From owner-freebsd-amd64@FreeBSD.ORG Wed May 25 23:53:48 2005 Return-Path: X-Original-To: amd64@freebsd.org Delivered-To: freebsd-amd64@FreeBSD.ORG Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id AEDC916A41C for ; Wed, 25 May 2005 23:53:48 +0000 (GMT) (envelope-from girgen@pingpong.net) Received: from melon.pingpong.net (82.milagro.bahnhof.net [195.178.168.82]) by mx1.FreeBSD.org (Postfix) with ESMTP id 48D7443D49 for ; Wed, 25 May 2005 23:53:47 +0000 (GMT) (envelope-from girgen@pingpong.net) Received: from localhost (localhost.pingpong.net [127.0.0.1]) by melon.pingpong.net (Postfix) with ESMTP id 238254AF56; Thu, 26 May 2005 01:53:46 +0200 (CEST) Received: from melon.pingpong.net ([127.0.0.1]) by localhost (melon.pingpong.net [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 44432-02-7; Thu, 26 May 2005 01:53:45 +0200 (CEST) Received: from [82.182.157.67] (1-2-8-5b.asp.sth.bostream.se [82.182.157.67]) by melon.pingpong.net (Postfix) with ESMTP id 7F80D4AF54; Thu, 26 May 2005 01:53:45 +0200 (CEST) In-Reply-To: References: <75f1b24e6dc7e145f7d36a874b825ab1@pingpong.net> Mime-Version: 1.0 (Apple Message framework v622) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: Content-Transfer-Encoding: 7bit From: Palle Girgensohn Date: Thu, 26 May 2005 01:53:45 +0200 To: Claus Guttesen X-Mailer: Apple Mail (2.622) X-Virus-Scanned: by amavisd-new at pingpong.net Cc: amd64@freebsd.org Subject: Re: Dual Xeon EM64T crashes reliably w/ 5.x amd64 X-BeenThere: freebsd-amd64@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Porting FreeBSD to the AMD64 platform List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 25 May 2005 23:53:48 -0000 2005-05-26 kl. 00.09 skrev Claus Guttesen: >> with identical hardware. His machine is not as loaded, so in his case >> moving from four CPUs (two "real" + HTT) to two real (shutting down >> HTT) was enough to stop the crashes. For me, I must run UP. > > What compile-options do you have in /etc/make.conf? Doing php I guess > it's a web-server, what other apps are running on the server? Is the > server located in a location with adequate cooling? cooling, yes. You can see my previous posts for more info, but in short, we run php apache-1.3, postgresql-8.0.3, perl-5.8.6 (amavisd), postfix, named, clamd. httpd is very busy. CPUTYPE?=nocona CFLAGS= -O -pipe COPTFLAGS= -O -pipe Kernel is generic except some small details, see http://lists.freebsd.org/pipermail/freebsd-amd64/2005-May/004949.html > May not apply any longer, but during the 5.1-days Scott Long advised > me to add the following lines to /boot/loader.conf: > > echo vm.kmem_size="450000000" >> /boot/loader.conf # sysctl vm.kmem_size vm.kmem_size: 419430400 > echo kern.maxvnodes="200000" >> /boot/loader.conf # sysctl kern.maxvnodes kern.maxvnodes: 100000 Both are lower, but I don't believe this would crash a 5.4 system? Much has happened since 5.1. > > That prevented my webservers from rebooting without any apparent > reason. Too many temp-files was the cause. Try purging temp-files more > often if the above lines do help. I don't have many temp files, doubt it is the problem. It can be an out-of-memory situation, possibly... I realize now it is swapping, 25% of swap used. Must get more memory, I guess... can the machine crash that hard when out of memory??? Problem is, I hardly dare to try anything right now. I'm running single-CPU, it works fine. If it starts crashing again soon, I'll start loosing customers. Would abandoning amd64 and installing a i386 system help? Probably yes? I'd rather not, that's a substantial amount time to reinstall everything... :( /Palle