|
CLUSTER USE
- About the Cluster
The managed clusters are:
These machines are the
frontend/headnode of the respective clusters. The ibicluster is the
production cluster for all users. The ibifast is the same as ibicluster,
except that the jobs running more than 24 hours will be cancelled. The
brdfcluster is for users who own nodes on brdfcluster. JP cluster is for
people from Dr. Jonathan Pritchard's group. For use of jpcluster, please get
explicit permission from Dr. Pritchard.
The ibicluster has 256 processors,
and ibifast has 64 processors. The production cluster has 256 nodes and the
fast cluster has 64 nodes. JP cluster is made of 32 bit machines (dual
processors) with varying RAM capacity. All the cluster machines run RHEL
v5.2. The clusters are managed using the ROCKS cluster distribution.
- Logging in
- Using the Cluster
'qsub' and 'qrsh' are the commands to be used
(Always!!) for submitting jobs to the clusters. Please avoid submitting to
individual nodes directly. Contact the support people if you have special
needs for your job.
- Job Submission
o $ qsub blast.sh o Your job 147606 ("blast.sh") has been submitted. o --Script-- o -------------------------------------------- o #!/bin/sh o #$ -cwd o #$ -S /bin/bash o #$ -o droso_ecoli.output o #$ -e blast_err.txt o blastall -p blastn -d /biodb1/blast/ncbi/ntseq/drosoph.nt -i /biodb1/blast/ncbi/ntseq/ecoli.nt o --------------------------------------------- o --Script explained-- o Lines starting with #$ are interpreted by the SGE interpreter. o line1: #!/bin/sh -> include this in all your shell scripts o line2: -cwd -> tells the SGE to run your script from the 'current working directory'. (recommended) o (default behaviour - is to run from home directory) o line3: -S /bin/bash -> to specify a specific shell other than the default.(recommended) o (ex: -S /bin/csh, /bin/tcsh, etc) o line4: -o -> to specify the output file (default - 'scriptname'.o????' created by the SGE) o line5: -e -> to specify the error file (default - 'scriptname'.e????' created by the SGE) o -rw-r--r-- 1 kiran ccuser 98 Jun 12 11:20 blast.sh.e147607 o -rw-r--r-- 1 kiran ccuser 0 Jun 12 11:20 blast.sh.o147607 o line6: the program to be executed
- Checking Status of Jobs
o o $ qstat o job-ID prior name user state submit/start at queue slots ja-task-ID o ---------------------------------------------------------------------------------------------------------- o 147604 0.25000 dim1-9.sh cbecquet r 06/11/2006 22:54:14 all.q@compute-0-23.local 1 o 121336 0.75000 wrap.sh babis r 06/06/2006 10:32:10 all.q@compute-0-9.local 1 o ******************************************************************************** o $ qstat -f|more o queuename qtype used/tot. load_avg arch states o ---------------------------------------------------------------------------- o nancy.q@compute-0-32.local BIP 0/1 0.00 lx26-amd64 o ---------------------------------------------------------------------------- o nancy.q@compute-0-33.local BIP 0/1 0.00 lx26-amd64 o ---------------------------------------------------------------------------- o all.q@compute-0-22.local BIP 0/2 0.00 lx26-amd64 o ---------------------------------------------------------------------------- o all.q@compute-0-23.local BIP 1/2 1.01 lx26-amd64 o 147604 0.25000 dim1-9.sh cbecquet r 06/11/2006 22:54:14 1 o ---------------------------------------------------------------------------- o all.q@compute-0-24.local BIP 0/2 0.00 lx26-amd64 o ---------------------------------------------------------------------------- o o ******************************************************************************** o $ qstat -u cbecquet o job-ID prior name user state submit/start at queue slots ja-task-ID o -------------------------------------------------------------------------------------------------------- o 147604 0.25000 dim1-9.sh cbecquet r 06/11/2006 22:54:14 all.q@compute-0-23.local 1 o o
- Other Commands
|
