When I try to run multinode MPPR, PAR issues a message that says it is starting a job on "node"; however, it seems to hang. When the process is cancelled, the following message appears:
"PING cannot reach node sandy. This message occurs when either the network or the node is down. This node is being removed from the active list. PAR done."
This happens because the Turns Engine uses PING to ensure the node is accessible. The software is hard-coded to expect a message back that is similar to:
NODE_NAME is alive
However, some UNIX systems may be set up differently, such that PING's default uses some options. An example of such an occurrence might be if PING is aliased to be "PING -S." This actually returns packet information and causes the Turns Engine to fail.
To avoid this, make sure that PING returns "NODE is alive" and no other information.