Re: [Xen-users] migrating with physical device as VBD

["Antonio Mauriello" <antoniomau@xxxxxxxxx>]

Hi at all,
as Jd knows, in my thesis project I'm developing a migration feature between not shared disks.
So, when I start migration between two physical nodes, it's copied also the virtual disk file (with the config file as ConVirt migration works).
At moment the feature works but it's needs many corrections (it needs a snapsot of state and restore in destination node yet).
Basically the feature copy the virtual disk file before ConVirt migration stars, so when migrations_check starts it finds virtual disk file and doens't go in error.

Regards,
Antonio Mauriello
student at Federico II University, Napoli, Italy.

On Tue, Apr 1, 2008 at 4:58 PM, Ross S. W. Walker <rwalker@xxxxxxxxxxxxx> wrote:

Jan Schulze wrote:
> Ross S. W. Walker wrote:
> > Jan Schulze wrote:
> >> Hi all,
> >>
> >> I am wondering, whether it's possible to migrate a DomU from one XEN Host to
> >> another *without* using shared storage. All HowTos that I found so far, are
> >> using some kind of shared filesystem or block device.
> >
> > Shared storage is mandatory, but it doesn't have to be traditional multi-host
> > scsi or fiber channel, with technology like iscsi and software like drbd
> > shared storage can be done on the cheap.
> >
> >> The DomU in question is using a physical device (partition /dev/sda1) that is
> >> exported as VBD from Dom0. If migration is possible with that setup, where will
> >> the DomU end up on the target system?
> >
> > It needs to be accessible there too at the same time, thus why shared storage
> > is mandatory. Check out drbd, the latest version does primary-primary which
> > is what you need for live migrations.
>
> DRBD seems really interesting. However, when using primary-primary, I
> would also need a cluster file system. As my existing DomU uses ext3, I
> won't be able to use DRBD in a primary-primary setup, right?

No not in this case because the domUs file systems don't come into play
here. The Xen servers will have access to the domUs disks at the same
time, but they will not write at the same time. Only 1 xen host has
write access and during live migration they coordinate the hand-off
of this access along with the network and running memory.

Question is whether the drbd can work quick enough in this fashion
to have the disks in sync by the time the hand-over is complete. You
may need to up the pause state from say 50ms to 100ms just to give
drbd enough time to complete the replication.

> >> In case, migrating this DomU won't work, I would have to copy the VM manually.
> >> Is it possible to access the contents of /dev/sda1 from Dom0? Simply shutting
> >> down the DomU and mounting /dev/sda1 in Dom0 does not work.
> >
> > There will be a partition table within that partition, so you can run
> > kpartx -a /dev/sda1, and access the first partition via /dev/mapper/sda11,
> > or mount it via a loopback mount to the offset of that partition.
> >
> > # mount -t <fstype> -oloop,offset=<part offset> /dev/sda1 /mnt
>
> Thanks, did not know that. Very useful.

Always happy to pass the knowledge along as it was passed to me.

-Ross

______________________________________________________________________
This e-mail, and any attachments thereto, is intended only for use by
the addressee(s) named herein and may contain legally privileged
and/or confidential information. If you are not the intended recipient
of this e-mail, you are hereby notified that any dissemination,
distribution or copying of this e-mail, and any attachments thereto,
is strictly prohibited. If you have received this e-mail in error,
please immediately notify the sender and permanently delete the
original and any copy or printout thereof.


_______________________________________________

Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users