Load Balancing Rule not working as expected

[shraddha_g]

Hi,

We have multinode setup for Qliksense Server (November 2018). We want to run all scheduling task on Rim node without Master-Slave Concept. We have written load balancing rule to achieve this by using Custom property. Max Concurrent Task on Rim node are 10 and On Central node it is set as 2.

When we start multiple tasks at a time, First 2 tasks are started on Central node and remaining tasks are on Development.

we want to do vice versa. first 10 tasks should always run on rim node and later ones are on Central.

Is there any modification in rule is required?

Load Balancing Rule - ((node.@NodeTask_CP=resource.@NodeTask_CP
))

(NodeTask_CP is Custom Property set for App and Nodes)

@rohitk1609 @Gysbert_Wassenaar @Michael_Tarallo

[rohitk1609]

@shraddha_g

Thanks for reaching out to me.

If i understood correctly, you  have created the custom property with apps and nodes resources ?

Now, solution I can give only if there is any problem.

If you balance one app with central and rim node then it would execute on any of the nodes by ROUND ROBIN ALGO, it means it will find best available server first and send the task to it.

so it is working as it is suppose to work.

Please correct me if I didn't get your point correctly 

Thanks,

Rohit

[shraddha_g]

Hi Rohit,

Thank you for your time.

Yes, I have created the custom property with apps and nodes.

I have a question based on the working you explained. How will ALGO check for Best available server(What will the criteria)?  I mean we don't want to run heavy loaded task on Central Node.

In our case, though Rim node resources are available, many tasks goes in Queue and wait for Production server to release resources.

we want tasks to run on Rim node only and if Rim node resources are consumed then only task should run on Central Node.

Is there anything we can do to achieve this?

Can we use Schedulers and Task/App combo in Custom property for this?

Thank you.

Regards,

Shraddha

[rohitk1609]

Hi Shraddha,

You may read how round robin works in detail:
https://en.wikipedia.org/wiki/Round-robin_scheduling

I have noticed Qlik gives priority to central node or master type, it might happens because it finds central first then look for other node.
Generally people avoid reloads on master node.
you can add one more scheduler node and balance the load among them and let the central node free.

My question to you, what is the problem if reload executes on central node ?

Second, Load balancing rule balance the load with Apps and Nodes, using task will not help you.

Thanks,
Rohit

[shraddha_g]

Hi Rohit,

We don't have option to have separate Rim node for Scheduler.

We have multiple tasks running concurrently which causing high Resource consumption on Central node leading to performance issue to Business.

In our case many tasks are getting in queue of central node though Development server is free to run task.

If you have any thought on this please let us know.

Regards,

Shraddha

[rohitk1609]

Hi Shraddha,
If I understand your Qlik site, there are only two servers , one is your central or master and same is your consumer node which is serving the apps to users.

Now there is another server which you call development server but handling reloads too.

1. so if you want to put more load to scheduler node or dev node, you may sync such apps which executes frequently to scheduler node or dev node and let other app run on central ?

2. You can increase the config of central server.

3. You may add one rim node which will serve the app and let central free.

See, Qlik recommends, 5 server at least in shared persistence , 1 central, 2 consumer and 2 scheduler nodes even if Qlik ask for separate server for file share and postgreDB.

but in your case there is only two server so I am afraid you have limited options or Qlik won't help you much in this case.

Thanks,
Rohit