[rabbitmq-discuss] Rabbitmq Crash

[Matthew Sackman]

On Thu, Jun 17, 2010 at 12:11:44PM +0100, Matthew Sackman wrote:
> On Thu, Jun 17, 2010 at 02:34:29PM +0500, Waqas Badar wrote:
> > We are using rabbitmq. When queues exceed 200000 messages then most of
> > the time rabbitmq crashes. Even rabbitmq denies to start untill we
> > delete the mnesia files. Its crash dump is attached with this mail. Has
> > any one also faced this issue? That is very regularly for us. Any help?
> 
> More useful to us is the rabbit logs and the rabbit-sasl logs. I suspect
> you've run out of memory. This is a well known issue with the current
> version of RabbitMQ and is caused by the fact that Rabbit always keeps
> all messages in RAM.

Also, which version of RabbitMQ are you using, and what clients are you
using? The last couple of releases of RabbitMQ have used "channel.flow"
to stop publishers from publishing messages when the broker starts to
run out of memory. However, some clients (some of the python ones) don't
support this and still continue to publish messages. The current release
(1.8.0 - released yesterday) takes a much sterner line with such clients
and closes their channels if they do not respond to channel.flow
messages in a timely manner.

Matthew