[zeromq-dev] ZeroMq 2.1.7 "memory leak"

Pieter Hintjens ph at imatix.com
Tue Feb 21 16:36:24 CET 2012

On Tue, Feb 21, 2012 at 9:00 AM, john skaller
<skaller at users.sourceforge.net> wrote:

> Whether or not using high water marks is a temporary solution,
> shouldn't you always have them?
> Everything in production code should be bounded, even if the
> bounds are large.

Indeed, and we switched 0MQ to using a default HWM of 1000, precisely
to prevent this kind of random memory leakage.

The right combination here is an appropriate HWM combined with correct
purging of messages (reading and discarding or processing). It's got
nothing to do with subscriber/publisher side filtering. These are
messages you subscribe to, and either you process them, or they build
up, or you ask 0MQ to discard them at some stage.


More information about the zeromq-dev mailing list