Tasks running slow at random times

[denwo2010]

Hi,

I have about 20 scheduled tasks, some run once a day some run every hour or so. There are few that just fail and then work fine when they run the next time.

I've had a look at the Task History for the tasks that run every hour and the Duration is random, it will say it completes in 2 minutes then the next time it runs it can take up to 20 minutes. Most of the tasks are connecting to the same server and there does not seem to be any issues on the server it connects to?

The client build 11.0.11414.0

Microsoft Windows Server 2003 R2

Enterprise x64 Edition

Service Pack

Six-Core AMD Opteron(tm)

Processor 8431

2.41 GHz, 63.9 GB of RAM

Any ideas why they would have random runnable times? Is there anything I should be checking out for?

Thanks

[marcus_sommer]

I assume that your server had to different times a different consumption of cpu+ram+network. Beside them your server is an AMD which could struggle on this what meant the server isn't always able to divide the task on all cores and is sometimes using only one core. AFAIK this is a known incompatibility between AMD and QV and using AMD as cpu is not recommended. Maybe you could improve this behavior with some server settings.

The link is quite old and not explicit AMD related but it could be a hint for you for what you should looking for:

Quick tips #8 - Server Settings For Best Performance

- Marcus

[#3]

One thing I found that helped when I had tasks randomly fail and then work the next time was to change the heap size in regedit.  It is recommended to change it from 768 to 2048.  Where and how to do this can be found on pg 113 of the latest QlikView server reference manual.

[Bill_Britt · Accepted solution]

Hi Denis,

To start you are on a release that is over two years old. Being you have only 6 cores, you can only run 5 processes at a time. Anything above that will cause you issues.

Bill

[Colin-Albert · Accepted solution]

Can you set your reloads so a task runs on completion of another task rather than running at the same time.

One other thought, if your scripts are accessing data from a database, could there be locks that cause the query to be held until a lock is released. Can you use a NOLOCK hint possibly.

As Bill bbt‌ mentioned, QlikView 11.0 is an old version and you should look at the options to upgrade.

[BillZrv]

Hello @denwo2010 did you managed to solve the problem? It seems I have the same!

Our System Settings:

QlikView 12.60.20200.0

16x Intel Xeon CPU E5-2630 v3 @2.40GHz

256 GB RAM

Windows Server 2012 R2

I tried to follow the advice of our friend above, who recommends to read the QlikView server reference manual and change the regedit size. Following the instructions, I figured out that on our system under:  System -> Setup -> Distribution Services... we have nothing set up there. Just a + sign button on the top right corner.  So, I didn't try this one yet since I am new to QlikView and I have to study what this does. 

And last but not least, in our case, we do accessing data from a database so, dear @Colin-Albert, if you can provide more clues on how to implement the NOLOCK it would be great!

On the SSMS we see an error of:  Async_Network_io suspended when we see the slowdown .

When things work normally (retrieve speed on the task manager 300-600Mbps) the task is completed in 5', and when we see the slow down the retrieve speed is weirdly steady at 100Mbps and the tasks need triple the time to complete.

One task is retrieving a table with a size about 17GB, but the thing is that we see the slow speeds happen with smaller tables too (around 700MB).

I am using QlikView for a month now for a project at work, and I have to say that I am really impressed by the QlikView community! Despite that I haven't fixed the problem yet, your support and advices I found in here helped already a lot!

Thanks!