• Introducing Dark Mode! Switch by clicking on the lightbulb icon next to Search or by clicking on Default style at the bottom left of the page!

Job queues (windows platform) | UBEs at random takes longer to complete

cncjay

Well Known Member
Hey Folks,
I am facing an issue, wanted to get your input.

I recall, there use to be an general rule of thumb: 1 CPU = 1 job queue. With the newest tools releases (in my case 9.1.5.7), does that still hold true ?

I have a batch server VM (8vcpu and 32gb RAM). At random, the jobs take a long time to complete. For example, an EDI job, that will complete in 45mins, will complete in 2+ hours, the next day.
We have looked that the database side of things to ensure no blocking is going on, etc. While there are so many areas I need to delve into, individually one by one so thought to start with the JDE job queues first.

How can you tell if I have reached some sort of a saturation point with the resources? The CPU and/or RAM never max out at 100%. We do run a lot of scheduled jobs overnight.

Any light on this topic will be of great help.

Thanks !!

E1 9.0 update 1 | tools 9.1.5.7
OS 2008R2 x64 | SQL 2008R2 | SP3
 

brandonkirsch

Active Member
Are your job queues singled threaded or multi-threaded? If they are single threaded, your jobs may appear to take longer due to a bunch of time spent in "Waiting" status.

If you are working with single-threaded queues, I have a SQL query I can share with you to detect when blocking/waiting is occurring in the queue.
 

brother_of_karamazov

Legendary Poster
Hey Folks,
I am facing an issue, wanted to get your input.

I recall, there use to be an general rule of thumb: 1 CPU = 1 job queue. With the newest tools releases (in my case 9.1.5.7), does that still hold true ?

I have a batch server VM (8vcpu and 32gb RAM). At random, the jobs take a long time to complete. For example, an EDI job, that will complete in 45mins, will complete in 2+ hours, the next day.
We have looked that the database side of things to ensure no blocking is going on, etc. While there are so many areas I need to delve into, individually one by one so thought to start with the JDE job queues first.

How can you tell if I have reached some sort of a saturation point with the resources? The CPU and/or RAM never max out at 100%. We do run a lot of scheduled jobs overnight.

Any light on this topic will be of great help.

Thanks !!

E1 9.0 update 1 | tools 9.1.5.7
OS 2008R2 x64 | SQL 2008R2 | SP3

The old rule of thumb was no more than 2 concurrently processing UBE's per CPU core. Not sure if this still holds. How many cores do you have available?


I also see that you have 8vCPU's. Take a look at this article and make sure you are NUMA optimized:

http://www.davidklee.net/2013/12/02/sql-server-virtual-machine-vnuma-sizing/


Check Processor Power Managment: http://jeffstevenson.karamazovgroup.com/2011/06/windows-2008s-magic-performance-button.html

As also pointed out, determine if actual processing time has changed. Do not rely on WSJ's Time Job Submitted and Time Job Activity, look in Row/Execution Detail/Audit Detail to find Job Process Time.


Are you sure everything on the database side is okay? Do you have index maintenance implemented?
 

gopal122

Member
Hi Guys,

I would like to share a similar experience for one of our clients . where we were often seeing cpu utilisation reaching max during month ends . eventually making log in slow, and erroring out . Apart from all things discussed above . In our case we had multiple sales download R4211 report processing simultaneously in a multi threaded queue . After lot of discussions and having issues over fortnight. We restricted sales download report to process one at a time from that batch queue. Things went fine .
 
Top