[infinispan-issues] [JBoss JIRA] (ISPN-9979) AbstractComponentRegistry.stop() can hang if called concurrently

[Philippe Julien (Jira)]

Philippe Julien created ISPN-9979:
-------------------------------------

             Summary: AbstractComponentRegistry.stop() can hang if called concurrently
                 Key: ISPN-9979
                 URL: https://issues.jboss.org/browse/ISPN-9979
             Project: Infinispan
          Issue Type: Bug
    Affects Versions: 9.4.3.Final
            Reporter: Philippe Julien



I believe that there is a bug in org.infinispan.factories.AbstractComponentRegistry.stop()

Our Wildfly 15 nodes often hang on shutdown on the following:

{noformat}
"MSC service thread 1-2" #37 prio=5 os_prio=0 tid=0x0000000003807000 nid=0xf32d in Object.wait() [0x00007f0012c6f000]
   java.lang.Thread.State: WAITING (on object monitor)
	at java.lang.Object.wait(Native Method)
	- waiting on <0x00000005cc795fa8> (a org.infinispan.factories.ComponentRegistry)
	at java.lang.Object.wait(Object.java:502)
	at org.infinispan.factories.AbstractComponentRegistry.stop(AbstractComponentRegistry.java:359)
	- locked <0x00000005cc795fa8> (a org.infinispan.factories.ComponentRegistry)
	at org.infinispan.cache.impl.CacheImpl.performImmediateShutdown(CacheImpl.java:1159)
	at org.infinispan.cache.impl.CacheImpl.stop(CacheImpl.java:1124)
	at org.infinispan.cache.impl.AbstractDelegatingCache.stop(AbstractDelegatingCache.java:520)
	at org.infinispan.manager.DefaultCacheManager.terminate(DefaultCacheManager.java:734)
	at org.infinispan.manager.DefaultCacheManager.stopCaches(DefaultCacheManager.java:786)
	at org.infinispan.manager.DefaultCacheManager.stop(DefaultCacheManager.java:762)
	at org.jboss.as.clustering.infinispan.subsystem.CacheContainerServiceConfigurator.accept(CacheContainerServiceConfigurator.java:114)
	at org.jboss.as.clustering.infinispan.subsystem.CacheContainerServiceConfigurator.accept(CacheContainerServiceConfigurator.java:70)
	at org.wildfly.clustering.service.FunctionalService.stop(FunctionalService.java:77)
	at org.jboss.msc.service.ServiceControllerImpl$StopTask.stopService(ServiceControllerImpl.java:1794)
	at org.jboss.msc.service.ServiceControllerImpl$StopTask.execute(ServiceControllerImpl.java:1763)
	at org.jboss.msc.service.ServiceControllerImpl$ControllerTask.run(ServiceControllerImpl.java:1558)
	at org.jboss.threads.ContextClassLoaderSavingRunnable.run(ContextClassLoaderSavingRunnable.java:35)
	at org.jboss.threads.EnhancedQueueExecutor.safeRun(EnhancedQueueExecutor.java:1985)
	at org.jboss.threads.EnhancedQueueExecutor$ThreadBody.doRunTask(EnhancedQueueExecutor.java:1487)
	at org.jboss.threads.EnhancedQueueExecutor$ThreadBody.run(EnhancedQueueExecutor.java:1364)
	at java.lang.Thread.run(Thread.java:748)
{noformat}

There are no other threads that are working with 0x00000005cc795fa8 in the thread dump.

I checked in a heap dump and the 0x00000005cc795fa8 org.infinispan.factories.ComponentRegistry object state is TERMINATED.

I think that the finally block of AbstractComponentRegistry.stop() is missing a notifyAll().

Shouldn't this:

{code}
...
      } finally {
         synchronized (this) {
            state = ComponentStatus.TERMINATED;
         }
      }
{code}

Be:
{code}
...
      } finally {
         synchronized (this) {
            state = ComponentStatus.TERMINATED;
            notifyAll();
         }
      }
{code}

This way, if two thread try to stop the same cache, the one that wins will notify the one that is waiting letting its stop complete.




--
This message was sent by Atlassian Jira
(v7.12.1#712002)