CentOS - Apr 2007 - Multipath-root (mpath) problems with CentOS 5

Hi list!

I have a server with dual port Qlogic iSCSI HBA. I set up the same LUN for
both ports, and boot the CentOS installer with "linux mpath". 

Installer detects multipathing fine, and creates mpath0 device for root disk.
Installation goes fine, and the system boots up and works fine after the
install from the multipath root device.

After install the setup is like this:

LUN 0 on port 1: /dev/sda
LUN 0 on port 2: /dev/sdb

and those form mpath0 device, which works OK.

But then comes the problem.. when I add another LUN for both ports, the
setup is like this:

LUN 0 on port 1: /dev/sda
LUN 1 in port 1: /dev/sdb

LUN 0 on port 2: /dev/sdc
LUN 1 on port 2: /dev/sdd

And now after rebooting the system, root device multipath (mpath0) doesn't
work anymore.. initrd seems to use/mount /dev/sda* partitions instead of mpath0
device. In fact the whole mpath0 is not started up.

The weird thing is that the other LUN is now using multipathing instead.. 
using mpath1 device. 

How can I fix that? I'd like to use multipathing for both LUNs, also for
root
device.  

If I regenerate initrd with verbose flag, it just sets up the mpath1, not
mpath0. I guess that's because mpath0 is not running when regenerating
initrd. 

I have edited /etc/multipath.conf and disabled the "blacklist" line,
and
added multipath-entries for both LUNs. So those should be fine..

Any ideas? 

Thanks!

-- Pasi

Pasi K?rkk?inen wrote:
> Hi list!
>
> I have a server with dual port Qlogic iSCSI HBA. I set up the same LUN for
> both ports, and boot the CentOS installer with "linux mpath". 
>
> Installer detects multipathing fine, and creates mpath0 device for root
disk.
> Installation goes fine, and the system boots up and works fine after the
> install from the multipath root device.
>
> After install the setup is like this:
>
> LUN 0 on port 1: /dev/sda
> LUN 0 on port 2: /dev/sdb
>
> and those form mpath0 device, which works OK.
>
> But then comes the problem.. when I add another LUN for both ports, the
> setup is like this:
>
> LUN 0 on port 1: /dev/sda
> LUN 1 in port 1: /dev/sdb
>
> LUN 0 on port 2: /dev/sdc
> LUN 1 on port 2: /dev/sdd
>
> And now after rebooting the system, root device multipath (mpath0)
doesn't
> work anymore.. initrd seems to use/mount /dev/sda* partitions instead of
mpath0
> device. In fact the whole mpath0 is not started up.
>
> The weird thing is that the other LUN is now using multipathing instead.. 
> using mpath1 device. 
>
> How can I fix that? I'd like to use multipathing for both LUNs, also
for root
> device.  
>
> If I regenerate initrd with verbose flag, it just sets up the mpath1, not
> mpath0. I guess that's because mpath0 is not running when regenerating
> initrd. 
>
> I have edited /etc/multipath.conf and disabled the "blacklist"
line, and
> added multipath-entries for both LUNs. So those should be fine..
>
> Any ideas? 
>
> Thanks!
>
> -- Pasi
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> http://lists.centos.org/mailman/listinfo/centos
>   
There seems to be a bug in the upstream version with mpath & root 
devices.....I asked RedHat support to help me with a similar problem but 
go the answer that i should wait until update 1....the current 
implementation of mpath for boot devices seems to handle adding 
paths/devices somewhat weird if you add/remove paths....

Bgrds,
Finnur