[Condor-users] Rescue DAG and environment

[Brian Candler <B.Candler@xxxxxxxxx>]

I'm not sure if the following is a bug or intentional behaviour, but I
thought I'd report it here anyway.

Suppose you run a dag using

    export FOO=BAR
    condor_submit_dag -import_env foo.dag

The environment FOO=BAR gets included as an environment line in the
foo.dag.condor.sub file, and hence is available to jobs.

Now suppose one of the jobs fails, so you want to run the rescue dag.  From
my reading of the manual, I thought all you had to do was

    condor_submit_dag foo.dag

However in this case, the environment settings are lost. The
foo.dag.condor.sub file appears to be rewritten, and the environment line
has only _CONDOR_DAGMAN_LOG and _CONDOR_MAX_DAGMAN_LOG

So as far as I can see, to restart the DAG you have to reset your local
environment to how it was before, and provide -import_env again to
condor_submit_dag.  There's obviously a risk here that you don't set up the
environment exactly the same.

ISTM it would be useful to maintain the existing .sub file. I notice at
http://research.cs.wisc.edu/condor/manual/v7.8/condor_submit_dag.html there
is an -update_submit option (which implies to me: the submit file should
*not* be updated if you don't give this flag)

I also tried running the rescue dag using

   condor_submit_dag -update_submit foo.dag

but that also wipes out the environment from foo.dag.condor.sub

Regards,

Brian.