On Thu, Feb 17, 2011 at 10:39 AM, Martin Sustrik <[email protected]> wrote:

> Fixed in master. When reconnection happens due to the peer closing the
> socket, it is delayed by RECONNECT_IVL (0.1 sec).
>
> This solves the problem at hand, however it still leaves 0MQ vulnerable to
> DoS attacks.

Though this is definitely an improvement to how the client behaves,
the fix does not actually solve the problem with the mailbox. Here is
a modified version of mailbugz that asserts with the same mailbox
overflow error: https://gist.github.com/831418

It happens after 604 reconnections, which I can make happen quickly by
reducing the reconnect interval. 604 is the same figure as before the
fix, meaning that delaying the reconnection has no effect, the mailbox
still grows, only more slowly.

Increasing the client's reconnect ivl makes the server survive longer
but it always crashes after exactly 604 reconnect attempts. That
suggests some flaw in the way the reconnection commands are managed in
the mailbox.

-Pieter
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Reply via email to