[
https://issues.apache.org/jira/browse/QPID-540?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Martin Ritchie resolved QPID-540.
---------------------------------
Resolution: Fixed
Prevent NPE when purging message from the main _message queue in the
ConcurrentSelectorDeliveryManager that have been delivered via a Subscribers
_messageQueue. Ensuring that any expired messages are still correctly handled.
i.e. the Queue size/depth is reduced and the message correctly dequeued from
the underlying store.
> Transient Broker throws NullPointerException and locks up.
> ----------------------------------------------------------
>
> Key: QPID-540
> URL: https://issues.apache.org/jira/browse/QPID-540
> Project: Qpid
> Issue Type: Bug
> Components: Java Broker
> Affects Versions: M2
> Reporter: Martin Ritchie
> Assignee: Martin Ritchie
> Fix For: M2
>
>
> During the testing with the transient configuration the broker threw the
> following.
> 2007-07-11 12:34:24,187 INFO [pool-1278-thread-1] handler.QueueBindHandler
> (QueueBindHandler.java:115) - Bind
> ing queue Queue(my_queue)@10656878 to exchange
> org.apache.qpid.server.exchange.DestNameExchange[amq.direct] wi
> th routing key my_queue
> Exception in thread "pool-1-thread-13" java.lang.NullPointerException
> at
> org.apache.qpid.server.queue.WeakReferenceMessageHandle.populateFromMessageMetaData(WeakReferenceMe
> ssageHandle.java:80)
> at
> org.apache.qpid.server.queue.WeakReferenceMessageHandle.loadMessageMetaData(WeakReferenceMessageHan
> dle.java:74)
> at
> org.apache.qpid.server.queue.WeakReferenceMessageHandle.getContentHeaderBody(WeakReferenceMessageHa
> ndle.java:64)
> at
> org.apache.qpid.server.queue.AMQMessage.getContentHeaderBody(AMQMessage.java:328)
> at
> org.apache.qpid.server.queue.AMQMessage.getSize(AMQMessage.java:906)
> at
> org.apache.qpid.server.queue.ConcurrentSelectorDeliveryManager.getNextMessage(ConcurrentSelectorDel
> iveryManager.java:461)
> at
> org.apache.qpid.server.queue.ConcurrentSelectorDeliveryManager.sendNextMessage(ConcurrentSelectorDe
> liveryManager.java:551)
> at
> org.apache.qpid.server.queue.ConcurrentSelectorDeliveryManager.processQueue(ConcurrentSelectorDeliv
> eryManager.java:684)
> at
> org.apache.qpid.server.queue.ConcurrentSelectorDeliveryManager.access$200(ConcurrentSelectorDeliver
> yManager.java:50)
> at
> org.apache.qpid.server.queue.ConcurrentSelectorDeliveryManager$Runner.run(ConcurrentSelectorDeliver
> yManager.java:856)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:650)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:675)
> at java.lang.Thread.run(Thread.java:595)
> This must be due to a message being dequeued but still existing on the queue.
> The problem is down to the use of selectors.
> If a message is consumed via a selector it is marked as taken and delivered
> to the consumer. If the reference count hits zero which it will if the
> message was only sent to the one queue. Then the message will be dequeued and
> any WeakReference Handles deleted. All as expected. HOWEVER,
> The message is still on the main queue. A warning message is logged saying
> "We could cleanup the main queue _messages here." It should say SHOULD.
> The message is referenced by the main queue and the next subscription
> (without a filter) that attempts to read from the queue will walk the
> _message queue. The getNextMessage() method will purge the messages that have
> expired or been taken. Such as this taken message from the filetered
> consumer. Problem is, here we attempt to reduce the _totalMessageSize which
> has already been done when the message was delivered. Thus in attempting to
> retrieve the size of the message we need to access the Headers that were on a
> WeakReferenceHandle and so are now null causing the NPE.
> The solution is to only attempt to reduce the _totalMessageSize when we are
> purging a message due to a time out. As this is the only path through the
> purgeMessage() method that state the message should be purge and the message
> has not already been delivered to a consumer of the queue.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
