[Condor-users] MPI jobs can not be run on Condor 6.8

[tracy_luofengji <tracy_luofengji@xxxxxxx>]

Dear all,

I used Condor 6.8 and mpich-1.2.7. I tested it on 2 nodes: one acts as master and the other acts as worker. On the worker node, I copied the content of the file $CONDOR_HOME/etc/examples/condor_config.local.dedicated.resource to the file /home/condor/condor_config.local, and changed the dedicated schedular to:

 

DedicatedScheduler = "DedicatedScheduler@xxxxxxxxxxxxxxxxxxxxx"

 

Then on the master node, I created a submission file as following:

universe = parallel
executable = /usr/local/mp1script
arguments = /usr/local/helloworld
machine_count = 1
should_transfer_files = yes
when_to_transfer_output = on_exit
transfer_input_files = /usr/local/helloworld

queue

 

When I submitted the job to condor, the job always kept idle, so I want to know the reason for it.

 

Thanks!

Regards,

Tracy

---

网易邮箱，中国第一大电子邮件服务商