Last weekend we've upgraded our ServerSoftware from QV 11 SR12 to QV11 SR17 because at some times we had memmory-problems. The memory went to 100% and the connection to the server via remote desktop and accesspont get lost.
After the upgrade the memory seems under control.
Only at some times the reloads are taking to much time.
Reloads of 7 sec are loading over 3 min.
We have a ot of reloads and if every reload is taking more time at a giving point in time there are to much reloads at the same time so all the reloads are giving an error (I've inserted for every reload a timeout).
The QDS+QVB combo uses memory not allocated by the QVS. Because the QVS usually grabs more and more RAM over time, fewer bytes will be left to perform reloads in. The QDS is not that picky about available RAM, it will just start swapping to disk until it has enough memory to complete are reload. However, this may slow the reload down to a crawl.
buy more RAM for your single server solution so that all services always have enough of it, or
put the publisher on its own machine so that no memory conflicts arise between the service that does the reloads and the service that does the displaying of QlikView documents.
The failures are self-generated, you were saying? The problem with reloads failing is that there is no residual load on either CPU or RAM. Because when a QVB job is aborted, all memory is released back to the free RAM pool.
Did you monitor your server load while tasks were starting to slow down and until the first few failures were triggered?