[Lukasz]

Hi,

I am running CFX on 64 A8 machines (=512 cores) on Microsoft Azure as follows:

\\hsransys.file.core.windows.net\main\apps\ANSYS Inc\v150\CFX\bin\cfx5solve" -def "%def_file%" -chdir "%workdirpath%" -start-method MSMPI2 -part %num_of_partitions%

and get the following error:

job aborted:
[ranks] message

[0-191] terminated

[192] fatal error
Fatal error in PMPI_Bcast: Other MPI error, error stack:
MPI_Bcast(buf=0x00007FF699BAC240, count=5000, MPI_PACKED, root=0, MPI_COMM_WORLD) failed
[ch3:nd] Send to 172.16.1.213:7 completed in error with 0xc00000b5

[193-511] terminated

---- error analysis -----

[192] on 10.10.13.159
mpi has detected a fatal error and aborted \\hsransys.file.core.windows.net\main\apps\ANSYS Inc\v150\CFX\bin\cfx5remote.exe

---- error analysis -----

We assume that the after MPI BROADCAST took too long and there should be a timeout parameter in the MPI stack that can be tuned so that the time needed for MPI Packets to come back can be extended?

Any suggestions how to set it up will be helpful.