Detecting Slow Queries in Production

In this lesson, you will learn how to set up automated alerts that notify your team the moment a database query exceeds your performance threshold.

The Concept of Latency Tracking

In production, your database performance can degrade silently due to growing datasets or suboptimal indexing. Instead of manually checking logs, we implement "Performance Hooks" in our Node.js persistence layer. By wrapping our database calls with a simple timer, we can detect when an operation takes longer than an acceptable threshold (e.g., 200ms) and trigger an automated alert.

Implementation Pattern

To keep this manageable, we create a wrapper function for our persistence calls. This acts as a middleware between your application logic and your database driver.

const SLOW_QUERY_THRESHOLD_MS = 200;

async function executeWithMonitoring(queryFn, queryName) {
  const start = Date.now();
  
  try {
    return await queryFn();
  } finally {
    const duration = Date.now() - start;
    
    if (duration > SLOW_QUERY_THRESHOLD_MS) {
      // Send alert to monitoring system (e.g., PagerDuty, Slack, or CloudWatch)
      console.warn(`[ALERT] Slow query detected: ${queryName} took ${duration}ms`);
      reportToMonitoring({ queryName, duration });
    }
  }
}

// Usage
await executeWithMonitoring(() => db.users.find({ status: 'active' }), 'findActiveUsers');

Why Automated Alerts?

The core benefit here is proactive identification. By logging the queryName and the duration, you gain a paper trail to investigate exactly which part of your application schema is struggling. If you notice a specific operation is consistently triggering alerts, you know it is time to revisit your indexes or perhaps implement a caching strategy for that specific read path.

This pattern is far more effective than general health monitoring because it highlights the specific culprit behind a bottleneck rather than just showing a generic spike in CPU or memory usage.

Key Takeaway: By wrapping database calls with execution timers and triggering alerts on latency thresholds, you turn silent performance regressions into actionable diagnostic events.