Fedora xen - Jul 2008 - Xen and disaster recovery / redundancy / failover

Hi All,

 

Forgive me if this was discussed last week I''m new...

 

I currently run two servers, one for mail, one for web + db, but will be
expanding

In future so a virtual architecture seems a good idea, especially if I can
run the VM''s

on any of the physical machines on the fly as load / failures dictate.
Ideally this would

all be automatic and the VM''s would share processor / storage
transparently
amongst

available machines, and I would also like to add extra storage as necessary
without having

to rebuild arrays or reinstall stuff.  

 

I''d like to replace the os on both physical machines with minimal
fedora 8 +
xen and

run the mail / web / db services inside VM''s.   I''ve got a F9
system running
inside a VM

on the mailserver now but don''t know how to get it running on both
machines
as I''m 

not sure what''s possible with exporting / sharing VM settings.

 

For now, What I need is to be able to run the image of the F9 system on my
web box

when the mail server crashes so that the mail server is still available
should the physical 

mail server break.

 

My thinking was if both machines just run a minimal system + the VM''s
then
the 

other disks in the machines can be running a raided gfs across both machines
to 

provide a redundant storage system that can survive a machine failure.   

Hopefully then should a box pack up I can still run all my services in
VM''s
on a single

machine since the VM images will be available to both machines on the shared
filesystem.

 

I don''t even know which parts of this are possible or whether its a bad
strategy, perhaps 

there''s a better approach to all of this, ideas welcome,

 

Thanks,

 

Jack

2008/7/3 Jack <xen@tequilasolutions.com>:
> I currently run two servers, one for mail, one for web + db, but will be
> expanding in future so a virtual architecture seems a good idea, especially
if I can
> run the VM''s on any of the physical machines on the fly as load /
failures dictate.
If you want to be able to failover VMs between hosts, you need the
storage for the virtual disks to either be replicated between all
hosts (drbd, cluster filesystem, etc), or to be stored on a external
device (iSCSI, NFS etc) without that each VM depends on the single
server where its disk resides.