Qlikview 11.20 SR19 CRITICAL TROUBLE

[barakovskikh]

Hi! Beginning with QV 11.20 SR 16 and upper application confuses fields and data during calculations.

1. Open the application

2. Reload document

3. = Bug (See 1 screnshot in attachment) <- Field 1 = Field 2 & '|' & Field 3

4. Do not close document and reload again (See 2 screnshot in attachment) = all good!

Bug actual for desktop (QV) and server (QVB) load

Spoiler

tmp:
NoConcatenate
LOAD * Inline [
x,y,z
];

Concatenate(tmp)
LOAD
RowNo() as x
,RowNo()*-1 as y
,RowNo() as z
AutoGenerate(100000);

DATA:
NoConcatenate
LOAD
x & '|' & y as KEY,
x,
y,
fabs(z) as z
Resident tmp;

drop Table tmp;

First load after open document

Second load after open document

[Brett_Bleess]

Hey guys, ok, new theory, the Windows update did not do the trick, and that makes sense given the other OS' things have been loaded upon, I should have thought that one through, but in the process of doing that, I may have stumbled across the actual culprit I think.  My laptop has 6 cores/12 hyperthreaded, and the VM Toni fired up has only 2 cores, and given I can replicate the issue, I think it must have something to do with the number of cores on the machine, the more cores, the more apt the issue is to occur...  This makes a lot more sense at this point, but I will have to see if we can get some more cores added to the VM, so we can try to get it to occur there, or I need to see if I can get the engineers on my laptop to have a look at things since I can replicate it.  This may take a full day or two at this point, so do not be surprised if nothing new until mid-week, but I think we may be on the right track at this point. 

Regards,
Brett

[Brett_Bleess]

Alexandr, can you try something for us?  I need you to edit your Settings.ini file and add the following line within the [Settings 7] area, be sure it is below that tag but above any other tag below, hopefully that makes sense! 🙂

EnableParallelReload=0

What this will do is force a single-thread condition on the reload, which if I am correct, should result in you not being able to replicate the issue any longer.  Note, there is a separate setting for QVD files, so this will not impact those, just FYI.  If you can confirm this 'fixes' things for you, then we know for sure the issue is in the multi-threading, and I should be able to get a defect report done, I may just have to get a different VM configured with enough cores to cause the issue...

Regards,
Brett

[barakovskikh]

Hi! I used the bat file to update the QV application with the /r key and save the results to qvd and the results amazed me

My first thought was - !!!OMG!!! @Brett_Bleess  you're really right and unfathomably cool. 

EnableParalellLoad = 0  ->>>>>> 100 launches = 0 errors in values

EnableParalellLoad = 1  ->>>>>> 100 launches > 5000 errors in values

This is a big step towards solving the problem

I have previously visited such thoughts, but I thought it was nonsense because of despair, I already assumed that the problem in software patches on Meltdown,Spectre

[marcus_sommer]

I think also that's really helpful because it excludes that the communication between QlikView and the OS is directly related with the issue else it's the communication between the multiple cores which conflicts in any way - probably caused from the above mentioned bugfix in SR16.

- Marcus

[Brett_Bleess]

Hey Alexandr, very nice, thanks for the thorough test, and just FYI, we are able to replicate in the VM after increasing CPUs to 12 as well, so I will be opening a case with R&D today, I just have to figure out the best way to go at this since you have no maintenance at this point.  This is likely going to take a week or two would be my best hunch, and I cannot guarantee the outcome either, as you know support for 11.20 is due to end next March, so it will be R&D's decision as to whether anything is done to address the issue.  I will circle back here once I get the case opened with R&D to leave that info for you, then we wait for the engineers to get to things and let us know if they will fix things or not.  The workaround of course is to use the setting I gave you, but the downside is that is going to force single-thread on the reloads, which will likely increase the time it takes to run things, but they should be stable at this point.  Let me know if you have further questions, otherwise I will shout back later today most likely or tomorrow at the latest to post the defect case info.

Regards,
Brett

[barakovskikh]

Thank you so much, we look forward to hearing from you in the future. At the moment, we rolled back to SR12, because it was planned to move to a newer one. I remind you that this error occurs in all recent versions starting with sr16 (inc. SR-16, SR17, SR18, SR 19). And judging by the 11.20 support, he still has a lot of customers. I think they also won't mind an update if a new stable version is released, because a lot of optimization and fault tolerance work has been done on your part in recent releases.

[Brett_Bleess]

Alexandr, sorry for the delay, I am waiting on of my team members to do some checking for me on how R&D wanted to proceed with things, and they got away from me before I could find out what was said on the call earlier today.  I will check in with them in the morning and then post an update on the plan.  The only solution right now is to use that setting to be sure things reload properly, but as I said, that is global across all reloads, and that can be added to the QlikViewBatch Settings.ini as well, i.e. Reload Engine/Publisher so it uses the same setting...

Regards,
Brett

[Brett_Bleess]

Alexandr, so Toni did submit things overnight for us, the defect ID is QVII-1752.  I know you have no access to things, so I will try to keep an eye upon things and let you know what the final outcome is once the engineers look things over.

Regards,
Brett

[barakovskikh]

Hi ! Do you have any good news for me, @Brett_Bleess ? 

P.S Forgive me for being a little impatient, but I'm very interested.

[Brett_Bleess]

No news to report currently, sorry, will shout as soon as there is any update.

Regards,
Brett