LitmusChaos Setup

Installing LitmusChaos and Running Chaos Monkey Experiments¶

LitmusChaos requires dedicated namespaces for running experiments and its own components. Follow these steps to set up the environment:

Create a namespace for chaos experiments:
```
kubectl create namespace chaos-testing
```

Set namespace context for subsequent commands:

kubectl config set-context --current --namespace=chaos-testing

Create the 'litmuschaos' service account (prerequisite for RB, install LitmusChaos components):
```
kubectl create serviceaccount litmuschaos -n litmus
```
Note: The litmus namespace is where LitmusChaos controller and related components are installed, while chaos-testing is used for experiment workloads.
Grant permissions (RBAC):
Ensure the litmuschaos service account has access to the chaos-testing namespace. Apply a role binding with restricted permissions:
```
kubectl create rolebinding chaos-runner-binding \
  --namespace=chaos-testing \
  --role=chaos-runner \
  --serviceaccount=litmus:litmuschaos
```
(Adjust the role and service account based on your cluster configuration. For production, use a more restrictive role than cluster-admin.)
Install LitmusChaos via Helm:
Add the LitmusChaos Helm repository and deploy:
```
helm repo add litmuschaos https://charts.litmuschaos.io
helm install litmuschaos litmuschaos/litmuschaos --namespace litmus
```
Alternatively, use Kubernetes manifests from the LitmusChaos GitHub repo for installation.

Run a Chaos Monkey experiment:
Create a Chaos Monkey experiment manifest (e.g., pod-delete.yaml):

apiVersion: litmuschaos.io/v1alpha1
kind: ChaosEngine
metadata:
  name: pod-delete
  namespace: chaos-testing
spec:
  engineName: chaos-mono
  appInfo:
    appNS: chaos-testing
    appLabels: {}
    appPods: []
  chaosServiceAccount: litmuschaos
  experiments:
    - name: pod-delete
      spec:
        engineName: chaos-mono
        workloadKind: pod
        workloadSelector:
          matchLabels:
            app: nginx
        chaosNamespace: chaos-testing
        metadata:
          annotations:
            "chaos.io/timeout": "30s"
        spec:
          signal: true
          signalType: pod
          podDelete:
            gracePeriod: 5
            preKill: true
            postKill: true

This example deletes a pod with label app: nginx in the chaos-testing namespace. Adjust labels and parameters based on your workload.

Execute and verify the experiment:
Apply the manifest and monitor the results:
```
kubectl apply -f pod-delete.yaml
kubectl get pods -n chaos-testing
```
Check logs and metrics to confirm the chaos effect and recovery.