Re: [Condor-users] About checkpointing mechanism in condor

[Daniel Forrest <forrest@xxxxxxxxxxxxx>]

Hi Tan,

> Hi, condor's checkpoint manual says that whenever a checkpoint
> signal is sent to a processe by condor, the condor's checkpoint
> handler (which is a signal handler implemented in condor's system
> call library) writes the checkpoint to network socket or a file. my
> question is what kind of signal does condor send if it wants the
> checkpoint handler to write the checkpoint to a network socket?

Writing to a network socket is dependent on the job running in the
Condor Standard Universe where remote system calls are used and all
files are read/written over the network.  The checkpoint file is just
like any other file with additional logic that allows the checkpoint
file to be read-from/written-to a checkpoint server.

If the job is not running in the Condor Standard Universe then there
are no remote system calls and the file is written directly to disk.

In both cases the same signals (TSTP and USR2) are used.

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