[GE users] Long delay when submitting large jobs

Sean Dilda agrajag at dragaera.net
Mon Jan 17 15:03:42 GMT 2005


    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Stephan Grell - Sun Germany - SSG - Software Engineer wrote:
> >>
> In an earlier email you stated that you use classic spooling over
> NFS.
> To say it simple: that is bad for the performance. :-)
> 
> Is there a reason for not using local BDB spooling? During job start
> are a lot of objects modified, and they are all spooled....

As someone else with a similar setup.. I use it primarily due to using 
shadow masters.

Also, in the past I've found bugs or been able to work around bugs in 
5.3 and 6.0 that wouldn't have been possible if I'd been using BDB 
spooling.  (one example is sge_qmaster doing way too much file I/O in 
one of the 5.3 releases, and comparing the files being constantly 
written to the ones that weren't gave the developers the clues needed to 
fix it.  Also, being able to work around the bug in 6.0u1 where 'qconf 
-de' crashes sge_qmaster)  Its due to these cases that I don't trust a 
binary spooling format, and won't trust it until SGE has a better track 
record.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list