[rabbitmq-discuss] rabbit cluster keeps crashing

Matthew Sackman matthew at lshift.net
Thu Apr 8 12:07:17 BST 2010


On Tue, Apr 06, 2010 at 02:29:48PM -0700, LoOoD wrote:
> Our servers are connected to gige switches and different subnets. The
> switches and routers that the servers are connected to are using less then
> 10% of their capacity. We don't see any other apps having any sort of
> network issues.  We send the rabbit cluster about 70 msgs/sec.

Ok, it's definitely not the network then ;)

> Another symptom we've noticed is the beam.smp process using 100% cpu. Is
> there any thing we can do, to help track down the exact cause? Some sort of
> increase debugging level? We're willing to do anything to figure out what
> going on and stabilize it.

Hmm, that's interesting. It suggests rabbit is very busy. Does this
happen before the nodedown messages start happening - if Rabbit is
sufficiently busy then it could be failing to schedule the net ticktimes
frequently enough.

70 msgs/sec is very low - Rabbit running on a decades-old PC will be
able to manage that. Thus what led you to setting up the cluster in the
first place and what topology / use cases do you have?


More information about the rabbitmq-discuss mailing list