From owner-freebsd-current@FreeBSD.ORG Thu Dec 21 12:21:23 2006 Return-Path: X-Original-To: current@freebsd.org Delivered-To: freebsd-current@FreeBSD.ORG Received: from mx1.freebsd.org (mx1.freebsd.org [69.147.83.52]) by hub.freebsd.org (Postfix) with ESMTP id 988E316A40F for ; Thu, 21 Dec 2006 12:21:23 +0000 (UTC) (envelope-from if@hetzner.co.za) Received: from mail1a.your-server.co.za (mail1a.your-server.co.za [196.7.18.227]) by mx1.freebsd.org (Postfix) with ESMTP id 21E9A13C45C for ; Thu, 21 Dec 2006 12:21:23 +0000 (UTC) (envelope-from if@hetzner.co.za) Received: from [192.168.2.25] (helo=hetzner.co.za) by mail1a.your-server.co.za with esmtpa (Exim 4.63) (envelope-from ) id 1GxM1J-0007Oj-5b for current@freebsd.org; Thu, 21 Dec 2006 13:23:29 +0200 Received: from localhost ([127.0.0.1]) by hetzner.co.za with esmtp (Exim 4.63 (FreeBSD)) (envelope-from ) id 1GxM1I-0000sq-Py for current@freebsd.org; Thu, 21 Dec 2006 13:23:28 +0200 To: current@freebsd.org From: Ian FREISLICH X-Attribution: BOFH Date: Thu, 21 Dec 2006 13:23:28 +0200 Message-Id: X-Authenticated-Sender: if@hetzner.co.za X-Virus-Scanned: Clear (ClamAV 0.88.4/2365/Thu Dec 21 11:39:44 2006) Cc: Subject: Polling bug (was Re: packets duplicated *massively* on transmit.) X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 21 Dec 2006 12:21:23 -0000 Ian FREISLICH wrote: > Hi > > I have two FreeBSD routers: > > In two reasonably busy datacenters. We're seeing packet loss that > we traced to a packet ariving on the world-facing interface being > retransmitted approximately every 10 microseconds or so for 1 to 5 > seconds out of the interface the client is on. Someone has responded in private mail, but re-reading this it's possible that I didn't explain the situation too well. One packet arrives on the uplink interface. This packet is then repeated between 60000 and 1500000 times out of the other interface to the exclusion of all other traffic. It doesn't do this for every packet, just once every 500000 or so. I've discovered that turning polling off on the interfaces stops these retransmission storms. Ian -- Ian Freislich