From owner-freebsd-stable@FreeBSD.ORG Mon Jan 7 12:26:05 2008 Return-Path: Delivered-To: stable@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 50BE716A419; Mon, 7 Jan 2008 12:26:05 +0000 (UTC) (envelope-from petefrench@ticketswitch.com) Received: from angel.ticketswitch.com (angel.ticketswitch.com [IPv6:2002:57e0:1d4e::1]) by mx1.freebsd.org (Postfix) with ESMTP id 15C1E13C45B; Mon, 7 Jan 2008 12:26:05 +0000 (UTC) (envelope-from petefrench@ticketswitch.com) Received: from smaug.rattatosk ([10.50.50.2]) by angel.ticketswitch.com with esmtps (TLSv1:AES256-SHA:256) (Exim 4.67 (FreeBSD)) (envelope-from ) id 1JBr3M-0009ob-FI; Mon, 07 Jan 2008 12:26:04 +0000 Received: from dilbert.rattatosk ([10.50.50.6] helo=dilbert.ticketswitch.com) by smaug.rattatosk with esmtp (Exim 4.67 (FreeBSD)) (envelope-from ) id 1JBr3L-0009vn-Jp; Mon, 07 Jan 2008 12:26:04 +0000 Received: from petefrench by dilbert.ticketswitch.com with local (Exim 4.68 (FreeBSD)) (envelope-from ) id 1JBr3L-0000Mf-JD; Mon, 07 Jan 2008 12:26:03 +0000 To: mharo@FreeBSD.org In-Reply-To: <4cd036390801070409l4d8f1cf9te86772f188f89dfe@mail.gmail.com> Message-Id: From: Pete French Date: Mon, 07 Jan 2008 12:26:03 +0000 Cc: stable@freebsd.org Subject: Re: kmem_map too small under ZFS X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 07 Jan 2008 12:26:05 -0000 > vm.kmem_size="512M" > vm.kmem_size_max="512M" I have similar to this in mine... vm.kmem_size=629145600 vm.kmem_size_max=629145600 which is about 600 meg - the machine has 2 gig of RAM. > vfs.zfs.prefetch_disable="1" > vfs.zfs.arc_max="150M" > kern.maxvnodes="400000" now these I havent got - I did see a reference to disabling the intent log, which is what I was about to try. I will give your settings a shot too. > Most of these settings came from various mailing list postings. It's > possible that some don't do anything useful, but I haven't had a zfs > related panic since. Worth a try - might see if I can find a way tro reliably trigger the opanic and then tyrn these bits back on one at a time too. thanks, -pete.