Azure Virtual Desktop Insights glossary
This article lists and briefly describes key terms and concepts related to Azure Virtual Desktop Insights.
Available sessions shows the number of available sessions in the host pool. The service calculates this number by multiplying the number of virtual machines (VMs) by the maximum number of sessions allowed per virtual machine, then subtracting the total sessions.
Client operating system (OS)
The client operating system (OS) shows which version of the OS end-users accessing Azure Virtual Desktop resources are currently using. The client OS also shows which version of the web (HTML) client and the full Remote Desktop client the users have. For a full list of Windows OS versions, see Operating System Version.
This item shows connection health. "Connection success" means that the connection could reach the host, as confirmed by the stack on that virtual machine. A failed connection means that the connection couldn't reach the host.
Daily active users (DAU)
The total number of users that have started a session in the last 24 hours.
The total number of alerts triggered each day.
Daily connections and reconnections
The total number of connections and reconnections started or completed within the last 24 hours.
Daily connected hours
The total number of hours spent connected to a session across users in the last 24 hours.
Diagnostics and errors
When an error or alert appears in Azure Virtual Desktop Insights, it's categorized by three things:
Activity type: this category is how the error is categorized by Azure Virtual Desktop diagnostics. The categories are management activities, feeds, connections, host registrations, errors, and checkpoints. Learn more about these categories at Use Log Analytics for the diagnostics feature.
Kind: this category shows the error's location.
- Errors marked as "service" or "ServiceError = TRUE" happened in the Azure Virtual Desktop service.
- Errors marked as "deployment" or tagged "ServiceError = FALSE" happened outside of the Azure Virtual Desktop service.
- To learn more about the ServiceError tag, see Common error scenarios.
Source: this category gives a more specific description of where the error happened.
Diagnostics: the service role responsible for monitoring and reporting service activity to let users observe and diagnose deployment issues.
RDBroker: the service role responsible for orchestrating deployment activities, maintaining the state of objects, validating authentication, and more.
RDGateway: the service role responsible for handling network connectivity between end-users and virtual machines.
RDStack: a software component that's installed on your VMs to allow them to communicate with the Azure Virtual Desktop service.
Client: software running on the end-user machine that provides the interface to the Azure Virtual Desktop service. It displays the list of published resources and hosts the Remote Desktop connection once you've made a selection.
Each diagnostics issue or error includes a message that explains what went wrong. To learn more about troubleshooting errors, see Identify and diagnose Azure Virtual Desktop issues.
"Input delay" in Azure Virtual Desktop Insights means the input delay per process performance counter for each session. In the host performance page at aka.ms/azmonwvdi, this performance counter is configured to send a report to the service once every 30 seconds. These 30-second intervals are called "samples," and the report the worst case in that window. The median and p95 values reflect the median and 95th percentile across all samples.
Under Input delay by host, you can select a session host row to filter all other visuals in the page to that host. You can also select a process name to filter the median input delay over time chart.
We put delays in the following categories:
- Good: below 150 milliseconds.
- Acceptable: 150-500 milliseconds.
- Poor: 500-2,000 milliseconds (below 2 seconds).
- Bad: over 2,000 milliseconds (2 seconds and up).
To learn more about how the input delay counter works, see User Input Delay performance counters.
Monthly active users (MAU)
The total number of users that have started a session in the last 28 days. If you store data for 30 days or less, you may see lower-than-expected MAU and Connection values during periods where you have fewer than 28 days of data available.
Performance counters show the performance of hardware components, operating systems, and applications.
The following table lists the recommended performance counters and time intervals that Azure Monitor uses for Azure Virtual Desktop:
|Performance counter name
|Logical Disk(C:)\Avg. Disk Queue Length
|Logical Disk(C:)\Avg. Disk sec/Transfer
|Logical Disk(C:)\Current Disk Queue Length
|Memory(*)\% Committed Bytes in Use
|PhysicalDisk(*)\Avg. Disk Queue Length
|PhysicalDisk(*)\Avg. Disk sec/Read
|PhysicalDisk(*)\Avg. Disk sec/Transfer
|PhysicalDisk(*)\Avg. Disk sec/Write
|Processor Information(_Total)\% Processor Time
|Terminal Services(*)\Active Sessions
|Terminal Services(*)\Inactive Sessions
|Terminal Services(*)\Total Sessions
|*User Input Delay per Process(*)\Max Input Delay
|*User Input Delay per Session(*)\Max Input Delay
|RemoteFX Network(*)\Current TCP RTT
|RemoteFX Network(*)\Current UDP Bandwidth
Potential connectivity issues
Potential connectivity issues shows the hosts, users, published resources, and clients with a high connection failure rate. Once you choose a "report by" filter, you can evaluate the issue's severity by checking the values in these columns:
- Attempts (number of connection attempts)
- Resources (number of published apps or desktops)
- Hosts (number of VMs)
For example, if you select the By user filter, you can check to see each user's connection attempts in the Attempts column.
If you notice that a connection issue spans multiple hosts, users, resources, or clients, it's likely that the issue affects the whole system. If it doesn't, it's a smaller issue that lower priority.
You can also select entries to view additional information. You can view which hosts, resources, and client versions were involved with the issue. The display will also show any errors reported during the connection attempts.
Round-trip time (RTT)
Round-trip time (RTT) is an estimate of the connection's round-trip time between the end-user’s location and the session host's Azure region. To see which locations have the best latency, look up your desired location in the Azure Virtual Desktop Experience Estimator tool.
The Sessions item shows the status of all sessions, connected and disconnected. Idle sessions only shows the disconnected sessions.
Severity 0 alerts
The most urgent items that you need to take care of right away. If you don't address these issues, they could cause your Azure Virtual Desktop deployment to stop working.
Time to connect
Time to connect is the time between when a user opens a resource to start their session and when their desktop has loaded and is ready to use. For example, for a RemoteApp, this is the time it takes to launch the application.
Time to connect has two stages:
- Connection, which is how long it takes for the Azure service to route the user to a session host.
- "Logon," which is how long it takes for the service to perform tasks related to signing in the user and establishing the session on the session host.
When monitoring time to connect, keep in mind the following things:
Time to connect is measured with the following checkpoints from Azure Virtual Desktop service diagnostics data. The checkpoints Insights uses to determine when the connection is established are different for a desktop versus a RemoteApp scenario.
For example, Insights measures the time for a desktop experience to launch based on how long it takes to launch Windows Explorer. Insights also measures the time for a RemoteApp to launch based on the time taken to launch the first instance of the shell app for a connection.
If a user launches more than one RemoteApp, sometimes the shell app can execute multiple times during a single connection. For an accurate measurement of time to connect, you should only use the first execution checkpoint for each connection.
Establishing new sessions usually takes longer than reestablishing connections to existing sessions due to differences in the "logon" process for new and established connections.
The time it takes for the user to provide credentials is subtracted from their time to connect to account for situations where a user either takes a while to enter credentials or use alternative authentication methods to sign in.
When troubleshooting a high time to connect, Azure Monitor will break down total connection time data into four components to help you identify how to reduce sign-in time.
The components in this section only show the primary connection stages. These components can run in parallel, which means they won't add up to equal the total time to connect. The total time to connect is a measurement that Azure Monitor determines in a separate process.
The following flowchart shows the four stages of the sign-in process:
The flowchart shows the following four components:
User route: the time it takes from when the user selects the Azure Virtual Desktop icon to launch a session to when the service identifies a host to connect to. High network load, high service load, or unique network traffic routing can lead to high routing times. To troubleshoot user route issues, look at your network paths.
Stack connected: the time it takes from when the service resolves a target session host for the user to when the service establishes a connection between the session host and the user’s remote client. Like user routing, the network load, server load, or unique network traffic routing can affect connection time. For this component, you'll also need to pay attention to your network routing. To reduce connection time, make sure you've appropriately configured all proxy configurations on both the client and session hosts, and that routing to the service is optimal.
Logon: the time it takes between when a connection to a host is established to when the shell starts to load. Logon time includes several processes that can contribute to high connection times. You can view data for the "logon" stage in Insights to see if there are unexpected peaks in average times.
The "logon" process is divided into four stages:
Profiles: the time it takes to load a user’s profile for new sessions. How long loading takes depends on user profile size or the user profile solutions you're using (such as User Experience Virtualization). If you're using a solution that depends on network-stored profiles, excess latency can also lead to longer profile loading times.
Group Policy Objects (GPOs): the time it takes to apply group policies to new sessions. A spike in this area of the data is a sign that you have too many group policies, the policies take too long to apply, or the session host is experiencing resource issues. One thing you can do to optimize processing times is make sure the domain controller is close to session hosts as possible.
Shell Start: the time it takes to launch the shell (usually explorer.exe).
FSLogix (Frxsvc): the time it takes to launch FSLogix in new sessions. A long launch time may indicate issues with the shares used to host the FSLogix user profiles. To troubleshoot these issues, make sure the shares are collocated with the session hosts and appropriately scaled for the average number of users signing in to the hosts. Another area you should look at is profile size. Large profile sizes can slow down launch times.
Shell start to shell ready: the time from when the shell starts to load to when it's fully loaded and ready for use. Delays in this phase can be caused by session host overload (high CPU, memory, or disk activity) or configuration issues.
The user report page lets you view a specific user’s connection history and diagnostic information. Each user report shows usage patterns, user feedback, and any errors users have encountered during their sessions. Most smaller issues can be resolved with user feedback. If you need to dig deeper, you can also filter information about a specific connection ID or period of time.
Users per core
This is the number of users in each virtual machine core. Tracking the maximum number of users per core over time can help you identify whether the environment consistently runs at a high, low, or fluctuating number of users per core. Knowing how many users are active will help you efficiently resource and scale the environment.
Windows Event Logs
Windows Event Logs are data sources collected by either the Azure Monitor Agent or the Log Analytics agent on Windows virtual machines. You can collect events from standard logs like System and Application as well as custom logs created by applications you need to monitor.
The following table lists the required Windows Event Logs for Azure Virtual Desktop Insights:
|Error and Warning
|Error, Warning, and Information
|Error, Warning, and Information
|Error and Warning
|Error, Warning, and Information
|Error, Warning, and Information
- To get started, see Use Azure Virtual Desktop Insights to monitor your deployment.
- To estimate, measure, and manage your data storage costs, see Estimate Azure Monitor costs.
- If you encounter a problem, check out our troubleshooting guide for help and known issues.
You can also set up Azure Advisor to help you figure out how to resolve or prevent common issues. Learn more at Introduction to Azure Advisor.
If you need help or have any questions, check out our community resources:
Ask questions or make suggestions to the community at the Azure Virtual Desktop TechCommunity.
To learn how to leave feedback, see Troubleshooting overview, feedback, and support for Azure Virtual Desktop.
You can also leave feedback for Azure Virtual Desktop at the Azure Virtual Desktop feedback hub