[rabbitmq-discuss] Durable + Exclusive queue disappeared

[Jonathan Schwietert]

Thanks again for the response Emile. We've given up on fixing the problem 
for now & will try, as you've noted, blowing away the DB in order to start 
fresh. After trying to duplicate the issue I can't find any way to 
replicate it. Multiple attempts of restarting the process, restarting the 
consumer in a running process all yield a fully working queue. 

The event which originally triggered this was our shovel had got into a bad 
state so we used a 'shutdown -r now' to simply restart the server which 
evidently failed at the point it was trying to shutdown the broker. At this 
point the broker was hard-rebooted and came back up with the error i posted 
here. If you're interested enough in me trying to replicate this issue 
again i wouldn't mind trying but it does seem like its a rather unique 
case. Our versions are: RabbitMQ *2.8.2* on Erlang *R15B01.*

On Thursday, April 4, 2013 9:11:00 AM UTC-6, Jonathan Schwietert wrote:
>
> Again, pasting in a response i received via email that isn't showing up 
> on this list, from *Emile Joubert:*
> *
> *Hi Jonathan,
>
> On 03/04/13 17:28, Jonathan Schwietert wrote:
>
> Thanks for the quick response Emile. I did happen to see that result in
>
> my searching but unfortunately our broker is standalone - so there are
>
> no other nodes to bring up.
>
>
> If your broker has only ever been standalone then the explanation that
> involves another broker does not apply, though that is the only way I
> know of to replicate the error condition. If you can recreate the error
> with a small sequence of steps then please post them. The only way to
> recover from the error is to move or delete the broker database directory.
>
> -Emile
>
> On Wednesday, April 3, 2013 10:28:46 AM UTC-6, Jonathan Schwietert wrote:
>>
>> Thanks for the quick response Emile. I did happen to see that result in 
>> my searching but unfortunately our broker is standalone - so there are no 
>> other nodes to bring up.
>>
>> On Wednesday, April 3, 2013 10:28:02 AM UTC-6, Jonathan Schwietert wrote:
>>>
>>> Pasting in a response i received via email that isn't showing up on this 
>>> list, from *Emile Joubert:*
>>> Hi,
>>>
>>> On 02/04/13 23:32, Jonathan Schwietert wrote:
>>>
>>> Hi, I have a queue which was declared with BOTH durable and exclusive
>>>
>>> (on accident) whose owner disconnected then reconnected and now
>>>
>>> receives: "NOT_FOUND - no queue 'test.queue' in vhost '/test.vhost'".
>>>
>>>
>>> See the documentation for the exclusive parameter:
>>>
>>> http://www.rabbitmq.com/amqp-0-9-1-reference.html#queue.declare.exclusive
>>>
>>> "Exclusive queues may only be accessed by the current connection, and
>>> are deleted when that connection closes."
>>>
>>> So your 'test.queue' should not longer exist. The presence of the string
>>> in the internal Mnesia files does not necessarily mean that the queue
>>> still exists since the name could appear in internal logs.
>>>
>>> The scenario you describe can occur if a queue was declared on a node
>>> which leaves a cluster. A "Not Found" error will result when attempting
>>> to redeclare or delete the queue on any remaining nodes in the cluster.
>>> The solution is for the node that left the cluster to rejoin. The use of
>>> mirrored queues will prevent this condition.
>>>
>>>
>>> -Emile
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20130404/cfec6280/attachment.htm>