Mismatch between .pvtu Pieces count and MPI visualization rank count

ParaView Support
1

Hi Everyone,

We have a solver that produces .pvd timeseries data where each element is a .pvtu that links together a bunch of pieces as _n.vtu that are output independently by each rank.

When we visualize this from one process everything works great, and when we launch a distributed paraview environment where the number of ranks is the same as the number of Pieces it also works fine.

The problem we have is when we try and launch a distributed paraview environment that has a mismatch between the number of ranks and the number of pieces in the .pvtu. Is this perhaps a FAQ for this kind of situation? Do we need to reprocess the entire timeseries input to distribute it to Paraview’s liking?

Any advice would be appreciated.

Thanks,
Luke

2

While ParaView will not redistribute accross more nodes unless you ask it to (Using RedistributeDataSet filter), it should still work, just not using all your nodes.

What problem do you encounter ?

Is this perhaps a FAQ for this kind of situation?

Way too advanced for FAQ, but part of the docs:
https://docs.paraview.org/en/latest/ReferenceManual/parallelDataVisualization.html

3

Thanks Mathieu,

I think we must have had some filter in the sequential pipeline that was broken for distributed. After rebuilding the pipeline from the ground up it seems to be working as expected.

So most likely “user error” somewhere.

Thank you,
Luke

1 Like