Re: [Xen-users] p2v migration + Xen

[Ralf Schenk <rs@xxxxxxxxxx>]

Dan Hawker schrieb:

> We have a machine in place that I would like to migrate, however am unsure
> as to what the best option would be. I am not going to get much time for
> the machine in question to be *not available*, so wondered if I could use
> rsync or similar to create a backup of the filesystem in question directly
> into a directory structure on the new server for the resultant Xen VM. I
> could do this whilst all things are still live. At the moment we'd like to
> swap over, we could then do a quick and final rsync, turn off the original
> server and bring the Xen VM up.

Hello !

I only wanted to agree to the other people commenting their migration
steps as I'm here in the middle of such a migration ( I'm in the phase
of testing the "old" productive server setup in a DOMU with a different
IP assigned at the moment.) Because the old server uses a bunch of IP's
for virtual webservers with SSL I can't test everything but thats enough
to say that I can do a last rsync with all relevant services disabled on
the production machine in a few minutes downtime and get it all up and
running in the DOMU on the twin server.

One thing will be different: I'm not using XEN with the goal to
consolidate servers but I want high availability (and flexibility).

My goal is to use the two servers (same hardware Supermicro DualXeon 3.2
  Ghz EM64T with 4 GB of RAM and SATA HW-RAID 1) with only 2 DOMU's and
a small DOM0 for management (256 MB or so).

The DOMU's will be mirrored crossover with DRBD (http://www.drbd.org)
between the two servers via direct gigabit crossover link. The servers
are LVM based but I needed to change the LVM layout a lot for this setup
(i.e. DRBD needs a place for metadata). Since one of the servers was
only running a test system so far but the other one is really in heavy
usage I was able to prepare the spare one from scratch.

When I'm done there will be two servers with only one active DOMU that
gets nearly all the ressources like CPU and RAM. I will be able to
switch on the other DOMU that represents the active DOMU of the other
server in a case of failure or for heavy upgrading of the base system.

This should also be combined with heartbeat to fail over automatically.

I would call this a "shared nothing active-active Webcluster" - with
minimal hardware-resources. In case of failure I will have only half the
maximum system performance - that is acceptable.

What I'm searching now:
If someone is doing something similar I would be interested in the
heartbeat configuration he/she uses.

Perhaps something to think of:
My servers were running DEBIAN Sarge (64bit) with the latest official
sarge-kernel which is 2.6.8 :-(. I tried to "xenify" with XEN 3.0.1 but
the 2.6.12-xen kernels that I used all crashed on this hardware in an
early stage of booting no matter wich noacpi, noapic, nousb etc.
parameters I used. I booted the system about 50-100 times with the help
of a serial console with no success. Even with the latest BIOS releases
it didn't work. So I switched to XEN-ustable with success. I was able to
boot but my initrd with root on LVM setup didn't work anymore because of
the missing devfs support. What really helped with minimal changes to
the base system (still original sarge glibc etc.) was to use a few
packages from http://www.backports.org. (udev, initramfs-tools, updated
module-init-tools). I've switched the DOMU's to udev too but I need no
initrd for the DOMU's since I compiled a customized 2.6.16.5-xen kernel
which I use for DOM0 and DOMU's.

-- 
__________________________________________________

Ralf Schenk
fon (02 41) 9 91 21-0
fax (02 41) 9 91 21-59
rs@xxxxxxxxxx

Databay AG
Hüttenstraße 7
D-52068 Aachen
www.databay.de

Databay - einfach machen.

_________________________________________________



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