2 Replies Latest reply: Mar 5, 2012 8:22 AM by pgac
pgac Level 1 Level 1

I used to be able to ssh on my Mac Pro 10.6.8 in school and submit jobs (running flow simulations) using the batch command.  The batch command starts a computational fluid dynamics code that is supposed to run for a number of iterations (typically this takes hours). 

 

When I submit the batch jobs now, the runs reach the number of iterations specified but then the run starts again.  The run is executed three times.  I tested this on different MacPro computers I have. 

 

I also tested with "at now" instead of "batch".  The result was the same:  the run was executed three times.

 

If I ssh and start the runs without batch, all is running fine.

 

I tested a small batch submission using line commands and the file batchtest:

 

ssh -X user@machine

batch

batchtest

control-D

 

where batchtest is:

 

ls -l codes

df -k > disks1

 

This worked fine, it did not loop.

 

I would appreciate any suggestions.

 

Thank you,

 

Paul


Mac Pro, Mac OS X (10.6.8)
Reply by twtwtw on Mar 4, 2012 9:31 PM Helpful

well, I think you've demonstated effectively that the batch utility itself is not at fault.  That implies something in the script you're batching (or the program you are running) has gone south.  Something obviously changed between the time it worked and the time it didn't - what might that be?  does the batch program work correctly when you run it directly rather than via ssh?

All replies

  • twtwtw Level 5 Level 5

    well, I think you've demonstated effectively that the batch utility itself is not at fault.  That implies something in the script you're batching (or the program you are running) has gone south.  Something obviously changed between the time it worked and the time it didn't - what might that be?  does the batch program work correctly when you run it directly rather than via ssh?

  • pgac Level 1 Level 1

    The problem is caused by the compiler I use.  I tried another compiler (gfortran) and the batch ends as it is supposed to.  The developers of the compiler that causes this issue were able to duplicate the problem and are working on fixing it.