From owner-freebsd-stable@FreeBSD.ORG Thu Mar 5 20:20:59 2015 Return-Path: Delivered-To: freebsd-stable@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [8.8.178.115]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id 0E5F7FF8 for ; Thu, 5 Mar 2015 20:20:59 +0000 (UTC) Received: from j006.host001.searchy.nl (j006.host001.searchy.nl [79.143.214.199]) by mx1.freebsd.org (Postfix) with ESMTP id C71671FA for ; Thu, 5 Mar 2015 20:20:58 +0000 (UTC) Received: from [192.168.5.21] (5418287B.cm-5-1a.dynamic.ziggo.nl [84.24.40.123]) (Authenticated sender: ppi@j006.host001.searchy.nl) by j006.host001.searchy.nl (Postfix) with ESMTPSA id E7CAE1E8C09 for ; Thu, 5 Mar 2015 20:15:08 +0000 (UTC) Message-ID: <54F8B94C.9050100@searchy.net> Date: Thu, 05 Mar 2015 21:15:08 +0100 From: "Frank de Bot (lists)" User-Agent: Mozilla/5.0 (X11; Linux i686; rv:34.0) Gecko/20100101 Firefox/34.0 SeaMonkey/2.31 MIME-Version: 1.0 To: freebsd-stable@freebsd.org Subject: Re: ZFS L2arc 16.0E size References: <54E1388C.3060602@searchy.net> In-Reply-To: <54E1388C.3060602@searchy.net> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.18-1 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 05 Mar 2015 20:20:59 -0000 Frank de Bot (lists) wrote: > Hello, > > I have a FreeBSD 10.1 system with a raidz2 zfs configuration with 2ssd's > for l2arc . It is running '10.1-STABLE FreeBSD 10.1-STABLE #0 r278805' > Currently I'm running tests before it can go to production, but I have > the following issue. After a while the l2arc devices indicate 16.0E free > space and it starts 'consuming' more than it can hold > I've tried to 'debug' with dtrace, I found out different things: - l2arc_write_buffers sometimes caused the vdev->vdev_stat.vs_alloc to grow larger than vdev->vdev_asize. - l2arc_dev->l2ad_end is larger than vdev->vdev_asize - At some point l2arc_eviction isn't doing anything, but l2ard_dev->l2ad_evict is higher than l2ard_dev->l2ad_hand . taddr is matching l2ard_dev->l2ad_evict . I would assume it should evict that space. l2arc_write_buffers will continue because there seems te be room enough, I guess this would be caused by vdev_asize - vs_alloc is negative and indicating a 16.0E freespace. It could be that I'm assuming wrong things or interpret things wrong. Please let me know. Frank de Bot