Status: Resolved (View Workflow)
Sometimes when the build node is stressed, the plugin terminates abnormally and kills the executor. Jenkins allows to restart the executor, and then the build process resumes as normal. Unfortunately, the stack trace below is the only evidence I have got:
java.lang.NullPointerException at hudson.plugins.throttleconcurrents.ThrottleQueueTaskDispatcher.buildsOfProjectOnNode(ThrottleQueueTaskDispatcher.java:121) at hudson.plugins.throttleconcurrents.ThrottleQueueTaskDispatcher.canTake(ThrottleQueueTaskDispatcher.java:80) at hudson.model.queue.QueueTaskDispatcher.canTake(QueueTaskDispatcher.java:101) at hudson.model.Queue$JobOffer.canTake(Queue.java:214) at hudson.model.Queue.pop(Queue.java:799) at hudson.model.Executor.grabJob(Executor.java:276) at hudson.model.Executor.run(Executor.java:206)
- is related to
JENKINS-15094 Parameters: NPE in canTake() procedures may kill all executors
This problem occurred only under high load on Jenkins, which makes me think it's related to some subtle race condition. It used to happen very seldom, about once in 1-2 months.
Plugin version was 1.6 at the time when I logged the issue. Now we're using 1.7.1 and the problem did not recur for a long time. I'll update this ticket if I see it again with the latest plugin version.
JENKINS-15094, which has been fixed in newest Jenkins core versions.
The plugin has been significantly modified during last years, this issue should not appear anymore. Marking as Fixed
What version of the throttle plugin are you seeing this with?