 |
 |
On Fri, 4 May 2007 [log in to unmask] wrote: > On Fri, 4 May 2007, Alexander Piavka wrote: > > > Maybe the JobWrapper can be changed to checkpoint itself if it fails to > > stagein the input-sanbox and then it fails to stageout output-sandbox. > > Then it would later be rerun by the pbs from a checkpointed stage. > > But what about other batch systems, that do not support that? Most of the sites run torque and sge , which can allow checkpointing as probably lsf does. Maybe Information System can publish if sub-cluster allows checkpointing, and JobWrapper would make use of that info. > The job wrapper cannot count on any fancy features. > Alex