Qlik Sense connection lost error

[jraomalla]

Hi all,

Our end users are having 'Connection lost' error after working some time on Qlik Sense QMC. Our Qlik Sense setup is as follows:

Qlik Sense single node - Aug 21 version

Qlik Sense running on AWS windows EC2 server with local Active Directory with around 1200 users.

From all these AD users, small number of engineers access Qlik Sense from safari in MacBook and majority of the users (end users) access it from Microsoft Edge browser in windows machines.

The issue is, engineers accessing it from MacBook are not having any issues but end users accessing it from Microsoft Edge browser are having 'Connection lost' error while they are actively working. Please note that users open Qlik apps in more than one tab.

We increased the Session timeout in virtual proxy from default 30 min to 120 min but still the issue persists. Please suggest any solution to resolve this issue.

[Andrew_Delaney]

You should not get connection lost errors in QMC, are you sure you are not referring to users working in apps instead?

If users are losing connection in the middle of work that suggests that the websocket we use for the engine connection is being terminated. Most often we see this done by an intermediary proxy or network load balancer. If it is happening for windows users and not mac users are both sets of users going through the same network infrastructure?

It may be worth talking to your network administrator about any proxy or load balancer timeout settings that are in place.

[Lauri]

I am investigating the same issue, which started this week after I upgraded to Qlik Sense Enterprise November 2021. I tested web sockets with the recommended tool and have no errors. I'll update you if I find a solution. I feel like this happened to me a couple years ago and I fixed it...

[Andrew_Delaney]

The websocket tester is will not reveal an issue where the socket is closed after a certain amount of time. It will only show situations where the websocket in unable to be established at all.

[jraomalla]

Hi Andrew_Delaney,

Thanks for your response. Sorry my bad I was referring to users working in apps. Load balancer has a value of 3600 seconds in idle timeout.

"If it is happening for windows users and not mac users are both sets of users going through the same network infrastructure?"

network infrastructure for both of these are different. We got our network team to analyse the traffic but they didn't find any issues.

you mentioned 'Websocket we use for the engine connection is being terminated" can we do any configuration change in Qlik QMC to keep it alive?

[jraomalla]

Thanks Lauri. will wait for your findings.

[Lauri]

Check out this article, @jraomalla :

https://community.qlik.com/t5/Knowledge/Enabling-TCP-Keep-Alive-Functionality-In-Qlik-Sense/ta-p/171...

I set the value to 30 and it appears to be working. I'm pretty sure that was my solution 2 years ago. (I need to keep track of these things...)

[jraomalla]

Thanks Lauri.

[Andrew_Delaney]

The article Lauri linked is the only further configuration that can be done on the Qlik side on this matter, otherwise have your network people look at timeouts enforced by proxies or loadbalancers they are using.

[jraomalla]

Thanks Andrew for your continuous support. We are using AWS load balancer which has 3600 seconds idle timeout. Users using transit gateway then route53 to reach aws application load balancer which forwards requests to Qlik Server. Our network team checked from their side and they didn't find any issues at timeouts.

Thanks