Load Test Infrastructure

Last updated on Mar 31, 2026

---

Load tests run on Linux Chaos Infrastructure with load testing enabled. This is the same Harness-managed agent used for chaos testing, installed on a Linux host, that also executes your Locust-based load tests and streams results back to the platform.

Overview

Each load test must be associated with an infrastructure. The infrastructure agent:

• Receives test execution commands from the Harness control plane
• Runs the Locust load generator process locally
• Streams real-time metrics and logs back during execution
• Supports multiple concurrent test runs depending on host capacity

Prerequisites

• A Linux host (bare metal, VM, or local VM via tools like Multipass or Lima) with network access to your target application
• The host must be able to reach Harness endpoints for agent registration
• sudo or root access to install the agent
• An environment configured in your Harness project (you will select it when creating the infrastructure)

Feature Flag

Load Testing requires the CHAOS_LOAD_TESTING_ENABLED feature flag. Contact your Harness sales representative to get it enabled for your account.

Set Up Load Test Infrastructure

1. Create an Environment

If you don't already have an environment for load testing:

1. During the infrastructure creation flow, click + New Environment
2. Enter a Name (e.g., load-test-env)
3. Select the Environment Type (Pre-Production or Production)
4. Choose Inline storage (stored in Harness) or Remote (stored in Git)
5. Click Save

2. Create a Linux Chaos Infrastructure

1. Go to Resilience Testing > Project Settings > Infrastructures
2. Select the Linux tab
3. Click + New Infrastructure

The setup wizard has three steps:

Step 1: Overview

• Enter a Name for the infrastructure
• Select the Environment you created
• Choose Root User for root configuration (required for load testing)

Step 2: Configure Options

• Log file max size (in MB): Maximum size per log file (default: 5)
• Log file max backups: Number of rotated log files to retain (default: 2)
• Experiment log file max age (in days): How long to keep experiment logs (default: 30)
• Infrastructure log file directory: Where logs are stored on the host (default: /var/log/linux-chaos-infrastructure)
• HTTP Proxy: (Optional) Proxy URL if the host requires one for outbound access
• HTTP Client Timeout: (Optional) Timeout in seconds for HTTP requests

Step 3: Deploy the Setup

Harness generates a one-time install command. Copy and run it on your Linux host:

curl https://app.harness.io/public/shared/tools/chaos/linux/<VERSION>/install.sh \
  | sudo bash /dev/stdin \
  --infra-id <YOUR_INFRA_ID> \
  --access-key <YOUR_ACCESS_KEY> \
  --account-id <YOUR_ACCOUNT_ID> \
  --server-url <YOUR_SERVER_URL> \
  --load

The --load flag enables load testing support and installs Locust alongside the chaos agent.

warning

The access key is one-time use. If the install fails, refresh the command in the Harness UI to generate a new access key before retrying.

3. Verify Connection

Once the script completes, the infrastructure status in Harness should change from Pending to Connected. If it remains in a pending state, check:

• Outbound connectivity from the host to Harness endpoints (the host must reach app.harness.io on port 443)
• Firewall or VPN rules blocking the agent's egress traffic
• That the install command ran without errors (check the script output)
• If a previous agent was installed on the same host, clean up leftover directories (/var/log/linux-chaos-infrastructure, /etc/linux-chaos-infrastructure, /opt/linux-chaos-infrastructure) before retrying

Infrastructure Sizing Guidelines

The right host size depends on your target load:

Virtual Users	Recommended Host	Notes
Up to 50	2 vCPU / 4 GB RAM	Light API testing
50–500	4 vCPU / 8 GB RAM	Typical API and web workloads
500–2000	8 vCPU / 16 GB RAM	High-throughput or complex scenarios
2000+	16+ vCPU / 32+ GB RAM	Consider distributed load generation

note

These are starting-point estimates. Actual resource usage depends on request complexity, response sizes, and scenario logic in your Locust script.

Managing Infrastructure

View Registered Infrastructure

Existing infrastructure is listed in the Infrastructure dropdown when creating or editing a load test. The dropdown shows the infrastructure name and current connection status. Only infrastructure installed with the --load flag appears in the load test infrastructure dropdown.

Multiple Infrastructure

You can register multiple Linux hosts in the same project. This lets you:

• Isolate staging and production load test environments
• Run tests from different network locations (e.g., on-premises vs. cloud)
• Dedicate infrastructure to specific teams or services

Remove Infrastructure

To deregister an infrastructure:

1. Navigate to Resilience Testing > Project Settings > Infrastructures > Linux tab
2. Select the infrastructure entry and delete it

To uninstall the agent from the host:

sudo systemctl stop linux-chaos-infrastructure
sudo systemctl disable linux-chaos-infrastructure
sudo rm -f /etc/systemd/system/linux-chaos-infrastructure.service
sudo systemctl daemon-reload
sudo rm -rf /etc/linux-chaos-infrastructure /opt/linux-chaos-infrastructure /var/log/linux-chaos-infrastructure
sudo rm -f /usr/local/bin/linux-chaos-infrastructure

Next Steps

• Get Started with Load Testing: Create and run your first load test
• Concepts: Understand virtual users, ramp-up, assertions, and load profiles