How to verify load balancing

[bryan_sng]

Hey guys,

I have setup 1 central node and 3 rim nodes using qlik sense june 2017 edition.

Under virtual proxies > node > load balancing, I have added all 4 nodes to each of the nodes (see photo below)

<snipped due to privacy issue>

<snipped due to privacy issue>

Now I assume activity in Qlik will be load balance across all 4 nodes (I did not configure load balancing rules)

However when I go to operations monitoring and did a manual reload, looking at the following photo,

<snipped due to privacy issue>

we can see that central node and rim 1 are clearly not in used.

rim 3 is also too heavily in used while rim 2 is used rarely.

Is operations monitor the correct method to verify that load balancing is indeed working?

And if so, the results seems to indicate load balancing is not working as per intended.

[rubenmarin]

Hi, I'm not sure how the default balancing (no balancing rules) works, just a few comments:

Nodes are assigned to sessions, if you open different apps in the same session all will go to the same node.

There is another monitoring app called Sessions monitor wich can show wich node has opened each app, you can import that app from %ProgramData%\Qlik\Sense\Repository\DefaultApps\

You can try setting load balancing rules to confirm central node can redirect sessions to each node. Also note that central node has a default rule to access everything, so maybe you need to remove central node from virtual proxy balance nodes to force some app to open on RIM nodes.

[bryan_sng]

Appreciate your reply Ruben!

We have previously imported sessions monitor and been reloading via task on a daily basis.

Nevertheless, I perform a manual reload on it and discover something astonishing.

<snipped due to privacy issue>

[Sessions Monitor last reload date]

<snipped due to privacy issue>

[Sessions Monitor Session Details]

It looks like the last session date is 2018-06-07 which is impossible!

I went ahead to open session details on operations monitor.

<snipped due to privacy issue>

[Operations Monitor Session Details]

It looks like operations monitor last session is correct but not the session details from Sessions Monitor.

With regard to the default load balancing rules, our central node is already redirecting sessions to each node but as per topic, it seems to be heavily sending sessions to rim 2 and neglecting itself, rim 1 and rim 3.

Hmm just to confirm, is my setup correct by assigning all nodes to each virtual proxy -> load balancing?

I does feel is correct because 2 of the servers has firewall (forgot to close) and we had customers complaining sometimes they can perform data reload (so happen they are load balance to central node and rim 1) and sometimes they cannot (load balance to rim 2 and rim 3).

[rubenmarin]

Hi Bryan, assigning all nodes to each virtual proxy should work and the central node is used, I've only configured a few multinode sites but from what I see the central node was the one heavily used while testing, I have to remove it from the balancing nodes to check redirection to other nodes, so my experience is opposite than yours with central node.
About sessions app data, have you configured the additional steps to allow accesing shared data (instead of the default 'localhost' data):
https://help.qlik.com/en-US/sense/November2018/Subsystems/Monitoring/Content/Sense_Monitoring/Introd...

[OguzcanA]

Hello Bryan,

You can try to open some random apps from reaching to your proxy which is running in centralnode ie. https://<QlikCentralNodeHostname>/hub 

Then in your RIM Nodes run taskmgr.exe and go to details tab to check CPU/RAM usage of Engine.exe.

If Engine.exe rises up in RIM nodes as you open apps ,your deployments seems ok.

You can temporarily stop QlikSense Engine service in your central node to forcing qliksense loadbalance to your rim nodes. If your setup is correct you should be able to open apps via reaching https://<QlikCentralNodeHostname>/hub while engine service is down on your central.

[bryan_sng]

Hmm, I would check this "localhost" configuration when I'm back on site, thanks!

[bryan_sng]

Thanks for the reply OguzancA!

Hmm, the processing are probably load balance to the nodes since Rim 3 is heavily loaded at 69% while Rim 2 at 2%.

I'm kind of curious why operations monitor report it that way if load balancing is configured and working correctly.

Do you guys have a somewhat distributed load across all rim nodes? like 23%, 27%, 26%, 24% across 4 nodes

[Anonymous]

Hi! did you solve this?

[bryan_sng]

sadly no, we are still trying various ways to see what can be done

[hjampani2019]

Bryan,

I am interested to know/learn about the updates on this communication.  Can you share me your findings?

Thanks

Hari