[ET Trac] [Einstein Toolkit] #618: damiana setup in simfactory is wrong

Einstein Toolkit trac-noreply at einsteintoolkit.org
Fri Oct 14 10:57:50 CDT 2011


#618: damiana setup in simfactory is wrong
----------------------------------------+-----------------------------------
  Reporter:  baiotti@…                  |       Owner:  eschnett         
      Type:  defect                     |      Status:  closed           
  Priority:  blocker                    |   Milestone:  ET_2011_10       
 Component:  SimFactory                 |     Version:                   
Resolution:  invalid                    |    Keywords:  error in database
----------------------------------------+-----------------------------------

Comment (by eschnett):

 Most of the space is used by the configurations. I estimate that a
 configuration uses about three to four times as much space as a source
 tree, if you exclude test results.

 Simfactory already ensures that a configuration is not used in the wrong
 way. Once you created a configuration, you can rebuild, submit, etc., and
 it will automatically remain self-consistent and will always submit to the
 same machine. You can manually (and possibly accidentally) change parts of
 the configuration, but that is not different from what you can to in any
 other situation. If you re-configure, Simfactory will notice this, and
 will enforce rebuilding everything from scratch.

 We can make Datura the default. You may even be able to do that by
 yourself with the command

 {{{
 echo datura > ~/.hostname
 }}}

 The underlying problem is that there are two separate machines sharing the
 same head node. Damiana/Datura are unique in this respect. Can this be
 changed? Can there be a second head node? Isn't there a second head node
 anyway that could be used? Don't you have that problem for every
 executable, having to remember for which machine you built it?

-- 
Ticket URL: <https://trac.einsteintoolkit.org/ticket/618#comment:6>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit


More information about the Trac mailing list