article Best Practices of Qlik Sense Multi-Node Setup, Load Balancing & Maintenance(Let's Discuss Here) in Qlik Sense Documents

Best Practices of Qlik Sense Multi-Node Setup, Load Balancing & Maintenance(Let's Discuss Here)

rohitk1609 — Thu, 06 Jun 2024 18:40:26 GMT

Best Practice of Qlik Sense Multi-Node Setup and Load Balancing

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

m_perreault — Wed, 02 Jan 2019 18:20:21 GMT

Hi Rohit,

Is it possible to create Load Balancing rules that are user based opposed to stream/app based. For example can I create a rule that users are only able to access Rim Node 1 and not Rim Node 2?

Additionally is it correct to say that each rim node will have a unique url?

Thanks,
Mark

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

rohitk1609 — Thu, 03 Jan 2019 13:06:02 GMT

Hi,

you can create load balancing rule but I don't think it will balance the load between proxies or RIM nodes. for try purpose, I created one load balancing rule where resource is User_asddcfgvvkdhykjshye and node is node1 but when I am opening any app on hub it is not bounding the apps for user whose id is asddcfgvvkdhykjshye to node1. it is working as per apps or stream rules.

Confusing point is when you create such rule no error comes out and no preview happens

but I can say, my answer is NO for user level load balancing rule.

Each RIM node who has proxy service enabled has a separate URL. but it depends how your users traffic come to you RIM nodes or proxy nodes. are you using network load balancer to bring users ?

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

axel — Thu, 24 Jan 2019 10:14:39 GMT

Hi!

We are in the process of adding another node to make a multi-node installation from our current single-node.

The reason we want a second node is not that the load is to high, but that there are resources that we cannot access from the central node that we want to connect to.

There is a firewall between the central node and the rim node, but we have opened all ports we could find information about in the documentation.

The new node is also in a different domain, in another AD, than the first node.

However, on the new node scheduler service and proxy service won't start.

Before we invest to much time in this I'd like to ask if you know if it is supported having machines from different ADs and domains in the same cluster.

Thanks!

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

rohitk1609 — Thu, 24 Jan 2019 10:23:38 GMT

User from different domain could work if user from other domain has required access.

Few questions for you

1. Can you ping central node server from rim node server by Hostname of central node server ?

2. the services user on rim node has full access of Share Folder ?If you have RDP of rim node, can you reach to share folder with service user of rim node \\centralserver\sharefolder

3. After installation, is certificates distributed in Node section ?

4. did you verify ports connectivity by powershell or telnet ?

Thanks,

Rohit

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

axel — Thu, 24 Jan 2019 10:32:40 GMT

Hello Rohit!

Thank you for your quick response!

1. Yes, ping is possible

2. Service user on the rim node is the same as on the central node, and access using Windows Explorer works as intended when using the FQDN of the central node. Therefore we made sure the FQDN was included in the path for the shared folders in the postgresql database. It is actually the servers that are part of different AD:s

3. Yes, certificates have been distributed using the instructions from qmc. It is possible to add the node in qmc, but it says the scheduler service has stopped.

4. The ports have been tested, but since we are not going to run this node as a proxy we have not opened 80 and 443. Could this be an issue for the scheduler service?

Thanks!

Axel

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

rohitk1609 — Thu, 24 Jan 2019 10:42:33 GMT

Few more questions,

1. did you add your service user in log on as service in group policies , might be name is same but due to different domain or machine its directory is different ?

2. did you restarted the rim node after installation, if no please do it once?

3. Did you tested all the ports in Overview section in https://help.qlik.com/en-US/sense/November2018/Subsystems/PlanningQlikSenseDeployments/Content/Sense_Deployment/Ports.htm

See this seems a connectivity issue, so the first point is check ports and especially check on both side, inbound and outbound.

Thanks,

Rohit

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

axel — Thu, 24 Jan 2019 14:32:51 GMT

Thanks!

1. Several of the qlik services start, it's just the proxy and scheduler service that do not start, but since this node is not configured to be a proxy that might be why that service doesn't even start. On the server the service user is in the local admin group and the qlik-specific group mentioned in the documentation. However if we have an issue with the firewall, authentication requests might still fail.

2. Node is restarted

3. We will test all the ports using telnet to verify that the communication works. Is it ok to keep 80 and 443 closed it if the node is not a proxy node? I will post here if the issue was in the firewall.

Thanks again for your help!

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

axel — Thu, 24 Jan 2019 14:39:19 GMT

Hi again,

I have another, related question.

When we get this node up and running as a part of the cluster I have a specific data connection that will only work if it runs on the rim node. I have other data connections that will only work if they run on the central node.

I have looked through the documentation and posts here, but I have not yet really found the info I am looking for, so I will ask here:

Is it possible to run just a specific reload task on a specific node, every time? This node will only be a reload node, not proxy.

Best regards,

Axel

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

rohitk1609 — Mon, 28 Jan 2019 05:17:43 GMT

Please find my inputs below:

1. For reloads or execute any app on rim node you have to make run scheduler and engine service. Repo is mandatory.

2. Auth of exe. file installation , log on as service and port communication are completely different things. Qlik needs ports to be opened even if it is working stand alone.

3. If possible start with all mandatory ports required , for proxy too, if you are intended to make your rim node as failover candidate , you have to run all services at rim node.

4. try to make rim node as failover candidate. its a great feature and makes your Qlik site highly available.

5. Yes it is definitely possible to bound any task or any app to reload on rim node only. you can set which reload will execute on which node.

You can control it by load balancing rules.

To investigate the connectivity issue between two Qlik servers, the first point is PORTS.

Keep positing your updates to fix it asap.

Thanks,

Rohit

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

npmurray — Wed, 06 Feb 2019 10:53:19 GMT

Hi,

I'm in the process of make a number of changes to our Qlik setup following a review that we got Qlik to do our environment. They made a number of recommendations;

1. Move the Repository DB to its own Server
2. Create a cluster file share
3. Add addtional engine nodes for capacity

We have completed items 1 & 2 and are working on item 3.

My quesiton is. We have create the new Engine node Server, installaed Qlik, added it via the QMC, create the proxy and virtual proxy. So now that I have done all that what config changes are needed to load balance the trafiic over both engine nodes?

Many thanks

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

rohitk1609 — Wed, 06 Feb 2019 13:09:14 GMT

Hi,

Thanks for reaching to us.

If I understood your concern correctly, load balance depends on two things

1. How you have added you node in QMC i.e services like engine , proxy

2. your load balancing rule.

Few questions for you:

1. Are you using network load balancer ?

2. how many total nodes in your Qlik site after adding new one ?

3. How you have planned to balance the load or distribute the load among nodes like you wana put reloads on one node and let use access apps on other node,

what it means, when you execute any app it will consume resources(RAM and CPU) from one Qlik machine and when user will open any app it will consume the resouce from other Qlik server.

Few people let central node free (1 Central, 1 scheduler and 1 Proxy) and put all reloads on one and let user consume apps by one proxy node.

in general we call proxy node as consumer node.

you don't need to activate proxy service on scheduler node but engine is must.

Load balance concept people use to solve a problem, so explain me in detail what you are trying to achieve ?

Thanks,

Rohit

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

npmurray — Wed, 06 Feb 2019 14:41:32 GMT

Thanks for the response.

No, we are not using a network load balancer and have no plans to use one.

After adding this new node, we will have 3, 1 central, 2 engine (consumer) nodes. As well as a separate Repository DB server and clustered File Share. We are planning to add a separate proxy node so by the end well have

1 Proxy node
1 Central/Scheduler node
2 Consumer nodes for user access
1 Consume node for development
1 Repository DB Server
1 Windows Clustered File share
1 NPRinting Server

We are looking to balance the load from users on the 2 consumer nodes an seperate out developmet on another node.

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

rohitk1609 — Thu, 07 Feb 2019 17:07:32 GMT

@npmurray

let me add my points so you can align with me.

Consider the image I am attaching in this coment.

1. Server 1, which your persistence layer, in Qlik language , where we have postgreDB and apps , we call it persistence layer. Very few firm has separate persistence layer, which is a good thing or I can say recommended by Qlik.

Generally people make that server a persistence server which is highly available that other servers so if your all server at one place or on cloud then I can say try to utilize this server resourcces in daily use like make it your central node or scheduler node.

NOTE: ALL SCHEDULER NODE OR FAILOVER CANDIATE ARE ALMOST EQUAL TO CENTRAL NODE EXCEPT SCHEDULER STATUS.

Since June 2017 we have flip flop central node concept where, your central node will be fliped to any other failover candiate when your current central or master is unavailable.

2. If you have no NLB it means your IT will make user traffic to your proxy node , my question is why do you have a dedicate proxy server, is that server will be at any other location ?

3. You can make repo DB and file share as one server, why are you using two servers separately ?

4. Yes I am with N printing on separate server .

See you are doing exactly what Qlik recomends but Qlik recommendation are based on some things.

Qlik fileshare should be on highly available server but if you are repoDB and Qlik file share are on same cloud cluster or location then why you split it to two machine ?

Please consider Fail-over candidate on high priority, I missed about it in your comments. You are gonna architect so big site and it is no worth without failover concept.

You can achieve load balancing rule solution on

https://help.qlik.com/en-US/sense/November2018/Subsystems/ManagementConsole/Content/Sense_QMC/configure-load-balancing-rules.htm

You can mail me to communicate better kumar.rohit1609@gmail.com

Thanks,

Rohit

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

axel — Mon, 11 Feb 2019 15:19:01 GMT

Hi again!

We have now made the also the scheduler service work inside the DMZ. We opened these ports (all TCP connections checked with telent):

business_server is the central node and dmz_server is the reload node.

business_server <- UPD 137 -> dmz_server
business_server <- UPD 138 -> dmz_server
business_server <- UPD 4748 dmz_server
business_server <- TCP 80 dmz_server
business_server <- TCP 139 -> dmz_server
business_server <- TCP 443 dmz_server
business_server <- TCP 445 -> dmz_server
business_server <- TCP 4239-> dmz_server
business_server <- TCP 4242-> dmz_server
business_server <- TCP 4243 -> dmz_server
business_server <- TCP 4432 dmz_server
business_server <- TCP 4444 -> dmz_server
business_server <- TCP 4747 -> dmz_server
business_server <- TCP 5050 dmz_server
business_server <- TCP 5151 dmz_server

However, this did not allow us to start the scheduler service. It looked like it wanted to access the Internet, which is not allowed from the DMZ. Against best practices we took the server out of the DMZ for a reinstallation. With the internet access all services started as expected, and it continued to work when the server was moved into the DMZ and internet access was removed. Do you know if the scheduler really requires internet access to be activated?

Thanks for helping us out!

Kind regards,

Axel

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

axel — Mon, 11 Feb 2019 15:17:16 GMT

Hi!

I still haven't completely solved my problem but since I now have the servers working I fell I am moving forward a bit. My current problem is that I have the central node which should display all applications and I have a separate reload node that should reload only one application. This application needs a data connection to a database that can only be reached from the reload node.

I assume I cannot create the data connection and thus not the application on the central node, since the central node cannot reach the database. Is this correct?
Thus I need a load balancing rule that allows me to connect to the reload node (which must also then have the proxy service running). I want to connect directly to the reload node; I don't want any connections to the reload node to be forwarded (by a load balancing rule) from the central node to the reload node. It looks like I still need some kind of load balancing rule for this. Right now when I try to connect I get the message "The proxy cannot perform load balancing on any of the engine services. Reload browser or contact system administrator". Can this be solved by a load balancing rule? If so, how?
How do I make the reload task run on only the reload node using load balancing rules?

Thanks a lot for any solutions or pointers you can give!

Kind regards,

Axel

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

rohitk1609 — Mon, 11 Feb 2019 17:08:15 GMT

Hi Axel,

Few points which will help you:

1. If I understood your Qlik site, you have two servers ,lets say A which is in DMZ zone, means outside of your network security and one which is non DMZ lets say B which is inside of security. My suggestion is make server B as your central node or master node and use server A to let user access your apps.

2. I do not think scheduler node needs internet, it might happen when you enable internet it opens certain ports which scheduler node needs to connect to master node.

3. If you wana put all your reloads on scheduler node , then go to QMC=>scheduler=> central= set it to master only, then no load will happen on central or master type.

4. In case if you make DMZ server as central then yes you can make connecton to database by taking RDP on scheudler node, login as https://localhost/hub and make connection. Please enable all services while adding scheudler node to central

please confirm DMZ mode or server is which is outside or security and publicly available ?

Thanks,

Rohit

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

axel — Mon, 18 Feb 2019 09:28:43 GMT

Hi Rohit,

1. I was unclear about the network structure, sorry; it looks like this:

https://docs.google.com/drawings/d/1LqBIGVrmJEfKdmAfCrhtBBt42QVwkS0RZ6hDt6aGSXw/edit?usp=sharing

2. During installation we had the reload node in the business network and the installation was successful. After moving the server to the DMZ the scheduler kept working. When we tried installing everything in the DMZ without internet the scheduler service would not start and we could see blocked connection attempts to a few places on the internet. This might all be a coincidence though, so we can leave it as it is.

3. Since the node in the business network needs to reload a lot of data sources that cannot be reached from the protected network DMZ, and since this is the node that will serve the apps to all users, it has to be the central node, and it has to be both master and slave scheduler.

4. Connecting to a terminal server in the DMZ allowed me to run the browser in the same network and connect to the DMZ Qlik server.

Thus what I am now facing seems to be to correctly configure Qlik to allow me to

Create an application that I can work with on the reload
Make this application visible on the central
Create a data connection to Oracle on the reload
Make reloads happen on the reload node for only this application

I been experimenting a little, so I have been able to create the application and make it visible on both nodes, but when creating a data connection on the reload node my options are much more limited than on the central node, for instance the Oracle connection wizard is not there, which is a problem.

I will not write what rules I have implemented, as it was only experimentation and it is probably pretty far from best practice.

So, if I could get some ideas on how to best solve the above problems I'd be very grateful!

Sincerely,
Axel

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

rohitk1609 — Tue, 19 Feb 2019 18:21:05 GMT

Hi Axel,

Sorry my friend for late reply!

Please find my points w.r.t. your points:

1. As I read DMZ zone is which is open for internet and available for external users but as you said your scheduler server is protected one and under DMZ zone. you are calling it reload node, fine.

You opened few ports for oracle and few to connect but I wana add a point here, Qlik has a large list of internal ports too which it needs to run smoothly. It might happen your server is installed but you never know which feature will not work unless you investigate and find it.

2. Qlik download and installs .net frame work as per its version so it might happen it was looking for internet to do it. Its very common client don't offer internet on Qlik server but Qlik server runs fine.

3. if you are calling server which is in business network CENTRAL so I hope repository and QlikShare folder is on same server. Now you open QMC or hub from any of server I hope you get same list of apps.

You can open HUB locally on reload node and create connection with oracle and make sure your all reload happens on this node, IN CASE IF YOU MAKE YOUR CENTRAL SERVER MASTER AND SLAVE , YOUR RELOADS WILL FAIL.

4. Make sure your service user can access share folder from reload server.

When you say "Make reloads happen on the reload node for only this application" how you did that ?

5. Did you make reload server as failover candidate. if no what will happen if your central is down ?

6. "Oracle connection wizard is not there," this is not at Central node ?

I am not sure what extacly your problem is but please make sure all possible aspects you covered.

mail me in case you need urgent help: kumar.rohit1609@gmail.com

Thanks,

Rohit

Re: Best Practices for Qlik Sense Multi-Node Setup & Load Balancing(Let's Discuss Problem here)

shiveshsingh — Wed, 27 Feb 2019 08:42:11 GMT

I have applications installed one one Qliksense server.

Now as per requirement, i need to migrate all the data to one SAN drive and other servers.

There will be one central node, one HA node for central, 1 Scheduler Node, 1 Scheduler Node HA, 3 Consumer nodes.

How can i proceed with the requirement, ?

I have looked on backing up the Qliksense site but will all the current data including repository will be stored in new SAN drive?