[torqueusers] Abaqus with PBS

Ashley Wright a2.wright at qut.edu.au
Mon Sep 12 04:10:31 MDT 2005


Hi Josh,

It is my experience that the compute nodes MUST be able to access the license server. We have set up NAT on the master nodes. (if you don't know NAT does some fancy things which allow a private network to talk to a public network). We have many software packages which use licence servers, and this is the only method I know of apart from installing the license server on the master node.

Hope this helps,
Ashley

  ----- Original Message ----- 
  From: group hpc 
  To: Ashley Wright 
  Sent: Monday, September 12, 2005 5:03 PM
  Subject: Re: [torqueusers] Abaqus with PBS


  Hi Ashley,

  Thanks for the info. We still unable to get the abaqus to work, I believe the problem is due to the compute nodes (Slave nodes) not be able to access to the license server directly. Only Master node able to access to the license server. Our clusters also have shared file system, Abaqus is installed in there and available for all nodes. Do I need to configure other things/environmet instead of "abaqus_v6.env"? I'm attached the following error from the abaqus output. 

  ------------------------------
  Server node is down or not responding
  See the system adminstrator about starting the server, or
   make sure the you're referring to the right host (see LM_LICENSE_FILE)
  Feature:       explicit
  Hostname:      cluster 
  License path:  27001 at cluster
  FLEXlm error:  -96,482
  For further information, refer to the ABAQUS Installation and Licensing Guide,
  or contact your ABAQUS Inc. representative. 
  -------------------------

  Pls let me know if you have an idea to fix this. Thank you for your time. :-)

  Best Regards,
  Josh

   
  On 9/12/05, Ashley Wright <a2.wright at qut.edu.au> wrote: 
    Hi Josh,

    I set up Abaqus (6.5-4) on our cluster last week. The first thing is the
    nodes in our cluster have a shared file system, we use for software 
    packages (for us this is called /pkg). Abaqus is installed here and so
    is available on all the nodes. We also have NAT setup on the master
    nodes so that Abaqus can access the license server (which is on another
    network).

    Assuming this is set up you need to create a site file 'abaqus_v6.env'
    most of the directives are specified in the manual but here are the
    important one for torque:

    #Set up your MPI environment, this is for LAM which we use at QUT. 
    #We also use ssh instead of rsh.
    mp_mpi_implementation = LAM
    mp_mpirun_path = {LAM:'/usr/local/bin/mpiexec',
                     MPICH:'/usr/local/mpich-1.2.5/bin/mpirun',
                     MPICHGM:'/usr/local/mpichgm- 1.2.6..14a/bin/mpirun.ch_gm'}
    mp_mpirun_options = '-boot'
    #mp_rsh_command = 'rsh -n -l %U %H %C'
    mp_rsh_command = 'ssh -x -n -l %U %H %C'

    #this is the location of the license server
    abaquslm_license_file=" 1712 at sirius.qut.edu.au"

    doc_root="http://auriga.qut.edu.au/abaqus/"

    #
    # QUT batch queue setup; abaqus_2cpus is the default queue 
    # (Using bash as the PBS interpreter is better than tcsh - fewer probs!)
    #
    run_mode = BATCH
    scratch  = "/tmp"


    max_cpus=8
    cpus=2


    queue_name=["abaqus_2cpus","abaqus_4cpus","abaqus_8cpus"] 
    queue_cmd="qsub -r n -q abaqus_nolimit -S /bin/bash -V -l nodes=1:ppn=2 %S"
    abaqus_2cpus="qsub -r n -q abaqus_nolimit -S /bin/bash -V -l
    nodes=1:ppn=2 %S"
    abaqus_4cpus="qsub -r n -q abaqus_nolimit -S /bin/bash -V -l 
    nodes=2:ppn=2 %S"
    abaqus_8cpus="qsub -r n -q abaqus_nolimit -S /bin/bash -V -l
    nodes=4:ppn=2 %S"



    Using this setup the user submits a job (from the master node) like so:
       'abaqus job=pillar queue=abaqus_8cpus cpus=8' 

    I hope this helps,
    Ashley



    group hpc wrote:

    > Hi,
    >
    > Does anyone know how to setup Abaqus application to work with Torque
    > and Maui in cluster environment? We are still struggling to get abaqus 
    > work on our cluster. We have installed abaqus on our cluster (Master
    > Node), but will grab the license from other license server. The job
    > able to run directly without go thru scheduler. However, when we tried 
    > to submit the job thru scheduler, the job won't run due to the license
    > problem. Does anyone know howto resolve this problem?
    > //
    > Thanks,
    > Josh
    >
    >
    >------------------------------------------------------------------------ 
    >
    >_______________________________________________
    >torqueusers mailing list
    >torqueusers at supercluster.org
    > http://www.supercluster.org/mailman/listinfo/torqueusers
    >
    >


    --
    Ashley Wright
    3864 9264
    a2.wright at qut.edu.au
    HPC and Research Support Group
    Queensland University of Technology (QUT)





  -- 
  Josh 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20050912/28b850a9/attachment-0001.html


More information about the torqueusers mailing list