How to uncache a QVW without reload or reboot? - Qlik Community

stevelord · ‎2017-01-19

Hi, is there a simple way in QVMC or moving files a special way to 'uncache' them from access point and free up physical memory without doing a reboot or reload of the QVWs in question?

I've been told I'm not allowed to reboot our server awhile pending resolution of another issue our cloud vendor is working on, so I can't do that to clear the cache. Also some of the dashboards are a bit big (or maybe our server's processing power is a bit humble) and take 30-40 minutes to reload which is a clever way I found to 'uncache' them to free up physical memory that I need for something else temporarily. I know how to set dashboards to precache themselves for quicker load time for the first user to login to them, but here I'd like a quick way to uncache them again when I have a temporary need for more physical memory during off hours if there is one.

Thanks!

PS> We'll be moving onto a parent company network in a couple of months, so 'get a more powerful server' is not in the cards right now either. If there's no simple trick, I'll just do my dashboard reload trick for now and hope cloud vendor fixes issue that happens with reboots in the near term.

Edit: And nevermind... I was going to do a trick I saw someone else suggest about putting an empty version of a qvw in the place of the full one in the document folder. So first I moved the full one out of the document folder. Then I opened qlikview to make a blank one of the same name to put in its place temporarily, but then I saw the physical memory drop from the 99% it was pegged at to 33%. When I looked back on QVMC, all but one of the qlikviews had dropped off the list of open documents with 0 sessions currently open. So... guessing maybe the change in the list of qvws in the document folder... something.. no idea why it would uncache the other documents.

marcus_sommer · ‎2017-01-20

AFAIK there is no inbuilt feature with them you could release the occupied RAM of an application within the server. The standard suggestion for cases like yours will be to increase the amounts of available RAM and/or to split your environment into several publisher and server-machines and not to use any "dirty" workarounds.

I must admit that I haven't never really thought about such workarounds although by us is RAM also a real bottleneck - in cases there aren't enough the qlikview services and/or the server itself goes down and I just re-start them again - it's not very funny but too little is just too little ...

Instead of your copying-approach of empty applications I would rather look for a way to reload the applications without data per qmc - maybe just with a variable-check at beginning:

if $(vClearCache) = true then

exit script;

else

... normal loading ...

end if

but I think it will be also have some drawbacks.

- Marcus

stevelord · ‎2017-01-23

Thanks Marcus, it was a temporary/one-off need so happy not to update script or reload dashboard to accomplish it. However, that piece of script is a cool idea.

We have a couple of Qlikviews people use every day, and some others that are pretty big but only used every month or few weeks, so are not needed in the cache constantly. Sometimes a colleague or I would be opening a big file to inspect something right on the server and it would get crowded at that point. We had setup a nightly 2:40AM reboot process to clear those infrequently used QVWs from the cache and keep space free. (And maybe boot a colleague who'd neglected to close a program when done looking at it, and maybe restart services if we pushed the server too hard and stopped one without realizing it, and do the OS updates for us. ) Nightly reboot helped a lot of things, but we had to stop that for awhile while the server vendor figures out another issue that caused our server to pull two ips semi-randomly on reboot. In the past we have upped our server processing power and RAM when it was strictly necessary to handle the volume of data we had after we'd optimized our script and processes as much as possible. Just now the volume of data doubled almost overnight the same time our overnight reboots had to be stopped, and we found ourselves a bit crunched again. We're transitioning onto our new parent company's infrastructure in a couple of months so not trying to spend any more money upgrading our stuff until after that time. Hanging in there!

marcus_sommer · ‎2017-01-23

You could further try to play with the document and/or session timeouts whereby I assume that you have already done it. Another important point is also the release which is used - earlier with QV 10 I used also regularly re-starts every 6 hours but now with QV 11.2 SR12 it's not necessary anymore.

- Marcus

stevelord · ‎2017-01-23

Thanks, I did forget to mention we're on QV 11.2 SR !2 as well. I heard SR13 was to fix an issue we're not having and QV12 had some kind of disagreement with Windows 7 systems, so we're camping on 11.2 SR12 for awhile. I don't remember ever having to reboot every 6 hours, but I do remember crashing every night for a week long ago, and showing a VP how I had delivered a whole new capability that's analyzing large amounts of data to provide new and important insights in exactly the most optimal way. And I needed the server to be at least as strong as the laptop they issued me. Back before the first server processor/ram upgrade. (Company was like 30 people back then. )