Qlik Sense load balancing not working properly

[aniketvasadkar_eq]

Hi All,

Recently we managed to create cluster with Qlik Sense May 2022 version with Patch4.

We are having 2 node 1 central and other is scheduler.

Central Node: All services are running and node purpose is BOTH >> Central scheduler is Manager and worker

Scheduler(RIM): Repository service, engine service & scheduler service is running and purpose is production.

Parallel thread settings on Central scheduler : 8

Parallel thread settings on scheduler : 10

Expecting 18 task in parallel but that happened to be just 10 on RIM. 

But when I see task executions, all tasks are getting executed on RIM node only except monitoring apps. We are expecting task to be distributed on both central and scheduler.

We did not change any load balancing rule. 

So essentially 10 task runs on RIM and then others Queued up.

Thanks

Aniket

[aniketvasadkar_eq]

I guess below article saying the same thing.

https://community.qlik.com/t5/Knowledge/Qlik-Sense-Scheduler-does-not-load-balance-properly-when-tas...

Do I need to enable legacy load balancing ?

FYI @Daniele_Purrone @Albert_Candelario @Bastien_Laugiero 

Thanks

Aniket

[Albert_Candelario]

Hello @aniketvasadkar_eq ,

The new algorithm should have improved the load balancing, hence what would be the point of returning to the old one?

You also need to take into account:
All chained (dependent) tasks will always be executed on the same node, regardless of delay. 

As a quick test, could you put the RIM node down and ensure the tasks are running properly on the central node. Plus as CENTRAL is both (dev and prod) is it possible that developers are executing manual reloads, if so, what is the load on that node, does it have enough capacity to execute scheduled reloads?

Cheers,

Albert

[aniketvasadkar_eq]

Hi @Albert_Candelario,

Thanks for the response.

Agreed, new algorithm is already improved and no point in going back.

Yes, I observed chained task which are going to same RIM node which is also fine.

But I observed today that even non chained task are still going to RIM and getting queued up and central do not have any task running and developers are hardly doing any local reloads(You can imagine on Sundays in absence of dev activities reloads were happening on RIM node only).

Also few doubt regarding local reloads by developers there are few scenarios as below. Are these all goes to central ?

1.  Developer reloads app by clicking load data button on upper right corner on load script window.

2. Developer right clicks on App on hub and reload which created manually triggered task on QMC

3. Developer triggers already created task from QMC task menu  

4. When we are reloading app in Work area by any of above means does it go to central always

5. Does published app reloads only eligible for scheduler load balancing ?

Thanks

Aniket 

[Albert_Candelario]

Hello @aniketvasadkar_eq ,

Thanks for your reply, I am currently reviewing it and I will provide a further update before end of the week 🙂

Cheers,

Albert