[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Condor-users] Running multiple jobs on the same computer
You can do this with the "Machine" ClassAd:
- Task A creates a file which contains the current machine it is running
on(i.e MachineName = "testServer").
- This file is transfered back to the submitting machine. A preScript (i.e
a perl script) of DAGman reads this file for the MachineName and writes it
in the requirements Job ClassAd of the tasks B1, B2,...
- see chapter 2.11 in the Condor manual and chapter 9 for information on
how to get the nachine name (condor_config_val).
Task A is running on Machine "testServer". Before it finishes it writes
"MachineName = testServer" to a file (name.txt).
A preScript reads name.txt to get the Value of MachineName and writes it to
the submit discription files of the Tasks B1, B2,... like:
' requirements = Machine == "testServer" '
> I'm trying to setup Condor on our network and would like to create a
> specific job hierarchy using Dagman:
> - Task A creates some data files that will be processed by multiple
> other jobs.
> Data files are created locally to have as low network traffic as can be.
> - Tasks B1, B2 and B3 should run on the same computer (and same
> processor) after Task A is done to be able to
> access the pre-computed data.
> How can I dynamically constrain a job to run on a specific
> processor? Is there any way to pass
> messages between tasks - from Task A - to edit the submit file of
> Task Bs before submission to
> constrain them to a specific computer?
> Thanks in advance.
> Horvátth Szabolcs
> Condor-users mailing list