[linux-lvm] scsi/fc multipath failover patch

[linux-lvm] scsi/fc multipath failover patch

[Michael E Brown]

All,
  I've been lurking on the list for a while, and I just found something
that would probably be of great value to everyone. While I've heard
people ask for it, I haven't seen anybody who knew of a patch to provide
multipath failover for Linux.  Linuxcare developed multipath failover
when they made the Sun T3 driver, although I haven't seen it publicized.
Their kernel patch provides a generic multipath-failover layer for
linux. The kernel patch is fairly small and clean, and comes with a
couple userspace tools to set everything up. The URLS are below

http://open-projects.linuxcare.com/t3/   <--Generic project homepage
http://open-projects.linuxcare.com/t3/download/sun-t3-0.3.1.tar.gz

cheers!
Michael Brown

Re: [linux-lvm] scsi/fc multipath failover patch

[Jos Visser]

Hi,

Does anyone know whether md, lvm and likeminded drivers are resistant
against this alternate pathing solution? I mean, when you connect a SCSI
(or FC disk for that matter) through two controllers, any application or
driver that does scanning of the attached disks and partitions will
encounter the same disk/partitions twice. 

In order to solve this, they must have some way of identifying the disk 
uniquely, and thus being able to conclude that they've seen it before.
HP-UX solves this by having a unique Pyshical Volume ID in the disk
header (PVID), thus it knows that a certain physical disk is an
alternate path for a disk that it has seen before.

Anyone any clues on how this is solved in a LinuxCare T3 patch/lvm/md
environment?

++Jos



And thus it came to pass that Michael E Brown wrote:
(on Mon, Nov 06, 2000 at 09:56:44AM -0600 to be exact)

> All,
>   I've been lurking on the list for a while, and I just found something
> that would probably be of great value to everyone. While I've heard
> people ask for it, I haven't seen anybody who knew of a patch to provide
> multipath failover for Linux.  Linuxcare developed multipath failover
> when they made the Sun T3 driver, although I haven't seen it publicized.
> Their kernel patch provides a generic multipath-failover layer for
> linux. The kernel patch is fairly small and clean, and comes with a
> couple userspace tools to set everything up. The URLS are below
> 
> http://open-projects.linuxcare.com/t3/   <--Generic project homepage
> http://open-projects.linuxcare.com/t3/download/sun-t3-0.3.1.tar.gz
> 
> cheers!
> Michael Brown

-- 
Success and happiness can not be pursued; it must ensue as the 
unintended side-effect of one's personal dedication to a course greater 
than oneself.

Re: [linux-lvm] scsi/fc multipath failover patch

[Andreas Dilger]

Joe Visser writes:
> Does anyone know whether md, lvm and likeminded drivers are resistant
> against this alternate pathing solution? I mean, when you connect a SCSI
> (or FC disk for that matter) through two controllers, any application or
> driver that does scanning of the attached disks and partitions will
> encounter the same disk/partitions twice. 

Heinz has said that in the upcoming 0.9 release of LVM, there will be
PVIDs (actually DCE UUIDs) so LVM will remove duplicate PVs from its
lists.  This can happen for reasons other than multipath, such as
/dev/md and /dev/{hd,sd} access to the same PV.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

Re: [linux-lvm] scsi/fc multipath failover patch

[Michael E Brown]

Andreas Dilger wrote:

> Joe Visser writes:
> > Does anyone know whether md, lvm and likeminded drivers are resistant
> > against this alternate pathing solution? I mean, when you connect a SCSI
> > (or FC disk for that matter) through two controllers, any application or
> > driver that does scanning of the attached disks and partitions will
> > encounter the same disk/partitions twice.
>
> Heinz has said that in the upcoming 0.9 release of LVM, there will be

Is there cvs repository or some such thing so that us mere mortals can
test/contribute to upcoming release? I've a lab full of machines that can be
used for the cause, several fibre-channel devices, lots of scsi external
storage, and lots of other neat toys I can do horrible experiments on without
whining about lost data.

>
> PVIDs (actually DCE UUIDs) so LVM will remove duplicate PVs from its
> lists.  This can happen for reasons other than multipath, such as
> /dev/md and /dev/{hd,sd} access to the same PV.
>
> Cheers, Andreas
> --
> Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
>                  \  would they cancel out, leaving him still hungry?"
> http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert