Dear rollout,
there is a problem at our site with some jobs:
Some jobs get into Waiting state in torque (we don't know the reason but
the job is started usually within one hour and finished successfully).
If the job is submitted via lcg-CE and the jobmanager notices the
Waiting state
it deletes it. This is described at
http://goc.grid.sinica.edu.tw/gocwiki/Unspecified_gridmanager_error
and it can be seen in the lcgpbs.pm perl script.
I have two questions:
1) What is this behaviour for and if I comment it out what will I
probably break?
2) Does the CREAM-CE behave the same way? (I am not able to check cream's
source).
Thank you,
--
Tomas Kouba
|