Metric Explorer in Monitoring
In Metrics, you can monitor various indicators of your infrastructure and applications in real time. For example, you can track RAM usage for a Compute Cloud VM instance or the number of requests to an Application Load Balancer virtual host.
Monitoring allows you to:
- Flexibly configure regular and derived metrics with custom formulas.
- Display one or multiple metrics on charts.
- View detailed data for each metric.
- Duplicate charts with a metric split by a certain parameter.
- Track changes over time and compare metrics for similar elements of your infrastructure.
With Metric Explorer, you can perform the following tasks:
-
Performance monitoring: Estimate system load, total allowed and denied requests to a host, and the number of errors.
-
Trend analysis: Find out how metrics changed over time, detect downtime and peak loads.
-
Issue tracking and troubleshooting: Spot problems in service performance by looking at suspicious indicators.
-
Incident investigation: View metric changes before the incident as well as the metrics and system elements related to the anomaly.
-
Configuring dashboard charts and adding alerts: Add charts to a dashboard to save them and monitor on a regular basis. Create alerts to get notifications about sudden changes in metrics.
To learn how to configure and use charts, see Viewing your services' metrics in Yandex Monitoring.
Chart repetition by parameters
The chart repetition feature based on a specific parameter allows you to build multiple similar charts with different values of this parameter. This helps you to analyze indicator anomalies and problems as you can split a metric into multiple ones and examine each of them separately.
Let’s say you have multiple VMs deployed in your cloud and observe a spike in CPU utilization. To track down the issue fast:
- Create a CPU utilization chart for all VMs.
- Enable chart repetition by VMs to create separate charts for each VM.
- When you spot a VM with high indicators, create CPU utilization charts for each of its vCPUs.
This data analysis method is known as drilling down, where you move from general to more detailed data to identify a problem or anomaly more precisely.
For more information about chart repetition, see Splitting a chart by a parameter.