Krane

Convert Google Compute Engine (GCE) autoscaling instance groups to Kubernetes autoscaling deployments.

Status

Proof of concept.

Install

Download the latest release from the releases page.

Usage

Create a Kubernetes cluster

gcloud container clusters create cluster0

Create an Autoscaling GCE Instance Group

Instances will leverage the Google Container VM and the Kubernetes Kubelet to start applications from Pod manifests. See the cloud-config.yaml for more details.

gcloud compute addresses create nginx

gcloud compute http-health-checks create nginx

gcloud compute target-pools create nginx --health-check nginx

gcloud compute instance-templates create nginx \
  --image-family gci-stable \
  --image-project google-containers \
  --machine-type n1-standard-1 \
  --metadata-from-file user-data=cloud-config.yaml

gcloud compute instance-groups managed create nginx \
  --base-instance-name nginx \
  --target-pool nginx \
  --size 2 \
  --template nginx

gcloud compute instance-groups managed set-autoscaling nginx \
  --max-num-replicas 10 \
  --target-cpu-utilization 0.25 \
  --cool-down-period 90

gcloud compute forwarding-rules create nginx \
  --ports 80 \
  --address $(gcloud compute addresses describe nginx --format 'value(address)') \
  --target-pool nginx

gcloud compute firewall-rules create nginx --allow tcp:80

Convert a GCE Instance Group to Kubernetes

krane -instance-group nginx | kubectl create -f -

deployment "nginx" created
horizontalpodautoscaler "nginx" created
service "nginx" created

To view the Kubernetes configs omit the pipe to kubectl: krane -instance-group nginx

Results

The Pod manifest from the instance group cloud-init metadata is converted to a Kubernetes deployment.

kubectl get deployments

NAME      DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
nginx     2         2         2            2           3m

kubectl get pods

NAME                     READY     STATUS    RESTARTS   AGE
nginx-2115999728-c5s11   1/1       Running   0          3m
nginx-2115999728-kbfj0   1/1       Running   0          3m

The autoscaling policy from the instance group is converted to a Kubernetes horizontal pod autoscaler.

kubectl get hpa

NAME      REFERENCE          TARGET    CURRENT   MINPODS   MAXPODS   AGE
nginx     Deployment/nginx   25%       0%        2         10        3m

If the instance group is part of a target pool which is referenced by a network load balancer as Kubernetes service will be created.

kubectl get svc

NAME         CLUSTER-IP     EXTERNAL-IP    PORT(S)   AGE
nginx        XX.XX.XXX.XX   XXX.XXX.XX.X   80/TCP    3m

kelseyhightower/krane