QVS is down while loading an application into memory

[yassinemhadhbi]

Good morning,

I have a problem that  constatly happening when an application is being loaded into memory by the QVS

I'm using QV 12.10 SR 10 

My applications are kinda big ( 10 go and 15 go applications )

When someone opens the application in the morning ( the first open after the reloading is finished) the applications starts loading into memory , it reaches its peak ( let's say 110 GO in RAM) the QVS goes down and then the RAM starts going down and settles in 60 GO.

When the QVS goes down it causes me a lot of trouble when there's another applications are being reloaded ( they fail)

[Chip_Matejowsky]

Hi @yassinemhadhbi,

High  memory usage when a very large QVW is opened for the first time is to be expected.  Based upon your description, it seems the QVS is experience resource exhaustion.  Do you see memory related warnings/errors in the QlikView Server Events log?  Do you have a QVS clustered environment or single QVS environment?  How much total RAM do you have on your QV server(s)?  How many 10GB - 15GB apps do you have? 

The attached Qlik Support Whitepaper QIX Engine Memory Management and CPU Utilization provides a great overview of how QlikView utilizes system resources.  Depending upon your current total system RAM, you may need to add more.  Also, probably a good idea to work on optimizing your apps and decrease their memory footprint as much as possible.

You can also try preloading your more popular apps, so that they are already loaded into memory prior to users opening them.  If you have Publisher, you'll do this in QMC > Documents > Source Documents > highlight the QVW and edit task > Server tab > Performance tab > Preload checkbox.  If you don't have Publisher, you'll do this in QMC > Documents > User Documents > highlight the QVW > Server tab > Performance tab > Preload checkbox.

Suggest that you upgrade to 12.40 or 12.50 from version 12.10 as it has lapsed for support assistance over two years ago.  And then you can engage directly with Qlik Support for issues such as this.

Best Regards

[yassinemhadhbi]

Good morning

I don't have cluster environment.

My QV server has 256 GO of RAM.

It doesn't matter how many big application there are , it happens when someone opens only one application. the QVS goes down while the application is being reloaded into RAM.

The two screenshot below describes what happenes 

[marcus_sommer]

If your application is compressed the real needed RAM footprint might be much larger as your mentioned 10/15 GB else maybe 10 times larger - and the de-compression will also need some resources and time. Therefore the installed RAM might be not enough for all your apps/user/tasks.

Beside this if your app contains synthetic keys, circular loops, many heavy tables without limiting selections on the landing page, section access, any OnOpen actions/macros or similar stuff it may need much further resources of CPU and RAM. I suggest to check this carefully because even much smaller apps could slow down big servers quite significantly if they aren't well built.

Further you mentioned the QVS goes down - do you mean the process crashed, is re-started or just remaining on 100% for a long while? Do you mean the qvs.exe (server) or the qvb.exe (publisher)?

- Marcus

[yassinemhadhbi]

I have more than sufficient RAM ( 256 GB ) when the application only needs 100 GB

It might bee the de-compress process , i won't be using compression and see

My application is fine, it doesn't have any of the mentioned problmes and it is perfectly developped by an expert.

THE qvs.exe goes down for 30 seconds ( without freeing the RAM )

[yassinemhadhbi]

The two screenshots above describes this behaviour perfectly 

[marcus_sommer · Accepted solution]

Not really, but I see now what you mean. The QMC message that the service is down does only mean that the management service and the server service couldn't communicate with each other in a certain time-frame which might be only temporary caused from the fact that the workload from the processor is on 100% and the OS isn't able to manage all called handles anymore. This doesn't mean that the process itself is down/crashed else it keep running - whereby quite often it's useless because the services-communication is a bit sensitive and may not be recovering if there are enough resources again.

- Marcus

[marcus_sommer · Accepted solution]

That the application worked fine while the development doesn't mean that's also currently true.

For example the data and/or the data-quality might have changed. Especially if there are solutions like: load * from source; you may not directly notice through any error that there are extra fields and/or missing ones and/or a different data-quality causing wrong/missing keys or similar stuff. Also differences between the creation/reloading/showing releases of the clients and server may end in a different datamodel.

Also the use of section access may reduce the datamodel and might causing any cartesian calculations by opening the app.

Usually this should be the first starting point in trouble-shooting - to ensure that really all tables/fields/associations/field-values are there like expected. For this it's often helpful to create a new app just without any UI objects and re-loading the app within the debugger with a reduced dataset which is step for step increased. If you do it with the desktop client directly on the machine on which the server runs and with the credentials of the services user you will exclude a lot of potential failure respectively you will notice it quite obviously on which point it breaks.

- Marcus