[Condor-users] Vanilla checkpointing and state files

[Matt Hope <matthew.hope@xxxxxxxxx>]

One of my users who is making use of the vanilla checkpointing
functionality would like to view the output from all runs of the job.

I suggested post run going to the spool directory and getting the
various files from the spool directories. This it seems is not
possible since the output files in the spool are overwritten after
each checkpoint rather than maintaing each checkpoint copy.*

Any suggestions on how to work round this within condor?

my own ideas are:

1) Change the program so output is also sent elsewhere (annying
network overhead)
2) Write a FileSystem watching app that spots a change to the file and
copies that instance away to a new directory with a unique name (bit
of a pain and making sure it runs is an issue, though I could make it
a service or a universe job)

Just wondered if there were any alternates

Thanks,
Matt

* This seems perfectly reasonable thing to do for space considerations
but being able to disable it would be nice