Menu:

Latest news:

Read more...

Frequently Asked Questions:


  1. How to access CSBL web server?
  2. How to access CSBL web mail, calendar, file sharing system?
  3. How to set up my e-mail account in outlook?
  4. What are my account?
  5. Where are my files?
  6. How to submit jobs?
  7. How to access my file from windows computer?
  8. How to send e-mail to all of our group members?
  9. What are the softwares I can use?
  10. How do I run I/O intensive jobs on the cluster?
  11. How do I setup our network printer on my personal computer?
  12. How to setup wireless network for my laptop?
  13. Where is the SSL certificate for our webserver and mail server?
  14. Where to download ISIS/DRAW?
  15. Why my Windows Secure Shell give an error message "No further authentication methods available?"
  16. How to construct and run a parallel job on the cluster?
  17. Where can I get the drivers for our printers?
  18. Endnote toolbar doesn't want to be displayed in Word XP...
  19. Acrobat toolbar doesn't want to be displayed in Word XP...
  20. How to access the TRANSFAC database installed at CSBL?
  21. How do I submit a job to our big cluster?
  22. What should I do and what should I not do when using computers?
  23. Additional head node and storage space at big cluster
  24. Lab backup policy
  25. How to create my homepage?
  26. How to migrate my homepage from the old server to the new server?
  27. How to update my homepage?
  1. How to access CSBL web server?
    http://csbl.bmb.uga.edu

    2. [top]

  2. How to access CSBL web mail, calendar, file sharing system?
    http://csbl.bmb.uga.edu/internal.html

    3. [top]

  3. How to set up my e-mail account in outlook?
    imaps server: csbl.bmb.uga.edu
    smtp server: csbl.bmb.uga.edu

    4. [top]

  4. What are my account?
    • NIS account: it covers all the Linux computers. To change your NIS password: (1) login any computer covered by NIS, (2) run "yppasswd".
    • Web server account: Some members of the lab have a different account than NIS account for the web server, some use their NIS account.
    • Webmail account: It's your web server account.
    • Samba account: Account used to access your files in "storage2" from your Windows computer. To change your samba password: (1) login to "storage2", (2) run "smbpasswd".
    • Calendar/egroupware account: same as your webmail account.
    • File sharing system/owl account: same as your webmail account.

    5. [top]

  5. Where are my files?
    • Your private files are stored at your home on the server. In most cases it is:
      /fileserver/home/staff/YourLoginName/
    • You have another directory on your local computer:
      /home/YourLoginName/
    • Some of you have another directory in /webserver/YourLoginName
    • Common databases and tools are stored in /fileserver1/
    • Scratch space in compute nodes is in /scratch

    6. [top]

  6. How to submit jobs?
    We use Sun Grid Engine as our job management system. Related commands are:
    qsub -> submit jobs
    qstat -> show job status
    qdel -> remove jobs
    We have three clusters and two file systems.
    The clusters are:
    • The test cluster with 8 nodes. Each node has 2 Xeon 3.06GHz processors and 2.5GB memory. The maximal job number for each node is 4 because I enabled Hyper Threading. Their names are from tnode001-tnode008. This cluster is located at the Life Science Building, and it is for test purpose and short jobs. It is Intel x86 archtecture. Its head node is qiang.
    • The Big Cluster with 148 nodes. Each node has 2 Xeon 3.06GHz processors and 2.5GB memory. Their names are from node001-node148. This cluster is located in the Boyd Building. Its head node is decoder.cc.uga.edu.
    • The JS20 cluster with 92 nodes. Their names are from pnode001-pnode068, pnode101-pnode124. pnode001-pnode068 have two 1.6GHz PowerPC 970 CPUs and 2.5G memory for each node, and pnode101-pnode124 have 2.2GHz PowerPC 970 CPUs and 2.5G memory for each node. It is located at Boyd Building. Its head node is decoder.cc.uga.edu.
    The file systems are:
    • The office file system. Located in the Life Science Building, it is connected to the office computers and test cluster. The access path is /fileserver and /fileserver1. The authorized users will have a HOME at /fileserver and access to /fileserver1.
    • The cluster file system. It is connected to the Big Cluster and the JS20 Cluster. The access path is /gpfs0 and /gpfs1. The authorized users will have a HOME at /gpfs0 and access to /gpfs1.
    For submitting a job the a cluster, you need login to its head node, write a simple script, use absolute path in your script, then run qsub. For details about Sun Grid Engine, please read its manual.

    7. [top]

  7. How to access my file from windows computer?
    Start -> run -> \\192.168.0.6, then input your samba account username and password to access your files. Mapping a network drive to your home directory in "storage2" is a good idea.

    8. [top]

  8. How to send e-mail to all of our group members?
    Just send an e-mail to "csblall@csbl.bmb.uga.edu".

    9. [top]

  9. What are the softwares I can use?
    These softwares are installed in /fileserver1/tools/:
    • maltab -> Mathematic Software package
    • gromacs -> Molecular Dynamics package
    • rasmol -> Molecular Graphics Software
    • meme -> For discovering and using protein and DNA sequence motifs
    • BioProspector -> A DNA seqence motif finding program
    • NCBI tools, blast, psiblast ...
    • dsspcmbi -> Secondary structure assignment
    • stride -> Secondary structure assignment
    • Acrobat reader 5 -> to Run, type "acroread"
    • grace -> Data analysis and graphics generating
    • Modeller 6.3 -> Homology Modelling
    • prospect -> Threading program developped in CSBL
    • SNNS -> Stuttgart Neural Network Simulator

    10. [top]

  10. How do I run I/O intensive jobs on the cluster?
    -->> What kind of problem will the I/O intensive jobs bring to our cluster system?
    When someone is running I/O intensive jobs (a lot of reading and writing to hard disk), if you are using your home - "/fileserver/home/staff/" for reading and writing, it will have our fileserver (computer - storage1) response very slow, possibly die and have to restart.

    -->> So, how can I run I/O intensive jobs without bothering the collegues?
    To run I/O intensive jobs, please use "/scratch" space, it is a local hard disk space in each of the computer nodes. That means if your program generates a lot of temporary files, please write them to "/scratch/". Attached is a csh script which successfully run for Youxing. This rule also applies to the big cluster.

    *******************sample job file*********************** 
    #!/bin/csh
set dir=/scratch/yqg_20
set cmd=/gpfs0/home/staff/youxing/bin/run_RDCmain_nosec_10_SCOP40_1.5_cluster.pl
set protein=1J6TA_NHAB
mkdir $dir
cd $dir
`$cmd  $protein`
rm -rf $dir
    ****************end of sample job file*******************

    11. [top]

  11. How do I setup our network printer on my personal computer?
    Here are the IP addresses for our three printers:
    • HP LaserJet P2055dn in room A110: 192.168.0.10 (duplex installed)
    • HP Color LaserJet 4600n at A110: 192.168.0.11
    • HP LaserJet 4200n at A134: 192.168.0.12 (duplex installed)
    To enable duplex in your Windows computer, Start->Control Panel->Printers and Other Hardware->Printers and Faxes->Right Click on the Printer->Properties->Device Settings->Installed Options->Duplex Unit->Installed.
    To print on both sides, click "Properties" in your "Print" dialog box, then click "Finishing" and check the Checkbox "Print On Both Side".

    To enable duplex on Linux, Settings->Printing Manager->Instances->Settings. In the "Duplex Printing" dialog, click on "Long side".
    When command-line, add the following option to lpr: 
    lpr -o sides=two-sided-long-edge FileToPrint.ps
    Except for final reports, please try to print as much as possible using double-side printing. Idem, do not print in color except if it's really needed. This not only saves money, but also paper and ink...

    12. [top]

  12. How to setup wireless network for my laptop?
    Because the campus wireless network (SSID: UGA) has covered Life Science Building, CSBL wireless network has stopped. You need MyID to access UGA wireless network(wifi).

    13. [top]

  13. Where is the SSL certificate for our webserver and mail server?
    Here they are. Download them and install them in "trusted root certificate".
    http://csbl.bmb.uga.edu/ssl/imapd.crt
    http://csbl.bmb.uga.edu/ssl/cacert.crt

    14. [top]

  14. Where to download ISIS/DRAW?
    here

    15. [top]

  15. Why my Windows Secure Shell give an error message "No further authentication methods available?"
    Follow this procedure:
    1. Download Windows Secure Shell Client 3.2.9 here
    2. Install it
    3. Start it
    4. Click menu "Edit->Settings"
    5. At tab "Authentication", add new authentication method "Keyboard Interactive"
    6. Connect via "Keyboard Interactive" again. Be a little patient, it takes a 5-10 seconds to be connected

    16. [top]

  16. How to construct and run a parallel job on the cluster?
    Lammpi is installed on the clusters, here is a sample job script and a c source file
    to do this. Kyle has the Copyright. :)

    17. [top]

  17. Where can I get the drivers for our printers?
    Here:
    HP LaserJet 4200n
    HPColor LaserJet 4600n
    HP LaserJet P2055dn

    18. [top]

  18. Endnote toolbar doesn't want to be displayed in Word XP...
    Click here for their official solution...

    19. [top]

  19. Acrobat toolbar doesn't want to be displayed in Word XP...
    Start Word->Help->About Microsoft->Disabled Items->PDFMaker->Enable->restart Word

    20. [top]

  20. How to access the TRANSFAC database installed at CSBL?
    http://csbl.bmb.uga.edu/cgi-bin/biobase/transfac/8.2/bin/start.cgi
    Please ask Fenglou Mao for the user name and password.

    21. [top]

  21. How do I submit a job to our big cluster?
    • Compilation

      First, all compilation and queuing should be run on mgt2, otherwise known as decoder.cc.uga.edu. If you want to compile a ppc application, you can ssh out to pnode001.

      Second, both architecture are now viewable in the same queue. So there are now about 90 new computers that you will have access to. The only catch is that they use a different type of CPU, so programs will need to be recompiled. If your entire job is perl code or java, you don't need to recompile.

    • $ARC: how to choose your architecture

      You can take advantage of both systems by using the $ARC variable. On x86 machine, $ARC is set to glinux, on PPC machines, it is set to ppc64linux.
      In your queue submission script, simply call a program with a path that looks like: $HOME/$ARC/bin/blast.
      On the x86 machine the program /home/you/glinux/bin/blast will be run, while on the PPC machines, /home/you/ppc64linux/bin/blast will be run. We will try to compile all commonly used applications in this manner. Ask us if you have any trouble. We should have all the blast tools up and running, so just use the line /gpfs1/$ARC/ncbi/bin/blastpgp , and you'll get blastpgp on any architecture.
      If you don't want to bother with the dual architecture thing, you can request your jobs be run on only one type of machine.
      You can request a specific type of machine with the flag -l arch=..., so you can put in your script, the line
      #$ -l arch=glinux
      and your program will only run on x86 machines.

    • Output and error output in your jobs

      Please keep in mind this is very important, if you don't follow the rules I list below, you will possibly crash our cluster. When you are submitting many jobs (say more than 100), add the two lines below in your job script to avoid screen output, which will crash our cluster if there are too many.
      #$ -o /dev/null
      #$ -e /dev/null

    • Parallel jobs

      We have also been trying to clean up the parallel environments. There are two essential types of parallel environments: lammpi and bare. The lammpi environment sets up the lammpi deamons so all you have to do is call mpirun. The bare environment simply allocates you the machines, and expects you to set up a communication environment. This is good for programs that use their own parallel communication environments, such as NAMD.

      In addition, there are different types of parallel environments that you can access by their 'geographic' location. The blade centers are comprised of 14 machine on the same switch, which means that the intercommunication between those machines has less latency. This is better for parallel jobs that require a lot of low latency communication, such as molecular dynamics. If you have a job that requires only occasional communication between parallel nodes, then you don't need to bother with this.

      The parallel enviroments are:
      lammpi (all nodes)
      lammpi-rsa1 to lammpi-rsa4 (sets of 12 machines from the x335s)
      lammpi-mm1 to lammpi-mm12 (sets of 14 machines from the Blades)
      bare
      bare-rsa1 to bare-rsa4
      bare-mm1 to bare-mm12
      So, if you want a large, distributed parallel environment, you would put something like:
      #$ -pe lammpi 50-150
      You can combine this with the architecture selection:
      #$ -l arch=glinux

      Or if you want to setup a PPC NAMD run, all machines connected by the same switch:
      #$ -pe bare-* 14-28
      #$ -l arch=ppc64linux

      Notice how the '*' is a wildcard, and will select what is on the blade chassis.

      For LAM/MPI jobs, you'll need to remember to set the LAM_MPI_SOCKET_SUFFIX number. This is used to differentiate multiple MPI sets that you might have running on the same machine. This can be done by adding the line:
      in BASH:
      export LAM_MPI_SOCKET_SUFFIX=$JOB_ID
      in TCSH:
      setenv LAM_MPI_SOCKET_SUFFIX $JOB_ID

       Also, remember, Grid Engine ignores the "#!/bin/bash" line at the beginning of the script, and use as defaults TCSH. In order to use BASH, add the line:
      #$ -S /bin/bash
      The LAM boot up process requires that you have your SSH public keys enabled (for password-less login between nodes). You can do this by:
      cd ~/.ssh
      ssh-keygen -t dsa (when it asks for the password, type nothing in, just press return)
      cp id_dsa.pub authorized_keys
      Then you need to make sure all the host certifications are current, I keep a copy of this in /gpfs1/tools, so type
      cp /gpfs0/tools/known_hosts ~/.ssh/.

    [top]

  22. What should I do and what should I not do when using computers
    • Don't occupy too much space, like more than 100GB in /fileserver, /fileserver1, /gpfs0 or /gpfs1.
    • Clear your home directory regularly.

    23. [top]

  23. Additional head node and storage space at big cluster
    We bought some new storage space in the big cluster. It is ready and available now.
    Total capacity: 12.67TB
    New head node for the cluster: decoder1.cc.uga.edu

    This new storage space occur as "/md3000" in all computers in big cluster, which include:
    decoder.cc.uga.edu (mgt2)
    decoder1.cc.uga.edu (stg5)
    all computational nodes (node001-node148,pnode001-pnode068,pnode101-pnode124)

    Everyone in the lab has an account in all these computers, but only two head nodes can be logged in from outside. (If I forgot to create an account for you, please tell me). And everyone in the lab has a directory in /md3000, either in "staff" or in "students", please find it by yourself.

    You can also submit jobs from stg5, old disks /gpfs0 and /gpfs1 can also be accessed from stg5.

    Performance issue: for /md3000, access it from stg5 is faster. For /gpfs0, /gpfs1 in principle it should be faster from mgt2, but based on my test if seems it is faster from stg5, so make your own choice.

    Because /md3000 is using XFS filesystem, quota at XFS is a little different from quota at other filesystem.

    Currently there are 3 parameters you need to know about quota at /md3000.

    soft limit: the space you occupied at /md3000 can not be more than "soft limit", or you will trigger a timer, but you can still use more space until you reach the "hard limit".

    hard limit: the space you occupied at /md3000 can never be more than "hard limit".

    timer: if you occupied space more than "soft limit", the timer will be triggered, when the timer is pop up, you can not use any more space at /md3000 unless you release some space to make your occupation become less then "soft limit".

    Currently "soft limit"=200G, "hard limit"=300G, Timer=7days.

    stg5 is 64 bit linux, the default executable file generated by gcc/g++ will be 64bit, which can not be run at 32bit linux. Our node001-node148 are 32bit linux, so you need 32bit executable files. You have two methods to do that:
    1. compile at mgt2 (decoder.cc.uga.edu, the old head node)
    2. compile at stg5 (the new head node) with option -m32

    eg. gcc -m32 xxx.c -o xxx
    The you can create 32bit executable file, which can be run at both 32bit and 64bit linux, but it can not utilize memory more than 4GB.

    24. [top]

  24. Lab backup policy
    • /fileserver & /fileserver1
      I will still use our old backup policy for /fileserver and /fileserver1: backup once at every weekend, and backup everything.
    • /gpfs0 & /gpfs1
      Please make a directory named "backup" at your personal directory, and copy your important files (such as source files, documents etc) into this directory, I will ONLY backup this directory. Currently I am planning to backup this directory every night. If it is too hard, I will go back to our old policy and backup it once a week (at weekend).
    • /md3000
      Same backup policy as /gpfs0 & /gpfs1.

    25. [top]

  25. How to create my homepage?
    I am using my username as an example below, replace "fenglou" with your username.
    1. ssh fenglou@192.168.0.4
    2. mkdir public_html
    3. chmod 755 public_html
    4. cd public_html
    5. copy whatever files you want to be your homepage here...

    26. [top]

  26. How to migrate my homepage from the old server to the new server?
    I am using my username as an example below, replace "fenglou" with your username.
    1. ssh fenglou@192.168.0.4
    2. rsync -avz fenglou@192.168.0.3:~/public_html .

    27. [top]

  27. How to update my homepage?
    I am using my username as an example below, replace "fenglou" with your username.
    1. ssh fenglou@192.168.0.4
    2. cd public_html
    3. copy your new files here or modify them by vi

    28. [top]

   Copyright © 2007 (CSBL team). Design Viklund.