Luciano Mollea wrote:
>
> Hi all.
> We're experiencing a strange situation here. It seems that randomly AMQ
> does not send a message, this happens depending on the load, this problem
> seems to happen more frequently as more threads concurrently send messages
> on the same queue.
>
> [...]
>
> [*] a thought about this: the send is not synchronized on the sender
> object (that is used multi-threaded) but on the sent message.. could this
> be the cause?
>
I reply myself as it seems we found the solution (and it is as we
suspected).
The send method of the sender object should be enclosed in a synchronized
block like
synchronized(sender) {
sender.send(message);
}
Yet, this is something that is not strictly required from the JMS specs
(ckeck for example here:
http://java.sun.com/javaee/5/docs/api/javax/jms/QueueSender.html):
[...] During the execution of its send method, a message must not be changed
by other threads within the client. If the message is modified, the result
of the send is undefined.[...]
It requires that the message be immutable for the send operation, it does
not strictly requires that only one send can be done at a time. To be
honest, it doesn't specify either that multiple sends may be done
concurrently.
I't some sort of "grey area" in the specs that may change from
implementation to implementation.
I'm not filing it as a bug, but I'd suggest that the send operations should
be made thread safe. If someone switches from an thread-safe implementation
of another JMS broker to ActiveMQ, he may consider ActiveMQ as "broken".
I think that it could be done by simply adding a synchronized declaration on
all the "send" (and maybe the "publish" of the TopicPublisher) methods.
Thanks the same, even to those that replied off-list.
--
View this message in context:
http://www.nabble.com/Messages-in-AMQ-are-randomly-not-sent--tf3572512s2354.html#a10011745
Sent from the ActiveMQ - User mailing list archive at Nabble.com.
