From owner-freebsd-current  Fri May 26 10:02:27 1995
Return-Path: current-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id KAA00351
          for current-outgoing; Fri, 26 May 1995 10:02:27 -0700
Received: from who.cdrom.com (who.cdrom.com [192.216.222.3])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id KAA00345
          for <current@freebsd.org>; Fri, 26 May 1995 10:02:26 -0700
Received: from bigdipper.iagi.net (bigdipper.iagi.net [198.6.14.10])
          by who.cdrom.com (8.6.11/8.6.11) with ESMTP id JAA08531
          for <current@freebsd.org>; Fri, 26 May 1995 09:55:04 -0700
Received: (from adhir@localhost) by bigdipper.iagi.net (8.6.8/8.6.6) id MAA11048; Fri, 26 May 1995 12:59:37 -0400
Date: Fri, 26 May 1995 12:59:37 -0400 (EDT)
From: "Alok K. Dhir" <adhir@iagi.net>
To: David Greenman <davidg@Root.COM>
cc: current@FreeBSD.org
Subject: Re: "mb_map full" ? 
In-Reply-To: <199505202358.QAA00163@corbin.Root.COM>
Message-ID: <Pine.BSI.3.91.950526125613.11002A-100000@bigdipper.iagi.net>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: current-owner@FreeBSD.org
Precedence: bulk

On Sat, 20 May 1995, David Greenman wrote:

> >  On a current system (supped and built this morning), I just received a 
> >"mb_map full" message in syslog just after starting a news feed to INN.  
> >Everything "appears" to be working though.  INN is using mmap().  I 
> >thought this bug was fixed long ago?  Or is something new?
> 
>    Fixed? It's not a problem that can be fixed other than trying to make it
> dynamic (which would be quite difficult given the current mbuf cluster
> allocation scheme). Someone suggested making it a function of maxusers; I
> might do this someday. I would guess that you don't have options GATEWAY in
> your kernel, and thus the number of mbuf clusters is small enough that you
> occasionally hit the in-use limit. You can increase the limit with options
> "NMBCLUSTERS=512" (or more if you need).

Along these lines - we are running -current with a full news feed.  As 
you may recall, I had posted some problems we were having with the system 
freezing (latest inn port, etc).

In any case, several people suggested we add the following to our kernel 
conf file:

#for News servers
options         "CHILD_MAX=128"
options         "OPEN_MAX=128"
options         "NMBCLUSTERS=512"

And someone also mailed me a patch for INN which apparently seals up a 
memory leak problem.  

I did all four things (applied patch and added above lines) and the 
machine seems much more stable now.  

I am mailing because I am wondering whether you can offer advice as to 
either additional kernel options I may need to optimize this machine's 
uptime, or values for the above options which may be more appropriate.

Thanks!


Alok K. Dhir
Internet Access Group, Inc.
adhir@iagi.net
(301) 652-0484  Fax: (301) 652-0649
http://www.iagi.net