This product or feature is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of theService Specific Terms. Pre-GA products and features are available "as is" and might have limited support. For more information, see thelaunch stage descriptions.

You can check the health of a Persistent Disk or Google Cloud Hyperdisk volume by reviewingthedisk performance status metric. This metric indicates whether the disk'sperformance is potentially affected by adverse events within Compute Engine.

An issue affecting the disk performance status might also be visible in yourproject'sPersonal Service Health(PSH) dashboard or theGoogle Cloud Service Healthdashboard.

This document discusses the disk performance status and how to use it to troubleshootperformance issues.

When to check a disk's health

If you notice a performance issue with a disk, check the disk's health by reviewingthe disk performance status metric. The disk performance status metric is updatedevery minute and represents disk performance over the entire previous minute.For steps to check the disk's health, seeview disk performance status.

The following table summarizes the possible values of the disk performance status.

If the performance status isn'tHealthy, seeUnderstand each statusfor next steps.

If the performance status isHealthy, the disk is functioning normally and youneed to check for other causes for the performance issue.You should check for application or operating system errors and make sure yourdisk is optimized correctly. For optimization guidelines,seeOptimize HyperdiskandOptimize Persistent Disk.

How the disk health relates to other disk performance metrics

The disk's health as indicated by the performance status metric shows theinternal status of the disk from Google's perspective. If a disk's status isDegraded orSeverely Degraded, the root cause is always within theCompute Engine infrastructure.

You generally can't change a disk's health by modifying the workload. However,in rare cases, a change to the workload might trigger an internal issue, so itmight be possible to mitigate an issue by modifying the workload.

To learn about the other available disk performance metrics, seeReview disk performance metrics.

Scenarios that don't affect the disk performance status

The disk performance status is unrelated to performance issues caused by thefollowing factors:

• Incomplete or insufficient disk optimization
• Performance limit associated with the disk and machine type (if the chosen machine type can't meet the performance requirements of your workload)
• Increased load on the disk due to workload traffic
• User, application, or operating system error
• Full or corrupted disks
• For Hyperdisk and Extreme Persistent Disk volumes, insufficientlyprovisioned IOPS or throughput.

In these situations, it is your responsibility to improve performance, such asby optimizing the disk, scaling up the workload, changingthe machine type, and provisioning more capacity, IOPS, or throughput.

View a disk's health in Cloud Monitoring

To view a disk's health, create a chart in Metrics Explorer.

Required roles and permissions

To get the permissions that you need to check the disk performance status metric, ask your administrator to grant you the following IAM roles on the project:

• Monitoring Viewer (roles/monitoring.viewer)
• To save a chart to a dashboard:Monitoring Editor (roles/monitoring.editor)

For more information about granting roles, seeManage access to projects, folders, and organizations.

You might also be able to get the required permissions throughcustom roles or otherpredefined roles.

Create a chart in Metrics Explorer

To create a chart, build a query with either the menu-driven interface orPromQL.

  last_over_time    (compute_googleapis_com:instance_disk_performance_status      {monitored_resource="gce_instance",        project_id ="PROJECT_ID",        device_name="DISK_NAME"}[${__interval}])

If you view the results in a chart, then there are 3 lines for each disk,one for each possible status. Similarly, if you view the query result in atable, then the table has 3 rows for each disk.

If you built the query with PromQL, then each row or line willhave a value of1 or0. For queries built with the menus, the values forwill be100% or0.

The disk's current health is represented by the row or line whose value is100%or1.

For example, the following screenshot shows the chart for a disk nameda-test-VM,whose status isHealthy:

If you view the query results as a table, the following table is an example ofthe results for a disk that'sHealthy:

The following screenshot shows the chart for a disk calledreplica-23509 whose status isDegraded:

For information about what each performance status means, seeUnderstand each status.After you create the chart, you cansave the chart to a dashboard for future use.

Fractional results

If your query includes fractional results like in the following table,this is typically because the selecteddisplay periodwas long. As a result Cloud Monitoring aggregated the data over time.A value of77% for theHealthy status means that the disk's status wasHealthy77% of the selected display period.

For a more granular view of a disk's health, use a display period of a few hoursor some number of minutes.

Understand each status

This section discusses what each status means and when you might need to takefurther action.

Decision tree

The following diagram illustrates how to proceed if a disk has a performance issueand summarizes the information in the preceding sections.

As shown in the flowchart, you should only contact support if there are no knownalerts in PSH and Cloud service dashboards and the disk status isSeverely Degraded. If the disk isDegraded, contact support only ifall of thefollowing conditions have been met:

• The disk has beenDegraded for more than 5 minutes
• You have ruled out a workload error or misconfiguration (such as networkingissues)
• No additional optimizations can be performed at the application, workload, ordisk level
• You've reviewed all the disk's metrics
• You've examined your workload and virtual machine (VM) logs

What's next

• Learn more aboutcreating charts with Metric Explorerand how to refine query results byadding filters to a chart.
• Check for active and past service health events in thePersonal Service Health dashboard and theGoogle Service Health
• For performance optimization guidelines, seeOptimize HyperdiskandOptimize Persistent Disk.