Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

Remote Desktop: "Remote Desktop Service is currently busy" when connecting

Technical article
Troubleshooting

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

Users trying to connect over the Remote Desktop Protocol (RDP) get the message "Remote Desktop Service is currently busy". The message appears while connecting, but Microsoft's page treats it as a symptom of something else on the system: a slow or unhealthy logon, a memory leak, a session leak, or Group Policy hardening. The first test is to sign in at the console, where RDP is not involved.

If a restart clears it for a while and then it returns, Microsoft's page points to a leak. Do not restart the Remote Desktop Services service or the server in working hours without telling users, because both disconnect people.

What the error means

Microsoft's page is "guidance on troubleshooting the 'Remote Desktop Service is currently busy' message encountered during Remote Desktop Protocol (RDP) connections", and it applies to Windows Server. It does not state a single cause. It describes how to narrow the cause down, and what to collect before contacting Microsoft support.

Microsoft's page works in layers. It says an important first test is to access the affected computer through the console, local or the hypervisor's VM console. If you meet other issues at the console, "it likely indicates that the 'Remote Desktop Service is currently busy' error is caused by another underlying issue on the system", and Microsoft's page recommends addressing that issue "before troubleshooting RDP".

For a virtual machine, Microsoft's page says to use the hypervisor's console. If you use the Hyper-V console, "make sure to use the Basic session to do the test. This is because the Enhanced session relies on RDP and you might receive the same 'Remote Desktop Service is currently busy' message."

Likely causes

Cause What you see How to confirm
Another fault on the system (slow logon, other errors) The same trouble when signing in at the console Console logon shows a different error, delay or odd behaviour
Memory leak in a service or process A restart helps for a while, then it returns A process's memory use grows over time in Task Manager or Get-Process
Session leak Resource use looks normal, because the leak is "under the hood" Restart clears it; memory grows after restart; Microsoft's page points at security software and other software that touches RDP sessions
Group Policy hardening, such as User Rights Assignment changes Started after a policy change Event logs for Group Policy; test with the policies removed
The Remote Desktop Services service itself Restarting only that service fixes it Check whether the service uses an unusual amount of memory

Prove the cause before changing anything

These steps change no settings; gpresult writes one report file.

  1. Console test. Sign in at the physical console or the hypervisor console, Basic session. Note any different error, slow logon or abnormal behaviour.

  2. Who is on the server. Record the sessions so you know whom a restart would affect.

query session
  1. Memory. List processes using a large amount of memory, to see whether any one stands out. The Get-Process page uses Get-Process | Where-Object WorkingSet -GT 20MB; change the number to suit your server.
Get-Process | Where-Object WorkingSet -GT 500MB

Run it twice, a few hours apart, and compare. A process whose figure keeps growing is the suspect Microsoft's page tells you to look for ("a certain process consumes an unusual amount of memory and ... the memory consumption grows over time").

  1. Group Policy. Look at the "GroupPolicy" event logs for policies that cannot be processed, or a problem with the Group Policy Client service (gpsvc); Microsoft's page says both "can be the cause of the RDP issue". Record the policies that apply. Create C:\Temp first if it does not exist:
gpresult /scope computer /h C:\Temp\gp-before.html

Example (illustrative): users of rds01 at contoso.com.au report the message every Monday morning. The console logon at rds01 works. Get-Process shows the endpoint protection agent at 400 MB on Friday and 2,000 MB the following Monday. A restart on the weekend brings it back to 400 MB. The pattern matches a leak in a non-Microsoft agent, not a fault in Remote Desktop.

Fix

Choose the step that matches what you found. Microsoft's page lists a restart first, offers restarting only the service as "A potential workaround without restarting the whole system", and then covers finding the leaking software and testing Group Policy.

Restart the computer. Microsoft's page says to try this first. If "restarting the computer temporarily solves the issue, but the issue reoccurs later, it might indicate a memory leak or other types of leaks, such as session leaks or token leaks."

Disruptive step: a restart ends every session and any unsaved work in it. Back-out: none; announce a time, ask users to save, and use the query session list you recorded to confirm everyone has left.

Restart only the Remote Desktop Services service, as a possible workaround without a full restart. Microsoft's page says "Restarting the RDS service disconnects all current users that are connected via RDP. The sessions are maintained, though, and users might reconnect to their sessions." If it solves the issue, check whether the service uses an unusual amount of memory.

Disruptive step: all RDP users are disconnected at once. Back-out: none for the restart; users reconnect to their kept sessions. Use the Services console and the service named Remote Desktop Services, and have the query session record to hand.

Stop or kill a leaking process. Microsoft's page says that "restarting the service or killing the process is also a good troubleshooting step and possible workaround."

Disruptive step: ending a process closes whatever it was doing, and a security agent stays unprotected until it is back. Record the process name and how it is started first. Back-out: start the service or program again.

Remove non-Microsoft software that touches RDP. Microsoft's page says that "A good approach to testing is gradually uninstalling different software and monitoring if the issue stops", and that, if the issue happens only through RDP, it is "advisable to uninstall any non-Microsoft software that interacts with the RDP stack". It notes that stopping software does not have the same effect as uninstalling it, "because there might still be modules loaded, hooks to the system's components".

Risk: uninstalling a security product leaves the server without it until you reinstall. Record the product, version and licence, keep the installer, and get the owner of security to agree. Back-out: reinstall from the saved installer. Update all remaining non-Microsoft software first.

Test without Group Policy hardening. Microsoft's page says to identify and remove the related GPOs to test, and, if possible, to try removing all of them without technical or security consequences.

Unlink the policy from the test scope; do not delete it. Deleting a GPO leaves nothing to link back, and the Restore-GPO page says that if the GPO no longer exists in the domain the cmdlet fails. It also restores a GPO's settings from a backup, not where it was linked.

Save first: record every link on the test scope, with its order and whether it is enabled and enforced, and back up the GPO. Microsoft's Backup-GPO page says the backup directory and GPO must already exist, so create C:\GpoBackups first.

(Get-GPInheritance -Target "OU=RDSH,DC=contoso,DC=com,DC=au").GpoLinks | Format-List *
Backup-GPO -Name "Server-Hardening" -Path C:\GpoBackups

The Get-GPInheritance page says its GpoLinks property lists the GPOs linked directly to the location, and shows only their names. The New-GPLink page shows link objects with Enforced and Order values. Check that the first command's output lists order, enabled state and enforced state for each link; if it does not, write them down from Group Policy Management before you unlink. Keep the record and the Backup-GPO output. The distinguished name is an example, so use your test organisational unit.

Risk: removing hardening policies lowers security while the test runs. Test one policy at a time, on the test scope only. The Remove-GPLink page says the cmdlet "does not delete the actual GPO or any other links between the specified GPO and other sites, domains, or OUs"; if the GPO is also linked higher up, it still reaches the test servers, so confirm with gpresult that the policy has gone before you read the test result. Unlink it:

Remove-GPLink -Name "Server-Hardening" -Target "OU=RDSH,DC=contoso,DC=com,DC=au"

Then run gpupdate /force on the test servers and repeat the connection test. Back-out: link it again at the same place with the recorded order, enabled state and enforced state, then run gpupdate /force on the servers:

New-GPLink -Name "Server-Hardening" -Target "OU=RDSH,DC=contoso,DC=com,DC=au" -Order 1 -LinkEnabled Yes -Enforced No

Use the recorded values instead of 1, Yes and No. The New-GPLink page shows -Target, -Order, -LinkEnabled Yes and -Enforced Yes. Keep Backup-GPO for the case where you edit the settings inside a GPO; for that, Restore-GPO -BackupId with the Id from the Backup-GPO output puts the settings back, while the GPO still exists.

Confirm the fix

  • Connect over RDP from a client and confirm the message does not appear.
  • Run the Get-Process query again after a day and a week and confirm the suspect process is not growing.
  • Check query session and confirm sessions are normal in number.

What to watch afterwards

  • The weekly growth of the process you suspected, as in the example.
  • Any new security or monitoring agent installed on the server.
  • New Group Policy hardening changes and the date they started.

Risks and limits

These commands have not been run or lab-tested.

If the problem is not solved, Microsoft's data-collection step for support is to download the TroubleShootingScript (TSS) tool from https://aka.ms/getTSS and extract it on both the RDP client and the server. On the client run .\TSS.ps1 -Scenario UEX_RDSCli, and on the server .\TSS.ps1 -Scenario UEX_RDSSrv, reproduce the error, and stop the trace when asked. In a full Remote Desktop Services deployment Microsoft's page says to collect from every system in the connection flow, including the RD Connection Broker and RD Gateway. Microsoft's page also asks for a complete memory dump while the error is on screen, which takes the server's time and storage.

For a virtual machine, Microsoft's page says to take a snapshot of the VM with the "full memory" option selected, if the hypervisor provides it, and to convert the snapshot to a complete memory dump with the hypervisor vendor's tools. It also says to keep the connection that shows the error open while the dump is taken, and to note which user account reproduced it.

Risk: a snapshot with memory on a production VM adds load while it is taken and uses datastore space until it is removed. Agree the time with the hypervisor owner, note the snapshot's name, and have the owner remove it once Microsoft support has the dump. Microsoft's page does not describe how to remove it, so follow your hypervisor's own instructions.

Microsoft's page does not say what number of sessions or memory level counts as busy, and it does not give a single fix.

Sources

  • Remote Desktop Service Is Currently Busy Message on Connecting — Microsoft Learn, page dated 2026-02-12, accessed 20 November 2026
  • Get-Process — Microsoft Learn, date not shown, accessed 20 November 2026
  • gpresult — Microsoft Learn, page dated 2024-11-01, accessed 20 November 2026
  • query session — Microsoft Learn, page dated 2024-11-01, accessed 20 November 2026
  • Backup-GPO — Microsoft Learn, date not shown, accessed 20 November 2026
  • Restore-GPO — Microsoft Learn, date not shown, accessed 20 November 2026
  • gpupdate — Microsoft Learn, page dated 2023-02-03, accessed 20 November 2026
  • Get-GPInheritance — Microsoft Learn, date not shown, accessed 20 November 2026
  • Remove-GPLink — Microsoft Learn, date not shown, accessed 20 November 2026
  • New-GPLink — Microsoft Learn, date not shown, accessed 20 November 2026