What's the accuracy of the QMF statistics maintained on the C++ broker?

[fadams <fr...@blueyonder.co.uk>]

Hi all,
I'm curious about the statistics maintained on the broker.

In particular I've been sending messages to the amq.match exchange and
binding to a queue. Now I've deliberately set it up such that my binding
only matches 2/3 of the messages I'm sending but looking at some of the
stats from QMF there's a bit of oddness.

Looking at the enqueues and dequeues these seem to match what I'd expect,
but looking at the exchange stats the number of messages looks wrong - both
the msg receives and messages routed look a bit low. The messages routed
ought to tie up to the number of messages enqueued but it doesn't.

Are there any gotchas I should be aware of.

I've been using RedHat's Cumin user interface to look at this stuff (just
started playing around with Cumin it looks pretty useful) I guess it could
be something to do with Cumin's polling but I refreshed for a while after I
stopped sending messages and things didn't change.

It would be good to be able to account for this as I'm tempted to go down
the Cumin route as some of my users just have to have some eye candy, but I
don't want to confuse matters with misleading stats.


On a related note I see that the exchange stats record msg drops, which
presumably is the number of "unroutable" messages due to there being no
matching binding - that has got me thinking. Is there any way to actually
deliver messages that don't match an explicit binding on an exchange to a
default queue (I'm particularly interested in this for the headers
exchange).


Finally although Cumin seems quite good (though a bit of a fight to get
running on Ubuntu) I was interested in other options for a GUI for the C++
broker is anything else out there? What's the approach actually recommended
by the Qpid community?

Cheers,
Frase








--
View this message in context: http://apache-qpid-users.2158936.n2.nabble.com/What-s-the-accuracy-of-the-QMF-statistics-maintained-on-the-C-broker-tp6758393p6758393.html
Sent from the Apache Qpid users mailing list archive at Nabble.com.

---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project:      http://qpid.apache.org
Use/Interact: mailto:users-subscribe@qpid.apache.org

Re: What's the accuracy of the QMF statistics maintained on the C++ broker?

[Gordon Sim <gs...@redhat.com>]

On 09/04/2011 02:56 PM, fadams wrote:
> Hi all,
> I'm curious about the statistics maintained on the broker.
>
> In particular I've been sending messages to the amq.match exchange and
> binding to a queue. Now I've deliberately set it up such that my binding
> only matches 2/3 of the messages I'm sending but looking at some of the
> stats from QMF there's a bit of oddness.
>
> Looking at the enqueues and dequeues these seem to match what I'd expect,
> but looking at the exchange stats the number of messages looks wrong - both
> the msg receives and messages routed look a bit low. The messages routed
> ought to tie up to the number of messages enqueued but it doesn't.

Is it higher or lower? Do the drops and routes add up to the receives?

> Are there any gotchas I should be aware of.

None that I am aware of and some trivial tests seem to give me the 
numbers I expect. Can you describe how to reproduce the discrepency you see?

> I've been using RedHat's Cumin user interface to look at this stuff (just
> started playing around with Cumin it looks pretty useful) I guess it could
> be something to do with Cumin's polling but I refreshed for a while after I
> stopped sending messages and things didn't change.
>
> It would be good to be able to account for this as I'm tempted to go down
> the Cumin route as some of my users just have to have some eye candy, but I
> don't want to confuse matters with misleading stats.
>
>
> On a related note I see that the exchange stats record msg drops, which
> presumably is the number of "unroutable" messages due to there being no
> matching binding - that has got me thinking. Is there any way to actually
> deliver messages that don't match an explicit binding on an exchange to a
> default queue (I'm particularly interested in this for the headers
> exchange).

You can set an 'alternate-exchange' for the exchange such that the 
broker attempt to re-route any unroutable messages through that 
alternate exchange. So e.g. you can create a fanout exchange as the 
alternate exchange and then any queue bound to that will get all dropped 
messages.

> Finally although Cumin seems quite good (though a bit of a fight to get
> running on Ubuntu) I was interested in other options for a GUI for the C++
> broker is anything else out there?

Not that I am aware of.

---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project:      http://qpid.apache.org
Use/Interact: mailto:users-subscribe@qpid.apache.org