From owner-freebsd-cluster@freebsd.org Mon Oct 12 15:13:15 2015 Return-Path: Delivered-To: freebsd-cluster@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id 825E2A1171C for ; Mon, 12 Oct 2015 15:13:15 +0000 (UTC) (envelope-from pjoubert@embl.fr) Received: from emblmta1.embl.fr (emblmta1.embl.fr [193.49.43.176]) by mx1.freebsd.org (Postfix) with ESMTP id 27A611F6D for ; Mon, 12 Oct 2015 15:13:14 +0000 (UTC) (envelope-from pjoubert@embl.fr) X-IronPort-AV: E=Sophos;i="5.17,673,1437429600"; d="scan'208";a="6437597" Received: from unknown (HELO [172.26.15.24]) ([172.26.15.24]) by emblmta1.embl.fr with ESMTP/TLS/DHE-RSA-AES128-SHA; 12 Oct 2015 17:12:04 +0200 To: freebsd-cluster@freebsd.org From: Pablo Joubert Subject: =?UTF-8?Q?HAST=c2=a0question_for_full_sync_replication?= X-Enigmail-Draft-Status: N1110 Message-ID: <561BCD8B.5000904@embl.fr> Date: Mon, 12 Oct 2015 17:11:07 +0200 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Icedove/38.2.0 MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit X-BeenThere: freebsd-cluster@freebsd.org X-Mailman-Version: 2.1.20 Precedence: list List-Id: Clustering FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 12 Oct 2015 15:13:15 -0000 Dear list, I have a setup with a failover HAST + ZFS cluster (CARP and auto switch scripts were deactivated long ago because of too much splitbrains). I didn't do the setup, and I do not have an extensive experience on HAST. I now have a drive failing (predictive failure) on the primary server that I want to change. I'd like to avoid to switch servers (mainly because zpool export/import is long and so there is an impacting service downtime). If I remove the disk on the primary host (or if it fails by itself), the secondary will be used to serve data. I'm so asking myself if I can do a replication from the secondary to the new primary. I know this will degrades performances, but I prefer that to a downtime. I saw on the table of https://wiki.freebsd.org/HAST (How does synchronization work?) that it will result to the "?! Full sync from secondary" synchronization type, so, I hope it is possible. I want to be sure that it won't result in a split brain situation. This is the procedure what I was thinking to follow: # remove the drive from raid controler: mfiutil fail 0 # change the disk, then recreate jbod for it mfiutil create -u0 create jbod E1:S0 # init hast device hastctl init mfid1 # set hast device as primary hastctl role primary mfid1 Is this the right thing to do ? Should I directly set the hast device to primary ? Will the replication start the right way (from secondary to primary) ? Best regards, -- Pablo Joubert EMBLGR IT Team