Cerebral is able to autoscale AWS Auto Scaling Groups (ASGs) by setting the desired
node count.
It does not modify the min/max bounds on an ASG as set in AWS (which may have been performed through e.g. kops
).
This means that a user should set the min
and max
fields on the Cerebral AutoscalingGroup CR to match the AWS ASG bounds in order to have the expected behavior.
It is expected that when the AWS ASG scales up, the label(s) associated with the nodeSelector
field in the AutoscalingGroup CR will be added to the new instance.
For example, the way to easily achieve this using kops
is to define nodeLabels
on the instance group that match the nodeSelector
.
The AWS engine does not require any configuration in the AutoscalingEngine CR itself.
It is expected that the well-known AWS environment variables (AWS_ACCESS_KEY_ID
, AWS_SECRET_ACCESS_KEY
, and optionally AWS_REGION
) are pulled in through the main Cerebral Deployment.
The CR is still required in order to inform Cerebral about the existence of the engine, as shown below.
apiVersion: cerebral.containership.io/v1alpha1
kind: AutoscalingEngine
metadata:
name: aws
spec:
type: aws