[ET Trac] [Einstein Toolkit] #860: Don't rmtree(scratchdir)

[Einstein Toolkit]

#860: Don't rmtree(scratchdir)
---------------------------+------------------------------------------------
 Reporter:  barry.wardell  |       Owner:  eschnett
     Type:  defect         |      Status:  new     
 Priority:  blocker        |   Milestone:          
Component:  SimFactory     |     Version:          
 Keywords:                 |  
---------------------------+------------------------------------------------
 The mdb in SimFactory has an entry for scratchdir. If scratchdir is a
 relative path, it is assumed to be relative to the restart output
 directory. It's not clear exactly what the purpose of this is, but
 SimFactory does the following things with it:

 1. Create the directory at the start of a simulation if it doesn't exist.
 2. Delete the entire directory using rmtree(scratchdir) during "cleanup".
 This happens, for example, when an existing simulation is restarted.

 The second of these seems very dangerous to me and I think it should be
 removed. In particular, if you happen to have scratchdir set to be the
 same for more than one simulation (by setting it to, e.g., /scratch), this
 scrachdir for all simulations will be erased as soon as one simulation
 does a cleanup.

 I recently had an even worse experience where I was using /scratch as my
 basedir and /scratch also happened to be set as the scratchdir in the mdb
 for that machine. All of my simulations on that machine had only a single
 restart, so all was well until I tried to restart one simulation and
 instead of doing so SimFactory erased all of my simulations on the
 machine.

 While I can admit to user error in this case (I also wrote the machine
 entry), I think the goal of SimFactory should be to err on the side of
 caution and shouldn't do this blind rmtree of whatever scratchdir is set
 to.

-- 
Ticket URL: <https://trac.einsteintoolkit.org/ticket/860>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit