currently I do external aerodynamics analysis for my thesis with OpenFOAM. For postprocessing I use paraview 5.4.1
I figured out how to do all the work in the GUI on parallel computing, with e.g. ‘mpiexec -np 40 pvserver --mpi --use-offscreen-rendering’. I connected Paraview to the server and the time to load Data and to switch between values for different field data is now reduced by at least factor 50!! It really is ultra fast now. This works pretty good so far.
I want to use a python script with pvbatch which creates slides in xyz directions, saving views for diffrent fields with mesh visible and without. as .png screenshots to my OpenFoam case folder.
Unfortunatly this process is super slow, because it uses only one! of my 44 cores available to get slices and screenshots done.
So far I tried to run ‘mpiexec -np 40 pvbatch --mpi --use-offscreen-rendering’ . The scripts starts and it creates all the slices as the previous single core run. But this time all 40 Cores are 100% load. The output in the terminal shows me something like:
So the input command changed something on the system, but there is no performance improvement to the process, it still takes a lot of time.
I dont want 40 times the same screenshot but 40 cores creating one screenshot and then the next one.
What can cause the problem? did i get something wrong about the usage for pvbatch? Do I need to connect to a pvserver with a command and execute the pvbatch later?
Can you help me to improve the performance on this? It will help me alot with my thesis.
So I just connectet to the pvserver with np 40, loaded my case, used the ProcessIdscalars, click apply, then I did an extract block to the internal mesh and used the vtkProcessId for visualization. It showed me colored Domain with 40 areas as you can see in the attached picture (DrivAer)
So now I tried to create slices for U magnitude with mesh (Surface with Edges). It shows the new slice instantly when hitting the apply button. I noticed something very strange!. I dont know why but this instant slice update works only for slices normal to y and z axis. When i try to create slices or update the position off the slice normal to x axis, the slice updates slowly and i get the following error message in the output messages window:
ERROR: In /home/buildslave/dashboards/buildbot/paraview-pvbinsdash-linux-shared-release_superbuild/build/superbuild/paraview/src/VTK/Common/DataModel/vtkDataObjectTree.cxx, line 377
vtkMultiBlockDataSet (0x11aee2a0): Structure does not match. You must use CopyStructure before calling this method.
The update is not as slow as with pvbatch but only slightly slower than y and z slices.
Why is that ?
So I tried to recreate this error with other cases but i was not able to do so. First thought was the following: case computet with 39 procs causes this issue when using pvserver with 40 procs, but the error is not showing up again either way, so maybe it was just one time thing.
regarding pvbatch i can send you my script directly or via mail if you want.
