On Wed, February 17, 2010 1:30 am, Vern Burke wrote:
> Matthew:
> The XCP cloud is currently 5 servers (2x Opteron single core with
> 4GB, soon to get a rolling upgrade to dual core Opterons and more
> memory). We're running about 40 virtual machines (all CentOS 5.4),
> practical limit is around 12 per. We leave the pool master open as extra
> capacity should a slave server croak.
> Networking is the standard XCP configuration using open vswitch.
> Public facing ports are static addresses, ports facing the storage area
> network are DHCP private addresses.
> Back end storage is a pair of the most rock solid NFS servers I
> could put together (2x Opteron, 4GB memory, 500GB drives on hardware
> RAID (soon to be upgraded to 1TB drives), oversize redundant power
> supply, etc) on a private Gigabit Ethernet storage network. I know this
> isn't the most bulletproof redundant storage configuration, but the
> MTBFs will hold us until I find something I'm happy with that doesn't
> require Rube Goldberg to make it work :).
> Problems were mostly confined to various bugaboos with XCP, such as
> leaving a critical file off the last distribution iso. All in all, I
> think XCP is one heckuva package, the bugs just come with the territory
> of being out on the front edge :).
Thanks, Verne
A couple more questions:
Are you using DRBD or similar to mirror the NFS data?
Do you aggregate links for more throughput?
This is the stuff we're mulling over at the moment. I might ask the same
question of this list to see what they have had success with in production
environments as we are trying to find the best bang for the buck in terms
of speed and cost.
Thanks,
Matt.
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
Home