Hi
I frequently face this problem. The simulation is finished at the cluster but the results are not copied to local directory. Now, to recover the simulation, it is suggested that the res file be copied back to scenario folder on local directory and run for zero iteration. However this is not robust and there are following limitations:
1) It says it cant find scd file and hence can't proceed. Sometimes, this can be mitigated by first generating a mesh by starting from 0 iteration and running for 0 iteration. Then copying the res file to folder and then run for the last iteration in res file for 0 iteration, to map the results. And sometimes it just refuses to map, and asks the cfz file to be submitted etc.
2) When it maps the results, you can't see the convergence plots. Thus you never know if it is infact converged.
Inherently you have no other way than just to re run the case.
I have already raised the request for robust cluster integration here. But it is unfortunately lost in many "hot ideas" with umpteen kudos(no offence, but many of these request merely cosmetic changes that have no bearing on performance!). But if that is too much to ask, at least there should be a robust workaround to retrieve the results.
Cluster integration is a non-negotiable aspect of a simulation software. Fortunately many of my simulations are steady and incompressible, so a day or 2 lost in re-running is bearable. But I can't imagine the outrage it may cause, if we were to do compressible transient analyses...
Hope someone is listening...
Omkar.
