Thanks a lot Robert! That helped a lot!

If I understand correctly if I don't want other flavors to end up on the specific hypervisor one way is to create a second aggregate with the rest hypervisors and the key set to false. Then modify the rest flavors so that the specific key is also set to false.

Instead of the above can I create a second aggregate with the rest hypervisors with another key set to true and then modify the rest flavors so that the second key is set to true? Wouldn't that mean that the rest flavors will be forced to use the remaining hypervisors so in such a way they are excluded from the first hypervisor?

Furthermore,

can I somehow setup the quota limits per hypervisor? Since all my nodes are not with the same specs can I somehow put quota limits for CPU and RAM per specific hypervisor?

Best regards,

George

On Tue, 25 Nov 2014 10:07:59 +0000, Robert van Leeuwen wrote:
Is it possible to configure OpenStack to boot a specific flavor only at a specific hypervisor and that hypervisor to accept only the specific
flavor and nothing else? If so how can I do it?

Yes, this is possible with host aggregates and metadata.
We did this for our database grade hypervisors which have large local
SSDs and have some other configuration settings:

First create an aggregate with the metadata with the hypervisor in it:
nova aggregate-create fastio
nova aggregate-set-metadata fastio ssdlarge=true
nova aggregate-add-host fastio hv1

Now create a flavor which uses this:
nova-manage instance_type set_key --name=m2.4xlarge --key=ssdlarge
--value=true

Note that this just makes sure this flavor always ends up on this Hypervisor.
To make sure all other flavors do not end up there you will need to
make a second aggregate with the other hypervisors with the same key
set to false + modify other flavors with this key set to false.


Cheers,
Robert van Leeuwen

--

_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to     : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack

Reply via email to