Open Stack

Hi,
one possible solution to "mitigate" this problem is to stacking instances
instead of spreading them.
This means that the scheduler will select the compute "most used",
maximizing in this way the space for large instances.
For that you need to set "ram_weight_multiplier" option with a negative
number. And depending in the size of your infrastructure do some tuning
with "scheduler_host_subset_size".

cheers,
Belmiro

On Wed, Sep 24, 2014 at 5:08 PM, Blair Bethwaite <blair.bethwaite at gmail.com>
wrote:

> Hi all,
>
> I'm trying to wrap my head around whether it's possible with the existing
> scheduler filters, to put a limit per host on the number of instances per
> instance-type/flavor? I don't think this is possible with the existing
> filters or weights, but it seems like a fairly common requirement.
>
> The issue I'm thinking of using this for is that of instance-to-host
> fragmentation in homogenous deployments, where there is a tendency as the
> zone approaches capacity to hit a utilisation ceiling - there are rarely
> any "gaps" large enough for high vcpu count instances. I'm guessing that
> limiting the number of smaller instances per host would help to alleviate
> this.
>
> Looks like knocking up such a filter wouldn't be too hard, just want to
> check whether there is another way...?
>
> --
> Cheers,
> ~Blairo
>
> _______________________________________________
> OpenStack-operators mailing list
> OpenStack-operators at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20140924/e79dd0a1/attachment.html>