Monitor resources in Azure
📘Microsoft Certified: Azure Administrator Associate (AZ-104)
1. What Are Metrics in Azure Monitor?
Metrics are numerical measurements collected at regular intervals from Azure resources.
They tell you what is happening right now or what happened recently.
Examples of metrics:
- CPU usage of a virtual machine
- Memory usage for an App Service
- Disk read/write operations
- Network throughput of a VM or Load Balancer
- Storage account request latency
- Number of HTTP requests handled by an Application Gateway
Key characteristics of metrics:
| Characteristic | Explanation |
|---|---|
| Near real-time | Metrics are updated every few seconds or minutes. |
| Numerical | Metrics are numbers (e.g., 70% CPU, 400 requests). |
| Lightweight | They are fast to collect and do not affect performance. |
| Stored in the Metrics database | Optimized for fast charting and near real-time analysis. |
2. Types of Metrics in Azure Monitor
Azure Monitor metrics fall under these major categories:
a. Platform Metrics
- Automatically collected by Azure.
- No configuration required.
- Examples: VM CPU, Network bytes, Disk operations, Storage transactions.
b. Resource Metrics
- Specific to the resource type.
- Example:
- Application Gateway → Total requests, Failed requests.
- Azure SQL → DTU percentage, CPU usage.
- Cosmos DB → Request Units (RU/s).
c. Custom Metrics
- Metrics sent from applications or scripts using the Azure Monitor API.
d. Multi-Dimensional Metrics
- Metrics with dimensions (extra data categories).
- Example:
Application Gateway request count grouped by:- HTTP status code
- Backend pool
- Listener
This allows more advanced filtering.
3. Metric Values: Aggregations
Metrics are collected at short intervals but can be summarized using aggregations.
The Azure exam expects you to understand these:
| Aggregation | Meaning |
|---|---|
| Average | Mean value over the selected time range. |
| Minimum | Lowest value during the interval. |
| Maximum | Highest value during the interval. |
| Total | Sum of values (useful for counts). |
| Count | Number of samples collected. |
| Percentile (P50, P95, P99) | Value below which X% of results fall; helps find outliers. |
Example:
For CPU usage, average and maximum are very important to detect performance problems.
4. Where to View Metrics
Metrics can be viewed in:
✔ Azure Portal → Metrics (under any resource)
This is the most common place.
✔ Azure Monitor → Metrics
For cross-resource monitoring.
✔ Dashboards
Pin charts to custom dashboards.
✔ Workbooks
Interactive reporting with combined metrics and logs.
✔ Azure CLI / Azure PowerShell
For automation.
5. How to Interpret Metrics (Exam-Focused)
For the exam, you must know how metrics help you identify performance and health issues.
Below are common Azure resources and the metrics you must recognize:
a. Virtual Machines (VMs)
| Metric | What it tells you |
|---|---|
| CPU Percentage | High CPU means processing load is high. |
| Network In/Out | Traffic going into/out of the VM. Useful for troubleshooting connectivity. |
| Disk Read/Write Operations per Second (IOPS) | Performance of the VM disk. |
| Disk Queue Length | High queue = disk bottleneck. |
VM Interpretation Example (IT context)
If a VM’s CPU stays above 85%, you may need:
- A larger VM size
- More VMs in a scale set
- Application optimization
b. Azure App Service (Web Apps)
| Metric | Importance |
|---|---|
| Requests | Number of incoming HTTP/HTTPS calls. |
| CPU Time | How much CPU the app uses. |
| Memory Working Set | Memory usage of the app. |
| Response Time | Slow response can indicate app or backend issues. |
| HTTP 4xx/5xx Errors | Shows user errors or server errors. |
Reading these metrics helps you determine:
- If your app is overloaded
- If autoscaling is needed
- If backend services are failing
c. Application Gateway
Important metrics:
- Total Requests
- Failed Requests
- Backend Latency
- Throughput
- Capacity Units Used
Interpretation:
- High backend latency = backend VM/app is slow
- High failed requests = misconfigured backend or health probe failures
- High capacity units = scale-up may be required
d. Azure Load Balancer
Key metrics:
- Data Path Availability
- Health Probe Status
- Byte Count
Interpretation:
- If health probes fail → backend VM is unhealthy
- Low availability → configuration issues or resource health problems
e. Azure Storage Accounts
Metrics include:
- Total Requests
- Ingress / Egress
- Success E2E Latency
- Availability
Interpretation:
- High latency = slow performance
- Low availability = service issues
- Abnormal request spikes = possible misconfiguration or attack
f. Azure SQL Database
Metrics:
- DTU % or vCore CPU %
- Data IO / Log IO
- Deadlocks
- Sessions
Interpretation:
- DTU/vCore usage high → scale up
- High deadlocks → query or indexing issues
- High IO → storage performance needs improvement
6. Metric Explorer – How to Use It
Azure’s Metrics Explorer allows you to visualize and analyze metrics.
Steps (important for the exam):
- Go to the resource → Metrics.
- Select Metric Namespace (if applicable).
- Choose Metric (e.g., CPU, Requests, Latency).
- Choose Aggregation (Average, Max, Total).
- Apply Filters (e.g., status code, region, backend pool).
- Apply Split by to view data per dimension.
- Select time range.
- Pin to dashboard if needed.
7. Alerts Based on Metrics (Exam MUST-Know)
Metrics can trigger alerts when values cross a threshold.
Examples:
- CPU > 80% for 10 minutes
- Storage latency > 20 ms
- Application Gateway 5xx errors > 100
Metrics alerts can notify:
- SMS
- Azure mobile app
- ITSM systems
- Automation Runbooks
- Logic Apps
AZ-104 expects you to know:
- The difference between metric alerts (real-time, faster) and log alerts (slower, KQL-based).
- Metric alerts are recommended for performance and health issues.
8. Common Metric Interpretation Scenarios for the Exam
Scenario 1: VM running slowly
- High CPU = CPU bottleneck
- High Disk queue = storage bottleneck
Scenario 2: Web App responding slowly
- High response time = performance issue
- High 5xx errors = app/server issue
Scenario 3: Storage account is slow
- High Success E2E latency = performance issue
- Low availability = service problem
Scenario 4: Load balancer not sending traffic to a VM
- Check Health Probe Status
- If failing → VM or application not healthy
Scenario 5: Application Gateway shows errors
- High backend failures = backend misconfiguration
- High capacity = scale out Application Gateway
9. Exam Tips for “Interpret Metrics in Azure Monitor”
✔ Understand what each common metric means
✔ Know which metrics apply to which Azure resource
✔ Know how to read charts and identify performance problems
✔ Understand dimensions and filters
✔ Know how to differentiate metrics vs logs
✔ Know how to create metric alerts
✔ Know that metrics are best for near real-time monitoring
Summary
Interpreting metrics in Azure Monitor is essential for maintaining healthy and well-performing Azure environments. For the AZ-104 exam, focus on:
- Understanding types of metrics
- Reading standard metrics for VMs, App Services, SQL, Load Balancers, Application Gateways, and Storage accounts
- Using Metrics Explorer
- Applying filters, dimensions, and aggregations
- Recognizing resource performance issues
- Creating metric-based alerts
If you understand how to analyze performance, identify bottlenecks, and interpret metric charts, you will be well-prepared for this exam topic.
