[dm-devel] multipath and san fabric

[Philip R. Auld]

Rumor has it that on Mon, Nov 13, 2006 at 02:39:08PM -0500 Edward Goggin said:
> You may want to also cross connect across the 2 target
> controllers (A,B) in case 2.  Doing so provides more
> redundancy for asymmetric arrays and better resource
> utilization in case of a failure of either an HBA or
> target controller which would otherwise prevent
> utilization of the other HBA or target controller
> for that LU when load sharing I/O across multiple
> paths.
> 
> So the I_T nexus's for the 4 paths are:
> 
> 	HBA1_TargetPort1
> 	HBA1_TargetPort3
> 	HBA2_TargetPort2
> 	HBA2_TargetPort4
> 

That's exactly what James's example 2 describes. It's just 
typo'd in the DM paths list. 



> > -----Original Message-----
> > From: dm-devel-bounces at redhat.com 
> > [mailto:dm-devel-bounces at redhat.com] On Behalf Of James Smart
> > Sent: Monday, November 13, 2006 2:18 PM
> > To: jslittl at hendricks.org; device-mapper development
> > Subject: Re: [dm-devel] multipath and san fabric
> > 
> > If the hba's are on different fabrics - why would you ever put an hba
> > that is not in the fabric in a zone in that fabric ?  (e.g. 
> > your second
> > option makes no sense).
> > 
> > If you are stating that the 2hba's are on different fabrics 
> > (and in zones
> > within those fabric) - but they are seeing the same Storage 
> > device on both
> > fabrics - then the dm config (note: I'm not talking about 
> > zoning anymore)
> > would have a path per I_T nexus. I state it this way to 
> > account for dual-ported
> > controllers, which may appear as separate targets, may exist 
> > within the fabric.
> > 
> > Example 1:
> >    Fabric A contains HBA1 and TargetPort1. Typically a zone 
> > would exist
> >      within Fabric A that contains these 2 ports.
> >    Fabric B contains HBA2 and TargetPort2. Typically a zone 
> > would exist
> >      within Fabric B that contains these 2 ports.
> >    The storage array contains 2 controllers with 1 port per 
> > controller, or
> >      is a single controller with 2 ports. Either way, the two 
> > ports are
> >      TargetPort1 and TargetPort2, are on the same array, and 
> > all storage
> >      can be seen from any port.
> >    --
> >    This would have 2 DM paths.
> >    The I_T nexus's are:
> >       HBA1_TargetPort1
> >       HBA2_TargetPort2
> > 
> > Example 2:
> >    Fabric A contains HBA1, TargetPort1, and TargetPort3. 
> > Typically a zone
> >      would exist within Fabric A that contains these 3 ports.
> >    Fabric B contains HBA2, TargetPort2, and TargetPort4. 
> > Typically a zone
> >      would exist within Fabric B that contains these 3 ports.
> >    The storage array contains 2 controllers, with 2 ports per 
> > controller.
> >      Meaning TargetPort1 and TargetPort2 are on controller A, 
> > and TargetPort3
> >      and TargetPort4 are on controller B. The ports are cross 
> > connected on
> >      the fabrics for redundancy. All storage can be seen from 
> > any port.
> >    --
> >    This would ahve 4 DM paths
> >    The I_T nexus's are:
> >       HBA1_TargetPort1
> >       HBA1_TargetPort2
> >       HBA2_TargetPort3
> >       HBA2_TargetPort4
> > 
> > -- james
> > 
> > John Little wrote:
> > > Hi all
> > > 
> > > Admittedly this is not the correct forum for this question 
> > but I have googled 
> > > this and the only people I have to ask around here are 
> > people who won't touch 
> > > it because it is Linux.
> > > 
> > > My question is this:  I'm using SLES10 with two emulex 
> > hbas, one connected to 
> > > fabric a and one to fabric b.  When zoning the switch for 
> > the hbas do I put 
> > > in only one path to the fabric per hba or do I put in two, 
> > one to each of the 
> > > separate fabrics?
> > > 
> > > Again I realize this is not exactly the correct forum but 
> > since I couldn't 
> > > find an answer anywhere else I figured you guys would be the most 
> > > knowledgeable.  If there are some docs somewhere that you 
> > could point me to I 
> > > would appreciate it.
> > > 
> > > John
> > > 
> > > --
> > > dm-devel mailing list
> > > dm-devel at redhat.com
> > > https://www.redhat.com/mailman/listinfo/dm-devel
> > > 
> > 
> > --
> > dm-devel mailing list
> > dm-devel at redhat.com
> > https://www.redhat.com/mailman/listinfo/dm-devel
> > 
> > 
> 
> --
> dm-devel mailing list
> dm-devel at redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel

-- 
Philip R. Auld, Ph.D.  	        	       Egenera, Inc.    
Software Architect                            165 Forest St.
(508) 858-2628                            Marlboro, MA 01752