[linux-lvm] Re: Problem of LVM on Suse

[Heinz J . Mauelshagen]

On Thu, Oct 23, 2003 at 06:50:29PM -0600, adimaths at softhome.net wrote:
> Hi, 
> 
> I am facing a problem in using LVM bundled with Suse Enterprise server 8 
> ,connected to storage area newtork
> System configuration:
> A storage area network is connected to a server with OS Suse Enterprise 
> server 8.
> So there are multiple scsi devices whose physical disk is identical. 
> 
> We are creating a Pseudo device file asscociated with identical scsi devices 
> using a driver that we have created.
> eg : /dev/pseudo-device -> /dev/sda , /dev/sdb 
> 
> Now using the bundled LVM on such pseudo device is a problem.
> problem details:
> pvcreate /dev/pseudo-device
> [result] ok. 
> 
> pvscan
> [result] it shows the scsi device associated with this pseudo device ie 
> /dev/sda and not our pseudo device. 
> 
> If LVM bypasses our pseudo device then the purpose of creating this device 
> and associating it with scsi is futile. 
> 
> The irony is, LVM works fine on RH 2.1 kernel = 2.4.9-e.3 and also with RH 
> 7.2
> However the above mention problem is observed on Suse Enterprise server 8 
> ,RH 2.1 kernel =2.4.9-e.25, RH 7.3 and RH 8.0. 
> 
> Would request you to clarify the following :
> 1>Could you please tell me how this problem can be rectified ? 

Change the sort-order (as you suggested below)

> 
> 2>Is it that Lvm that is bundled with the OS can't be used for
> the above mentioned purpose ? 
> 
> 3>If it can't be , then what design feature
> of the current LVM prevents it from supporting the above task? 
> 
> 4> Can any change in our product help us to solve this problem? 
> 
> 5>Moreover I think LVM on RedHat is working just by chance.
> I think pvscan depends on the order of the devices in the
> /proc/partitions , and hence when our pseudo devices are
> at the end of the /proc/paritions , pvscan shows our devices
> as active else it shows scsi as active. 
> 
> Is my inference correct? 

Yes.

LVM2 is much more suitable to cover such vendor-specific configurations,
because it has configurable device name filters. LVM1 needs code changes
to support additional device name spaces.

> 
> Regards
> Aditya Vasudevan 

-- 

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
Mauelshagen at Sistina.com                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-