Cluster document preload in all node and balance "preloaded document"

[korsikov]

Hi all.

We have:

Qlik Cluser with QVS1 and QVS2.

Two app: appA and appB

I go to:

In QMC:

• Documents
• User Documents
• Select your document
• Server tab
• Performance tab
• Document Control
• Select Customize
• Choose the Cluster node
• Set oprion Preload in Cluster node

In web server chosen by balancing "Preloaded document". When both nodes are available all works well. But when one node a  failure of such a situation comes.

Example:

AppA reload in QVS1. All node run

If we connect to AccessPoint appA open in QVS1.

QVS1 falure

appA reconnect and open in QVS2.

QVS1 go

appA avalable preloaded in QVS1 and QVS2. load balance "Preloaded document"

in this case, on which server the document opens?

The experiments showed that opens on the server where this point was open session or most open sessions. And all new connection open appA in QVS2.

But then that did not work out as we need to. All documents are to work on the second server.

how to fix it?

[ergustafsson]

Hi Alexander,

You need to explain a bit more on how you are doing this.

What clients are you using?

Are you saying that preloaded does not?

If so, have you tested it in the latest version? There was some issues with it previously.

Regards,

Erik

[korsikov]

Hi Erik!

Need: Appendix A to always run on the QVS1, but if the QVS 1 failure was available on other servers in the cluster.

All node in cluster use 11.20.12354.0 SR6.

I have two questions.

1 Balancing mechanism which uses a web server if it is selected method "Preloaded document" and the document has already preloaded on both cluster noda's

2. after the failed server 1 and was again started to work. How to force it to open new sessions on the first server, without reloading the second server.

[ergustafsson]

Hi Alexander,

1. Then it is randomly.

2. You can set documents to preload on the server. Or have the cluster mechanism set to "CPU with RAM overload".

Regards,

Erik

[korsikov]

Interestingly running the random number generator in the performance click the cluster. Always opens on the server where there is already an open session

Let's describe in detail the problem again

How to do that would crash all Annex A session open on the QVS1, without changing the parameters of the cluster without restarting the service.

If it is possible, of course.

To put the session timeout is very small that would have closed the session on the QVS2. And then in a regular situation, the new session to be opened Annex A to the QVS1. This is not an option, the application constantly and actively used.

[ergustafsson]

Good to know. I guess it is using resources that are already in memory - good!

What is "Annex A"? A document?

Do you want to close all the sessions?

Please elaborate a bit more.

[korsikov]

sorry for my bad English.

annex A = application A = some QVW

I would want that appA is always executed on the QVS1.

But after a QVS1 failure , an application is opened on the QVS2, and there remains work even when the server is back in QVS 1.

[ergustafsson]

Hi Alexander,

If the QVS nodes are clustered and QVS node 1 is down, users will get directed to node 2.

You can set inside Source/User documents preload options for each node, see attached.

Regards,

Erik

[korsikov]

I can not rightly pointed out, but I did it in the beginning, described in the first post

[shane_spencer]

I think the issue is that in normal course of running Alexander wants Document A on Server 1 only and Document B on Server 2 only. This is to avoid the use of 1 document affecting the performance of another by having both documents open on a single server (during normal running)

However if Server 1 fails he wants Document A to automatically become available on Server 2.

Is this possible?