A job configuration file provides a description of the job to the job manager. Such a description includes the number of tasks to create, the path names of the executables, command line arguments to the tasks and some additional information described below. A job configuration file is essential for every job. Configuration information is specified as attribute-value pairs and takes the form:
attribute = < value >
attribute = < value >
....
....
More than one attribute-value pair may be specified on a single line. For example a config file containing
t = foo a = -opt1 a = arg
n = 3
d = /home/srao/prm/bin
c = false
s = zephyr.isi.edu
tells the job manager to start up 3 copies of the executable foo
residing in the directory /home/srao/prm/bin/$ARCH, with
command line arguments -opt and arg, on nodes acquired
from the system manager running on host zephyr.isi.edu and that
the tasks cannot be checkpointed/migrated.
Note that the $ARCH component of the pathname was not specfied in the config file. PRM assumes that all executables reside in a directory whose name is the same as the architecture name of the host machine. Therefore you must place sparc executables in a directory sun4 and HP pa-risc executables in a directory hp700.
Some examples of job configuration files can be found in $PRM_TOP/demo. A complete listing of attributes and the values they take are provided in the man pages for jobmngr.
Starting with Version 1.1, the default configuration of the node manager expects all application binaries to reside in $PRM_TOP/bin/$ARCH. This places restrictions on the set of applications available to a remote user. The PRM administrator can override the restriction by starting node managers with the -r option. As this option also allows remote users to transfer programs from remote locations and execute them on the node, the administrator is advised to consider security implications before running node managers with the -r option.