Experiencing slow response time after moving to Windows 2019

[sledison]

Has anyone else seen a dramatic slowdown is reload times since moving to Windows Server 2019?  We are in the process of upgrading all our Qlikview servers to Windows Server 2019 (from 2012).  The server that run our QMC and QDS services are reloading tasks extremely slow, some 3-5 times slower than the older server.  We are seeing huge amounts of tasks being queued and many are failing due to 45 minute timeout waiting for an engine.  

Old servers had 4 CPU's  with 10 core each and 20 threads

New servers are 2 CPU's with 24 cores and 96 threads

I had the engines set to 16, but after the upgrade changed it to 14 and still having multiple issues. 

I don't want to continue with the server rollovers into production, if I will see the same slowdown there.

[Chip_Matejowsky]

Hi @sledison,

Note, this thread was posted in the Qlik Sense > Deployment & Management forum but looks to be for QlikView. I'll move it to the QlikView > Administration forum so that it reaches it's intended audience.

At first read through, it seems that you may need to increase the Windows Desktop Memory Heap size as detailed in Qlik Support KB article How to increase and how many reload engines (QVB) to allow in an environment.

Another potential cause could be settings in the individual QlikView services configuration files that were present in the old environment weren't carried over to the new environment. If you still have access to the old environment, suggest that you check the following files to ensure the new servers have the same entries:
%ProgramFiles%\QlikView\ManagementService\QVManagementService.exe.config
%ProgramFiles%\QlikView\DistributionService\QVDistributionService.exe.config

Also, consult the Qlik whitepaper Scaling the QlikView Publisher for other potential configuration changes that may improve performance.

Best Regards

[marcus_sommer · Accepted solution]

Beside the QMC and tasks configurations and assuming that there is any issue it might be useful to take a look on the various document-logs and comparing them.

Which one are executed with an approximately run-time and which one are needing n times more? How do they differ in action? The aim would be to detect any pattern in regard to the network/storage performance and the measures which may need a certain amount of RAM and/or CPU resources which may not available at this time. In further steps this might be combined with a look on the event/application/performance logs from QlikView and the OS.