User Profile Disk reset at Logon

Question

For a small numbers of users, we are encountering a problem where their UPD is “reset” upon login, which means all their profile data are gone. In these cases, we currently just restore their UPD (the actual VHD file) from backup, but we’d like to find the root cause to be able to prevent it from happening again.

Our infrastructure is entirely Windows Server 2019 based and the relevant servers are the following ones:

• Five VMs with RDS Role
• One VM with AD Role
• One VM with File Server Role
• One VM with Broker Role

Users log on to the Broker, which distributes their session to one of the five RDS servers based on capacity (number of users). Their profile data are stored in a UPD on the File Server, which is mounted on the respective RDS server.

For users who are affected, their “real” UPD (which may be several GB in size) is removed from the File Server and replaced with a bare-bones UPD (barely 300 MB), which does not contain any of their personal profile data. Interestingly, the creation date of the "bare-bones" UPD is identical to the "real" UPD.

The problem occurs for some, but not all, users. The users for who this is happening are in the same OU and groups, but other members of the same OU and groups do not have this issue.

Has anyone encountered this before? Do you have any suggestions as to which role, component or configuration could cause this behavior?

Answer 1

@Phoebe Lu (Shanghai Wicresoft Co,.Ltd.) Thanks for your response!

1. All five RDS VMs are in the same collection and the same GPOs apply to them.
2. I assume there will be no difference between gpresult for affected users and other users, seeing as all of them have the same GPOs applied, but we will check it.
3. No temp profiles are generated that I know of, since the RDS server is able to mount a UPD for the user (not knowing that it's an "empty" UPD instead of the "real" one). I will double-check this.
4. The user habits are very similar, they all use software from a small set of applications that we support on the RDS server.

We are not yet sure whether it's something more of a configuration problem (GPOs, ...) or infrastructure (connectivity to AD or file server, some process interfering with the stored UPDs, ...).

We'll check your suggestions and continue our search, maybe we can find some useful events around the time the issue happens. In the meantime of course we'd be happy with any further suggestions :)

Answer 2

Thanks @Phoebe Lu (Shanghai Wicresoft Co,.Ltd.) . On the RDS under Microsoft > Partition > Diagnostic we found:

time1, \server\path-to-vhd\TEMP-UVHD-SID.vhdx, Capacity: 4294967296
time2, \server\path-to-vhd\TEMP-UVHD-SID.vhdx, Capacity: 0
time3, \server\path-to-vhd\TEMP-UVHD-SID.vhdx, Capacity: 4294967296
time4, \server\path-to-vhd\TEMP-UVHD-SID.vhdx, Capacity: 0
time5, \server\path-to-vhd\UVHD-SID.vhdx, Capacity: 4294967296

The server mounts a TEMP- file instead of the actual profile. This happens twice, once with correct capacity and once with 0 bytes. After those, it mounts a VHD file with the proper file name and capacity (this is the “empty” VHD file).

We found a thread where a user reported the same issue (user "Groß, Roman" in the thread https://social.technet.microsoft.com/Forums/office/en-US/da4610ea-bb6a-4fa8-9892-d0dc512f6176/upd-user-profile-disk-overwritten-at-logon?forum=winserverTS), but no solution has been documented in that thread either.

Does anyone know what could cause these mount events?