Monitoring Azure: Advisor, Service Health & Monitor

AZ-900 AZ-900 · Describe Azure management and governance

By Abu Anns · Last updated 20 November 2026Exam version: AZ-900 (skills measured as of July 20, 2026)

Three monitoring tools, three questions

Azure has three monitoring services, and the exam tests which one answers which question. Azure Advisor answers "how can we do better?" — personalised recommendations for your resources. Azure Service Health answers "is Azure itself healthy?" — the status of Azure services and regions. Azure Monitor answers "how are our resources performing?" — metrics, logs, and alerts for your own workloads. Mix them up and the monitoring questions go wrong; keep the three questions apart and they answer themselves.

Azure Advisor: personalised recommendations

Azure Advisor analyses your resources and configuration and gives personalised, actionable recommendations in five categories: reliability (for example, enable replication for a single-instance VM), security (enable MFA, close open management ports — working with Defender for Cloud), performance (resize an under-provisioned resource), operational excellence (design for manageability, use availability sets), and cost (delete idle resources, right-size over-provisioned VMs, buy reservations). Each recommendation explains the benefit and often offers a one-click fix. Advisor is the exam's answer whenever a scenario asks "how do we find out what to improve?" or "which service recommends cost savings?" — it is a recommendations engine, not a monitoring dashboard and not a status page.

Azure Service Health: is Azure itself OK?

Azure Service Health reports the health of the Azure platform itself. It has three parts. Azure status shows the global health of Azure services — whether anything is down or degraded right now. Service issues reports problems affecting your subscriptions and regions specifically. Planned maintenance tells you about upcoming maintenance that could affect your resources. Related to it is Resource Health, which shows the health of your individual resources (whether a specific VM is available, for example) and helps diagnose whether a problem is yours or Azure's. The key distinction: Service Health is about Microsoft's platform; Azure Monitor is about your resources. If a question asks "where do I check whether Azure is having an outage in my region?", the answer is Service Health.

Azure Monitor: metrics, logs, and alerts for your resources

Azure Monitor is the comprehensive monitoring service for your Azure (and on-premises) resources. It collects two kinds of data. Metrics are numerical measurements over time — CPU percentage, memory usage, request counts, disk latency. Logs are records of events — what happened, when, and to what. Monitor stores logs in a Log Analytics workspace, where you query them with the Kusto Query Language (KQL) to investigate issues, build dashboards, and find patterns across many resources.

Monitor turns data into action through alerts. You define a rule — a metric crossing a threshold, a log query returning results, or an activity-log event (such as a resource being deleted) — and when the rule fires, Monitor notifies you or triggers automation through an action group (email, SMS, webhook, runbook). The exam tests the chain: data in, rule evaluated, alert fired, action group responds.

Azure Monitor Application Insights is Monitor's application performance monitoring (APM) feature: it tracks a live application — request rates, response times, failure rates, dependencies, and exceptions — and tells you how the application behaves for real users, including availability tests that ping it from around the world. If a scenario asks "why is our web application slow, and which requests are failing?", Application Insights is the answer — not Advisor, not Service Health, and not plain Monitor metrics alone.

Worked example: a slow website, three tools

A company's website is slow. The investigation uses all three services in order. First, Service Health: is Azure having a problem in this region? No — the platform is healthy, so the problem is the company's own. Second, Application Insights: which requests are slow? The data shows database queries taking ten times longer than usual. Third, Azure Monitor: the database's metrics show CPU at 100% — the database is under-provisioned. The fix: scale up the database. Afterwards, Advisor recommends a reservation for the now-larger database to cut its cost, and the team sets a Monitor alert to fire if database CPU passes 80% again. Four services, one incident, each doing the job the others cannot.

Metrics and logs in practice

It helps to picture what Monitor actually collects. Metrics are numbers sampled over time: a virtual machine's CPU percentage every minute, a storage account's ingress bytes, a web app's request count, a database's deadlocks. Metrics are lightweight, collected automatically for most resources, and kept for a limited retention period — they answer "how much?" and "how fast?" Logs are detailed records: every sign-in, every resource creation, every error with its message and stack trace, every firewall decision. Logs are verbose and queryable — they answer "what happened, and why?" The two work together in an investigation: a metric shows you when something went wrong (CPU spiked at 14:32), and the logs show you what went wrong (a deployment at 14:31 changed a configuration). The exam tests the distinction with scenarios: a question about tracking CPU over time is metrics; a question about finding the error behind an outage is logs.

Alerts, action groups, and workbooks

An alert has three parts, and the exam tests the chain. The signal is what is watched: a metric (CPU above 90%), a log query result (more than ten failed sign-ins in five minutes), or an activity-log event (a resource was deleted). The condition says when the alert fires: a threshold, a duration, a frequency. The action group says what happens next: send an email or SMS, call a webhook, trigger a Logic App, or run an automation runbook that restarts the service. Without an action group, an alert fires and nobody knows — the action group is what turns detection into response. For presenting all of this, Azure Workbooks provide interactive, customisable reports that combine metrics, logs, and queries into dashboards for operations reviews — the visualisation layer on top of Monitor's data. Workbooks are a supporting feature rather than a headline exam topic, but knowing they exist completes the picture: Monitor collects, alerts respond, and workbooks present.

Quick comparison

ServiceQuestion it answersKey features
Azure AdvisorHow can we improve?Recommendations: reliability, security, performance, operational excellence, cost
Azure Service HealthIs Azure healthy?Azure status, service issues, planned maintenance, Resource Health
Azure MonitorHow are our resources doing?Metrics, logs (Log Analytics), alerts, action groups
Application InsightsHow is our application behaving?Requests, response times, failures, dependencies, availability tests
Advertisement

Exam traps to watch for

Practice questions

Test yourself on this guide. Answers are checked instantly, with a full explanation for each question.

Q1.A company's website has stopped responding. The team first wants to check whether Azure itself is experiencing an outage in their region. Which service should they check?

Report an error

Q2.A company wants personalised recommendations to reduce its Azure spending, such as deleting idle virtual machines and right-sizing over-provisioned ones. Which service provides these recommendations?

Report an error

Q3.A team wants to be notified automatically when a virtual machine's CPU usage stays above 90% for more than five minutes. Which Azure Monitor feature should it configure?

Report an error

Q4.A development team wants to know which requests to its web application are failing, how long responses take, and whether the application is available from around the world. Which service should it use?

Report an error

Q5.An administrator wants to query logs from many Azure resources in one place to investigate why a deployment failed. Where are Azure Monitor logs stored and queried?

Report an error

Q6.Which of the following is one of the five categories of Azure Advisor recommendations?

Report an error

Q7.A company wants advance notice of planned Azure maintenance that could affect its virtual machines. Where should it look?

Report an error

Q8.What are the two main types of data that Azure Monitor collects?

Report an error