Course contentsShow
← Cloud and Infrastructure
Lesson 645 of 2,98912. Monitoring FoundationsFree lesson

Composite Alarms

Combine multiple alarms using AND/OR logic to reduce noise and create intelligent alerting rules.

Composite Alarms

What you'll learn: How to combine multiple alarms into a single intelligent alert using logical operators to reduce alert noise.

The Problem with Individual Alarms

Imagine you're monitoring a web application. You have separate alarms for high CPU usage, memory saturation, and elevated error rates. If all three fire simultaneously, you receive three different alerts — but they're all symptoms of the same underlying problem. This creates alert fatigue and confusion about which issue to tackle first.

What Are Composite Alarms?

Composite alarms let you combine multiple individual alarms using AND and OR logic to create smarter alerting rules. Instead of reacting to every metric individually, you define conditions that reflect real operational problems.

Think of it like a smoke detector and a heat sensor working together: you only want the fire alarm to sound when both detect danger, not just one or the other.

How They Work

A composite alarm evaluates the state of other alarms (OK, ALARM, or INSUFFICIENT_DATA) rather than raw metrics:

  • AND logic: All specified alarms must be in ALARM state
  • OR logic: Any one of the specified alarms triggers the composite
  • Complex expressions: Combine AND/OR to model sophisticated scenarios

Common Use Cases

Service Health Check: Trigger when HighCPU AND HighMemory AND HighLatency all fire — indicating a genuine service degradation, not just a temporary spike.

Multi-Region Redundancy: Alert only when Region-A-Down OR Region-B-Down (but silence if both regions fail, since that might need a different escalation path).

Cost Guard: Fire when BudgetThreshold-80% AND UnusualSpend both alarm, preventing false positives from expected seasonal costs.

Key Takeaway: Composite alarms reduce alert noise by combining multiple individual alarms with logical operators, helping you identify true operational problems rather than isolated metric spikes.