Open Stack

Hello,

I have an improvement idea for the instance group API, the design of 
which is described here:

https://wiki.openstack.org/wiki/InstanceGroupApiExtension

Currently, the anti-affinity policy is defined such that each member of 
the group shall be deployed on a distinct physical host, limiting the 
damage caused by a single host failure to a single member of the group.

In some cases, the availability requirements are not that strict. 
Sometimes it is acceptable to lose more than one instance at the same 
time. It would be good if the anti-affinity policy accepted the maximum 
number of instances per host as an argument, which would default to 1. 
Alternatively, the limit could be specified proportionally to the number 
of group members.

This is relevant when the size of the group exceeds the number of 
eligible hosts. This sceario might easily realize for those who want to 
minimize power consumption by suspending and shutting down as many hosts 
as possible and thus maximizing the utilization level of the remaining 
hosts.

BR.
Kaarle