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: 
adamdavi3s
Master
Master

Server timing sessions out (using via F5)

Hi All,

We're deployed with all of our timeout settings set to 30mins, but users are still getting dropped after 5 mins:

20161125T151313.000Z20161128T104431.000Z4700InformationSession stop reason: Socket closed by client
20161125T151313.000Z20161128T104431.000Z4100NoticeCAL usage: Named CAL session for user "DOMAIN\ADAM.DAVIES" stopped

We do access via an F5 which we think is the issue but just want to confirm that the above message is Qlik thinking the user has disconnected and not Qlik disconnecting the user?

Labels (1)
21 Replies
Miguel_Angel_Baeyens

Probably you are already doing it, but make sure all services are down before modifying any config file so changes can persist. It could well be some service restart which overwrote that value to the original in the QMC.

adamdavi3s
Master
Master
Author

yep, stop everything and IIS reset (although not on this particular server) to belt and braces make sure the new config gets picked up

It'll be me and my second child sleep deprivation having not saved or something!

adamdavi3s
Master
Master
Author

Ok so:

Tweaked those settings to be 1800, rebooted both servers, confirmed the settings had taken- didn't work

Double checked all browsers- All the same I am afraid

Will talk to networks about the reverse proxy.....

Miguel_Angel_Baeyens

Sleep deprivation: you earned even more respect with that (I only have one for the time being)

With the new setting 1800 are the config files also showing 1800 secs timeout after some time of the services up and running?

adamdavi3s
Master
Master
Author

Two under two.... short term pain, long term gain!

Yep the new times remain just fine.

I've just switched our internal webserver on to QVWS to ensure it isn't an issue that IIS was working around and that one is not timing out so I think it has to be something in the network somwhere

adamdavi3s
Master
Master
Author

Ok so....

We now have QVWS running on our main server and the stand alone web front end in the DMZ

The internal server does not time out

The external one (via F5) does

We pointed the F5 to the internal server, and this does not time out

So it is something to do with the server in the DMZ but I have no idea what to look at.

The only thing I can think is we are using a specific domain name to connect and something isn't working as it hops between the two ip addresses

Miguel_Angel_Baeyens

If the internal works but the external does not the question is -and sorry for the obvious: what are the differences between both? Things I would recheck:

  • software (QVWS config vs IIS recycling or session timeout in the service),
  • network,
  • configuration of the probes in the balancer,
  • configuration of QlikView webserver in the QMC,
  • internal has a single web server and external has more than one web server,
  • DMZ uses a proxy which is applying the timeout

After all, it is not that it does not work, just that it times out earlier than expected and from what I can see, always the same timeout (not erratic like 30 seconds now and 60 seconds one hour later and so).

adamdavi3s
Master
Master
Author

Hi Miguel,

I can't thank you enough for your time on this, its really got me foxed!

Software- should be exactly the same now the internal is running QVWS

Network- could be it

Balancer- Not sure we have one as we only use the one front end, the one on the server I purely use for testing.

Qlik Server- Should be the same as they both point to the same QVS

DMZ- could be it

I completely agree, it does work and it works well.. its just the user consistently gets kicked after 5 mins of idle time, the issue is that we don't want to deploy this to snr management as they obviously take a lot of phonecalls etc and we don't want them to loose their sessions.

I am sure it has to be something in the connection between the QVWS in the DMZ and the QVS. However because the DMZ has a read only AD then we're not doing anything fancy with ticketing and the service just runs under the same domain account... it just works nicely. I am sure the answer lies within our network team but they're looking for some pointers I think

Miguel_Angel_Baeyens

Not the best method and probably too cumbersome but still useful is to use the Scalability Center tools (basically JMeter) to emulate users longer than 5 minutes and see what the logs record, which exact error does JMeter get when it is timed out and which system is sending that message.

ON a second thought, you could still do it without all the JMeter hassle by using the developer console in the browser, all IE, Firefox and Chrome have good, built-in consoles/developer tools which can help you troubleshooting where the issue is.

adamdavi3s
Master
Master
Author

Thats a very valid point, I am always pointing people to the dev tools to debug extension scripts and often forget they can debug network as well!