Jump to content

Logo

- - - - -

Solved : CDJ engine taking a lot of time to start


4 replies to this topic

#1 Cyrille

    Member

  • Members
  • PipPip
  • 41 posts

Posted 26 July 2010 - 03:42 PM

Hi,

we are experiencing some problems with our main server when restarting $U: all the services and engines start straigh away except the CDJ service which takes around 45mins to start. During that time the server uses 100% of the CPU as the CDJ process takes as much resource as it can and we cannot see the jobs running on the server in Univiewer.

I believe this is dependant on the reorganisation politic and the size of the history files however we run an online reorganisation every night and we keep "only" 6 month history (220k jobs) which is a lot but not enough for me to explain this duration. On the server the execution history data file u_fmhs50.dta is about 700Mb.
This weekend we did a maintenance work on the server and in the meantime I stopped $U and ran an offline reorganisation which improved the CDJ start by 15mins but it is still too long.

Does anyone know exactly why it takes so much time for the CDJ engine to be operational and if we can do something about that?

Thanks,
Cyrille
(this is a Windows 2003 server with DUAS 5.3.3 FX24800A installed)

#2 men

    Hero Member

  • Root Admin
  • PipPipPipPip
  • 4,389 posts
  • Gender:Male
  • Location:Europe

Posted 26 July 2010 - 04:31 PM

Hello,

As the problem seems more on the technical side, I had a quick discusssion with Tech Support.

Their recommendation is to apply FX24848 (latest patch in CA mode for V533) which improves the efficiency of the online reorganization.

Should the problem persist, they recommend that you open a ticket with the result of a uxtrace -cf19 BEFORE launching any type of reorganization.

The size of dynamic files such as u_fmhs50, u_fmsb50, u_fmcx50 impacts the duration of data retrieval by the CDJ, but it clearly should not take that long!

Keep us posted.

Michel

#3 Cyrille

    Member

  • Members
  • PipPip
  • 41 posts

Posted 26 July 2010 - 05:09 PM

Hi Michel, thanks for your quick answer (as usual)!

It is going to be difficult for me to apply a CA patch on our main Live server soon as this is not a critical problem.
I ran an offline reorg last weekend and it did not really improve the CDJ start so I do not think an online reorganisation even improved will be better than an offline one.

Also I cannot find anything relevant in the log files. Corrupted files, network misconfiguration?
This server is a virtual one, do you think it could change something?

#4 men

    Hero Member

  • Root Admin
  • PipPipPipPip
  • 4,389 posts
  • Gender:Male
  • Location:Europe

Posted 27 July 2010 - 10:29 AM

Hi Cyrille,

I don't think the server being virtual changes things.

I think the best thing is to take the uxtrace -scf19 next time you start the Company / CDJ so we get a complete snapshot of the system and of Dollar Universe when the issue is present (the option 19 allows to gather both the data and index files of DU). This way Tech Support will have all required data to work upon the case.

One more question though, when the issue occurs, is the Graphic Job Monitor window active? If yes, can you try without opening the DU interface?

Michel

#5 men

    Hero Member

  • Root Admin
  • PipPipPipPip
  • 4,389 posts
  • Gender:Male
  • Location:Europe

Posted 22 October 2010 - 12:30 PM

Hi,

Another question: How many CPUs have you on your virtual machine?

Michel





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users