kops-autoscaler-openstack

The purpose of this application is to provide capability to scale cluster up/down in case of need. This application supports currently two use-cases:

if kops instancegroup minsize is larger than current instances in openstack -> scale up
if kops instancegroup maxsize is smaller than current instances in openstack -> scale down

This application will detect the need of change by running kops update cluster <cluster>. Scaling means that this application will execute kops update cluster <cluster> --yes under the hood.

This application makes it possible to use kops rolling-update <cluster> command in openstack kops.

Provide autoscaling capability to kops openstack
Usage:
  kops-autoscaling-openstack [flags]
Flags:
      --access-key string        S3 access key (default "test")
      --custom-endpoint string   S3 custom endpoint
  -h, --help                     help for kops-autoscaling-openstack
      --load-balancer-metrics    collect load balancer metrics (default false)
      --loglevel                 log level (default 2)
      --name string              Name of the kubernetes kops cluster (default "")
      --secret-key string        S3 secret key (default "")
      --sleep int                Sleep between executions (default 45)
      --state-store string       KOPS State store (default "")

Prometheus metrics are automatically exposed in port 2112.
These metrics contains information about current Kops cluster OpenStack instances.

Example:

# HELP openstack_instance Openstack instance
# TYPE openstack_instance gauge
openstack_instance{id="18b8451c-a7f0-4d26-b251-79080ff2518c",name="nodes-1-test-k8s-local",status="ACTIVE"} 1
openstack_instance{id="1f8f641f-2f95-4f42-bc69-70fbb9a7eed0",name="nodes-2-test-k8s-local",status="ACTIVE"} 1
...

How to install

See Examples

Create secrets, set env variables that you usually use when doing things with kops.

How to contribute

Make issues/PRs

Debug in Production

This binary do have pprof debugging options available in localhost:6060/debug/pprof/ by default.
When this binary is executed inside kubernetes cluster, port-forward that port to your localhost and start debugging.
Command for that is kubectl port-forward <pod> 6060:6060