SpotWorkload

README

SpotWorkload

Spot-Workload Representation and Management Component of the Technion's SpotOS Project

This repository contains the API for SpotWorkload CRD and the set of its managing controllers.

Getting Started

SpotWorkload CR

A SpotWorkload CR defines a workload configuration as expected to input in CloudCostOptimizer, and reflects the workload's scheduling decisions and status as it progresses.

SpotWorkload Controllers

The controllers are responsible for:

• Translating the workload's scheduling decisions into tolerations and node affinity rules

TODO: enhance docs, add examples

Build and push the image to docker registry

1. Set the REGISTRY environment variable to hold the name of your docker registry:

   $ export REGISTRY=...
   

2. Set the IMAGE_TAG environment variable to hold the required version of the image.
   default value is latest, so in that case no need to specify this variable:

   $ export IMAGE_TAG=latest
   

3. Run make to build and push the image:

   $ make push-images
   

Deploy on a cluster

1. Set the REGISTRY environment variable to hold the name of your docker registry:

   $ export REGISTRY=...
   

2. Set the IMAGE environment variable to hold the name of the image.

   $ export IMAGE=$REGISTRY/$(basename $(pwd)):latest
   

3. Run the following command to deploy the spot-workload-controller controller to your cluster:

   envsubst < deploy/spot-workload-controller.yaml.template | kubectl apply -f -
   

Cleanup from a cluster

1. Run the following command to clean spot-workload-controller from your cluster:

   envsubst < deploy/spot-workload-controller.yaml.template | kubectl delete -f -