New Relic — Complete API Performance Monitoring Guide

In this tutorial, you will learn about New Relic. We cover key concepts, practical examples, and best practices to help you master this topic.

New Relic is a cloud-based observability platform that provides application performance monitoring, distributed tracing, and infrastructure monitoring. It offers a unified view of your entire API stack.

What You'll Learn

You'll learn how to instrument your API with New Relic, use distributed tracing, and analyze Transaction performance.

Why It Matters

New Relic pioneered APM and remains a leading platform. Its AI-powered error grouping, distributed tracing, and customizable dashboards make it powerful for API monitoring.

Real-World Use

A SaaS platform uses New Relic to monitor its Python API. When an error occurs, New Relic groups similar errors, shows the affected transaction trace, and correlates the error with the deployment that introduced it.

Implementation

import newrelic.agent

newrelic.agent.initialize("newrelic.ini")

app = Flask(__name__)

@app.route("/api/users")
@newrelic.agent.background_task()
def get_users():
    with newrelic.agent.FunctionTrace("database_query"):
        users = db.fetch_all("SELECT * FROM users")
    return jsonify(users)

@newrelic.agent.function_trace()
def process_payment(amount):
    if amount > 10000:
        newrelic.agent.record_custom_event("LargePayment", {"amount": amount})
    return payment_service.charge(amount)

; newrelic.ini
[newrelic]
license_key = your-license-key
app_name = My API
log_level = info
monitor_mode = true
transaction_events.enabled = true
distributed_tracing.enabled = true
cross_application_tracer.enabled = true
error_collector.enabled = true
browser_monitoring.enabled = false

Common Mistakes

| Mistake | Fix | |---------|-----| | Not setting app_name | Transactions grouped under "Python Application" | Set descriptive app name per service | | No custom attributes | Cannot search by important fields | Add user_id, order_id as custom attributes | | Not using error groups | Hundreds of individual errors to triage | Let AI group similar errors | | No deployment markers | Cannot correlate performance with deploys | Use REST API to mark deployments | | Excessive transaction volume | Cost overruns | Set sampling rules for high-volume endpoints |

What's Next

Learn about alerting with PagerDuty and Slack.