Open Stack

Hi Daneyon,


> I am looking for any recommendations on implementing nova-volume at scale.
>  Is it best to have dedicated nodes that run nova-volume or to run
> nova-volume on compute nodes?  Also, does it make sense to have the
> nova-volumes Volume Group shared with the compute instance directory?
> Any insight would be greatly appreciated.
>

I am interested in hearing best practices about this as well. In the
meantime, I can explain what we're doing with regards to this - and who
knows, maybe it's already a best practice.

We run nova-volume and nova-compute on the same node. This is so we can
most efficiently use the hardware we have purchased. For example, if you
have a node with 32 cores, 128gb of RAM, and 20TB of disk space, dedicating
the whole node to nova-compute will utilize all cores, RAM, but maybe only
2TB of disk. So we add nova-volume to the node to make up for that.

The alternative would be to purchase two types of servers with vastly
different hardware configurations. IMO, the cost and management overhead
with this is not worth it.

We have not looked into a single storage device such as a NAS or SAN nor
have we looked into distributed block filesystems such as Sheepdog or Ceph.
When we do, it is quite possible that our architecture will change.

With regards to the nova-volumes VG + compute instances, we create a LV
slice of nova-volumes called "nova-instances" and mount it at
/var/lib/nova/instances.

Thanks,
Joe


-- 
Joe Topjian
Systems Administrator
Cybera Inc.

www.cybera.ca

Cybera is a not-for-profit organization that works to spur and support
innovation, for the economic benefit of Alberta, through the use
of cyberinfrastructure.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20120523/339ff52c/attachment-0002.html>