Check robustness of parallel HDF5 by opening/closing each time step

[baagaard-usgs]

We have seen writing HDF5 in parallel to be fragile. Matt reports that opening/closing each time step is more robust.

We should test the robustness of parallel HDF5 with opening and closing the file each time step.

[baagaard-usgs]

@knepley Is there a way to tell the PETSc HDF5 viewer to open or close the HDF5 file or do I have to open and destroy the viewer each time I want the file HDF5 open or closed?

[knepley]

@baagaard-usgs Open and close happens on creation and destruction. The idea is that there is no state to preserve, with the exception of the filename.