I threw this together fairly quickly after doing a quick setup on
a spare machine.  The single system case should work exactly as
shown, I just copied the output I got.  Multiple systems I didn't
have available to copy output, but they should work as well.

Note that Torque is really picky about the hostname.  If you
get errors during pkg_config() set PBS_SERVER_NAME in
/etc/make.conf to resolve to your machine.

Thanks for testing :)




Testing Torque on a single system.
    You'll need USE="server" to build pbs_server and pbs_sched.
    Emerge and use the pkg_config().  This sets up enough that
    you may start the server, mom and sched processes and begin
    to submit jobs.

== Verify the batch queue now exists.
jbronder@serac ~ $ qstat -q

server: serac.cold-front

Queue            Memory CPU Time Walltime Node  Run Que Lm  State
---------------- ------ -------- -------- ----  --- --- --  -----
batch              --      --       --      --    0   0 --   E R
                                               ----- -----
                                                   0     0


== Verify your local machine is available.
jbronder@serac ~ $ pbsnodes $(hostname)
serac
     state = free
     np = 1
     ntype = cluster
     status = opsys=linux,uname=Linux serac 2.6.23-gentoo-r3 #1 Wed Jan 9 07:21:15 EST 2008 x86_64,sessions=5051,nsessions=1,nusers=1,idletime=3848,totmem=1220396kb,availmem=929576kb,physmem=963364kb,ncpus=? 15201,loadave=0.00,netload=115122588,state=free,jobs=,varattr=,rectime=1207296315


== Start an interactive job on one node.  (Cannot be done as root)
jbronder@serac ~ $ qsub -I -lnodes=1:ppn=1
qsub: waiting for job 5.serac to start
qsub: job 5.serac readyA
jbronder@serac ~ $ exit
qsub: job 5.serac completed


== Start a normal job, verify it's in the queue and that it gives
== output.  We also verify the tm interface with pbsdsh.
jbronder@serac ~ $ cat pbs.test 
#!/bin/bash
#PBS -N testjob
#PBS -l nodes=1:ppn=1
#PBS -o output
#PBS -e errors

sleep 20

# Get hostname from all nodes.
pbsdsh hostname

# Get some output in the error file.
echo "look ma, errors" 1>&2

jbronder@serac ~ $ qsub pbs.test
4.serac
jbronder@serac ~ $ qstat
Job id                    Name             User            Time Use S Queue
------------------------- ---------------- --------------- -------- - -----
4.serac                   testjob          jbronder               0 R batch          

jbronder@serac ~ $ sleep 20
jbronder@serac ~ $ cat output 
serac
jbronder@serac ~ $ cat errors 
look ma, errors
jbronder@serac ~ $ 



Testing Torque with multiple machines.
    Setting up the server:
       You'll need USE="server" only on this machine.  Use
       pkg_config() again to get a basic setup.  Then modify
       /var/spool/torque/server_priv/nodes and add the hostnames
       of your additional machines and the number of cores they
       have.  For instance, node1 with 2 cores and node2 with 4
       cores would look like:
       
       node1 np=2
       node2 np=4

       The number of cores should be reflected in the pbsnodes
       output that is checked later.

       Start pbs_server and pbs_sched.

    Setting up the nodes:
        Emerge without the server flag.  Use pkg_config() and
        modify /var/spool/torque/mom_priv/config and change
        $pbsserver to point to your server machine.  Start 
        pbs_mom.


    At this point you can do the same tests as were done in the
    single machine case.  The only difference is you may now
    increase the number of nodes you request (X) and the number
    of processors per node (Y) in the resource string
    "nodes=X:ppn=Y".