Dear OpenMPI users
I have a problem with openmpi codes hanging in MPI_BCast ...
All our nodes are connected to one LAN. However, half of them
also have an interface to a second private LAN. If the first
openMPI process of a job starts on one of the dual-homed nodes, and
a second process from the same job starts on a single-home, it
will hang on MPI_BCast. It works if the first process is on a
single-homed node, or for all other combinations.
It also works if I disable the private interface. Otherwise there
are no network problems. I can ping any host from any other.
openmpi programs without MPI_BCast work OK.
I guess that this is something to do with openmpi passing IP
addresses between processes. Is there some setting that can be
done to override this?
Has any seen anything like this, or have any ideas??
Thanks in Advance
CODAS & IT Department, UKAEA Culham
Culham Science Centre, Abingdon, OXON, OX14 3DB, UK
Voice: +44(0)1235-46-4569, Fax: 4404
Work email: David.Robson_at_[hidden]
Home email: David.W.Robson_at_[hidden]