Why monitoring apps are only visible in 'failover' node?

[Huiying]

Hi

Our system is Qlik Sense 2019 June version, multi-node environment. We have 1 central, 2 engine, 2 scheduler and 1 proxy. PostgreSQL has own independent server. Engine node 1 is failover node.

Last week We upgraded from 2018Nov to 2019June version, and enabled monitoring apps.

I was adding load balancing to proxy and virtual proxies during weekend.

Today, to my surprise,  the monitoring apps are invisible from central node, but visible from engine node 1.

Could you tell me what caused this?

Best regards,

Susan

[Giuseppe_Novello]

What it looks like, your Engine 1 because Central or it failed over, you can confirm this by going to QMC > nodes > top right next to "actions" there's a "column selector" drop-down > select "central node", the one with "yes", it means it is your current Central. 

BR

Gio

[Huiying]

Hi Gio

You are right! the Engine 1 is not the real central node. Do you know where to check to fix it?

Central node is running normally, I don't see why engine node has to be the central...

Best regards,

Susan

[Huiying]

Hi Gio

Now some more confusing information.

We have 2 identical QS clustered environment, all have 1 central, 2 engine, independent postgre, 2 scheduler, 1 proxy.

Environment 1 is our lab, 2019June version. Real central node is Central node. Monitoring app shows only in Engine node 1.

Environment 2 is our production environment, 2018Nov version. Real central node is Engine 1. Monitoring app shows in Central node and Proxy node.

Somewhere misconfigured?

Environment 2 central node is running normally. I have no idea why the real central node is Engine 1.

Do you know where I should check?

Best regards,

Susan

[Giuseppe_Novello]

Env1 --> possible a load balancing rules was modify, go to QMC> load balancing rules > you should only have  "ResourcesOnCentralNode" and ResourcesOnNonCentralNodes, for "ResourcesOnNonCentralNodes" the condition should be "((node.iscentral="false" and resource.stream.id!="a70ca8a5-1d59-4cc9-b5fa-6e207978dcaf"))" >> this means that the Operation monitor should only show to the cental nodes and not in the RIM nodes. 

Env2 --> to make it back to be Central = Central, you should only stop Engine1 and other RIMs and keep running the actual central node, after 10 mins it will roll back to any available fail-over candidate ( in this case the real central) and once that happen, central will be central once again. 

[Huiying]

Hi Gio

Just checked the load balancing rule for Env1, condition is correct, same as you posted.

Do you know where else to check? 

Is it something related to proxy or virtual proxy?

The monitoring apps were in central node before I changed proxy settings (I changed settings of a few proxy + virtual proxy, added load balance nodes, white listed site). But Env2 has same proxy settings. so now not sure what could be the reason...

Best regards,

Susan

[Huiying]

Hi Gio

I did as you advised on Env2. Now its central node is real central. Its monitoring apps shows also only in Engine node 1.

I tested changing proxy, but no help. So probably not related to proxy.

I also check load balancing rule, all correct.

Could it be something in our configuration overrules load balancing rule? 

Best regards,

Susan

[Huiying]

Now I find something interesting, but don't understand why.

We have the default load balancing rule:

For ResourcesOnNonCentralNodes, we have: ((node.iscentral="false" and resource.stream.id!="a70ca8a5-1d59-4cc9-b5fa-6e207978dcaf"))

For ResourcesOnCentralNode, we have: ((node.iscentral="true"))

But still monitoring apps are in the failover node, Engine1.

Today I found, in the virtual proxies, I found Engine1 proxy has both central and Engine2 as load balancing nodes. If I remove central node from there, then monitoring apps disappear totally.

Why? 

When searching for instruction, I found this page: https://help.qlik.com/en-US/sense/June2019/Subsystems/ManagementConsole/Content/Sense_QMC/add-load-b...

If I understand correctly from the article, we should 
1) all nodes should have 1 virtual proxy
2) Every virtual proxy should have load balancing node

Then questions are:
1) Central should not be any node's load balancing node, right?

2) Is there rule which node should loadbalance which node? Or does it only depend on how powerful the machine is?

Best regards,

Susan