[rabbitmq-discuss] 404, "NOT_FOUND - no queue 'foo' in vhost '/'"

[Matt Pietrek]

Some additional information I've learned since posting the original, below.

After shutting down the primary node, the queue causing the 404 error has
actually vanished from the broker. Keeping an eye on the Management Web UI,
I literally see it disappear.

In addition, in the rabbit@<nodename>.log, I see a whole bunch of spew,
including many instances of this:

=WARNING REPORT==== 20-Dec-2011::14:09:17 ===
Non-AMQP exit reason '{{{function_clause,
                            [{gb_trees,lookup,
                   
[<<150,83,52,247,194,247,140,35,130,167,168,
                                    35,224,203,183,186>>,
                                  {dict,0,16,16,8,80,48,
                   
{[],[],[],[],[],[],[],[],[],[],[],[],[],
                                       [],[],[]},
                                      {{[],[],[],[],[],[],[],[],[],[],[],[],
                                        [],[],[],[]}}}]},
                             {rabbit_amqqueue_process,
                                 '-confirm_messages/2-fun-0-',2},
                             {lists,foldl,3},
                             {rabbit_amqqueue_process,confirm_messages,2},
                             {rabbit_amqqueue_process,next_state,1},
                             {rabbit_amqqueue_process,noreply,1},
                             {gen_server2,handle_msg,2},
                             {proc_lib,wake_up,3}]},
                        {gen_server2,call,
                            [<0.1298.0>,
                             {basic_get,<0.2542.0>,false},
                             infinity]}},
                       [{gen_server2,call,3},
                        {rabbit_misc,with_exit_handler,2},
                        {rabbit_channel,handle_method,3},
                        {rabbit_channel,handle_cast,2},
                        {gen_server2,handle_msg,2},
                        {proc_lib,init_p_do_apply,3}]}'

Again, this is a stock 2.7.0 install.

From:  Matt Pietrek <mpietrek at skytap.com>
Date:  Tue, 20 Dec 2011 13:04:50 -0800
To:  <rabbitmq-discuss at lists.rabbitmq.com>
Cc:  Matt Pietrek <mpietrek at skytap.com>
Subject:  404, "NOT_FOUND - no queue 'foo' in vhost '/'"

I'm encountering a strange situation that I don't think has been covered by
other threads.

I'm running RabbitMQ 2.7.0 on Ubuntu 10.04 LTS with the stock PIKA 0.9.5
client, and taking advantage of mirrored queues.

My scenario:
* I have two apps that ping-pong messages to each other using two queues (a
pseudo-RPC type of thing.)
* The apps talk to RabbitMQ via a keepalived monitored Virtual IP.
* Before starting the apps, I predeclare both queues.
* In the RabbitMQ management Web UI, I see both nodes and the Queues tab
shows both queues as "HA D", so I believe they're mirrored.
* Start up the two ping-ponging apps, let run for few seconds. Note messages
flowing back and forth.
* Stop the primary node (rabbitmqctl stop)
* Keepalived fails over the Virtual IP
* After a few seconds I see one of the apps resume processing messages as
expected.
* Shortly thereafter the other app gets an exception from
channel.basic_get(), which is the 404 "no queue found" exception.
* I don't believe it's a connection error, as during my queue reads, my
logic catches AMQPConnectionErrors exceptions and retries after a short
wait.
I've looked at my code carefully, and there's no calls to queue_declare
other than the original creation (item 3, above.)

Any thoughts on what could be wrong?

Thanks much,

Matt



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20111220/b90fc2e5/attachment.htm>