Open MPI / MPICH - What happens if a node terminates?

Question

I would like to know what happens if an OpenMPI / MPICH2 cluster node terminates? Is there some kind of mechanism that is tolerant of this case and continues to execute?

Thanks for your answers Heinrich

Answer 1

Note that the function that existed since MPI 1.x is that you can install an error handler: for example,

http://www.mpi-forum.org/docs/mpi-11-html/node148.html

, MPI_ERRORS_ARE_FATAL ( ), ( , ).

; MPI .

MPI - http://icl.cs.utk.edu/ftmpi/ ( MPI 1.2). http://osl.iu.edu/research/ft/cifts/ , OpenMPI , / , BLCR, .

MPI-3 API MPI, .

Answer 2

, MPI . , , , , , . .