QlikView job fails when CPU is 100% instead of going into Que

[Troy1]

QlikView version: 12.1 SR8

QDS Settings:  is 85% CPU

We have cluster (2) publisher server. What i have noticed lately is job fails within few seconds of start. When i check on the server where job failed they were running at 100% CPU. 

• My question is shouldn't QlikView put those jobs in Que instead of failing it and run them once server load goes down?
• How can we avoid this situation, is there a way publisher server sends jobs to less busy server instead of round robin.

[Miguel_Angel_Baeyens · Accepted solution]

To be able to put any task on "Queue" status, QlikView must start the task first. If utilization of the server is 100% and the task is never really started by the QDS, it will never get into that status and the error should be closer to "no engine available".

Details on QDS groups: https://support.qlik.com/articles/000027134

Load Balancing algorithms: https://help.qlik.com/en-US/qlikview/April2020/Subsystems/Server/Content/QV_Server/QlikView-Server/Q...

I'd like to add that, if the server is really undersized or working under heavy stress, (e.g.: not just peaks but consistently high RAM and CPU consumption over several minutes) the server should be considered unstable and review its capacity and resources, number of concurrent tasks, type of tasks, length and complexity of QlikView scripts, other software running at the same time, etc.

RAM and CPU peaks are somehow to be expected with QlikView, but if that peak jeopardizes the environment, a deeper analysis must be conducted.

[Brett_Bleess · Accepted solution]

I just wanted to add one further point on this post, the 12.10 track is no longer supported as well, so you should be planning for an update to a supported track, which is currently 12.20 or higher, but 12.20 support will be ending this coming November, so I would likely target 12.30 or higher, so you are not having to do two upgrades so closely together.  Support for 12.10 ended November 2019.

I agree with Miguel's points, but the one other thing would be if you had increased the number of Engines for Distribution higher than 8, you may be running out of Desktop Heap Space, here is another Article reference regarding QMS/QDS best practices and settings in the config files:

QMS/QDS config settings 

There is also a link in that article to an R&D whitepaper regarding scaling of Publisher you likely want to read over as well to ensure you are following all best practices related to the QDS cluster you have configured.  

Regards,
Brett