From owner-freebsd-performance@FreeBSD.ORG Mon Dec 20 18:24:51 2004 Return-Path: Delivered-To: freebsd-performance@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 3866216A4CE for ; Mon, 20 Dec 2004 18:24:51 +0000 (GMT) Received: from fury.noc.clara.net (fury.noc.clara.net [195.8.70.97]) by mx1.FreeBSD.org (Postfix) with ESMTP id 80E5D43D39 for ; Mon, 20 Dec 2004 18:24:50 +0000 (GMT) (envelope-from dave@uk.clara.net) Received: from dave (helo=fury.noc.clara.net) by fury.noc.clara.net with local-esmtp (Exim 4.34) id 1CgSDB-000CXI-Gq for freebsd-performance@freebsd.org; Mon, 20 Dec 2004 18:24:49 +0000 To: freebsd-performance@freebsd.org Date: Mon, 20 Dec 2004 18:24:49 +0000 From: Dave Williams Message-Id: Sender: Dave Williams Subject: Odd news transit performance problem X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 20 Dec 2004 18:24:51 -0000 Folks, we're seeing a somewhat puzzling scenario with a FreeBSD 4.9RC box of an October '03 vintage kernel. The box is currently running innd for news transit, and in circumstances of heavy traffic load performance on the machine drops through the floor. Various reporting tools (systat, vmstat, top, etc) report that the box is idle - there's no significant contention for memory, disk, network, etc. that we can see and actually bouncing the box seems to bring performance back up to speed again for a period - restarting innd doesn't have the same effect. The machine is a 2.8GHz Xeon, Intel 10/100 and 1000 NICs, is running with two onboard LSI 53C{mumble} scsi controllers, device polling enabled, HZ=1000, and MAXDSIZ set to 1GB and MAXSSIZ/DFLDSIZ. We're a bit stumped as to where to look at the moment. :( Thanks Dave -- Dave Williams dave@uk.clara.net