Azure Monitor Metrics and Alerts

Azure Monitor Metrics is the time-series database at the heart of Azure monitoring. It collects numerical data from Azure resources at regular intervals, enabling real-time dashboards, trend analysis, and automated alerting. Understanding metrics and alerts is fundamental to operating any Azure workload effectively.

---

What Are Azure Monitor Metrics?

Metrics are lightweight, numerical values collected at regular intervals (typically every minute). They describe the state and performance of a resource at a point in time.

Metric Properties

Property	Description	Example
Timestamp	When the value was recorded	2025-03-15T10:30:00Z
Name	The metric identifier	Percentage CPU
Value	The numerical measurement	73.5
Dimensions	Key-value pairs for filtering	Instance=web-01, Region=uksouth
Aggregation	How values are summarised	Average, Sum, Count, Min, Max

Platform Metrics vs Custom Metrics

Type	Source	Cost
Platform metrics	Automatically collected from Azure resources	Free (standard retention)
Custom metrics	Sent from your application or agents	Charged per metric time-series ingested

Platform metrics are available immediately when you create a resource — no configuration required.

---

Metrics Explorer

Metrics Explorer is the built-in tool for visualising and analysing metrics in the Azure Portal:

Key Features

• Chart types — line, area, bar, and scatter charts
• Time ranges — from 30 minutes to 30 days
• Aggregations — average, sum, count, min, max, and percentiles
• Splitting — break down metrics by dimension (e.g., CPU per VM instance)
• Filtering — focus on specific dimension values
• Pinning — pin charts to Azure dashboards for quick access

Common Metric Queries

Resource	Metric	Use Case
Virtual Machine	Percentage CPU	Detect high CPU utilisation
Virtual Machine	Available Memory Bytes	Detect memory pressure
App Service	Http Server Errors	Monitor 5xx error rates
App Service	Response Time	Track application latency
SQL Database	DTU Percentage	Monitor database capacity
Storage Account	Transactions	Track request volume
Load Balancer	Health Probe Status	Monitor backend health

---

Alert Rules

Alert rules automatically evaluate conditions and trigger actions when thresholds are met or anomalies are detected.

Alert Rule Components

Component	Description
Scope	The target resource(s) being monitored
Condition	The signal and logic that triggers the alert (e.g., CPU > 80%)
Action Group	The notification and automation targets
Alert Rule Name	A descriptive name for identification
Severity	0 (Critical) to 4 (Verbose)

Alert Types

Type	Signal Source	Example
Metric alert	Azure Monitor Metrics	CPU > 80% for 5 minutes
Log alert	Log Analytics workspace	More than 10 errors in 15 minutes
Activity log alert	Azure Activity Log	Resource deleted or VM stopped
Service Health alert	Azure Service Health	Outage in UK South region

Static vs Dynamic Thresholds

Approach	Description	Best For
Static	Fixed threshold (e.g., CPU > 80%)	Well-understood workloads with predictable patterns
Dynamic	Machine learning-based, adapts to historical patterns	Workloads with variable baselines (e.g., weekday vs weekend)

Dynamic thresholds learn the normal pattern of a metric over time and alert only when behaviour deviates significantly — reducing false positives.

---

Action Groups

An action group defines what happens when an alert fires: