If Jenkins is restarted whilst a workflow is running using slaves from mansion cloud then the result is an unkillable build on the jenkins ui.
this is easiest to reproduce cloudbees Dev@Cloud as the mansion cloud is already setup there.
1) create a simple workflow that allocates a node and sleeps for a resonable amount of time.
2) start the build and verify it completes
3) start the build and then restart jenkins.
- when the master comes back up it reconnects to the salve that was running the task and resumes.
- when the master comes back up it reconnects to the salve that was running the task and fails
- when the master comes back up it keeps attempting to reconnect to the salve that was running the task and the user can manually abort the workflow at this time.
the master comes back up and the workflow resumes - but it does not appear to reconnect to the slave, and nor can you kill the build with the red "x" button in the jenkins ui,
The only way to kill the build is to delete the build and restart jenkins again.
The problem with this workaround is that it is likely another workflow will have been killed when you do this restart and the whole cycle of kill -> reboot -> kill reboot starts again.