Skip to main content
Announcements
Have questions about Qlik Connect? Join us live on April 10th, at 11 AM ET: SIGN UP NOW
cancel
Showing results for 
Search instead for 
Did you mean: 
Anonymous
Not applicable

How does QS services in a multi-node environment work?

1) I would like to understand how Qlik Sense Scheduler service on a Reload/scheduler node in a multi node environment work:


Where does data reloading into QVF takes place? Is it on Reload/scheduler node? If so, does that mean that there will be a copy of the QVF file under C:\ProgramData\Qlik\Sense\Apps folder of Reload/scheduler node? And the QVF will be synchronized to central and engine nodes after reload?


2) If central node is down, would engine nodes continue to serve the user, i.e, would users be able to access the apps on HUB? Would Reload/scheduler node be able to reload data into apps?


Thanks,





6 Replies
Anonymous
Not applicable
Author

Hi Bilge,

1) Reloading of apps take place on nodes that have the Scheduler setup as slave. You can find out which Schedulers that are slaves by looking at their settings in the QMC. A slave Scheduler uses the Engine on the same node to perform the reload. Anything stored under C:\ProgramData\Qlik\Sense\Apps is handled by the Engine but does not necessarily have anything to do with a reload. Files could be placed there due to an incoming sync.

The meta data of a reloaded app will be sync'ed to the central node, and then further to other rim nodes. As soon as a node receives the meta data about the reloaded app and if it should have that app according to sync rules, it will download it (the binary data, the QVF) in chunks from any other node that have chunkes of the same "version" of it.

An app stored by the Engine under C:\ProgramData\Qlik\Sense\Apps does not have to be the same size on different nodes even though it is the same "version". This is because the QVF file is a virtual file system which can expand but will not shrink under normal production circumstances.

2) Every node will continue to work independently and will continue to serve users if configured to, when the central node is offline. However, no sync will occur as that always goes through the central node. An app can still be reloaded (if the script itself does not depend on the offline machine) on a node but will not be sync'ed to other nodes.

/Mattias

Anonymous
Not applicable
Author

Hi Mattias,

Thanks for the detailed info, it is very helpful.

One point is still not so clear to me:

You said "Anything stored under C:\ProgramData\Qlik\Sense\Apps is not necessarily related to reload", because it may be due to a sync. But, still the reload should take place on a QVF file which is in C:\ProgramData\Qlik\Sense\Apps on a Reload/scheduler node (slave node), is that right?

Bilge

Anonymous
Not applicable
Author

As far as I know, yes.

Anonymous
Not applicable
Author

Hi Everyone,

Do we have any concrete answer for the above post. I would like to know more on high availability implementation in QlikSense.

As part of Qlik Architecture, we have main components, which will be hosted as servers.

1.       Proxy

2.       Engine

3.       Scheduler

4.       Central node.

In the scenario of high availability how many servers we have to have and how it works.

Assume a multimode with 4 servers (each one server for the above components), with high availability do we require 8 nodes? This looks not correct to me. Can you help on this?

My question is How High Availability is enabled on a multi node environment?



jbarrantesg
Partner - Contributor
Partner - Contributor

Hi Mattias,

I tested in a multi-node architecture in Qlik Sense 3.0.3. I configured a secondary node with Production purpose, I activated all the services and I  authorized the certificate. Then, I configured the load blancing in every node.

The syncrhonization works but when the central node is offline, I can't access to the hub  from the secondary node.

Do you know what configuration I am missing?

Thanks in advance,

dionverbeke
Luminary Alumni
Luminary Alumni

I think when the central node is down, everything is down...

it is load balancing , not total reliability.

Dion