Missing rendering in ParaView? Exodus output corrupted?

[mangerij]

Hi,

I am looking at the exodus outputs of a time-dependent simulation in ParaView v5.9.0. At some point yesterday, I got a strange error that printed to the terminal during a MOOSE simulation, killed the simulation, and now every simulation does not visualize properly. I did not save the exact error but I noticed something similar about a month ago. That error is attached in this screenshot:

I am unsure that this is related to the visualization issues, but after this error I was unable to recompile without first using git clean -xfd

As for the visualization, I see the variables at time step 0, but they get cut off somehow after. See the two images below.

I verified this on two different machines that both have ParaView v5.9.0. Another wrinkle to this puzzle is that the exodus outputs from the cluster from the same version of MOOSE/Ferret can be visualized correctly in both installed ParaView applications.

I am worried something got corrupted somehow but I have done a complete recompile and still have the same problem.
thanks

[loganharbour]

Can you run this in dbg and get us a backtrace on that MPI assertion?

[cticenhour]

Might also be worth going backward and trying an older release (5.8 might be sufficient) in order to use the "legacy" reader.

[mangerij]

I can make it fairly small, 8800 dof but then ParaView is reading the information correctly it seems. It seems only when I start to scale it up (nothing crazy to 88000 dof) I encounter the visualization issues. @GiudGiud Please send me your email and I will send the input file

EDIT: I have not tried different versions of ParaView, but I can check this

[GiudGiud]

guillaume.giudicelli@inl.gov

[mangerij]

Small update: @GiudGiud @cticenhour @loganharbour

The visualization error was a mistake on my end. Not sure why this was not showing up before but my mesh moves a little and it can depend on the size of the simulation. Therefore PlotOverLine was glitching because it couldn't plot all of it because the mesh moved.

Regarding the MPI bugs: if they arise again, I will get a representative input file to you guys for valgrinding.

[loganharbour]

Ahh great! This makes much more sense 😆