Hi
We manage a few virtualized windows systems running a mix of windows server 2012R2 and windows server 2019
These are rather small deployments with 2-12 users.
All servers are deployed equally and have the following software installed
- Active Directory Domain Services (The server acts as a domain controller for the users on the server)
- Remote Desktop Services Connection Broker
- Remote Desktop Services Session Host
- Microsoft SQL server
- Microsoft Office 2019 (SPLA)
- Microsoft OneDrive for Business (Server 2019 VMs only)
Each customer has their own virtual server running the above software and the server runs AD for that specific customer only
I do know that this kind of setup is not officially supported by Microsoft, but it worked fine for Windows SBS solutions previously.
The installed versions of SQL server is 2012 SP4, except for one of the server 2019 instances where we have SQL server 2017 installed. The SQL server 2017 is a standard edition, while the others are express editions.
What we are experiencing, especially on the VM with server 2019 and SQL 2017 is that when windows has more that 300 active processes, it starts to give periodic slow downs where basic tasks like sending email, saving excel sheets or invoicing from their accounting software becomes unresponsive. A couple of minutes later, everything is back to normal for 5-10 minutes after which it all starts over again.
As the issue is most easy to reproduce on the server 2019 with SQL 2017, we have focused our current troubleshooting to here. In Microsoft terms this is still a rather small deployment with only 12 active users on the same server at once.
Until now we have tried to increase the resources given to the VM, which normally runs absolutely fine with 4 CPU cores and 24GB of RAM. As we saw a lot of processes sitting and waiting for response for either Exchange Online or Sharepoint Online, we recently increased this to 6 CPU cores to have additional resources available to handle these processes. With either 4 or 6 CPU cores, the users had these slow downs for about 15-25 minutes, but as of yesterday we increased to 12 CPU cores and are now at the cases experienced earlier with contious slow downs over a couple of minutes each.
We have already ruled out a disk performance problem as all virtual machines are running on mirrored nVME setup. From monitoring the storage we can see that there is no queue and no delays on the disks when the slow downs occur.
My theory at the moment is that Windows has a hard time dealing with the 300+ processes and throwing more CPU resources at the problem would not solve anything. But at the same time it seems a bit overkill to split a 12 user system into more servers
Does anyone have similar issues or experience with comparable strange slow downs?