Possible performance problems with reload after upgrading from QV server 11.20 to QV server 12.1 SR 6

[Anonymous]

Hi!

I just wonder if someone have bad experiences with slow reload of documents after upgrading from QV Server 11.20 to QV Server 12.1 SR 6?

Someone has told me that QV Server 12 might be slow during reload.  In fact 50% slower than using 11.20?

We have a New Publisher-server with 512 Gb memory.  There will be quite many reload tasks on the server, but with 512 Gb memory we don't expect any problems.  We shall anyway to some tests in our test-environmet.

It would be interesting to know if someone can confirm that reload using QV 12.1 really is much slower than using QV 11.2.

Best regards

Torunn Midtsjo

[pritamdeb]

Hi Torun,

This is not a solution. However, if it of any help, we are also facing quite a few problems with the upgrade from 11.2 SR 12 to 12.1 SR6.

Although the Publisher tasks are not taking any longer than before, we are seeing a drop in the front end user experience -  the objects render faster on 11.2 than on 12.1. Also, some of the scripts that run just fine on 11.2 are throwing General Script and Internal Inconsistencies (A/F/CPV) errors with variable sets of records - 1 out of 4 times, the reload with 5k records fails, even with scrambled data.

At this point, we are debating whether to go down to SR5/SR4 but Qlik support do not suggest us doing that.

Hope you have found some solutions to your problems - there are probably some under-the-hood settings that need tweaking or simply that 12.1 SR6 has some bugs that we have unfortunately stumbled into. This could also be a result of the multithreading to singlethreading change in the new QIX engine.

Nevertheless, it is difficult to explain to stakeholders why performance would drop after an upgrade.

Best Regards

Pritam

[pritamdeb]

Hi,

Hope you did get to a solution to your problems.

We upgraded to 12.1 SR7 and the inconsistencies and reload failures went away. There was a bug fix in the new release which says 'Lookups may fail'.

Now we have a situation where certain transform jobs are taking 2-3 times longer than on 11.2. In one of the jobs, I commented out a particular line of code which has nested lookups and the job then ran along the same time as on 11.2. With that line of code active, the transformation ran fine up to 2.8 mil records post which it started crawling and was fetching 4-10 records per second.

I think that the new engine does not like lookups or applymaps of a certain kind, although I am not able to identify which.

Hopefully there is a fix to these soon as it is very difficult to explain to users why things are 2x slower after upgrading the hardware (as per white listed servers on the Scalability lab release) as well as the software.

Regards

Pritam

[Anonymous]

Hi Pritam!

Thanks a lot for sharing your experiences with QV server 12 with me! I appreciate very much to receive such information.

We are planning to start using QV server 12 in production in the beginning of September. During August we are planning to do a lot of testing and it is very good to be aware of possible problems.

Best regards

Torunn Midtsjø