openstack team mailing list archive

[Vishvananda Ishaya <vishvananda@xxxxxxxxx>]

On Nov 6, 2012, at 10:03 AM, "Dugger, Donald D" <donald.d.dugger@xxxxxxxxx> wrote:

> The compute code currently defines a set of capabilities (info about disk & memory consumption, cpuinfo and a few other things) that are periodically reported up to the scheduler.  You can then use the extra_specs to specify key/value pairs for a flavor that will be used by the scheduler’s compute_capabilities’ filter to make decisions based upon these capabilities.  Note you can do relational comparisons on these capabilities, look at:
>  
>                 http://docs.openstack.org/developer/nova/devref/filter_scheduler.html
>  
> for details.
>  
> Currently, you are limited to the capabilities that are hard coded into the compute node code.  I’m thinking about, and intend to create a blueprint in the not too distant future, adding a plugin into the compute code so that you can add support for new capabilities.

You should be able to add in capabilities via host aggregates as well if you enable this AggregateInstanceExtraSpecs filter. This means you can change them at run time.

Vish