Re: [HTCondor-users] Help with condor_wait (stuck in Windows using HTCondor 8.4.5 and 8.4.6)

[Martà Coma Company <mcoma@xxxxxxxxxxxxx>]

Hello,

We have been testing further and found out that if we call "condor_wait 
log.condor" in a loop for each job, the condor_wait command works as 
expected every time. To do that we edited the submission file:

from: Log    = ../log.condor
to:   Log    = log.condor

We understand that may be some error in the condor_wait command when 
working with log files containing more than one job in Windows.

Someone could tell us the proper way to address this issue?

Thanks,

Marti Coma
CIMNE Aerospace Engineering Group

> Date: Tue, 17 May 2016 18:05:52 +0200
> From: Mart? Coma Company <mcoma@xxxxxxxxxxxxx>
> To: htcondor-users@xxxxxxxxxxx
> Subject: [HTCondor-users] Help with condor_wait (stuck in Windows
> 	using HTCondor 8.4.5 and 8.4.6)
> Message-ID: <ef2d4c4d7008e02601b6cd289c58dbcf@xxxxxxxxxxxxx>
> Content-Type: text/plain; charset=UTF-8; format=flowed
>
> Hello,
>
> We are new to HTCondor, have been configuring a little pool since the
> last course in Barcelona. The tests we have conducted submitting jobs
> manually have been successful, now we are trying to interact with
> HTCondor from an in-house optimization code written in C++. We have
> encountered problems running in Windows 7 (so far, in Linux is working
> as expected).
>
> The problem is that after the "condor_submit submit.condor" system call
> from our code, we call "condor_wait log.condor" and SOMETIMES it gets
> stuck in the condor_wait (we call the submit and wait commands in a
> loop). The log.condor shows that all jobs are terminated, condor_q
> returns no jobs at all and the results of all calculations are there 
> and
> correct. If we kill the condor_wait from the task manager the process
> continues without problems, until it gets stuck again in another loop
> iteration. We have been waiting for several hours for condor_wait to
> return.
>
> We use initialdirs and relative paths in the submit file, so all jobs
> are logged in the same file:
>
> ################################
> #       Condor submit file     #
>
> Universe   = vanilla
>
> Executable = mathcasesexe_windows.exe
>
> Log    = ../log.condor
> Output = out.condor
> Error  = err.condor
>
> initialdir = condorInd$(Process)
>
> should_transfer_files = YES
> transfer_input_files = Eval.DVs, ../prob.dat
> when_to_transfer_output = ON_EXIT
> transfer_output_files = Eval.individual, Cons.individual
>
> Queue 24
> ################################
>
> We have tested both HTCondor 8.4.5 and 8.4.6 in the submission node 
> with
> the same issues. We have also tried to delete the log.condor file
> between loop iterations, but the problem remains.
>
> We found the same problem for a Linux user in version 7.2.5 and solved
> in version 7.4.3 in 2010
> (https://lists.cs.wisc.edu/archive/htcondor-users/2010-June/msg00221.shtml).
> Could be a bug in condor_wait for Windows?
>
> How can we solve this problem? Any help would be much appreciated.
>
> Thanks in advance,
>
> Mart? Coma
> CIMNE Aerospace Engineering Group
>
> End of HTCondor-users Digest, Vol 30, Issue 14
> **********************************************