[Pacemaker] Making utilizations dynamic

[Michael Smith]

Michael Schwartzkopff wrote:

> Imagine the consequences for a cloud cluster consisting of 30 nodes hosting 
> 100 virtual machines. All machines would be migrated to the least possible 
> number of real machines during the night when there no work to do. In the next 
> morning when work starts virtual machines would be migrated to free machines 
> when the work load gradually increases.

Hi Michael,

My needs are much simpler, but related: I have a high-availability 
cluster hosting Xen instances, and in case of failure or planned outage 
I want to make sure available RAM is considered when deciding where 
instances are migrated.

The available RAM on a destination node would need to be evaluated just 
before moving the VM. It's complicated by the fact that, during live 
migration, the destination node's available memory drops slowly until 
the instance is fully migrated.

I think it might require some custom agent to track expected memory 
usage on each node, accounting for pending migrations; the RA could 
query the agent and fail the action if not enough memory is available. 
It wouldn't really help the cluster decide where to put instances, but I 
guess it would tell it where it *can't* put instances.

Mike