[jboss-dev-forums] [Design of Messaging on JBoss (Messaging/JBoss)] - JBMESSAGING-565 - Race condition on QueueManagementTest

[clebert.suconic@jboss.com]

This thread is to discuss the racing condition between getMessageCount and Delivery Thread.


Just for convenience I have copied the JIRA description which has some relevant information for this discussion.

anonymous wrote : 
  | QueueManagementTest::testMessageCountOverFullSize has a race condition between getMessageCount and Deliver Thread.
  | 
  | PagingChannelSupport::messageCount is an arithmethic expressions of messageRefs.size() + deliveries.size() + nextPagingOrder - firstPagingOrder
  | 
  | deliveries.size is being controlled by Delivery Thread, determined by DeliveryRunnable that runs as-synchronously on Destination.
  | 
  | While getMessageCount might diverge momentarily it will eventually be fixed as soon as DeliveryRunnable has the appropriate time to send the result to the collection deliveries.
  | 
  | To fix this we would need to synchronize delivery and messageCount somehow, probably adding global lock, what we don' t want to add I assume. So, if this is correct the test has to be fixed somehow.
  | 
  | I have changed the test in such way it will make easier to identify the racing condition. Use tracing in your tests and you will see a retry routine that will eventually pass.

View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=3975072#3975072

Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=3975072