From stuart at bmsi.com  Mon Jan 30 17:11:19 2006
From: stuart at bmsi.com (Stuart D. Gathman)
Date: Mon, 30 Jan 2006 17:11:19 -0500 (EST)
Subject: [Pymilter] Memory Leak?
Message-ID: <Pine.LNX.4.44.0601301708020.26409-100000@bmsred.bmsi.com>

There seems to be a slow memory leak (or data cancer) introduced. Memory
use grows until milter is restarted, by about 8Megs/40000 connections.  
There are some caches which are not clear (CBV cache), but memory growth
occurs even with very few CBVs.  I have eliminated the experimental gossip
support. I wonder if a leak has been introduced into miltermodule.c?

-- 
	      Stuart D. Gathman <stuart at bmsi.com>
    Business Management Systems Inc.  Phone: 703 591-0911 Fax: 703 591-6154
"Confutatis maledictis, flamis acribus addictis" - background song for
a Microsoft sponsored "Where do you want to go from here?" commercial.



From stuart at bmsi.com  Mon Jan 30 22:11:16 2006
From: stuart at bmsi.com (Stuart D. Gathman)
Date: Mon, 30 Jan 2006 22:11:16 -0500 (EST)
Subject: [Pymilter] Re: Memory Leak?
Message-ID: <Pine.LNX.4.44.0601302207550.30998-100000@bmsred.bmsi.com>

I added a HELO GC trigger for gc.collect().  Also, by stopping sendmail,
pymilter will quickly revert to minimal size.  I added 10K to size of ctx
and 30K to size of bmsMilter for testing.  Under basic operation, there is
no leak.  Will start checking production sites closely.

-- 
	      Stuart D. Gathman <stuart at bmsi.com>
    Business Management Systems Inc.  Phone: 703 591-0911 Fax: 703 591-6154
"Confutatis maledictis, flamis acribus addictis" - background song for
a Microsoft sponsored "Where do you want to go from here?" commercial.