Use OpenStack Senlin, Heat templates to enable autoscaling
As it is with other cloud platforms, autoscaling in OpenStack is important to meet changing workload demands. Here's how to enable that process with Senlin and Heat templates.
The OpenStack platform constantly evolves, and as a result, some processes that worked in the past are no longer valid -- which is now the case for autoscaling. Before, users set up OpenStack autoscaling with the Ceilometer service and Heat templates. Now, users need a new tool.
OpenStack Senlin is a service that lets admins create and manage clusters of related cloud resources to simplify orchestration. It integrates with Heat and Ceilometer to perform autoscaling tasks, and these tools work together to ensure that a cluster scales in and out based on workload demands.
When you use OpenStack Senlin with Heat, you use specific Heat cluster resources. Use YAML code to define the cluster, as well your scale-in and scale-out policies. The core code of such a configuration might look as follows, where Listing 1 defines the cluster and Listing 2 defines the scale-in and scale-out policies. Both sample configurations are from OpenStack documentation.
Listing 1: Define the Senlin cluster in Heat
cluster:
type: OS::Senlin::Cluster
properties:
desired_capacity: 2
min_size: 2
profile: {get_resource: profile}
Listing 2: Define the scale-in and scale-out policies
scale_in_policy:
type: OS::Senlin::Policy
properties:
type: senlin.policy.scaling-1.0
bindings:
- cluster: {get_resource: cluster}
properties:
event: CLUSTER_SCALE_IN
adjustment:
type: CHANGE_IN_CAPACITY
number: 1
scale_out_policy:
type: OS::Senlin::Policy
properties:
type: senlin.policy.scaling-1.0
bindings:
- cluster: {get_resource: cluster}
properties:
event: CLUSTER_SCALE_OUT
adjustment:
type: CHANGE_IN_CAPACITY
number: 1
While the resources in Listing 1 and 2 define the basis of the OpenStack Senlin cluster, you need to do more to make it fully operational. A load-balancer policy is a mandatory part of the configuration, and the OpenStack Neutron service will create one automatically. Based on the load-balancer policy, you need to define resources to manage the scale-in and scale-out process, which completes the basic framework you define in Heat.
The last required element is the alarm trigger. You must include Ceilometer alarms in the Heat template so that the alarm action defines which Heat trigger to pull -- either the receiver_scale_in or the receiver_scale_out, based on what needs to be done.
