activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Danielius Jurna <daniel...@elitnet.lt>
Subject Broker state and racing conditions
Date Tue, 05 Dec 2006 06:58:50 GMT

While analyzing http://issues.apache.org/activemq/browse/AMQ-724 issue it
seems that broker state is spread across several objects and modification of
this state is not synchronized. On connection add/remove modification is
made in many objects (you can see it in AbstractConnection), so on heavy
load it can leave broker in inconsistent state. Bug mentioned above exactly
shows the problem. And the fix is only dealing with consequences, not with
the cause. This bug never occures when synchronization is made in
AbstractConnection on add/remove operations. So it seems that current
architecture is more towards performance than roubustness.
Any thoughts?
-- 
View this message in context: http://www.nabble.com/Broker-state-and-racing-conditions-tf2759642.html#a7694274
Sent from the ActiveMQ - Dev mailing list archive at Nabble.com.


Mime
View raw message