Genbeo
Home | Project | People | News | For Team | Calendar | Library |
Genbeo
Genbeo (genbeo.genomecenter.ucdavis.edu) is a computer cluster located on the UCDavis campus and is administered by Micheal Lewis.
Submitting a job
Submitting a job to Genbeo requires the use of qsub. It will send your job to an available node and assign it a job ID.
Example :
qsub -b y -V /share/eisen/lab/bin/blastp /share/eisen/lab/ComboDB/combo.pep test.pep -o test.blastp
You can track your job by typing
qstat
It will list your job IDs, the node they are running on and their status. If you need to delete a job :
qdel jobID
using a wrapper script for qsub
When running multiple jobs, it's often a good idea to create wrapper scripts for each individual job. The script would look like the following with a .sh extension.
#!/bin/bash ## #$ -cwd #$ -p -10 #$ -M emailaddress #$ -N GENEMAPPER.0 #$ -notify #$ -S /bin/bash #$ -o out #$ -e err /share/eisen/lab/bin/blastp /share/eisen/lab/ComboDB/combo.pep test.pep -o test.blastp ###
The submission to the queuing system would look like :
qsub -V myfile.sh
Things to keep in mind
Our home directories are stored on an NFS server. The compute nodes are connected to the NFS server via an ethernet network link. If your program running on a compute node accesses data on the NFS server, it must traverse the network to get the data and it is generally slow. If you are running many jobs, and all of them are accessing data on NFS (your home directory, or /share/eisen), it overwhelms the NFS server and your (and everybody else's) programs will run really, really slow. The solution is to copy your data to the compute node where the job is running as part of your qsub job script. Something like the following should work when added to the wrapper script:
mkdir -p /state/partition1/eisenlab/myjobname cp /home/myname/mydataset /state/partition1/eisenlab/myjobname cd /state/partition1/eisenlab/myjobname myprogram mydataset cp myresults /home/myname rm -rf /state/partition1/eisenlab/myjobname
GUI for Cluster monitoring [Ganglia]
You can use Ganglia to
- Check you job status
- Look for available nodes
- Look for specific nodes with certain # of CPUs or RAM
- Check Job queues
- And much more
Open Terminal Type,
ssh -L 59000:genbeo.genomecenter.ucdavis.edu:80 uname@genbeo.genomecenter.ucdavis.edu
or
ssh -L 59000:128.120.243.34:80 uname@128.120.243.34
uname is your genbeo user name
Enter your password
Open a web browser and goto
localhost:59000/ganglia
Go To
- Job Queue: To find you jobs and its id
- Cluster Top: Find the status nodes and the jobs running on them
- Click on your node's/nodes' graph: To get its detailed status