From owner-freebsd-geom@FreeBSD.ORG Thu May 10 15:13:32 2012 Return-Path: Delivered-To: freebsd-geom@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [69.147.83.52]) by hub.freebsd.org (Postfix) with ESMTP id 3E6FC106566B for ; Thu, 10 May 2012 15:13:32 +0000 (UTC) (envelope-from trent@snakebite.org) Received: from exchange.liveoffice.com (exchla3.liveoffice.com [64.70.67.188]) by mx1.freebsd.org (Postfix) with ESMTP id 1EF868FC08 for ; Thu, 10 May 2012 15:13:31 +0000 (UTC) Received: from EXHUB02.exchhosting.com (192.168.11.214) by exhub11.exchhosting.com (192.168.11.109) with Microsoft SMTP Server (TLS) id 8.3.213.0; Thu, 10 May 2012 08:13:31 -0700 Received: from EXMBX10.exchhosting.com ([fe80::9c37:32f6:a508:a44f]) by exhub02.exchhosting.com ([fe80::311c:a4c3:90a7:3e53%12]) with mapi; Thu, 10 May 2012 08:13:31 -0700 From: Trent Nelson To: "mj@feral.com" Date: Thu, 10 May 2012 08:13:29 -0700 Thread-Topic: Teaching gmultipath about path cost/priority Thread-Index: Ac0uv3Vbm6rmwL8aTjSkezjxn9IgHg== Message-ID: <4A6AB54F-A0B3-40A6-A8E4-EE9992B5C52D@snakebite.org> Accept-Language: en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: acceptlanguage: en-US Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Cc: "freebsd-geom@freebsd.org" Subject: Teaching gmultipath about path cost/priority X-BeenThere: freebsd-geom@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: GEOM-specific discussions and implementations List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 10 May 2012 15:13:32 -0000 Hi again Matt, I'm sure I'll eventually find a subsystem relevant to me that you *didn= 't* author, but for now, I have some gmultipath questions. Basically, the entire approach is perfect for me. As all my SAN target= s are JBODs (i.e. I don't have any targets that could saturate a single H= BA and thus would benefit from active/active multipathing), active/passive is exactly what I want. But... in my case, all paths are not equal. My fabric has four switche= s=20 and is fully meshed; a host may have up to four HBAs, each one connecte= d=20 to a separate switch. From a redundancy and failover perspective, I've made sure there are an abundance of paths available to a single disk. However, unless something has failed, there's only one path I want to u= se, as I've pre-planned that path to be the most optimal. It sounds like the order gmultipath uses to select paths is derived fro= m=20 the order it tastes disks -- something I have very little control over. It'd be ideal if there was a way of teaching gmultipath about a path cost/priority, so that it can make an informed decision about which path to choose a) when first initializing, and b) during failover. Thoughts? Regards, =20 Trent.=