[rabbitmq-discuss] Preventing uncontrolled crashes due to file descriptor exhaustion

Tony Spataro tony at rightscale.com
Wed Jul 14 17:54:24 BST 2010

Thanks Marek. I'll retry my tests with a 1.8 broker and verify that it
doesn't crash when its number of connections approaches its FD limit.

For a use case like mine (where the broker is listening for connections on a
public IP) I don't have the luxury of guaranteeing an upper bound on the
number of open sockets; but if the broker stops accepting connections once
it reaches its threshold, that should be sufficient to stop any new FDs from
being allocated.


2010/7/13 Marek Majkowski <majek04 at gmail.com>

> On Tue, Jul 13, 2010 at 11:24, Marek Majkowski <majek04 at gmail.com> wrote:
> > In the new persister branch "bug21673" we've made some magical tweaks.
> > In this branch we try to make sure that there are always enough file
> descriptors
> > for erlang.
> Good news, apparently the "magical tweaks" were merged to the main
> branch some time ago. From what I see, they are included in 1.8.0.
> Quick experiments show that rabbit always leaves about 80 file descriptors
> for internal use.
> Cheers,
>    Marek
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20100714/c5054243/attachment-0001.htm>

More information about the rabbitmq-discuss mailing list