Re: [Condor-users] using the standard universe (checkpointing) between different arch/os types

[Erik Paulson <epaulson@xxxxxxxxxxx>]

On Thu, Oct 05, 2006 at 02:47:58PM -0400, Adam Bazinet wrote:
> Hi,
> 
> We have a heterogeneous Condor pool with a fair number of INTEL/LINUX machines.  We would like to be able to submit checkpointing jobs to those, but in our current configuration (which is not easily changed, trust me), one may only submit jobs from the collector node, which happens to be an OS X machine (and thus has no concept of the standard universe).  Obviously, in the normal situation checkpointing happens between the submit machine and an execute node, both of which are linux machines.  Is there some way (possibly with a separate checkpoint server?  or by transferring input files to the execute node?) to submit a checkpointing job from that OS X machine?  Open to wacky ideas, thanks.
> 

You can use the 'condor' grid type to submit jobs from the MacOS X schedd to
a Linux schedd.  

At some point, in order to run a checkpointable job, it has to get into a
schedd that runs on a platform that supports checkpointing.

-Erik