[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Condor-users] (no subject)

I'm a new user of Condor.
I have to run a program which is like that:
an initial part
a first_part (called openWP_first_part) which can be run be by different processes in parallel
and a third_part (called openWP_third_part) which can be run be by different processes in parallel

The first_part depend on the initial part and the third part depends on the first_part.
I have to run the program whith condor. The dag file is like that:

JOB init init.cmd
JOB openWP_first_part openWP_first_part.cmd
JOB openWP_third_part openWP_third_part.cmd
JOB end end.cmd
PARENT init CHILD openWP_first_part
PARENT openWP_first_part CHILD openWP_third_part
PARENT openWP_third_part CHILD end

According to the environment I have to use (named OpenWP), I have to set a node as a master. Then I launch the program by: condor_submit_dag dag
The problem I have is with the parallel jobs (first_part and third_part). When the process on the master node has finished the openWP_first_part, the openWP_third_part starts even if other processes of the first_part still run (at this moment the processes of the openWP_first_part which haven't finished yet are run in parallel with some processes of the openWP_third_part).
So it looks like there is not a "wait" between jobs.
My question is, is this a normal behaviour or is is there a way to oblige the processes of the third_part to wait until all the processes of the first_part have finished before starting.