Received: (at submit) by bugs.debian.org; 28 Apr 1997 17:56:58 +0000
Received: (qmail 30280 invoked from network); 28 Apr 1997 17:56:57 -0000
Received: from hur-s0.fuller.edu (HELO waterf.org) (mail@206.1.27.36)
  by master.debian.org with SMTP; 28 Apr 1997 17:56:56 -0000
Received: from clameter by waterf.org with local (Exim 1.61 #1)
	id 0wLufe-0002A7-00 (Debian); Mon, 28 Apr 1997 10:56:42 -0700
From: clameter@debian.org
Subject: exim: Segfaults under high load
To: submit@bugs.debian.org
X-Mailer: bug 2.93
Message-Id: <E0wLufe-0002A7-00@waterf.org>
Sender: Christoph Lameter <clameter@waterf.org>
Date: Mon, 28 Apr 1997 10:56:42 -0700

Package: exim
Version: 1.61-1

As discussed on the exim mailing list:

Under high load exim will sometimes segfault freezing messages that are
affected.

Philip Hazel has suggested that the problem is due to libdb instabilities.
Perhaps exim should switch back to use gdbm?

(I am responsible for putting libdb support in exim I know. And I did it
because there was the perception on the list that DB was more reliable than
GDBM).

I get these frozen messages on two systems running bo.

1. Cyrix 166 Kernel 2.0.30 relaying about 60.000 messages a day.
   (exploder for .com domain for vger mailing lists)
   Around 200 freezes a day.

2. 486DX66 Kernel 2.0.29 relaying around 30.000 messages a day.
   (exploder for .edu domain for vger mailing lists)
   perhaps one message frozen a day.

I have solved this by setting an autunfreeze after 1 hour in /etc/exim.conf
so that delivery is retried which then causes the messages to be properly
delivered.

-- System Information
Debian Release: 1.3
Kernel Version: Linux waterf 2.0.30 #1 Tue Apr 22 22:34:09 PDT 1997 i486 unknown

Versions of the packages exim depends on:
libc5	Version: 5.4.23-2