Critical Handle Leak and High RAM Usage in Qlik Sense Service Dispatcher (Multi-Node Environment)
Hi everyone,
I’m reaching out to see if anyone has encountered a specific issue regarding the Qlik Sense Service Dispatcher in a multi-node environment.
Environment Setup:
Nodes: 12-node multi-server deployment.
Configuration: Development is disabled on most nodes (view-only); 2 isolated nodes are dedicated to specific tasks.
Monitoring: We track metrics via Zabbix.
QlikSense MAY 2024 SR 11
The Problem:
On several servers, the QlikSenseServiceDispatcher.exe begins accumulating an extreme number of handle counts over time. This is a cumulative leak that usually spans 3 to 5 days, with handle counts reaching between 4 million and 9 million.
As the handle count increases, RAM usage climbs in parallel. Once it hits 99%, the affected node becomes completely unresponsive. The symptoms vary depending on the node's role:
Engine Node: Reports fail to open.
Proxy Node: Users cannot log in.
Central Node: The QMC becomes inaccessible.
Current Findings & Temporary Fix:
We opened a ticket with Microsoft, and their analysis confirmed that the Service Dispatcher is the root cause of the handle leak. Currently, restarting the service immediately flushes the handles and clears the RAM, but this is not a sustainable solution. We don't have the resources for 24/7 manual monitoring and intervention.
Questions:
Has anyone experienced a handle leak of this magnitude (6M+) with the Service Dispatcher?
Are there specific sub-services (Data Profiling, Broker, etc.) within the Dispatcher that are known to cause this?
What is the best way to debug exactly which sub-process or task is triggering the handle leak?
Any guidance on how to identify the root cause or configuration changes to prevent this would be greatly appreciated.
Thanks for your reply, in addition we find out there are so many log files generated related to NL-App-Search (which is located under ServiceDispatcher log folder) which is like 9GB. We will try to disable the nlappsearch/qliknotifier/mobilityregistrat on all servers then watch the response of the server for a while. I'll let you know the results if it works 🙂
For anyone dealing with the same thing, we tracked the system for 3 months after disabled the nlappsearch/qliknotifier/mobilityregistrat settings. No memory leaks or weird CPU Handle count spikes occured(which caused by qliksenseservidedispatcher.exe). We still dont really get why Qlik was consuming system resources so aggressively but this solved the issue for us.
