reboot itself!

Critical: The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly.

How did you install docker?
Please follow this guide
https://docs.docker.com/install/windows/docker-ee/

That is the method we used for install.

The curious thing is that sometimes the bat calls work. We have some large and complex pipeline scripts in which many or even most of the bat calls will work but some still fail and the hello world example above fails ever time.

I suspect that something else in the larger pipelines are inadvertently sidestepping / “fixing” the issue in real time. So there might be some kind of context in which the calls work and another (like the hello world) where they don’t. Haven’t been able to narrow it down yet.

Wish I could help you but we haven't experienced the issue