[torqueusers] Gaussian 03 and Torque

P Spencer Davis psdavis at bsu.edu
Thu Feb 22 16:11:10 MST 2007


I have recently installed torque 2.1.6 on a cluster running the 2.6.9
kernel RHEL4. I have an extremely minimal default queue set :
Queue batch
        queue_type = Execution
        total_jobs = 0
        state_count = Transit:0 Queued:0 Held:0 Waiting:0 Running:0
Exiting:0
        resources_default.nodes = 1
        resources_default.walltime = 01:00:00
        mtime = Thu Feb 15 14:33:22 2007
        resources_assigned.mem = 0b
        resources_assigned.nodect = 0
        enabled = True
        started = True

and the server has the following config:

Server someserver.org
        server_state = Active
        scheduling = True
        total_jobs = 0
        state_count = Transit:0 Queued:0 Held:0 Waiting:0 Running:0
Exiting:0
        managers = msdaemon at someserver.org,auser at someserver.org
        operators = auser at someserver.org
        default_queue = batch
        log_events = 511
        mail_from = adm
        resources_default.mem = 4mb
        resources_assigned.mem = 0b
        resources_assigned.nodect = 0
        scheduler_iteration = 600
        node_check_rate = 150
        tcp_timeout = 6
        pbs_version = 2.1.6

Everything but Gaussian 03 works fine. When I attempt to submit a
Gaussian job it starts up ok, but fails with the following message as it
hands execution to l1.exe:

l1.exe: bsd/mdutil.c:1703: ntrext1_: Assertion `(lenbuf >= (64 *
statfsbuf.f_bsize))' failed.

The same input, same node runs just fine from ssh or rsh within the
cluster. I'm not getting any error messages in the server or mom logs
(which are still set to the default) and when I've requested e-mail
conformation of the jobs status, I get this:

PBS Job Id: 30.someserver.org
Job Name:   STDIN
Execution terminated
Exit_status=0
resources_used.cput=00:00:00
resources_used.mem=0kb
resources_used.vmem=0kb
resources_used.walltime=00:00:28

So Torque seems to think that everything is fine.

Any ideas? I know that lots of sites are using torque to manage clusters
running Gaussian jobs, but I seem to be missing something here.

              Thanks for the help,
                       Spencer Davis


More information about the torqueusers mailing list