From owner-freebsd-fs@freebsd.org Tue Jun 19 08:11:15 2018 Return-Path: Delivered-To: freebsd-fs@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2610:1c1:1:606c::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id 03A431005653 for ; Tue, 19 Jun 2018 08:11:15 +0000 (UTC) (envelope-from Albert.Shih@obspm.fr) Received: from mx-p1.obspm.fr (mx-p1.obspm.fr [145.238.193.20]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (Client CN "*.obspm.fr", Issuer "TERENA SSL CA 3" (verified OK)) by mx1.freebsd.org (Postfix) with ESMTPS id 8620C723D5 for ; Tue, 19 Jun 2018 08:11:14 +0000 (UTC) (envelope-from Albert.Shih@obspm.fr) Received: from io.chezmoi.fr (io-p2.obspm.fr [145.238.197.205]) (authenticated bits=0) by mx-p1.obspm.fr (8.14.4/8.14.4/DIO Observatoire de Paris - 15/04/10) with ESMTP id w5J8B5cB350026 (version=TLSv1/SSLv3 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT) for ; Tue, 19 Jun 2018 10:11:07 +0200 Date: Tue, 19 Jun 2018 10:11:24 +0200 From: Albert Shih To: freebsd-fs@freebsd.org Subject: Re: ZFS send problem Message-ID: <20180619081124.GA7228@io.chezmoi.fr> References: <20180614175445.GA1781@io.chezmoi.fr> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <20180614175445.GA1781@io.chezmoi.fr> User-Agent: Mutt/1.10.0 (2018-05-17) X-Virus-Scanned: clamav-milter 0.99.3 at mx-p1.obspm.fr X-Virus-Status: Clean X-BeenThere: freebsd-fs@freebsd.org X-Mailman-Version: 2.1.26 Precedence: list List-Id: Filesystems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 19 Jun 2018 08:11:15 -0000 Le 14/06/2018 à 19:54:45+0200, Albert Shih a écrit > Hi, > > I still got a very serious problem with zfs send (not recv). > > After I take a snapshot and try to send it, the server begin to send the > snapshot (I can see on the target the incomming data) and after sometime > (5=10m) on the server the zfs command freeze, I can launch zpool command > either. > > I check arc size, and if I let the max amount of memory to the > vfs.zfs.arc_max (~ Max RAM ~ 192Go) the zfs command freeze simultaneously > with the arc reach 100.3% of the max (but I cannot make lots of test, so > I'm not sure it's related). > > If I set le vfs.zfs.arc_max at half of the RAM ( = 96Go) I can see thought > zfs-stats the arc come to 100.3% and empty to 98-99% and raise again. > Either way the zfs command not really freeze totally but take lot of time > to answer (~10-20 seconds when normally it's < 1sec). > > The only « special » configuration (vs my other server doing same thing > without problem) is I got two zpool on this one. I've no idea if that can > be a problem (one with 2 ssd, the other with 4 vdev of 7 disks) > > I check dmesg, /var/log/all.log and don't find any information or error > message. > So I eventually find a solution to my problem, I don't know if it's a bug (I can do a bug report), any way, I would love if anyone can give me a explanation To fix the problem with the zfs send I need to disable prefetch by adding sysctl -w vfs.zfs.prefetch_disable=1 with that option the zfs didn't hang anymore. Regards -- Albert SHIH Observatoire de Paris xmpp: jas@obspm.fr Heure local/Local time: Tue Jun 19 10:07:47 CEST 2018