Wednesday, February 01, 2012 8:31 PM
Here is what I am trying to do....
I have a cluster that has 80 cores. 16 of these cores are on a stand alone tower machine. The other cores are in our sever rack. I have created a cluster with all the machines together. I would like to create a job in HPC that has 2 tasks; the second task is dependent on the first.
I want to get a solution for the first task using the 16 cores on machine_x26, and then wait till this completes. I know how to wait for this to complete so no help needed there. I then want to perform another "task" within this job utilizing machine_x01 to machine_x09. I can get the first task to run on machine_x26, but then the second task also runs on machine_26 even though I have specified /nodes:blah, blah, blah. How do I get the second task to use a different set of resources?
Tuesday, February 07, 2012 12:27 AMModerator
The most direct method of addressing this with the current release is for the first task to submit a second job. The second job would containing what you describe as the second task. Since they are now seperate jobs, it should be straight forward to describe the different set of resources.
-Colin, Microsoft HPC
- Proposed As Answer by Colin WatsonModerator Tuesday, February 07, 2012 12:27 AM