I have a setup with a main Jenkins server and two static VMs configured as agents. One VM runs Ubuntu with Docker installed. The other static VM runs Windows 10 so that I can run tests on Windows.

The following pipeline hangs on the bat step in the 2nd stage. I've let it run overnight with no progress.

def staticVMWithDockerLabel = // label for the Ubuntu static agent with Docker
def staticVMWindowsLabel = // label for the Windows static agent.

pipeline {
  agent {
    docker {
      label staticVMWithDockerLabel
      image 'ubuntu:20.04'
      alwaysPull true
    }
  }
  stages {
    stage('Main Agent') {
      steps {
        echo 'On main agent'
        sh 'echo script ran on main agent'
        echo 'End main agent'
      }
    }
    stage('Windows Agent') {
      agent { label staticVMWindowsLabel }
      steps {
        echo 'On Windows agent'
        bat 'echo script ran on Windows agent'
        echo 'End Windows agent'
      }
    }
    stage('Main Agent Again') {
      steps {
        echo 'On main agent again'
        sh 'echo script ran on main agent again'
        echo 'End main agent again'
      }
    }
  }
}
 

When this pipeline runs:

1. The Docker instance is started as expected.
2. The first stage, 'Main Agent' runs successfully within the Docker instance. The console includes the expected text.
3. The second stage starts successfully. But then it hangs on the bat step. The console output ends with:
   [Pipeline] stage
   [Pipeline] { (Other Agent)
   [Pipeline] node
   Running on <agent name> in <workspace>
   [Pipeline] {
   [Pipeline] echo
   On Windows agent
   [Pipeline] bat
   No additional content in console output

Alternatives:

• If using dockerfile instead of docker the same hang occurs. This is my actual use case; using docker in the above example makes it easier to reproduce.
• If I stop using Docker and instead use the static VM agent directly (i.e., If I replace the pipeline level agent with "agent { label staticVMWithDockerLabel }") then everything works as expected. However, I lose the ability to control agent configuration per project that Jenkins' Docker integration gives me.
• If I change the pipeline level to "agent none" and put the Docker agent on stages 1 & 3, it works. However, I lose the ability for stages 1 & 3 to run in the same container instance.