openstack team mailing list archive

[Steven Hardy <shardy@xxxxxxxxxx>]

On Mon, Jul 01, 2013 at 08:48:24AM +0000, John Ashford wrote:
> I am a newbie and trying to understand how to scale. If i have a web server on a VM within Openstack is it done by simply assigning more cores to the web server VM? And is there a way to do so automatically within Openstack? And when you reach the maximum number of cores on one physical machine, can you spread load across multiple physical machines without a load balancer in front? 

There are several ways you could approach this problem, but I'll mention
the AutoScaling capability provided by heat, as it does exactly what you
want AFAICT.

We provide orchestration capabilities which allow you to spin up groups of
instances, e.g running web server applications, and balance requests via
another instance running a simple haproxy based loadbalancer (so you don't
need an actual physical loadbalancer appliance).

You can use the AutoScaling functionality without any loadbalancer resource
if you wish, such as in the OpenShift demo linked below (where the broker
node is doing the load balancing)

The trigger for scaling out can either be manual (using the InstanceGroup
resource combined with stack updates), or automatic (using the
AutoScalingGroup resource combined with alarms.

There are example templates in the heat-templates repository:

https://github.com/openstack/heat-templates/blob/master/cfn/AutoScalingMultiAZSample.yaml

Here are some demo videos which describe/demonstrate this feature in more detail:

http://www.youtube.com/watch?v=9lLSZi6njq4

http://www.youtube.com/watch?v=nEzK-2lsEVg&feature=youtu.be&t=34m35s

HTH,

Steve