ishki

That's a good tip, we do actually have some quite old Appvolumes packages (bundled with not-too-fresh versions of Visual C++ 2015 Redistributable, etc.) that may be the cause of the issue. Thanks!

I've experienced a situation when Horizon GUI was not any longer showing system status properly, however connecting directly to the database showed that events were taking place as usual. Apparently the GUI was somehow stuck in half-working status. Do you have a chance to restart the connection server(s) and check if the admin UI will be showing the connections correctly after that?

We upgraded to 2503 soon after release, everything else was fine but there are issues when users are reconnecting back to previously opened and then disconnected Windows 10 instant clone sessions. Our W10 master is 22H2 Ent with latest WSUS patches. There will be a number of access violation errors like shown below on screen (in this example affected process is rdeSvc, but others indicate host-nnn as well). Rolling back to agent 2412 while everything else remains at 2503 level seems to resolve the issue. Any idea what might be the root cause here? I could not figure out any explanation from local logs.

We're running ESXi 8.0.3 on HPE gen9 hardware on a secondary cluster, with VSAN, GPU:s and all, main cluster being supported gen10 hardware. No problems so far, so the software seems to be not too picky about the underlying hardware, but I guess if there were any need to contact support, we'd be advised to call again after a hw update.

Hi, Sharing this in case someone else runs into same issue: We got a new Horizon host that has Nvidia L40s GPUs, and after setting it up I created a master image and an Instant Clone pool of GPU machines to be run on it. All good so far. Provisioning stopped, however, with the following error message in Horizon: "[03/07/2025, 12:41 AM] Error during Provisioning Cloning of VM [vm-name removed for privacy] has failed: Fault type is UNKNOWN_FAULT_FATAL - No GPU capable host is available for provisioning [vm-name removed for privacy] with profile nvidia_l40s-8q. Please refer to this KB - https://kb.omnissa.com/s/article/59271 for more details." OK, the link mentioned above instructs "Download the latest graphic-profiles file attached to this article" but contains no attached file, and a quick look around in the support portal downloads section does not help either. Luckily this was simple enough to be resolved by looking into the file that should be replaced which indeed missed L40s configuration (but had L40-with-no-s) and by adding into it the following paragraph (we don't need other than q-type VM:s): # Q Series Virtual GPU types for NVIDIA L40s nvidia_l40s-1q=48 nvidia_l40s-2q=24 nvidia_l40s-3q=16 nvidia_l40s-4q=12 nvidia_l40s-6q=8 nvidia_l40s-8q=6 nvidia_l40s-12q=4 nvidia_l40s-16q=3 nvidia_l40s-24q=2 nvidia_l40s-48q=1 Save, exit, restart CS:s and provisioning those GPU VM:s works again. We're using Horizon 8 2412 so it was a bit surprising the latest CS version is still by default unprepared for this card, as it's been around for a while now.

Has anybody else noticed issues with HTML Access after upgrading to 2412? In our case, VM doesn't seem to be receiving all keyboard input correctly, more specifically dropping out key presses of special characters that need regular key+AltGr combination. Sometimes a character is echoed on the VM but sometimes those are dropped off. Changing browser seems not to help. If we're connecting to the same VM with Hz Client instead of HTML Access, there's no such issue.

I've observed higher CPU usage in case of a CS that's running on a Windows Server 2022, compared to ones running on Server 2019. However, as we only upgraded one Win server to that newer version, I don't have much statistical evidence of that correlation. You may want to test toggling off settings "Cloud delivered protection" and "Automatic Sample Submission" from Windows Security/Virus and Threat protection/Manage Settings on the windows server your CS is running on. Also, you can define under "Exclusions" that Defender should not scan your Connection Server folders.

Hello, Are there any plans to offer NSX (with full firewalling features) as an optional OEM component to Horizon license bundle? It's not too clear now what and from which counter we should purchase if we intend to upgrade Horizon licenses and keep using and updating NSX in a Horizon-only environment as before. It would be way less messy and frustrating if we could avoid making separate deals with Broadcom for that purpose.

Right now I can see and use all necessary licenses but I'm quite worried what awaits by the end of the year, when we must renew those. We use both VSAN and NSX in our Horizon setup and if I'm not mistaken, it's just very restricted "sample" VSAN capacity that comes bundled with Horizon offering by default, as it's apparently basically fixed to Horizon & VVF. A huge concern is whether if that's an dead end wrt. said extras and we practically would have to throw away the bundled VVF licenses and purchase new VCF ones at full price in order to be able to open path for decent capacity VSAN and NSX (+get additional licenses for NSX firewalling, as the VCF-bundled version of course now lacks by default that key part of NSX functionality). All in all in that case the price tag may get so high we have no other choice than to start looking for alternatives.

For me Desktop Pool view is probably the most actively used, due to frequent need to monitor pool usage and push image updates, etc. It would be great if the pool error status indicator which is now a tiny red dot could be changed to more visually noticeable alert (perhaps highlighting the whole row, for example). If I may suggest another thing for consideration; in each Pool the Stop Provisioning on Error option is now strictly an on/off choice. I'd guess sometimes an error might be caused by just a momentary glitch, so if there would be an option to define number of errors (for instance, 5 tries ) to be tolerated before stopping, that might save users from basically unnecessary VM outage esp. considering out-of-working hours situation when manual admin intervention takes time.

Has anyone else noticed increased Conncetion Server CPU load once upgrading to 2406? In our test environment, CS1 and CS2 still show load at 0,5GHz or less but CS3 and CS4 have constantly jumped to around 3GHz even though the test environment has hardly any users connected. Looking at the process list on affected CS reveals that TomcatService.exe is the main consumer of CPU resources. Rebooting the servers drops it down for some time but does not have a lasting effect.