Quickstart

Run your first chaos experiment in minutes with Harness Chaos Engineering

Use this guide to set up and run your first chaos experiment on target infrastructure.

Before you begin

Review these resources and requirements:

• Key concepts: Understand chaos engineering concepts.

• What's supported: Confirm supported platforms and capabilities.

• Target infrastructure: Access a Kubernetes cluster, Linux machine, or cloud environment.

You need the following:

• Harness account: Sign up if you do not have one.

• Target infrastructure: Use a Kubernetes cluster with kubectl access or a Linux machine with administrator privileges.

• Permissions: Get administrator access to install chaos agents.

Access Harness Chaos Engineering

To access Harness Chaos Engineering:

1. Sign up or log in to your Harness account.

2. In the left navigation, navigate to the Chaos Engineering module.

3. Create a project or ask your administrator to add you to an existing project.

Create an environment

A chaos experiment is executed in an infrastructure that is associated with an environment.

To create an environment:

1. Navigate to Environments, then click New Environment.

2. Enter the environment name, description, and tags.

3. Select Production or Non-Production.

4. Click Create.

Create New Environment

Select an existing environment from the list when one is available.

Set up chaos infrastructure

After you create an environment, add infrastructure to it.

Kubernetes

To add Kubernetes infrastructure:

1. In your environment, click New Infrastructure.

2. Select Kubernetes.

3. Select an installation mode:

   • Cluster-wide access: Target resources across all namespaces.

   • Specific namespace access: Restrict chaos injection to a namespace.

4. Copy and run the installation command in your cluster:

# Example installation command (use the one provided in UI) <a href="#example-installation-command-use-the-one-provided-in-ui" id="example-installation-command-use-the-one-provided-in-ui"></a>
kubectl apply -f https://app.harness.io/chaos/delegate/manifest/...

1. Wait until the infrastructure shows the CONNECTED status.

Linux

To add Linux infrastructure:

1. Click New Infrastructure, then select Linux.

2. Download and install the chaos agent:

# Download the agent <a href="#download-the-agent" id="download-the-agent"></a>
curl -O https://app.harness.io/chaos/linux-agent
chmod +x linux-agent

# Install with your infrastructure ID and access key <a href="#install-with-your-infrastructure-id-and-access-key" id="install-with-your-infrastructure-id-and-access-key"></a>
sudo ./linux-agent --install --infra-id=<YOUR_INFRA_ID> --access-key=<YOUR_ACCESS_KEY>

Create your first chaos experiment

Create and run your first chaos experiment. Start with Pod Delete because it has a small blast radius.

Identify your target

Identify the target for your experiment:

1. Identify the microservice that you will target.

2. For Kubernetes, select a pod from your application.

3. Pod delete: Start with this experiment.

Chaos Experiment Steps

Create the experiment

Create the experiment:

1. Navigate to Chaos Experiments, then click New Experiment.

2. Select Blank Canvas or a Template.

3. Configure the following fields:

   • Name: My First Pod Delete Experiment

   • Description: Testing pod resilience

   • Tags: Enter tags that organize the experiment.

Add a chaos fault

Add and configure the chaos fault:

1. In the experiment builder, click Add Fault.

2. Select Kubernetes → Pod → Pod Delete.

3. Configure the following fields:

   • Target Pods: Select pods or use label selectors.

   • Chaos Duration: Enter 30 seconds.

   • Force: Set this to false for graceful deletion.

Add resilience probes

Probes validate your hypothesis during the experiment. Add an HTTP probe:

1. Click Add Probe.

2. Select HTTP Probe.

3. Configure the following fields:

   • URL: Enter your application endpoint.

   • Method: Select GET.

   • Success Criteria: Enter response code 200.

   • Run Properties: Select execution during chaos.

Run your first experiment

Run and monitor the experiment:

1. Review the experiment configuration.

2. Save the experiment.

3. Click Run.

4. Monitor the following:

   • The experiment timeline.

   • Probe results.

   • System metrics and logs.

Experiment Execution

Analyze results

After the experiment completes, review the following:

1. Resilience Score: Review the overall system resilience rating.

2. Probe Results: Check health-check results during chaos.

3. Timeline: Review experiment execution phases.

4. Logs: Review execution logs for troubleshooting.

Experiment Results

Understand results

Use these results to assess system resilience:

• Passed probes: Your application handled the chaos.

• Failed probes: Identify areas that need improvement.

• Resilience score: Higher scores indicate greater resilience.

Quick onboarding options

Harness Chaos Engineering offers two onboarding methods:

Automated onboarding

Automated onboarding provides the following:

• One-click setup: Creates an environment, discovers services, and runs experiments.

• Minimal decisions: Harness Chaos Engineering handles most configuration.

• Resilience scores: View resilience scores after experiments run.

Guided onboarding

Guided onboarding provides the following:

• Step-by-step guidance: Walk through each step with options.

• More control: Customize each step while receiving guidance.

• Learning focus: Understand the process while you configure it.

To access onboarding, navigate to Chaos Engineering → Overview, then select Select a cluster.

Common first experiments

Use these experiments to begin resilience testing:

Pod Delete for Kubernetes

Purpose: Test application resilience to pod failures.

• Fault: Pod Delete.

• Duration: 30 seconds.

• Success Criteria: The service remains available, and new pods start quickly.

CPU Stress for Linux and Kubernetes

Purpose: Test application behavior under CPU pressure.

• Fault: CPU Stress.

• CPU Load: 80%.

• Duration: two minutes.

• Success Criteria: Application performance degrades gracefully.

Network latency

Purpose: Validate application behavior under network delays.

• Fault: Network Latency.

• Latency: 100ms.

• Duration: one minute.

• Success Criteria: Timeouts are handled gracefully.

Next steps

You have run your first chaos experiment. Use these resources next:

• Chaos faults: Try different types of chaos faults.

• Probes: Configure monitoring.

• GameDays: Organize chaos engineering events.

• CI/CD integration: Automate chaos testing in pipelines.

Next step

Build more resilient systems through systematic chaos engineering practices.