Could anybody explain the relationship between Catalyst Live and offscreen rendering

Now I use off screen rendering and in-situ visualization in ParaView 5.4.1 in my Regional Earth System Simulation, in which MPI parallelization was used. I get confused about the rendering process and mpi_abort happened. The NVIDIA Quadro GPU card is installed in my computer. So I need your help because I am green hand at using the related functions in ParaView.
(1) I compiled ParaView with EGL and without X11 and Qt5-GUI, the necessary Catalyst files were configured well in the tutorial. When I run the executable using only 1 process, the atmosphere model runs normally, but the in-situ animation cannot come out in the screen? How should I compile Paraview with the necessary Cmake switch? I don’t understand the mechanism of off screen or headless rendering.
(2) My computer has one Intel CPU with 8 cores and has one NVIDIA GPU card, when I run mpirun -np 8 ./regesm, can the Catalyst Live work? How can I choose the Qt5 client GUI, VTK_USE_OFFSCREEN, VTK_USE_EGL,…?
(3) If I have a cluster with 2 node, every node was installed with one CPU having 8 core and one NVIDIA GPU card. How can I install paraview in every node? Should I install ParaView with off screen rendering (VTK_USE_EGL) on every node?
I hope the expert can explain for me, this will help me!

This may help:

The live visualization should be enabled. Please check the Catalyst examples in the python script for command like : coprocessor.DoLiveVisualization(datadescription, "localhost", 22222)