[one-users] Using shared fs and ssh TM

Fri Jul 27 09:02:28 PDT 2012

Somehow the checkpoint did not get copied. Can you try to stop a VM
and check if the checkpoint is transfered back to the frontend? Just
to see if the tm is working correctly.

On Fri, Jul 27, 2012 at 11:49 AM, Andreas Calvo <andreas.calvo at scytl.com> wrote:
> No,
> System datastore (where all QCOW2 delta are stored) is not shared, it relays
> on the node local filesystem.
> However, datastore 1 (where all images are stored) it is stored and shared
> on all nodes under the same path.
>
> To use the same directory logic, the paths have been linked.
>
> In this case:
> /var/lib/one is shared on all nodes
> /var/lib/one/datastores/0 is linked to /one/datastores/0, which is local
> /one/datastores/1 is linked to /var/lib/one/datastores/1 (which is shared)
>
> If I'm not wrong, the system datastores holds the incremental changes,
> whereas the other datastores hold images.
> El 27/07/12 11:22, Javier Fontan escribió:
>
>> Is your system datastore (0) shared and mounted in all your nodes?
>>
>> On Wed, Jul 25, 2012 at 3:56 PM, Andreas Calvo <andreas.calvo at scytl.com>
>> wrote:
>>>
>>> Hello again,
>>> I've tried to reuse the SSH MV script, but it fails.
>>> Output is:
>>>
>>> Wed Jul 25 15:51:59 2012 [VMM][I]: Command execution fail:
>>> /var/tmp/one/vmm/kvm/restore /var/lib/one//datastores/0/5220/checkpoint
>>> cloud13 5220 cloud13
>>> Wed Jul 25 15:51:59 2012 [VMM][E]: restore: Command "virsh --connect
>>> qemu:///system restore /var/lib/one//datastores/0/5220/checkpoint"
>>> failed:
>>> error: Failed to restore domain from
>>> /var/lib/one//datastores/0/5220/checkpoint
>>> Wed Jul 25 15:51:59 2012 [VMM][I]: error: Failed to create file
>>> '/var/lib/one//datastores/0/5220/checkpoint': No such file or directory
>>> Wed Jul 25 15:51:59 2012 [VMM][E]: Could not restore from
>>> /var/lib/one//datastores/0/5220/checkpoint
>>> Wed Jul 25 15:51:59 2012 [VMM][I]: ExitCode: 1
>>> Wed Jul 25 15:51:59 2012 [VMM][I]: Failed to execute virtualization
>>> driver
>>> operation: restore.
>>> Wed Jul 25 15:51:59 2012 [VMM][E]: Error restoring VM: Could not restore
>>> from /var/lib/one//datastores/0/5220/checkpoint
>>> Wed Jul 25 15:51:59 2012 [DiM][I]: New VM state is FAILED
>>>
>>>
>>> Any thought?
>>> I have to check but I expect to have the same problem when
>>> stopping/resuming
>>> VMs.
>>>
>>> El 11/07/12 18:38, Javier Fontan escribió:
>>>
>>> You are right. I've overlooked the driver. In qcow the mv driver is
>>> dummy as it expects the qcow image to be shared. You can just copy mv
>>> script from ssh tm to qcow remotes directory. I have not tested that
>>> but it should work. The qcow image will me moved on stop to the
>>> frontend and on resume back to a node. The backing storage path should
>>> be the same in your setup.
>>>
>>> On Wed, Jul 11, 2012 at 6:07 PM, Andreas Calvo <andreas.calvo at scytl.com>
>>> wrote:
>>>
>>> The shared storage is mounted in the same place in all the nodes.
>>>
>>> The directory structured is as follows:
>>> /var/lib/one/datastores (shared storage)
>>> -- 0 -> link to /one/datastores/0
>>> -- 1
>>> -- 100
>>>
>>> /one/datastores/ (local storage)
>>> -- 0
>>> -- 1 -> link to /var/lib/one/datastores/1
>>> -- 100 -> link to /var/lib/one/datastores/100
>>>
>>> When a virtual machine is stopped, its delta is stored in
>>> /one/datastores/0/$VMID.
>>> When another node tries to resume the virtual machine, expects to have in
>>> the same place the delta (and some other files, like the checkpoint),
>>> but,
>>> as it is not in its local disk, it fails.
>>>
>>> Same thing happen when a virtual machine is migrated.
>>> I think the TM should be tweaked to be part qcow and part SSH.
>>>
>>> El 11/07/12 17:51, Javier Fontan escribió:
>>>
>>> If the shared datastore is mounted in the same place in both nodes
>>> there wont be any problem. The base image will be accessible from both
>>> nodes (shared) and the qcow delta is moved to the host.
>>>
>>> On Wed, Jul 11, 2012 at 12:59 PM, Andreas Calvo <andreas.calvo at scytl.com>
>>> wrote:
>>>
>>> Javier,
>>> Thanks for your quick reply!
>>> The only problem is that when a stopped virtual machine is resumed from
>>> another host in the cloud (different from which it was stopped), as it
>>> cannot find the deployed (in this case, linked) image and the deltas.
>>>
>>> El mar 10 jul 2012 11:46:55 CEST, Javier Fontan escribió:
>>>
>>> Sure it is possible. You need to have the datastore that holds the
>>> images shared and mount it in every node. Then you'll make the system
>>> datastore(0) local and configure it to use qcow tm drivers. Make sure
>>> that /var/lib/one/datastores/0 is not mounted from the shared storage
>>> in the nodes as this is where the deltas will be written.
>>>
>>> On Mon, Jul 9, 2012 at 11:59 AM, Andreas Calvo <andreas.calvo at scytl.com>
>>> wrote:
>>>
>>> Hello,
>>> Is there any way to mix TM in an environment?
>>> We currently have a shared FS with GFS2 using qcow, but when a lot of
>>> VMs
>>> are launched, writing changes becomes a I/O bottleneck.
>>> We were thinking of a mixture where the image is shared and the
>>> incremental
>>> changes (qcow) are write locally.
>>>
>>> Is it possible?
>>>
>>> Thanks
>>>
>>> --
>>> Andreas Calvo Gómez
>>> Systems Engineer
>>> Scytl Secure Electronic Voting
>>> Plaça Gal·la Placidia, 1-3, 1st floor · 08006 Barcelona
>>> Phone: + 34 934 230 324
>>> Fax: + 34 933 251 028
>>> http://www.scytl.com
>>>
>>> NOTICE: The information in this e-mail and in any of its attachments is
>>> confidential and intended solely for the attention and use of the named
>>> addressee(s). If you are not the intended recipient, any disclosure,
>>> copying,
>>> distribution or retaining of this message or any part of it, without
>>> the
>>> prior
>>> written consent of Scytl Secure Electronic Voting, SA is prohibited and
>>> may be
>>> unlawful. If you have received this in error, please contact the sender
>>> and
>>> delete the material from any computer.
>>>
>>> _______________________________________________
>>> Users mailing list
>>> Users at lists.opennebula.org
>>> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>>>
>>>
>>>
>>>
>>> --
>>> Andreas Calvo Gómez
>>> Systems Engineer
>>> Scytl Secure Electronic Voting
>>> Plaça Gal·la Placidia, 1-3, 1st floor · 08006 Barcelona
>>> Phone: + 34 934 230 324
>>> Fax:   + 34 933 251 028
>>> http://www.scytl.com
>>>
>>> NOTICE: The information in this e-mail and in any of its attachments is
>>> confidential and intended solely for the attention and use of the named
>>> addressee(s). If you are not the intended recipient, any disclosure,
>>> copying,
>>> distribution or retaining of this message or any part of it, without the
>>> prior
>>> written consent of Scytl Secure Electronic Voting, SA is prohibited and
>>> may be
>>> unlawful. If you have received this in error, please contact the sender
>>> and
>>> delete the material from any computer.
>>>
>>>
>>> --
>>> Andreas Calvo Gómez
>>> Systems Engineer
>>> Scytl Secure Electronic Voting
>>> Plaça Gal·la Placidia, 1-3, 1st floor · 08006 Barcelona
>>> Phone: + 34 934 230 324
>>> Fax:   + 34 933 251 028
>>> http://www.scytl.com
>>>
>>> NOTICE: The information in this e-mail and in any of its attachments is
>>> confidential and intended solely for the attention and use of the named
>>> addressee(s). If you are not the intended recipient, any disclosure,
>>> copying,
>>> distribution or retaining of this message or any part of it, without the
>>> prior
>>> written consent of Scytl Secure Electronic Voting, SA is prohibited and
>>> may be
>>> unlawful. If you have received this in error, please contact the sender
>>> and
>>> delete the material from any computer.
>>>
>>>
>>>
>>>
>>>
>>>
>>> --
>>> Andreas Calvo Gómez
>>> Systems Engineer
>>> Scytl Secure Electronic Voting
>>> Plaça Gal·la Placidia, 1-3, 1st floor · 08006 Barcelona
>>> Phone: + 34 934 230 324
>>> Fax:   + 34 933 251 028
>>> http://www.scytl.com
>>>
>>> NOTICE: The information in this e-mail and in any of its attachments is
>>> confidential and intended solely for the attention and use of the named
>>> addressee(s). If you are not the intended recipient, any disclosure,
>>> copying,
>>> distribution or retaining of this message or any part of it, without the
>>> prior
>>> written consent of Scytl Secure Electronic Voting, SA is prohibited and
>>> may be
>>> unlawful. If you have received this in error, please contact the sender
>>> and
>>> delete the material from any computer.
>>>
>>>
>>>
>>
>>
>
> --
> Andreas Calvo Gómez
> Systems Engineer
> Scytl Secure Electronic Voting
> Plaça Gal·la Placidia, 1-3, 1st floor · 08006 Barcelona
> Phone: + 34 934 230 324
> Fax:   + 34 933 251 028
> http://www.scytl.com
>
> NOTICE: The information in this e-mail and in any of its attachments is
> confidential and intended solely for the attention and use of the named
> addressee(s). If you are not the intended recipient, any disclosure,
> copying,
> distribution or retaining of this message or any part of it, without the
> prior
> written consent of Scytl Secure Electronic Voting, SA is prohibited and
> may be
> unlawful. If you have received this in error, please contact the sender
> and
> delete the material from any computer.
>
>
>

-- 
Javier Fontán Muiños
Project Engineer
OpenNebula - The Open Source Toolkit for Data Center Virtualization
www.OpenNebula.org | jfontan at opennebula.org | @OpenNebula