-
Bug
-
Resolution: Duplicate
-
Minor
-
None
-
Jenkins 1.625.3
Windows slaves (JNLP)
Deadlock situations similar to those described in remoting#36
However, in these cases the other side of the stack trace was:
"NioChannelHub keys=3 gen=41: Computer.threadPoolForRemoting [#2]" id=224 (0xe0) state=BLOCKED cpu=76%
- waiting to lock <0x28a9d2ba> (a hudson.remoting.Channel)
owned by "Computer.threadPoolForRemoting [#5] for XXXXXXX" id=249 (0xf9)
at hudson.remoting.Channel.terminate(Channel.java:833)
at hudson.remoting.Channel$1.terminate(Channel.java:509)
at hudson.remoting.AbstractByteArrayCommandTransport$1.terminate(AbstractByteArrayCommandTransport.java:71)
at org.jenkinsci.remoting.nio.NioChannelHub$NioTransport.abort(NioChannelHub.java:208)
at org.jenkinsci.remoting.nio.NioChannelHub.run(NioChannelHub.java:637)
at jenkins.util.ContextResettingExecutorService$1.run(ContextResettingExecutorService.java:28)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
That is, an abort caused by an CancelledKeyException (see JENKINS-24050).
The already merged solution in remoting#36 does not seem to cover all cases, as if there are not writable bytes and no one is reading (the channel is in an abnormal situation) the loop may keep going forever, maintaining the deadlock.
As abort starts by closing the ends of the NIO channel, additional closed state change checks can be introduced in the loop providing a way out.
- duplicates
-
-
- Resolved
-
- is related to
-
-
- Resolved
-
- links to