We also just recently started seeing this issue within the last couple of weeks I believe.
In my configuration, I am loading all 5 executors with the same job via remote triggering using a curl statement without any delay. This job spawning configuration has not been changed in several months. Recently, however, I have noticed this defect's error turn up and all jobs fail.
I am thinking there is contention against the destination maven installation folder.
Environment details... Ubuntu Jenkins LTS 1.642.2 (will be 1.642.3 this weekend) Windows 2008 Slave 5 Executors Maven 3.0.4 JDK 8
While my scenario is around concurrently executing a single maven project, I expect any combination of multiple Maven projects could result in this condition. This is unless Jenkins has some built in throttle mechanism which is not being respected while using a single project concurrently.
Unfortunately, the Windows slave in my case is effectively dead to Maven projects after this and I need to kill the Java process on the system before it will serve Maven projects again.
I agree with the Critical status of this Defect.
|