[mpich-discuss] mpdboot handshake problem

[Qi Ying]

Hi All,

Recently I had trouble starting mpd using mpdboot on my cluster. This
seems caused by a single node in the system (n002). The following is
the debugging output. However, I can manually start mpd on n001 and
n002 (and have them join the ring), and there is no problem. Any
insights or suggestions?

Thanks,

Qi Ying

[qying at n001~] $ mpdboot -n 2 -f ~/mpd.hosts --rsh=/usr/bin/rsh -v -d

running mpdallexit on n001.newwolf.edu
LAUNCHED mpd on n001.newwolf.edu  via
debug: launch cmd= /opt/mpich2/bin/mpd.py   --ncpus=1 -e -d
debug: mpd on n001.newwolf.edu  on port 55059
RUNNING: mpd on n001.newwolf.edu
debug: info for running mpd: {'ncpus': 1, 'list_port': 55059,
'entry_port': '', 'host': 'n001.newwolf.edu', 'entry_host': '',
'ifhn': ''}
LAUNCHED mpd on n002  via  n001.newwolf.edu
debug: launch cmd= /usr/bin/rsh -n n002 '/opt/mpich2/bin/mpd.py  -h
n001.newwolf.edu -p 55059  --ncpus=1 -e -d'
debug: mpd on n002  on port 54319
mpdboot_n001.newwolf.edu (handle_mpd_output 385): failed to handshake
with mpd on n002; recvd output={}