Irrespective of you are using document chaining or any other way, QlikView is designed to not free memory until it reaches the minimum limit as set up in the QMC, System, Setup, QlikView Servers, Performance, Working Set.
And this does make sense: QlikView works in memory, so the more memory available, the better QlikView will perform, without crashing the computer. In memory means that information is in memory, but as well that each click is cached so any user can benefit from what other users did before, therefore making the UX faster.
Hope this makes sense,
I have in my qlikview document - settings/document properties/server tab, Refresh mode when document is updated on the server to -> "Client initiates refresh. If old data not kept in server or client too old, session will be disconnected".
My observation is this setting does not manage memory in the server optimally. The usage keeps growing and I think when it reaches peak working set, terminates the session causing 'No connection' to clients.
If I change the setting to "Client initiates refresh. If old data not kept in server or client too old, refresh automatically"
Then the server manages memory and users don't get 'No connection'. This setting is good for us from server perspective. But the problem is users are refreshed automatically when document is updated on the server, which is not desirable to our users. So my question is how do I keep this setting (refresh automatically) but override this in client by not automatically refreshing. - ( Is there a browser setting that prevent server to autorefresh? in Ajax client. Any other suggestions?
The key to your problem is "If old data not kept in server". This means for this to work you would have to uncheck the box "Allow Only One Copy of Document in Memory " in the QVS setting under the document tab.
What this does is allow more than one copy of the document in memory (old document = old data). This would also mean that you could have the same document loaded several time in memory base on the reloads and users logged in.