[ https://issues.apache.org/jira/browse/POOL-337?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16433251#comment-16433251 ]
Phil Steitz edited comment on POOL-337 at 4/11/18 4:03 PM: ----------------------------------------------------------- Sorry, I missed the unit test in the updated PR. In my testing I did not configure a listener to get the swallowed exceptions. I think the analysis is correct and the patch should fix the issue. It might be good to consider a little refactoring that would allow the GOP to hold a reference to the task that it needs to cancel (i.e. change the type of the Evictor) so you would not need to maintain the map in the EvictionTimer. was (Author: psteitz): Sorry, I missed the unit test in the updated PR. What I am still not getting is how the stack trace at the top can happen. I can see how the memory leak is a problem and I think there is definitely a bug here, but I would like to be able to reproduce this directly from the GOP API, which I can't do. The underlying Evictor task should get cancelled while the eviction lock is held, but somehow the executor is still firing it in the trace at the top. > EvictionTimer does not remove cancelled tasks from the executor, leading to > an IllegalStateException when the evictor attempts to evict > --------------------------------------------------------------------------------------------------------------------------------------- > > Key: POOL-337 > URL: https://issues.apache.org/jira/browse/POOL-337 > Project: Commons Pool > Issue Type: Bug > Affects Versions: 2.5.0 > Reporter: Reinald Verheij > Priority: Major > Attachments: EvictionTimer.java, > EvictionTimer.java.original-from-2.5.0.java > > > EvictionTimer does not remove cancelled tasks from the executor, leading to > an IllegalStateException when the evictor attempts to evict. > > EvictionTimer::schedule() adds eviction tasks to the executor, but the cancel > does not remove it. This is asymmetric and leads to the following exception: > {noformat} > java.lang.IllegalStateException: Pool not open > at > org.apache.commons.pool2.impl.BaseGenericObjectPool.assertOpen(BaseGenericObjectPool.java:713) > at > org.apache.commons.pool2.impl.GenericObjectPool.evict(GenericObjectPool.java:721) > at > org.apache.commons.pool2.impl.BaseGenericObjectPool$Evictor.run(BaseGenericObjectPool.java:1077) > at > java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) > at java.util.concurrent.FutureTask.runAndReset(FutureTask.java:308) > at > java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$301(ScheduledThreadPoolExecutor.java:180) > at > java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:294) > at > java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) > at > java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) > at java.lang.Thread.run(Thread.java:748){noformat} > I think the cancel would need to remember the future which returned from > {{executor::scheduleWithFixedDelay()}} in {{schedule()}} and then do > something like this (see [^EvictionTimer.java] compared to original > [^EvictionTimer.java.original-from-2.5.0.java] ) > {code:java} > if (futures.containsKey(task)) { > futures.remove(task).cancel(false); > executor.purge(); > }{code} -- This message was sent by Atlassian JIRA (v7.6.3#76005)