[MPICH] Heterogenous cluster MPICH2

Philip Sydney Lavers psl02 at uow.edu.au
Mon Jun 20 21:10:22 CDT 2005


I am experimenting with extending the current cluster of four
AMD athlon64/opteron  boxes under FreeBSD.

Paul1 is one of these 


proton is a sun sparc64 machine with Solaris9 and I just put
python and MPICH2 onto it.

Here is what happens:

proton:
bash-2.03$ mpd &
[1] 1302
bash-2.03$ mpdtrace -l
proton_33368

Paul1:
-bash-3.00$ mpd -h proton -p 33368 &
[1] 18549
-bash-3.00$ Paul1_65089 failed ; cause: invalid challenge msg: {}
    traceback: [('/usr/local/bin/mpd', '1158',
'_enter_existing_ring'), ('/usr/local/bin/
mpd', '175', '_mpd_init'), ('/usr/local/bin/mpd', '1398', '?')]

and,
proton:
bash-2.03$ proton_33368 (_handle_rhs_challenge_response 1010):
INVALID msg for rhs response msg=:{}: from host=Paul1 port=65089

proton runs single node jobs OK. Similar things happen if I
start mpd on Paul1 first as host.

Can anyone suggest what is going wrong?

regards,

Phil Lavers




More information about the mpich-discuss mailing list