Slow message dispatch

[Mohit Anchlia <mo...@gmail.com>]

I am looking at an issue where I am losing about 120 ms between posting a
message to the queue and receiving a message on the queue. Is there a way
to debug and see where it might be spending the extra time?

Re: Slow message dispatch

[Mohit Anchlia <mo...@gmail.com>]

Broker is non-persistent. I am sending very small messages and the measure
is only between when I send the message and at the other end when I receive
the message. Consumers and producers are not busy at this point.

I am not re-creating sessions, I am using thread local and I am caching
sessions. I am also not re-creating temp queues, that also gets cached
using thread local.

Should I profile activemq? Is there an easy tutorial on how to profile
activemq?


On Thu, Aug 28, 2014 at 7:32 PM, artnaseef <ar...@artnaseef.com> wrote:

> First, if the broker is using persistence, try turning off persistence
> (persistent=false in the broker tag).  If that eliminates the delay, it's
> related to storage.  If not, storage is not involved.
>
> Using a java profiler, such as Yourkit, also comes to mind.
>
> Is there any benchmark against which the 120ms is compared to determine it
> is a "long time"?  Also, are the producer and consumer both on the same
> server as ActiveMQ to rule out network latency?  And, is this delay seen
> consistently across a large number of produced and consumed messages?
>
> BTW - another thing to check - make sure the producer and consumer are not
> closing and recreating connections; ActiveMQ (like all JMS) connections are
> intended to be long lived, processing large numbers of messages in a single
> connection lifetime.
>
>
>
> --
> View this message in context:
> http://activemq.2283324.n4.nabble.com/Slow-message-dispatch-tp4685115p4685123.html
> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>

Re: Slow message dispatch

[artnaseef <ar...@artnaseef.com>]

First, if the broker is using persistence, try turning off persistence
(persistent=false in the broker tag).  If that eliminates the delay, it's
related to storage.  If not, storage is not involved.

Using a java profiler, such as Yourkit, also comes to mind.

Is there any benchmark against which the 120ms is compared to determine it
is a "long time"?  Also, are the producer and consumer both on the same
server as ActiveMQ to rule out network latency?  And, is this delay seen
consistently across a large number of produced and consumed messages?

BTW - another thing to check - make sure the producer and consumer are not
closing and recreating connections; ActiveMQ (like all JMS) connections are
intended to be long lived, processing large numbers of messages in a single
connection lifetime.



--
View this message in context: http://activemq.2283324.n4.nabble.com/Slow-message-dispatch-tp4685115p4685123.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.