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

Re: [Condor-users] About condor's standalone checkpoint library



Hi Tan,

> to use condor's standalone checkpoint library, we donot use
> condor_submit.  Instead we need to run the application directly like
> - ./my_app.

Why do you need to run the application directly?

> Does anyone know if I follow this method and use condor's standalone
> library, whether the application is submitted to condor like the
> same way as would have been using condor_submit? Or the machine
> where I am running the executable becomes the execution machine?

It is a bit of a misnomer to talk about "condor's standalone library."
A program is compiled for Condor by using condor_compile.  Such a
program can be run as a Condor "Standard Universe" job and take
advantage of Condor's remote system calls, checkpointing, and process
migration between machines.  Or it can be run locally (standalone)
and be checkpointed by the user sending it the TSTP or USR2 signals.
It's the same executable in either case.

In the case of running it locally, it is just like any other program
and Condor knows nothing about it.  Condor only knows about it if it
is in the Condor job queue in which case it will be running remotely
as the child of a condor_starter on an execution machine and talking
to a condor_shadow on the submit machine.

So to answer your question, no, the application is not submitted to
Condor the way it would be with condor_submit (i.e. it is not in the
job queue).  And, yes, the machine where you run the executable is in
effect the execution machine, as it would be with any other program.

The question again becomes, what are you trying to do that you can't
use condor_submit, but you still seem to want remote execution?

-- 
Daniel K. Forrest	Laboratory for Molecular and
forrest@xxxxxxxxxxxxx	Computational Genomics
(608) 262 - 9479	University of Wisconsin, Madison