False alarm here. I had subtly and inadvertently introduced some threading with Pika doesn't like too much.<br><br>Once those issues are fixed, things seem to work as expected.<br><br>Thanks for the quick response Matthias!<br>
<br><div class="gmail_quote">On Tue, Aug 14, 2012 at 9:29 AM, Matt Pietrek <span dir="ltr"><<a href="mailto:mpietrek@skytap.com" target="_blank">mpietrek@skytap.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Hmm... I'm using Pika. Perhaps it's feeling overloaded and dropping the connection. At least I have a new direction to pursue.<br><br>Thanks!<div class="HOEnZb"><div class="h5"><br><br><div class="gmail_quote">On Tue, Aug 14, 2012 at 8:59 AM, Matthias Radestock <span dir="ltr"><<a href="mailto:matthias@rabbitmq.com" target="_blank">matthias@rabbitmq.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Matt,<div><br>
<br>
On 14/08/12 01:25, Matt Pietrek wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
I'm hitting an issue where a queue disappears from underneath me in a<br>
high volume federated scenario.<br>
<br>
Background: I'm on RabbitMQ 2.8.5 on top of Ubuntu 10.04.<br>
<br>
I have three brokers bi-directionally federated similarly to the "Small<br>
complete Graph" example in <a href="http://www.rabbitmq.com/federation.html" target="_blank">http://www.rabbitmq.com/<u></u>federation.html</a>. My<br>
exchange is called "Skytap", and is a topic exchange. One of the routing<br>
key 'words' directs messages to the appropriate broker.<br>
<br>
In my example, I have a single master sending "requests" to two slaves<br>
which reply back to the master, simulating an RPC scenario. A client app<br>
on the 'master' broker reads from an exclusive queue called 'master",<br>
while client apps on the slave brokers read from an exclusive queue<br>
called "slave'.<br>
<br>
In the low volume case, everything works as expected: A steady stream of<br>
request messages seen on the slaves and a steady stream of reply<br>
messages seen on the master.<br>
<br>
However, if I write "request" messages as fast as possible, in short<br>
order I see the 'master' queue disappear. (I know this because I have<br>
mnesia logging turned on.)<br>
</blockquote>
<br></div>
Since the 'master' queue is exclusive, and there is nothing unusual in the rabbit logs indicating any rabbit-internal problems (you've checked the sasl log as well, right?) the only way that queue can vanish is for it to be either deleted explicitly by the creating connection or for the connection to have been closed.<br>
<br>
Can't see how federation would have anything to do with this since it is not touching user queues.<span><font color="#888888"><br>
<br>
Matthias.<br>
</font></span></blockquote></div><br>
</div></div></blockquote></div><br>