MPICH error messages
The following message appears in the errlog file when launching my job in parallel:
mpdrun_wilkes.jpl.nasa.gov: cannot connect to local mpd (/tmp/mpd2.console_name);
possible causes:
1. no mpd is running on this host
2. an mpd is running but was started without a "console" (-n option)
~
~
This message means that the MPI (Message Passing Interface) server, called mpd, is not running.
Therefore, no parallel jobs can run on the cluster. To solve this issue, just type, at the command
prompt on the server side (if for example your cluster has 8 cpus):
mpd --ncpus=8 &
This will launch the MPI server to manage 8 cpus on the cluster.
