Construct chaos experiment YAML without ChaosCenter
Chaos Experiment is a set of different operations coupled together to achieve desired chaos impact on a Kubernetes Cluster.
A basic chaos experiment consists of these steps:
- Install ChaosExperiment CR
- Install ChaosEngine CR
- Cleanup Chaos resources
Before we begin​
To construct a Chaos Experiment without ChaosCenter, make sure you are aware of Chaos Experiment, ChaosEngine CR and the different steps present in it.
Steps to construct a chaos experiment​
LitmusChaos leverages the popular GitOps tool Argo to achieve this goal. Argo enables the orchestration of different chaos faults together in the form of a single chaos experiment which is extremely simple and efficient to setup and use.
The structure of a chaos experiment is similar to that of a Kubernetes Object. It consists of the mandatory fields like apiVersion, kind, metadata, spec.
Few additional terms in an Argo chaos experiments are:
- Template : It consists of different steps with their specific operations.
templates:
- name: custom-chaos
steps:
- - name: install-chaos-experiments
template: install-chaos-experiments
- - name: pod-delete
template: pod-delete
- - name: revert-chaos
template: revert-chaos
- Steps : It is a single step inside a chaos experiment which runs a container based on the input parameters. These can also be sequenced parallely.
steps:
- - name: install-chaos-experiments
template: install-chaos-experiments
- - name: pod-delete
template: pod-delete
- name: pod-cpu-hog
template: pod-cpu-hog
- - name: revert-chaos
template: revert-chaos
- Entrypoint : The first step that executes in a chaos experiment is called its entrypoint.
entrypoint: custom-chaos
Here, the template with the name custom-chaos will be executed first.
- Artifacts : Artifacts are defined as the files saved by the containers in each step.
- name: install-chaos-experiments
inputs:
artifacts:
- name: pod-delete
path: /tmp/pod-delete.yaml
raw:
data: >
apiVersion: litmuschaos.io/v1alpha1
description:
message: |...
Once the chaos experiment is constructed, it should look like this:
apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
name: pod-delete-experiment
namespace: litmus
spec:
arguments:
parameters:
- name: adminModeNamespace
value: litmus
entrypoint: custom-chaos
securityContext:
runAsNonRoot: true
runAsUser: 1000
serviceAccountName: argo-chaos
templates:
- name: custom-chaos
steps:
- - name: install-chaos-experiments
template: install-chaos-experiments
- - name: pod-delete
template: pod-delete
- - name: revert-chaos
template: revert-chaos
- name: install-chaos-experiments
inputs:
artifacts:
- name: pod-delete
path: /tmp/pod-delete.yaml
raw:
data: >
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: 3.0.0
spec:
definition:
scope: Namespaced
permissions:
- apiGroups:
- ""
- apps
- apps.openshift.io
- argoproj.io
- batch
- litmuschaos.io
resources:
- deployments
- jobs
- pods
- pods/log
- replicationcontrollers
- deployments
- statefulsets
- daemonsets
- replicasets
- deploymentconfigs
- rollouts
- pods/exec
- events
- chaosengines
- chaosexperiments
- chaosresults
verbs:
- create
- list
- get
- patch
- update
- delete
- deletecollection
image: litmuschaos/go-runner:3.0.0
imagePullPolicy: Always
args:
- -c
- ./experiments -name pod-delete
command:
- /bin/bash
env:
- name: TOTAL_CHAOS_DURATION
value: "15"
- name: RAMP_TIME
value: ""
- name: FORCE
value: "true"
- name: CHAOS_INTERVAL
value: "5"
- name: PODS_AFFECTED_PERC
value: ""
- name: LIB
value: litmus
- name: TARGET_PODS
value: ""
- name: SEQUENCE
value: parallel
labels:
name: pod-delete
app.kubernetes.io/part-of: litmus
app.kubernetes.io/component: experiment-job
app.kubernetes.io/version: 3.0.0
container:
args:
- kubectl apply -f /tmp/pod-delete.yaml -n
{{workflow.parameters.adminModeNamespace}} | sleep 30
command:
- sh
- -c
image: litmuschaos/k8s:latest
- name: pod-delete
inputs:
artifacts:
- name: pod-delete
path: /tmp/chaosengine-pod-delete.yaml
raw:
data: |
apiVersion: litmuschaos.io/v1alpha1
kind: ChaosEngine
metadata:
namespace: "{{workflow.parameters.adminModeNamespace}}"
generateName: pod-delete
labels:
instance_id: 86a4f130-d99b-4e91-b34b-8f9eee22cb63
spec:
appinfo:
appns: default
applabel: app=nginx
appkind: deployment
jobCleanUpPolicy: retain
engineState: active
chaosServiceAccount: litmus-admin
experiments:
- name: pod-delete
spec:
components:
env:
- name: TOTAL_CHAOS_DURATION
value: "30"
- name: CHAOS_INTERVAL
value: "10"
- name: FORCE
value: "false"
- name: PODS_AFFECTED_PERC
value: ""
container:
args:
- -file=/tmp/chaosengine-pod-delete.yaml
- -saveName=/tmp/engine-name
image: litmuschaos/litmus-checker:latest
- name: revert-chaos
container:
image: litmuschaos/k8s:latest
command:
- sh
- -c
args:
- "kubectl delete chaosengine -l 'instance_id in
(86a4f130-d99b-4e91-b34b-8f9eee22cb63, )' -n
{{workflow.parameters.adminModeNamespace}} "
Install Experiment​
ChaosExperiment CR:​
The
install-experimentstep consists of ChaosExperiment CR in its artifact. ChaosExperiment CR is the heart of LitmusChaos and contains the low-level execution information. They serve as off-the-shelf templates that one needs to "pull" (install on the cluster) before including them as part of chaos run against any target applications (the binding being defined in the ChaosEngine). The experiments are installed on the cluster as Kubernetes custom resources and are designed to hold granular details of the experiment such as image, library, necessary permissions, chaos parameters (set to their default values). Most of the ChaosExperiment parameters are essentially tunables that can be overridden from the ChaosEngine resource.ChaosEngine CR:​
The ChaosEngine is the main user-facing chaos custom resource with a namespace scope and is designed to hold information around how the chaos experiments are executed. It connects an application instance with one or more chaos experiments while allowing the users to specify run level details (override experiment defaults, provide new environment variables and volumes, options to delete or retain experiment pods, etc.,). This CR is also updated/patched with the status of the chaos experiments, making it the single source of truth with respect to the chaos.
Resources​
- The ChaosExperiment CR and ChaosEngine CR of different experiments are available at ChaosHub.