[rabbitmq-discuss] RabbitMQ crash in production - help diagnose from log
Bryan Alves
bryanalves at gmail.com
Sun Jul 24 18:30:06 BST 2011
Sorry for taking so long to get back, my answers inline
On Tue, Jul 5, 2011 at 5:26 PM, Matthias Radestock <matthias at rabbitmq.com>wrote:
> Bryan,
>
> On 02/07/11 13:13, Matthias Radestock wrote:
>
>> It looks like the crash might be triggered by a rare race
>> condition. [...] Nothing obvious jumps out at me from the
>> code, so this may take a while to track down.
>>
>
> As feared, a few hours of poking the code hasn't thrown up any dirt.
>
> Do you still have the rabbit.log file from around the time of the crash? If
> so, please send it to us.
>
Sorry I don't have this anymore.
>
> Also, it would be useful to get some more details on how you are using
> rabbit...
>
> How many queues are there? How much queue churn is there (i.e. queues being
> created and deleted)?
>
There's about 3 durable queues. 0 churn
>
> Do published messages only ever get routed to a single queue or can they
> get routed to multiple queues?
>
Messages very often (>90%, estimate) get routed to multiple queues
>
> How big are the messages?
>
< 1k bytes
>
> How long (#messages) are the queues, and how does that figure change over
> time?
>
Consumers almost always keep up with load, so the queue size is usually 0,
occasionally there will be spikes if we manually dump a lot of messages into
the queue. Spikes in this sense is a few thousand messages, up to maybe
50,000 messages. Queues will gradually make their way back to 0 after a few
minutes after these instances.
>
>
> Regards,
>
> Matthias.
>
--
Bryan Alves
Software Engineer
Blue State Digital
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20110724/0a063e62/attachment.htm>
More information about the rabbitmq-discuss
mailing list