When a Pipeline is triggered while the master node has been put temporarily offline, a OneOff (flyweight) executor is leaked.

As I'm unsure whether the issue lies in core or in the pipelines at put both components.

Environment

• Version: 2.121.3
• workflow-aggregator:2.5
• workflow-cps:2.54

 
Steps to reproduce

• Create a simple Pipeline Job, eg:

node('master') {
    sh 'echo hello'
}

• Put the master node temporary offline
• Trigger the Pipeline build
• Put the master node back online
• Observe the leaked executor

Observations

This executor has the following characteristics:

Executor #-1(-1)
 Active? false
 Likely Stuck? false
 Interrupted? false
 Busy? false
 Owner? hudson.model.Hudson$MasterComputer@228777fd
 Current Work Unit? null
 Causes Of Interruption? []
 Idle Start Milliseconds? 1534847122799
 Progress: -1
 Asynchronous Execution: null
 Executable: null
 executorOwner: hudson.model.Hudson$MasterComputer@228777fd
  tiedJobs: []

Jenkins logs show:

Executor #-1 for master grabbed hudson.model.queue.WorkUnit@4891a81f[work=hello] from queue but master went off-line before the task's worker thread started. No termination trace available.

Pipeline logs show nothing:

Démarré par l'utilisateur admin
Running in Durability level: MAX_SURVIVABILITY
[Pipeline] node
Running on Jenkins in /Users/pierrebeitz/cbsupporthome/jenkins-home/workspace/hello
[Pipeline] {
[Pipeline] sh
[hello] Running shell script
+ echo hello
hello
[Pipeline] }
[Pipeline] // node
[Pipeline] End of Pipeline
Finished: SUCCESS

Workaround

A restart of Jenkins will cleanup the executors.