When a run aborts

Details of TRANSP Processing

From the user's perspective, the process of running transp has several steps:
  pretr - initial processing of the namelist
  trdat - read in input data from the mdsplus tree, check for 
          inconsistencies and process the data for transp
  transp - actual transp run for computing the evolution of the plasma
  mdswrite - write the results back to the user's mdsplus tree
  finishup - copy any other result files to their final resting place

This process can break at any stage. If the job submittal dies in pretr due to a namelist formatting error or trdat because of input data problems it will show up as a stopped job in pretr and the user should receive an email describing what caused the failure. Because people do not generally reuse or delete jobs which fail in pretr, we will take a quick look at it, make sure it is not a problem with the code then delete it. If there is not enough information in the emails to resolve the problem you should let us know. There is not much we can do with a run which does not succeed at this stage.

If you get through trdat, then the job will start running transp. At this point, you can view how transp processed your input and you can monitor the run from the fusion grid monitor web page (click on your runid, click on view data, click on input data or monitor data -- there are ways to customize this). If the run crashes during transp it will show up as a stopped job in runtrx. We usually look at these more closely since it could possibly be caused by a bug in the code. If it is a bug then we can sometimes fix it and restart your job. If it crashed because of physical causes, we will ask if you want the job deleted or completed as though it had finished successfully.

If you make it through transp, the output data will be written back to your mdsplus tree. This can fail if the mdsplus server is down or the globus certificate has expired. Put a long expiration time on your globus certificate in case the run crashes and it takes a while for us to get it going again.

To look at output

Aborted runs are automatically written into MDSplus

TRANSP Production System
Last modified: Tue Nov 30 18:01:06 EST 2010