-
Bug
-
Resolution: Fixed
-
Major
-
None
-
2 * E5-2690 (16 real cores, 32 inc hyperthreading)
112 MB RAM
-Xmx24g -Xms16g
The slow request checker pegs the JDK VM_Thread at 100% cpu as it is in a constant loop grabing thread dumps.
Getting full thread dumps on a largish system (18 slaves, 5 polls, several web requests) every 3 seconds is not a great idea.
There is no way to tune this and the only way to fix this is to uninstall the support core plugin and restart jenkins.
depending on the JDK you may be able to terminate the Slow request thread with the following
import jenkins.util.Timer import java.util.concurrent.* def timer = Timer.get() def tasks = timer.queue.toArray() def field = FutureTask.class.getDeclaredField("callable") field.setAccessible(true) for (task in tasks) { if (field.get(task).task.class == com.cloudbees.jenkins.support.slowrequest.SlowRequestChecker.class) { println(task.toString() + " <-----========== SlowRequestChecker") println(task.cancel(true)) println(timer.queue.remove()) } }
Once the requestChecker is removed the UI response is near instantanious.