autoscale.md

Run Autoscaling job on your local machine

This documentation shows an example to run two jobs on a local kubernetes cluster and see the job scaling status.

Prerequisites

• install minikube
• install kubectl

Run local Autoscaling job

1. Start a local minikube cluster.

   minikube start --kubernetes-version v1.6.4

2. Run the following commands to create sample training workspace and data.

   mkdir /path/to/workspace
   cp $REPO_PATH/doc/autoscale_example/*.py /path/to/workspace
   mkdir -p /path/to/workspace/data/
   cp -r $REPO_PATH/doc/autoscale_example/uci_housing/ /path/to/workspace/data/

3. Mount the workspace folder into Minikube:

   minikube mount /path/to/workspace:/workspace

   The minikube mount command will block, so start a new terminal to continue the tutorial.

4. Start controller and a example job:

   cd $REPO_PATH/k8s/controller
   kubectl create -f controller.yaml
   kubectl create -f trainingjob_resource.yaml
   kubectl create -f autoscale_job/
   kubectl get pods

5. Start another job simulating cluster load, then you can observe the scale process using kubectl get pods:

   kubectl create -f autoscale_load/
   kubectl get pods