parallel, remote server fails (and soon faults) rendering vts/ pvd files
This issue was created automatically from an original Mantis Issue. Further discussion may take place here.
I have a dataset that is showing crazy behavior with ParaView in parallel. Here is how to replicate:
-
Linux, master, Local server.
-
Open G1WakeGrid.pvd. Apply. There are 6 timesteps. (Note that the .pvd file has many more timesteps. Don't go beyond about step 4, or you get an error from the inconsistent .pvd file.)
-
Step forward a few timesteps. Step back. <> This case looks good.
-
Linux, master, Remote server, 2 processes.
-
Open G1WakeGrid_0.vts. Apply. <> This case looks good.
-
Open G1WakeGrid_1.vts. Apply. <> This case looks good.
-
Linux, master, Remote server, 2 processes.
-
Open G1WakeGrid.pvd. Apply.
-
Step forward one timesteps.
<> This case looks bad. Half of the model will go away. If you go to the information tab, every step forward or backward will halve the number of cells/ points in the dataset. Just step forward/ back to see this crazy behavior.
This will eventually progress until there is a seg fault, trying to malloc a few trillion bytes.
Utkarsh, ask me for this dataset.
Alan - G1WakeGrid, gpfs1.