Litmus chaos-operator for injecting chaos experiments on Kubernetes

Litmus chaos operator is used by Kubernetes application developers and SREs to inject chaos into the applications and Kubernetes infrastructure in a managed fashion. Its objective is to make the process of validation and hardening of application workloads on Kubernetes easy by automating the execution of chaos experiments. A sample chaos injection workflow could be as simple as:

Install the Litmus infrastructure components (RBAC, CRDs), the Operator & Experiment custom resource bundles via the operator manifest
Annotate the application under test (AUT), enabling it for chaos
Create a ChaosEngine custom resource tied to the AUT, which describes the experiment to be executed

Benefits provided by the Chaos Operator include:

Standardised chaos experiment spec
Categorized chaos bundles for stateless/stateful/vendor-specific
Test-Run resiliency
Ability to chaos run as a background service based on annotations

What is a chaos operator and how is it built?

The Chaos Operator is a Kubernetes Operator, which are nothing but custom-controllers with direct access to Kubernetes API that can manage the lifecycle of certain resources or applications, while always trying to ensure the resource is in the "desired state". The logic that ensures this is commonly called "reconcile" function.

The Chaos Operator is built using the popular Operator-SDK framework, which provides bootstrap support for new operator projects, allowing teams to focus on business/operational logic.

The Litmus Chaos Operator helps reconcile the state of the ChaosEngine, a custom resource that holds the chaos intent specified by a developer/devops engineer against a particular stateless/stateful Kubernetes deployment. The operator performs specific actions upon CRUD of the ChaosEngine, its primary resource. The operator also defines a secondary resource (the engine runner pod), which is created & managed by it in order to implement the reconcile functions.

What is a chaos engine?

The ChaosEngine is the core schema that defines the chaos workflow for a given application. Currently, it defines the following:

Application info (namespace, labels, kind) of primary (AUT) and auxiliary (dependent) applications
ServiceAccount used for execution of the experiment
Flag to turn on/off chaos annotation checks on applications
Chaos Experiment to be executed on the application
Attributes of the experiments (overrides defaults specified in the experiment CRs)
Flag to retain/cleanup chaos resources after experiment execution

The ChaosEngine is the referenced as the owner of the secondary (reconcile) resource with Kubernetes deletePropagation ensuring these also are removed upon deletion of the ChaosEngine CR.

Here is a sample ChaosEngineSpec for reference: https://v1-docs.litmuschaos.io/docs/getstarted/#prepare-chaosengine

What is a litmus chaos chart and how can I use it?

Litmus Chaos Charts are used to install "Chaos Experiment Bundles" & are categorized based on the nature of the experiments (general Kubernetes chaos, vendor/provider specific chaos - such as, OpenEBS or application-specific chaos, say NuoDB). They consist of custom resources that hold low-level chaos(test) parameters which are queried by the operator in order to execute the experiments. The spec.definition.fields and their corresponding values are used to construct the eventual execution artifact that runs the chaos experiment (typically, the litmusbook, which is a K8s job resource). It also defines the permissions necessary to execute the experiment.

Here is a sample ChaosEngineSpec for reference:

apiVersion: litmuschaos.io/v1alpha1
description:
  message: |
    Deletes a pod belonging to a deployment/statefulset/daemonset
kind: ChaosExperiment
metadata:
  name: pod-delete
  labels:
    name: pod-delete
    app.kubernetes.io/part-of: litmus
    app.kubernetes.io/component: chaosexperiment
    app.kubernetes.io/version: latest
spec:
  definition:
    scope: Namespaced
    permissions:
      - apiGroups:
          - ""
          - "apps"
          - "batch"
          - "litmuschaos.io"
        resources:
          - "deployments"
          - "jobs"
          - "pods"
          - "configmaps"
          - "chaosengines"
          - "chaosexperiments"
          - "chaosresults"
        verbs:
          - "create"
          - "list"
          - "get"
          - "patch"
          - "update"
          - "delete"
    image: "litmuschaos/go-runner:latest"
    imagePullPolicy: Always
    args:
    - -c
    - ./experiments -name pod-delete
    command:
    - /bin/bash
    env:

    - name: TOTAL_CHAOS_DURATION
      value: '15'

    # Period to wait before/after injection of chaos in sec
    - name: RAMP_TIME
      value: ''

    - name: FORCE
      value: 'true'

    - name: CHAOS_INTERVAL
      value: '5'

    ## percentage of total pods to target
    - name: PODS_AFFECTED_PERC
      value: ''

    - name: LIB
      value: 'litmus'    

    - name: TARGET_PODS
      value: ''

    ## it defines the sequence of chaos execution for multiple target pods
    ## supported values: serial, parallel
    - name: SEQUENCE
      value: 'parallel'
    labels:
      name: pod-delete
      app.kubernetes.io/part-of: litmus
      app.kubernetes.io/component: experiment-job
      app.kubernetes.io/version: latest

How to get started?

Refer the LitmusChaos documentation litmus docs

How do I contribute?

You can contribute by raising issues, improving the documentation, contributing to the core framework and tooling, etc.

Head over to the Contribution guide