[rabbitmq-discuss] RabbitMQ running at 100% CPU.

[Martin Sustrik]

> The results indicate that this is probably a bug in Leopard, or a bug in 
> the Erlang runtime that only manifests itself on Leopard.
>
> The trigger appears to be a tcp client closing the socket when the 
> server is in the middle of writing to its peer.
>
> The next step is to construct a test case that doesn't involve rabbit.
>   
Matthias,

I'm not sure whether this will help, but anyway: Leopard (and presumably 
OS X in general) handles connection closed by the other peer by 
returning POLLHUP from the poll rather than POLLIN as is common on most 
platforms. Can that confuse Erlang/Rabbit?

Martin