Troubleshoot machines
Note:
Citrix Health Assistantis a tool to troubleshoot configuration issues in unregistered VDAs. The tool automates a number of health checks to identify possible root causes for VDA registration failures and issues in session launch and time zone redirection configuration. The Knowledge Center article,Citrix Health Assistant - Troubleshoot VDA Registration and Session Launchcontains theCitrix Health Assistanttool download and usage instructions.
TheFilters > Machinesview in the Monitor tab displays the machines configured in the site. The Multi-session OS Machines tab includes the load evaluator index, which indicates the distribution of performance counters and tooltips of the session count if you hover over the link.
Click theFailure Reasoncolumn of a failed machine to get a detailed description of the failure and actions recommended to troubleshoot the failure. The failure reasons and the recommended actions for machine and connection failures are available in theCitrix Director Failure Reasons Troubleshooting Guide.
Click the machine name link to go to theMachine Detailspage.
The Machine Details page lists the machine details, infrastructure details, and details of the hotfixes applied on the machine.
Machine-based real-time resource utilization
TheMachine Utilizationpanel displays graphs showing real-time utilization of CPU and memory. In addition, disk and GPU monitoring graphs are available for sites with VDA versions7.14or later.
Disk monitoring graphs, average IOPS, and disk latency are important performance measurements that help you monitor and troubleshoot issues related to VDA disks. The Average IOPS graph displays the average number of reads and writes to a disk. SelectDisk Latencyto see a graph of the delay between a request for data and its return from the disk, measured in milliseconds.
GPU Utilization
SelectGPU Utilizationto see the percentage utilization of the GPU, the GPU memory, and of the Encoder and the Decoder to troubleshoot GPU-related issues on multi-session and single-session OS VDAs.
Supported GPU versions:
- NVIDIA Tesla M60 GPUs running Display Driver version 369.17 or later. For more information, seeNVIDIA vGPU Software.
- 一个MD Radeon Instinct MI25 GPUs and AMD EPYC 7V12(Rome) CPUs. For more information, see一个MD Drivers and Support.
Drivers:
The appropriate drivers or extensions must be installed on the VDAs.
For NVIDIA GPUs, install GRID drivers manually or via extensions. For more information, seeNVIDIA vGPU Software.
- Note that for NVIDIA, only GRID drivers are supported. CUDA drivers do not work with the NVadsA10 v5-series and are not supported.
- For a sample process to install Nvidia Grid GPU drivers via extensions on Azure based machines, seeNVIDIA GRID drivers. NVIDIA GPU Driver Extension - Azure Windows VMs - Azure Virtual Machines.
- For a sample process to install Nvidia Grid GPU drivers manually, see一个zure N-series NVIDIA GPU driver setup for Windows - Azure Virtual Machines.
For AMD GPUs, install AMD graphics drivers manually or via extensions. For more information, see一个MD Drivers and Support.
- For a sample process to install AMD GPU drivers via extensions on Azure based machines, see一个MD GPU Driver Extension - Azure Windows VMs - Azure Virtual Machines.
- For a sample process to install AMD GPU drivers manually on Azure machines, seeInstall AMD GPU drivers on N-series VMs running Windows.
Usage Notes:
- The GPU Utilization graphs are available only for VDAs running 64-bit Windows.
- AMD GPU利用图只提供for VDAs running Citrix Virtual Apps and Desktops 7 2212 or later.
- The VDAs must have HDX 3D Pro enabled to provide GPU acceleration. For more information, seeGPU acceleration for Windows Single-session OSandGPU acceleration for Windows Multi-session OS.
- When a VDA accesses more than one GPU, the utilization graph displays the average of the GPU metrics collected from the individual GPUs. The GPU metrics are collected for the entire VDA and not for individual processes.
- For AMD, encoder and decoder usage are not supported separately. Any encoding/ decoding workload using the GPU will be reported as the general 3D load on GPU usage.
- Ensure that you install the NVIDIA WMI during installation. This window is available only during manual installation.
- If drivers are installed but Director does not detect GPU
- Check Task Manager. If drivers are installed properly, the GPU should show up in Task Manager.
- Check if the machine is registered. Sometimes machines may take some time to be detected as online.
- If the GPU usage shows no activity in Director, make sure that the workload you are running is using the GPU. For graphics workloads, this can be enabled from Settings > System > Display > Graphics Settings > Choose the app to set preference. Make sure to turn on High Performance. Sometimes, Windows defaults to using the CPU for graphics workloads when this is set to system default or power saving, based on other settings.
- The data is updated every minute and the data visualization starts within a minute of selectingGPU Utilization.
Machine-based historical resource utilization
In theMachine Utilizationpanel, clickView Historical Utilizationto view the historical usage of resources on the selected machine. The utilization graphs include critical performance counters of CPU, memory, peak concurrent sessions, average IOPS, and disk latency.
Note:
The Monitoring policy setting,Enable Process Monitoring, must be set to Allowed to collect, and display data in the Top 10 Processes table on the Historic Machine Utilization page. The collection is prohibited by default.
The CPU and memory utilization, average IOPS, and disk latency data is collected by default. You can disable the collection by using theEnable Resource Monitoringpolicy setting.
- From theMachine Utilizationpanel in theMachine Detailsview, selectView Historical Utilization.
In theHistorical Machine Utilizationpage, setTime Periodto view usage for the last 2 hours, 24 hours, 7 days, month, or year.
Note:
一个verage IOPS and disk latency usage data are available only for the last 24 hours, month, and year ending now. Custom end time is not supported.
- Click一个pplyand select the required graphs.
- Hover over different sections of the graph to view more information for the selected time period.
For example, if you selectLast 2 hours, the baseline period is the 2 hours prior to the selected time range. View the CPU, memory, and session trend over the last 2 hours and the baseline time. If you selectLast month, the baseline period is the previous month. Select to view the Average IOPS and disk latency over the last month and the baseline time.
- ClickExportto export the resource utilization data for the selected period. For more information, seeExport reportssection in Monitor Deployments.
Below the graphs, the table lists the top 10 processes based on CPU or memory utilization. You can sort by any of the columns, which show Application Name, User Name, Session ID, Average CPU, Peak CPU, Average Memory, and Peak Memory over the selected time range. The IOPS and Disk Latency columns cannot be sorted.
Note:
The session ID for system processes is displayed as “0000”.
- To view the historical trend on the resource consumption of a particular process, drill into any of the Top 10 processes.
Machine Console access
You can access the consoles of Desktop and Multi-session OS machines hosted on XenServer Version 7.3 and later directly from Monitor. This way, you don’t require XenCenter to troubleshoot issues on XenServer hosted VDAs. For this feature to be available, the XenServer hosting the machine must be of Version 7.3 or later and must be accessible from the Monitor.
To troubleshoot a machine, click theConsole链接对应的机器详细信息面板。一个fter authentication of the host credentials you provide, the machine console opens in a separate tab using noVNC, a web-based VNC client. You now have keyboard and mouse access the console.
Note:
- This feature is not supported on Internet Explorer 11.
- If the mouse pointer on the machine console is misaligned, seeCTX230727for steps to fix the issue.
- Console access is launched on a new tab, ensure that your browser settings allow pop-ups.
- For security reasons, Citrix recommends that you install SSL certificates on your browser.
Microsoft RDS license health
You can view the status of Microsoft RDS license in the Machine Details panel in theMachine Detailsand theUser Detailspage for Multi-session OS machines.
One of the following messages is displayed:
- License available
- Not configured properly (warning)
- License error (error)
- Incompatible VDA version (error)
Note:
The RDS license health status for machines under grace period with valid license displays aLicense availablemessage in green. Renew your license before they expire.
For warning and error messages, hover over the info icon to view additional information as given in the following table.
Message Type | Messages in Monitor |
---|---|
Error | 一个vailable for VDAs version 7.16 and later. |
Error | New RDS connections are not allowed. |
Error | RDS licensing has exceeded its grace period. |
Error | 一个License Server is not configured for the required OS level with the Per Device Client Access licensing type. |
Error | The configured License Server is incompatible with the RDS Host OS level with the Per Device Client Access licensing type. |
Warning | Personal Terminal Server is not a valid RDS licensing type in a Citrix Virtual Apps and Desktops deployment. |
Warning | 远程桌面管理不是一个有效的licensing type in a Citrix Virtual Apps and Desktops deployment. |
Warning | 一个n RDS licensing type is not configured. |
Warning | The Domain Controller or License Server is unreachable with the Per User Client Access RDS licensing type. |
Warning | With the Per Device Client Access licensing type, the Client Device license could not be determined since the license server for the required OS level is unreachable. |
Note:
This feature is applicable only for Microsoft RDS CAL (Client Access License).
PVS target device metrics
You can view the status of PVS target devices for single-session and multi-session OS machines on theMachine Detailspage in Director. Several metrics forNetwork,Boot, andCacheare available on this panel. These metrics help you monitor and troubleshoot PVS target devices to ensure that they are up and running.
Network:
Network Bandwidth Utilization: Average bandwidth utilization across all NICs.
Server Reconnect Count: Number of times the server has reconnected due to network issues or server rebalancing or shutdowns and restarts of the Citrix Provisioning Stream Service.
Total UDP Retry Count: Number of times the provisioning target device has tried to reconnect to the provisioning server using UDP. This metric helps you to know if there are any network issues in the Citrix Provisioning Stream Service (for example, bad switch configurations).
Boot:
Boot Bytes Read MB: Bytes read while booting.
Boot Byes Written MB: Bytes written while booting.
Boot From: Boot medium (vDisk, local disk, and so on).
Boot Retry Count: Number of retries to boot the machine.
Boot Time: Time taken to boot the machine, in seconds. By default, there is a 5 second delay between retries. If this delay grows into double digits, there is a significant increase in boot time. Check your provisioning configuration to resolve this issue.
Target Software Version: Version of the Provisioning target device software.
vDisk Name: vDisk from which the Provisioning target device is booting.
Cache:
Write Cache Type: vDisk can be set to different types of cache. For more information, see Knowledge Center articleCTX119469.
Write Cache Volume Drive Letter: Drive letter for write cache types involving drives.
Write Cache Volume Size MB: Total configured volume size for write cache.
Cache File Size MB: Current cache file size (cache on device RAM with overflow on hard disk).
Ram Cache Usage MB: Current RAM cache size (cache on device RAM with overflow on hard disk). Use Overflow to disk only if necessary. This metric is useful when setting or optimizing the proper size of RAM cache.
For more information, seeUsing the Status Tray on a target device.
Provisioning target device metrics is available only on:
Provisioning machines.
Provisioning target device version 7.19 and later.
VDA version 2003 and later.
Note:
Metrics for Server Reconnect Count and UDP Retry Count are available only for Provisioning target version 1912 CU2 and later.