[linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[Romeo Theriault]

I apologize if this has been asked before but I was unable to find any
references to it in the mailing list archives and googling around
isn't helping much.

Is there a way to directly mount, bypassing LVM, a ext3 partition that
resides in a LVM LV and VG? It resides on one PV.

The problem is that I have read-only copies of VM's that are backed up
via a SAN based snapshot/backup tool. I'm trying to create a method to
allow the VM admins to restore their files from the snapshot backed up
VM's. I'm at the point where I can access the partitions and can mount
the ext3 partitions fine but I'm having trouble with the LVM volumes
because I can't import the pv/vg because they have the same name/uuid
as the existing VG (it's really the same system) and I can't change
them with something like vgimportclone because the backed up vm image
files are read-only.

Thanks for any help.

--
Romeo Theriault

Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[Ray Morris]

   In general, no, though in your specific case it may be
that all of the extents are stored contigously and it might
work. You would probably need to use losetup -o and specify
the offset.  This would be more of a last ditch data recovery
effort than something you'd design into a production system,
though.

   You might have better results altering one of these issues
that is causing you to consider such action:

>  I can't import the pv/vg because they have the same name/uuid
> as the existing VG (it's really the same system) and I can't change
> them with something like vgimportclone because the backed up vm image
> files are read-only.

   In other words, you would back up to one of the following questions:
How can I change the names and UUIDs of the backups?
How can I import an LV which has a conflicting UUID?
How can I use vgimport with a read only source?

   Specifically, you might be able to ignore the meta data on
the backup volumes with pvmetadatacopies  = 0 and use "dirs"
in lvm.conf, so you can change the working meta data even
though they are read only.
--
Ray Morris
support@bettercgi.com

Strongbox - The next generation in site security:
http://www.bettercgi.com/strongbox/

Throttlebox - Intelligent Bandwidth Control
http://www.bettercgi.com/throttlebox/

Strongbox / Throttlebox affiliate program:
http://www.bettercgi.com/affiliates/user/register.php


On 05/31/2010 02:21:36 PM, Romeo Theriault wrote:
> I apologize if this has been asked before but I was unable to find any
> references to it in the mailing list archives and googling around
> isn't helping much.
> 
> Is there a way to directly mount, bypassing LVM, a ext3 partition that
> resides in a LVM LV and VG? It resides on one PV.
> 
> The problem is that I have read-only copies of VM's that are backed up
> via a SAN based snapshot/backup tool. I'm trying to create a method to
> allow the VM admins to restore their files from the snapshot backed up
> VM's. I'm at the point where I can access the partitions and can mount
> the ext3 partitions fine but I'm having trouble with the LVM volumes
> because I can't import the pv/vg because they have the same name/uuid
> as the existing VG (it's really the same system) and I can't change
> them with something like vgimportclone because the backed up vm image
> files are read-only.
> 
> Thanks for any help.
> 
> --
> Romeo Theriault
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
> 
> 

Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[Hull, Brett (MSE)]

Hello,

  Can you restore these backups to another system?  Such as one designated to do this job?  Maybe using a virtual machine? You could then create the environment restore the data and move the "recovered files" to the production box. I do not know your environment, but you might not want to change the backed up data because that is what you will use to restore if there is a full loss of data on your production system.

Best regards,
Brett

-----Original Message-----
From: linux-lvm-bounces@redhat.com [mailto:linux-lvm-bounces@redhat.com] On Behalf Of Ray Morris
Sent: Monday, May 31, 2010 2:45 PM
To: LVM general discussion and development
Subject: Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

   In general, no, though in your specific case it may be
that all of the extents are stored contigously and it might
work. You would probably need to use losetup -o and specify
the offset.  This would be more of a last ditch data recovery
effort than something you'd design into a production system,
though.

   You might have better results altering one of these issues
that is causing you to consider such action:

>  I can't import the pv/vg because they have the same name/uuid
> as the existing VG (it's really the same system) and I can't change
> them with something like vgimportclone because the backed up vm image
> files are read-only.

   In other words, you would back up to one of the following questions:
How can I change the names and UUIDs of the backups?
How can I import an LV which has a conflicting UUID?
How can I use vgimport with a read only source?

   Specifically, you might be able to ignore the meta data on
the backup volumes with pvmetadatacopies  = 0 and use "dirs"
in lvm.conf, so you can change the working meta data even
though they are read only.
--
Ray Morris
support@bettercgi.com

Strongbox - The next generation in site security:
http://www.bettercgi.com/strongbox/

Throttlebox - Intelligent Bandwidth Control
http://www.bettercgi.com/throttlebox/

Strongbox / Throttlebox affiliate program:
http://www.bettercgi.com/affiliates/user/register.php


On 05/31/2010 02:21:36 PM, Romeo Theriault wrote:
> I apologize if this has been asked before but I was unable to find any
> references to it in the mailing list archives and googling around
> isn't helping much.
> 
> Is there a way to directly mount, bypassing LVM, a ext3 partition that
> resides in a LVM LV and VG? It resides on one PV.
> 
> The problem is that I have read-only copies of VM's that are backed up
> via a SAN based snapshot/backup tool. I'm trying to create a method to
> allow the VM admins to restore their files from the snapshot backed up
> VM's. I'm at the point where I can access the partitions and can mount
> the ext3 partitions fine but I'm having trouble with the LVM volumes
> because I can't import the pv/vg because they have the same name/uuid
> as the existing VG (it's really the same system) and I can't change
> them with something like vgimportclone because the backed up vm image
> files are read-only.
> 
> Thanks for any help.
> 
> --
> Romeo Theriault
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
> 
> 


_______________________________________________
linux-lvm mailing list
linux-lvm@redhat.com
https://www.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[Romeo Theriault]

Hi,

> �Can you restore these backups to another system? �Such as one designated to do this job?

Yes, I've thought about this a bit too. The problem that I see with
this is that I'll only be able to have one restore at a time
happening, since almost all our linux vm's have the same named VG's
and LV's and they are on read-only media so I wouldn't be able to
rename them....

I've going to have to keeping thinking about this one.

Thanks for the pointers.

Romeo

Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[Ray Morris]

> I'll only be able to have one restore at a time happening,
> since almost all our linux vm's have the same named VG's
> and LV's

    Fedora used to by default name the VG for every machine
volgroup001 or something like that.  They then changed to
naming it after the machine, I believe.  You are discovering
one of the reasons why having them all named the same is a
bad idea.  When a SAN is involved, one can of course envision
several other problems limits on future usage.  Your best course
of action may be to eliminate the root cause and rename the
source VGs.
--
Ray Morris
support@bettercgi.com

Strongbox - The next generation in site security:
http://www.bettercgi.com/strongbox/

Throttlebox - Intelligent Bandwidth Control
http://www.bettercgi.com/throttlebox/

Strongbox / Throttlebox affiliate program:
http://www.bettercgi.com/affiliates/user/register.php


On 06/01/2010 08:27:22 AM, Romeo Theriault wrote:
> Hi,
> 
> > �Can you restore these backups to another system? �Such as one  
> designated to do this job?
> 
> Yes, I've thought about this a bit too. The problem that I see with
> this is that I'll only be able to have one restore at a time
> happening, since almost all our linux vm's have the same named VG's
> and LV's and they are on read-only media so I wouldn't be able to
> rename them....
> 
> I've going to have to keeping thinking about this one.
> 
> Thanks for the pointers.
> 
> Romeo
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
> 
> 

Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[Romeo Theriault]

On Tue, Jun 1, 2010 at 22:54, Ray Morris <support@bettercgi.com> wrote:
>> I'll only be able to have one restore at a time happening,
>> since almost all our linux vm's have the same named VG's
>> and LV's
>
> � Fedora used to by default name the VG for every machine
> volgroup001 or something like that. �They then changed to
> naming it after the machine, I believe. �You are discovering
> one of the reasons why having them all named the same is a
> bad idea. �When a SAN is involved, one can of course envision
> several other problems limits on future usage. �Your best course
> of action may be to eliminate the root cause and rename the
> source VGs.

Hmmm, interesting solution. I like it, though it's going to take some
work to get all the VM's changed...

Thanks!

Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[K. Richard Pixley]

[-- Attachment #1: Type: text/plain, Size: 4054 bytes --]

You might be able to make those partitions available using nbd from the 
server, importing them on the vms.

--rich

Hull, Brett (MSE) wrote:
> Hello,
>
>   Can you restore these backups to another system?  Such as one designated to do this job?  Maybe using a virtual machine? You could then create the environment restore the data and move the "recovered files" to the production box. I do not know your environment, but you might not want to change the backed up data because that is what you will use to restore if there is a full loss of data on your production system.
>
> Best regards,
> Brett
>
> -----Original Message-----
> From: linux-lvm-bounces@redhat.com [mailto:linux-lvm-bounces@redhat.com] On Behalf Of Ray Morris
> Sent: Monday, May 31, 2010 2:45 PM
> To: LVM general discussion and development
> Subject: Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?
>
>    In general, no, though in your specific case it may be
> that all of the extents are stored contigously and it might
> work. You would probably need to use losetup -o and specify
> the offset.  This would be more of a last ditch data recovery
> effort than something you'd design into a production system,
> though.
>
>    You might have better results altering one of these issues
> that is causing you to consider such action:
>
>   
>>  I can't import the pv/vg because they have the same name/uuid
>> as the existing VG (it's really the same system) and I can't change
>> them with something like vgimportclone because the backed up vm image
>> files are read-only.
>>     
>
>    In other words, you would back up to one of the following questions:
> How can I change the names and UUIDs of the backups?
> How can I import an LV which has a conflicting UUID?
> How can I use vgimport with a read only source?
>
>    Specifically, you might be able to ignore the meta data on
> the backup volumes with pvmetadatacopies  = 0 and use "dirs"
> in lvm.conf, so you can change the working meta data even
> though they are read only.
> --
> Ray Morris
> support@bettercgi.com
>
> Strongbox - The next generation in site security:
> http://www.bettercgi.com/strongbox/
>
> Throttlebox - Intelligent Bandwidth Control
> http://www.bettercgi.com/throttlebox/
>
> Strongbox / Throttlebox affiliate program:
> http://www.bettercgi.com/affiliates/user/register.php
>
>
> On 05/31/2010 02:21:36 PM, Romeo Theriault wrote:
>   
>> I apologize if this has been asked before but I was unable to find any
>> references to it in the mailing list archives and googling around
>> isn't helping much.
>>
>> Is there a way to directly mount, bypassing LVM, a ext3 partition that
>> resides in a LVM LV and VG? It resides on one PV.
>>
>> The problem is that I have read-only copies of VM's that are backed up
>> via a SAN based snapshot/backup tool. I'm trying to create a method to
>> allow the VM admins to restore their files from the snapshot backed up
>> VM's. I'm at the point where I can access the partitions and can mount
>> the ext3 partitions fine but I'm having trouble with the LVM volumes
>> because I can't import the pv/vg because they have the same name/uuid
>> as the existing VG (it's really the same system) and I can't change
>> them with something like vgimportclone because the backed up vm image
>> files are read-only.
>>
>> Thanks for any help.
>>
>> --
>> Romeo Theriault
>>
>> _______________________________________________
>> linux-lvm mailing list
>> linux-lvm@redhat.com
>> https://www.redhat.com/mailman/listinfo/linux-lvm
>> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>>
>>
>>     
>
>
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>   

[-- Attachment #2: Type: text/html, Size: 5648 bytes --]

Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[Romeo Theriault]

On Tue, Jun 1, 2010 at 05:45, Ray Morris <support@bettercgi.com> wrote:
>
> �In general, no, though in your specific case it may be
> that all of the extents are stored contigously and it might
> work. You would probably need to use losetup -o and specify
> the offset. �This would be more of a last ditch data recovery
> effort than something you'd design into a production system,
> though.


Alright, was hoping for a life saver here but it doesn't sound like
it's this way and you're right it doesn't sound production worthy.


>
> �You might have better results altering one of these issues
> that is causing you to consider such action:
>
>> �I can't import the pv/vg because they have the same name/uuid
>> as the existing VG (it's really the same system) and I can't change
>> them with something like vgimportclone because the backed up vm image
>> files are read-only.
>
> �In other words, you would back up to one of the following questions:
> How can I change the names and UUIDs of the backups?
> How can I import an LV which has a conflicting UUID?
> How can I use vgimport with a read only source?
>
> �Specifically, you might be able to ignore the meta data on
> the backup volumes with pvmetadatacopies �= 0 and use "dirs"
> in lvm.conf, so you can change the working meta data even
> though they are read only.

Hmm, this sounds interesting I'll have to spend some time trying to
figure out how this might be possible.

Thanks for the ideas.


Romeo

Re: [linux-lvm] Is it possible to bypass LVM and mount contained partition directly?

[Bryn M. Reeves]

On 05/31/2010 08:21 PM, Romeo Theriault wrote:
> I apologize if this has been asked before but I was unable to find any
> references to it in the mailing list archives and googling around
> isn't helping much.
> 
> Is there a way to directly mount, bypassing LVM, a ext3 partition that
> resides in a LVM LV and VG? It resides on one PV.

If you know how the volume group is laid out on the PVs you can manually
construct device-mapper devices using the dmsetup command to get access
to the LVs in this VG. This is a bit fiddly (need to convert from
extents to sectors to pass as arguments to device-mapper) but it is
possible.

> The problem is that I have read-only copies of VM's that are backed up
> via a SAN based snapshot/backup tool. I'm trying to create a method to
> allow the VM admins to restore their files from the snapshot backed up
> VM's. I'm at the point where I can access the partitions and can mount
> the ext3 partitions fine but I'm having trouble with the LVM volumes
> because I can't import the pv/vg because they have the same name/uuid
> as the existing VG (it's really the same system) and I can't change
> them with something like vgimportclone because the backed up vm image
> files are read-only.

If you're not able to use vgimportclone then another alternative would
be to create a new copy of /etc/lvm somewhere and modify the
etc/lvm/lvm.conf that it contains so that it has a filter that excludes
everything apart from the devices containing this VG. Then you can set
LVM_SYSTEM_DIR to point to this directory and have a "sandbox" that is
restricted to just seeing the "clone" devices.

You will probably still run into problems activating LVs this way since
e.g. the device names and the VG directory in /dev will collide with the
ones used by the system VG but at least you can run the tools verbosely
this way and see the maps that would be generated for device-mapper
without having to work them out by hand.

Regards,
Bryn.