http://www.notur.no/news/archive/20100113.html

PDC machines (Beskow and Tegner)

[7 Oct 2015, Illa]

Saving data

These machines share the same filesystems, AFS and CFS.
  • The homes are mount on a AFS filesystem (/afs), which is reachable from any computer without even having to login in the machine.
    More info on AFS here.
    The quota in this file system is only 500 Mb, but it is backup.
  • Lustre ( /cfssystem) is meant for the actual computing and big data storage.
    It has two different partitions, both of them without backup
  • Post-processing data

    They strongly encourage users to avoid using the login nodes for any kind of post-processing. Instead, one should use one of the nodes of Tegner of Beskow.
    In both cases the allocation commands are the same, but Tegner has a different allocation system that does not go through slurm and it is allowed to run scripts for even 10 days there. Also Tegner has nodes with much more ram, in case you need to work with big files.
    Requesting allocation:
    salloc --nodes=1 -t 1:00:00 -A 2014-11-34 will allocate time in 1 node for only 1 hour. You can go up to 10 days in Tegner, only 24 hours in Beskow.
    Some special features for the allocation:
  • Sometimes you have to wait a lot for getting an allocation time, so it is also possible to set the starting time. So, if you run the command the night before, you will probably be allocated in the morning.
    salloc --begin=9:00:00 --nodes=1 -t 8:00:00 -A 2014-11-34
    Ask for 1 node, for 8 hours, starting at 9 in the morning.
  • Asking for a specific type of node. If you need a node with a big amount of ram memory to work with big data, you can ask for this specific type of node:
    salloc -t 1:00:00 -N 1 --mem=1000000 Ask for 1 node with 1 TB of ram.
    It is also possible to specify the minimum number of cpus (--mincpus), and so on.
  • Using the allocated time
    Once you are granted the allocation, you will get a message similar to:
    salloc: Granted job allocation 9097
    salloc: Waiting for resource configuration
    salloc: Nodes t03n16 are ready for job
    In this case, I got the node t03n16 (from now on, change the numbers with your allocated node).
    You can use this node in different ways:
  • Running a script non-interactively: example
    aprun -n 1 ./src/read_all_videofiles.x
    In this case you do not need to specify the name of the node. The script will run non-interactively there
  • Running a script interactively, i.e. idl. In this case you have to login into the node:
    ssh -X t01n01.pdc.kth.se Go to the directory and start your idl session.
  • Important remarks
  • The nodes CANNOT read the home directory. Only the files in the Public folder.
    You can go around this by linking your configuration files, like .cshrc, or placing everything in the /cfs partition.
    What works better for me is linking configuration files from the home and having all idl related programs and stuff in the /cfs part.
    Anyway, you have to be sure your path is correct.
  • When you are allocating a node, you have the whole node for yourself. It means all the processors. If you want to run some non-mpi programs, you can just login into the node several times and run each program in one ssh session. I.e. you can run several ./src/read_all_videofiles.x or idl sessions, but you have to be careful with the ram memory usage.
  • Changing the cvs repository, etc

    [18 Apr 2015]

    To check out the new pencil code web pages, say:
      svn checkout https://github.com/pencil-code/website/trunk/ www
    

    [8 Mar 2014]

    CVS fresh checkout
    If you want to set it from fresh, say
      setenv CVSROOT :pserver:$USER@130.242.129.93:/var/cvs/brandenb
    
    In bash, you'd say
      export CVSROOT=:pserver:$USER@130.242.129.93:/var/cvs/brandenb
    

    [3 Nov 2013, 17:00]

    We are now almost back to normal with the cvs repository. For those who need to switch back from 52 to 51, just say:
      change_CVS :pserver:$USER@130.242.129.93:/var/cvs/brandenb
    
    The only problem is that the email notification does not yet work.

    [3 Nov 2013]

    During today's afternoon, Wolfgang plans to move CVS back to the old machine. This means that at some moment, CVS on 52 will simply become unavailable, and, after additional notification, 51 will be available. For those who never switched, it should work just like before 11 October. For the others, they just need to follow revised instructions that I will post later this evening on this website. If people have changes that are not checked in, there should be no problem either. They should be able to check in the changes when they have issued the corresponding change_CVS command. More details will follow after the change.

    [16 Oct 2013]

    It takes longer getting norlx51 back up. We have new disks, but the machine doesn't boot correctly. In the interim, we'll use /Users/brandenb/backup on 130.242.129.60 and when norlx51 is back up, we'll copy the content of /Users/brandenb/backup to its normal location.

    [12 Oct 2013]

    Our usual cvs machine is currently not booting up (we were in the progress of installing new disks). We hope to get it up early in the week. If you need to use cvs in the mean time, you can use it from the backup. But be careful: changes that you checked in on 11 Oct 2013 between 14:48 and later that day are not included. If this is the case, "cvs up" (even after fixing the CVSROOT) may result in "cvs [update aborted]: could not find desired version 1.157 in /Users/brandenb/backup/...", so move those files into a new temporary directory (mkdir 1; mv paper.tex 1), then say "cvs up", and finally "mv 1/* ." and then "cvs diff" and "cvs ci".

    To change your CVSROOT, say:
      change_CVS :pserver:$USER@130.242.129.60:2401/Users/brandenb/backup
    
    If you want to set it from fresh, say
      setenv CVSROOT :pserver:$USER@130.242.129.60:2401/Users/brandenb/backup
    
    In bash, you'd say
      export CVSROOT=:pserver:$USER@130.242.129.60:2401/Users/brandenb/backup
    
    Of course, you'd need to say "cvs login" on each machine once.

    [23 Mar 2010]

    To print from norosx52, note that the printer name is now Nordita_Xerox_Astro_Building, so you should set one of the two (for bash and tcsh, respectively):
      export PRINTER=Nordita_Xerox_Astro_Building
      setenv PRINTER Nordita_Xerox_Astro_Building
    

    [22 Jan 2010]

    If you don't have it already, get the file change_CVS, put it in your bin (so that it is in your path), make it executable, and say in each old CVS tree:
      change_CVS :pserver:$USER@130.242.129.93:/var/cvs/brandenb
    
    provided $USER corresponds to your userid on cvs. If you want to set it from fresh, say
      setenv CVSROOT :pserver:$USER@130.242.129.93:/var/cvs/brandenb
    
    In bash, you'd say
      export CVSROOT=:pserver:$USER@130.242.129.93:/var/cvs/brandenb
    

    Setting up your printer

    Printing help on Nordita help computing pages

    Notes written by Andrew Duncan while visiting Nordita (4 May 2011)

    1) Select Start>Devices and Printers then click on "Add Printer"
    2) Select "Add a Network, wireless or bluetooth printer"
    3) This will initiate a search for printers...
    4) Select printer with IP address = 130.242.129.248 and click on "Next"
    5) Select the defaults to complete the installation
       (driver will be downloaded and installed automatically)
    6) Print test page if desired.
    

    Items to be filled in on the computing form:

  • Follow the link to forms.html.
  • Support emails:
        support@hpc2n.umu.se
    

    Other items:

    README

    winclean.html

    norosx52.html

    memstick.html

    hyper3_info.html

    VirtualBox_Help.html

    My infamous sys directory

    Virtual Box help

    linear_density

    ssh_tunnel.html

  • Axel's CVS repository
    setenv CVSROOT :pserver:$USER@norlx51.nordita.org:/var/cvs/brandenb
    setenv CVSROOT :pserver:nyokoi@130.242.129.93:/var/cvs/brandenb
    cvs co ap  #(to get the astrophysics directory)
    
    After you checked in changes, I will need to approve the changes and update the changes. If you run bash (check by saying "echo $SHELL"), say
    export CVSROOT=:pserver:$USER@130.242.129.93:/var/cvs/brandenb
    
    and don't forget the colon here! To set your correct editor, say *one* of the following:
    setenv CVSEDIT vi
    setenv CVSEDIT emacs
    setenv CVSEDIT nano
    
    If you are sitting on the Nordita network, cvs.nordita.dk is not interpreted correctly. If this is a problem, try:
    pc_cvs_chroot :pserver:$USER@cvs.nordita.dk:/var/cvs/brandenb :pserver:$USER@norlx51.nordita.org:/var/cvs/brandenb
    
    If your latex compilation complains about some missing inputs file, try
      export TEXINPUTS=$HOME/tex/inputs
    
    or, in tcsh, with even more useful stuff,
      setenv TEXINPUTS .::$PENCIL_HOME/doc/texinputs//:$HOME/tex/inputs//:./texinputs:./fig:./figs:./png
    
    This of course assumes that your $HOME/tex/inputs contains the missing files.
  • computing reports for SNIC

    report10.pdf, prop10.pdf

    report11.pdf, prop11.pdf

    report12.pdf, prop12.pdf

    report13.pdf, prop13.pdf

    report14.pdf, prop14.pdf

    report15.pdf, prop15.pdf

    report16.pdf, prop16.pdf


    Axel Brandenburg
    $Date: 2017/04/05 20:43:36 $, $Author: brandenb $, $Revision: 1.41 $