[GE issues] [Issue 3237] finishing a tightly integrated job takes too long

joga Joachim.Gabler at sun.com
Wed Feb 3 12:39:39 GMT 2010


http://gridengine.sunsource.net/issues/show_bug.cgi?id=3237



User joga changed the following:

                What    |Old value                 |New value
================================================================================
                  Status|NEW                       |STARTED
--------------------------------------------------------------------------------




------- Additional comments from joga at sunsource.net Wed Feb  3 04:39:38 -0800 2010 -------
SUGGESTED FIX

Make sure the master task job end report is sent not only in the load report interval, but more often.
When a master task job end report arrives,
send a special ACK back to execd, which triggeres flushing of the job report.
It will then get sent again as early as possible.

This could in theory be once a second.
Should there be significant load on qmaster, processing of the master task job report will be delayed, 
which will delay sending the special ACK to execd, 
and delay resending of the job report,
effectively preventing qmaster overload due to the suggested fix.

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=36&dsMessageId=242909

To unsubscribe from this discussion, e-mail: [issues-unsubscribe at gridengine.sunsource.net].



More information about the gridengine-users mailing list