From owner-freebsd-stable Tue Jan 29 11:20:48 2002 Delivered-To: freebsd-stable@freebsd.org Received: from petra.vif.com (ip216-239-64-165.vif.net [216.239.64.165]) by hub.freebsd.org (Postfix) with ESMTP id 785AC37B41B for ; Tue, 29 Jan 2002 11:20:27 -0800 (PST) Received: (from squid@localhost) by petra.vif.com (8.11.6/8.11.6) id g0TJKQ851080 for stable@freebsd.org; Tue, 29 Jan 2002 14:20:26 -0500 (EST) (envelope-from talist@vif.com) X-Authentication-Warning: petra.vif.com: squid set sender to talist@vif.com using -f Received: from 216.239.71.137 ( [216.239.71.137]) as user talist@mail.vif.com by email.vif.com with HTTP; Tue, 29 Jan 2002 14:20:26 -0500 Message-ID: <1012332026.3c56f5fa347a9@email.vif.com> Date: Tue, 29 Jan 2002 14:20:26 -0500 From: talist@vif.com To: stable@freebsd.org Subject: RE: Crashing Freebsd + Dummynet + heavy traffic + workaround References: <1D8129375B07124580A9FB73321C3D8D01244441@pdxexch01.netiq.local> In-Reply-To: <1D8129375B07124580A9FB73321C3D8D01244441@pdxexch01.netiq.local> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit User-Agent: Internet Messaging Program (IMP) 4.0-cvs Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG This is another possible workaround. But I think there is here a good oportunity to fix Dummynet since we can reproduce the problem. Quoting Dustin Harris : > I had similar problems and found that by doing an ipfw flush before making > changes to the pipe I avoided the frequent lockups. > > -----Original Message----- > From: talist@vif.com [mailto:talist@vif.com] > Sent: Tuesday, January 29, 2002 5:22 AM > To: stable@freebsd.org > Subject: Re: Crashing Freebsd + Dummynet + heavy traffic + workaround > > I noticed that the box can be made much more resilient by boosting the > following > kernel variable: > options HZ=1000 # granularity was 100 > > After rebuilding the kernel, freebsd did not freeze but I am sure this is > only a > workaround. Dummynet should not be able to lock-out the machine. > > > > I am using a freebsd box as a gateway in order to throttle down bandwidth > for > > some particular traffic. > > The box freezes under high load and the only way to recover is through a > power > > cycle. > > The crash does not generate any logs, but I have ruled out hardware > causes. I > > can repeat the lock-out with 2 different hardware and with xl(4) and > dc(4) > NIC's > > > > I am suspecting a problem with DUMMYNET's PIPE and QUEUE. > > > > I have collected all the config information about the machine with a > small > > > script that helps in crashing the box after a few minutes of alternating > the > > dummynet pipe's size. > > http://www.geocities.com/dummynetus/ipfw.txt > > > > In summary, I have tried the following: > > - Boost maxusers to 128 > > - Boost NMBCLUSTERS to 10240 > > - monitor netstat -m gives normal readings even before freeze > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-stable" in the body of the message > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message