From owner-freebsd-questions  Thu Feb  6 11:14:10 1997
Return-Path: <owner-questions>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id LAA22961
          for questions-outgoing; Thu, 6 Feb 1997 11:14:10 -0800 (PST)
Received: from news.interworld.net (news.interworld.net [206.124.224.6])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id LAA22940;
          Thu, 6 Feb 1997 11:14:04 -0800 (PST)
Received: (from pete@localhost) by news.interworld.net (8.7.5/8.7.3) id LAA26773; Thu, 6 Feb 1997 11:14:03 -0800 (PST)
From: Peter Carah <pete@news.interworld.net>
Message-Id: <199702061914.LAA26773@news.interworld.net>
Subject: ENOBUF and netstat -m
To: questions@freebsd.org, isp@freebsd.org
Date: Thu, 6 Feb 1997 11:14:02 -0800 (PST)
X-Mailer: ELM [version 2.4 PL25]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-questions@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

I'm persistently getting ENOBUF on the news machine, usually from
ctlinnd.  I haven't yet upgraded to 1.5.1; we're running unoff4.
netstat -m always has a comfortable number of clusters and also
always reports 0 requests delayed or denied.  This seems inconsistent :-)

A 1.5.1 upgrade is due today; it is a bit complicated by the change
to the file layout :-(

(this started when we added a feed that tends to open 6 streams at
a time; it apparently has to do with thenumber of sockets open
at once.  INND is indeed behind reading from the streams but since 
netstat -m always reports lots of free memory I can't figure out where 
the particular ENOBUF is coming from - there are about a hundred
occurences of this error in a quick grep of the kernel source.)

Is there a kernel tweak other than the listen count and nmbclusters 
that may apply here?  Both of those have been increased greatly.

Also, is active file mmap safe to use in 2.1.5 or do I need to use 
'read'?  I got burned once (in irix 4.0.5) where the innd config said 
to use mmap but innd cored a lot until I converted back to read.
This isn't happening here but innd is grossly behind, at only 1-2
articles/sec.

Thanks in advance,
-- Pete