Monitor Azure SQL Database with metrics and alerts

Applies to: Azure SQL Database

You can use Azure Monitor metrics to monitor database and elastic pool resource consumption and health. You can use alerts to send notifications when metric values indicate a potential problem.

Metrics

A metric is a series of numeric value measured at regular time intervals, often using units such as count, percent, bytes, etc. Depending on the nature of the metric, you can use aggregations such as total, count, average, minimum, maximum to calculate metric values over a duration of time. You can split some metrics by dimensions. Each dimension provides an additional context to the numeric values.

Examples of available Azure SQL Database metrics are: CPU percentage, Data space used, Deadlocks, Tempdb Percent Log Used.

In Azure SQL Database portal, several commonly used metrics are charted on the Monitoring tab of the Overview page. The metrics let you assess resource consumption and health of a database or an elastic pool at a glance.

Under Key metrics, select See all metrics or anywhere inside the chart to open metrics explorer. In the Metrics page, you can view all other available metrics for the database or elastic pool resource. In metrics explorer, you can change the time range, granularity, and aggregation type for the chart, change the type of chart, expand the scope to include metrics from other Azure resources, create alert rules, etc. You can also open metrics explorer by selecting Metrics menu item, under Monitoring in the resource menu.

Use metrics to monitor databases and elastic pools

You can use metrics to monitor database and elastic pool resource consumption and health. For example, you can:

• Right-size the database or elastic pool to your application workload
• Detect a gradual increase in resource consumption, and proactively scale up the database or elastic pool
• Detect and troubleshoot a performance problem

The following table describes commonly used metrics in Azure SQL Database.

Alerts

You can create alert rules to notify you that the value of one metric or multiple metrics is outside of an expected range.

You can set the scope of an alert rule in multiple ways to suit your needs. For example, alert rule scope can be set to:

• A single database
• An elastic pool
• All databases or elastic pools in a resource group
• All databases or elastic pools in a subscription within an Azure region
• All databases or elastic pools in a subscription within all regions

Alert rules periodically evaluate aggregated metric values over a lookback period, comparing them to a threshold value. You can configure the threshold value, evaluation frequency, and lookback period.

If an alert rule is triggered, you are notified according to your notification preferences, which you specify in the action group linked to the alert rule. For example, you can receive an email, an SMS, or a voice notification. An alert rule can also trigger actions such as webhooks, automation runbooks, functions, logic apps, etc.

To learn more about Azure Monitor alerts, see Azure Monitor alerts overview. To get familiar with metric alerts, review Metric alerts, Manage alert rules, and Action groups.

Recommended alert rules

The metrics and optimal thresholds to use in alert rules vary across the wide spectrum of customer workloads in Azure SQL Database.

The recommended alerts in the following table are a starting point to help you define the optimal alerting configuration for your Azure SQL Database resources. Depending on your requirements, your configuration might differ from this example. You might use different thresholds, evaluation frequencies, or lookback periods. You might choose to create additional alerts, or use different alert rule configurations for different applications and environments.

Here are examples of typical alert rule configurations.

Some of the recommended alert rules use dynamic thresholds to detect anomalous metric patterns that might require attention. Alert rules based on dynamic thresholds do not trigger until sufficient historical data has been collected to establish normal patterns. For more information, see Dynamic thresholds in metric alerts.

By default, metric alerts are stateful. This means that once an alert rule is triggered, the alert is fired only once. The alert remains in the fired state until it is resolved, at which point a resolved notification is sent. An alert rule triggers a new alert only once the previous alert is resolved. Stateful alerts avoid frequent notifications about an ongoing condition. For more information about stateful and stateless alerts, see Alerts and state.

Related content

• Azure Monitor metrics overview
• Azure Monitor metrics aggregation and display explained
• Azure Monitor alerts overview
• Tutorial: Create a metric alert for an Azure resource
• Best practices for Azure Monitor alerts
• Troubleshooting problems in Azure Monitor alerts
• Monitoring and performance tuning in Azure SQL Database and Azure SQL Managed Instance
• Configure streaming export of Azure SQL Database and SQL Managed Instance diagnostic telemetry