netcdf5 Exodus files failing in parallel
Some Sierra codes are moving from netcdf4 to netcdf5, used to make Exodus datasets. Netcdf4 works properly with ParaView. Netcdf5 works properly builtin server, but fails remote server. I am including the netcdf4 file for comparison. As this blocks Sierra from moving to netcdf5, I'm marking as a showstopper. This will also require us to make a point release for Sandia this fall.
Here is how to replicate.
- Linux, 5.13.1, builtin server.
- Open netcdf5_simple.e. This opens properly.
- Linux, 5.13.1, remote server.
- Open netcdf5_simple.e. This throws error messages and the dataset won't open.
Message is as follows: ERROR: In vtkIOSSReaderInternal.cxx, line 328 vtkIOSSReader (0xd9b4ee0): Error in UpdateTimeInformation: Failed to open database /snip/netcdf5_simple/result.e
ERROR: In vtkExecutive.cxx, line 729 vtkPVCompositeDataPipeline (0xef5f9f0): Algorithm vtkIOSSReader (0xd9b4ee0) returned failure for request: vtkInformation (0xef72560) Debug: Off Modified Time: 151344 Reference Count: 1 Registered Events: (none) Request: REQUEST_INFORMATION ALGORITHM_AFTER_FORWARD: 1 FORWARD_DIRECTION: 0
Looking at the code, my wild guess is that there are different paths for builtin server and remote server for UpdateTimeInformation. The remote server one is breaking with the slightly different code path processing the netcdf5 file?
Passing the files to Cory. OK to pass to anyone at Kitware. @spiros.tsalikis