Chaos Operator

K8s operator for supervised chaos engineering.

Supervised Chaos Engineering

By that term we understand chaos engineering without "random" part. Sometimes it's usable to be able to apply different chaos actions in controlled way.

Installation

Either by applying single manifests from crds directory or by helm (easier). To install by helm, add repository:

helm repo add <repository name> https://davenury.github.io/Chaos-Operator

and then install release:

helm install <release name> <repository name>/chaos-operator

You can also customize your deployment, by providing file with values. Given a file named values.yaml in your current directory:

namespace:
  name: chaos-operator    # namespace to deploy operator and samples, if applied
  enabled: true            # whether to create namespace

operator:
  deployment_name: chaos-operator         # name of deployment 
  image: ghcr.io/davenury/chaos-operator:latest   # image of chaos operator, for specific versions, head to: https://github.com/users/Davenury/packages/container/package/chaos-operator

serviceAccount:
  name: chaos-operator      # name of the service account
  enabled: true              # whether to create the service account
  clusterRole:
    name: chaos-operator    # name of the cluster role

samples:                    # whether to create a sample applications to demonstrate the capabilities of the operator
  deployment: true          # creates sample-deployment, used by default scenario in the first phase
  single_pod: true          # creates nginx pod, used by default scenario in the second phase
  service: true             # creates sample-service, used by default scenario in the third phase

you can pass new values while installing the release:

helm install <release name> <repository name>/chaos-operator -f values.yaml

Helm chart will create chaos-operator deployment, as well as needed resources (those can be disabled by modifying values.yaml file). To execute sample scenario, enter the crds directory and:

kubectl apply -f sample.yaml

Scenario definition

Sample definition is like this:

apiVersion: lsc.davenury.github.com/v1
kind: Scenario
metadata:
  name: sample
  namespace: chaos-operator
spec:
  phases:
    - duration: PT60S
      actions:
      - namespace: chaos-operator
        resourceType: deployment
        resourceName: sample-deployment
        action: scale
        value: 2
    - duration: PT30S
      actions:
        - namespace: chaos-operator
          resourceType: pod
          resourceName: nginx
          action: delete
    - duration: PT20S
      actions:
        - namespace: chaos-operator
          resourceType: service
          resourceName: my-service
          action: delete

Spec is composed of phases. A phase is a set of actions applied at the same time throughout the same span of time. A phase is composed of duration (ISO-8601 compatible) and actions - a list of action. action refers to single action performed on cluster, e.g. deletion of a pod or scaling of a deployment. Each action requires:

• namespace - the namespace of the resource
• resourceType - k8s resource type (e.g. deployment, pod, service)
• resourceName - the name of the resource
• action - a verb that describes what should happen (e.g. scale, delete) Additional param for scale action: value - describes the desired value of the scale action.

Available actions

• delete pod
• delete service
• scale deployment