From owner-freebsd-questions Mon Nov 29 18:21:25 1999 Delivered-To: freebsd-questions@freebsd.org Received: from implode.root.com (root.com [209.102.106.178]) by hub.freebsd.org (Postfix) with ESMTP id 2F0C215749 for ; Mon, 29 Nov 1999 18:18:55 -0800 (PST) (envelope-from dg@implode.root.com) Received: from implode.root.com (localhost [127.0.0.1]) by implode.root.com (8.8.8/8.8.5) with ESMTP id SAA28138; Mon, 29 Nov 1999 18:17:34 -0800 (PST) Message-Id: <199911300217.SAA28138@implode.root.com> To: Doug Barton Cc: Colin Campbell , freebsd-questions@FreeBSD.ORG Subject: Re: netstat -m confusion In-reply-to: Your message of "Mon, 29 Nov 1999 18:09:11 PST." From: David Greenman Reply-To: dg@root.com Date: Mon, 29 Nov 1999 18:17:34 -0800 Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG >On Mon, 29 Nov 1999, Colin Campbell wrote: > >> Hi, >> >> I just had one box reboot because it ran out of mbufs, so I thought I'd >> check another one and saw something quite confusing, namely: >> >> 826/4724/4608 mbuf clusters in use (current/peak/max) >> >> How can the peak be > max? (FreeBSD 3.2 off the WC CD set). > > I've seen this myself in really high-load situations. The only >answer I've ever come up with is that the value of "max" is slightly more >conservative than what is really available in the kernel, and/or that it's >rounded off to some factor of 8 which can be less than what's actually >there. Notice for instance that 4608/1024=4.5, whereas 4724 is an >"odd" number. I suspect that if you did the MAXUSERS multiplication that >your actual number of mbufs is 4608 < x < 5120, but I haven't looked at >the source to confirm my theory. > > According to the figures that DG quoted way back in the 2.2.x days >(by my recollection/experience anyways) the max should always be >= 1.5 x >peak, so the point to all of this is that you need lots more >NMBCLUSTERS. :) The max doesn't include additional space that was allocated for regular mbufs (which we recently changed to be allocated out of the same kernel VM map space), so if the peak is greater than the max, then this is telling you that you ran out and used some other space as well that wasn't part of NMBCLUSTERS. Anyway, increase the parameter to solve the problem. -DG David Greenman Co-founder/Principal Architect, The FreeBSD Project - http://www.freebsd.org Creator of high-performance Internet servers - http://www.terasolutions.com Pave the road of life with opportunities. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message