Minnesota Supercomputing Institute
MSI Tutorial Resources
Link to the semester long LATIS Tutorial
http://latis.umn.edu/services-and-programs/research-support/2016-research-workshop-series
Link to broader list of MSI Tutorials
https://www.msi.umn.edu/tutorials/current
Stratus
How to add a second user to a given instance:
# Add the user:
sudo useradd -m -d /home/<username> -s /bin/bash <username>
# Now setup the SSH key for the new user:
sudo mkdir -p /home/<username>/.ssh
sudo chmod 700 /home/<username>/.ssh
sudo vi /home/<username>/.ssh/authorized_keys
# [... paste in your PUBLIC (*.pub) key, save and quit ...]
sudo chown -R <username>:<username> /home/<username>/.ssh
sudo chmod 600 /home/<username>/.ssh/authorized_keys
# If desired, give the user sudo privileges
sudo passwd <username> #give the user a password, maybe the same as their username
sudo adduser <username> sudo
Mesabi
Common PBS commands
MSI uses a PBS scheduler. It's easy to find a lot of information online about how to use PBS, but here are some commonly used functions.
qsub mypbsscript.pbs # submit a PBS script to schedule a job
qsub -t 1-22 arrayscript.sh # submit an array of jobs
qstat # check on job status
showstart <jobid> # check when your scheduled job is due to begin (note this is always
# an overestimate, as it depends on walltimes for all running
# and queued jobs
acctinfo # see total account service unit allocation, service units used by
# each person, and fairshare status.
Example code to submit jobs on Mesabi
First, submitting a job is easiest with a PBS batch script. The following script, called kinship.pbs
, runs a program called vcf2kinship to create a kinship matrix on MCTFR genotypes. It requests a single node and 12 processors on that node. It requests 25 GB of memory.
#!/bin/bash -l
#PBS -l walltime=10:00:00,nodes=1:ppn=12,mem=25gb
/home/vrie0006/hyoung/software/rvtests/executable/vcf2kinship --inVcf /home/vrie0006/hyoung/genotypes.vcf.gz \
--bn \
--thread 10 \
--out /home/vrie0006/hyoung/kinship
That script can then be submitted for scheduling to the mesabi short queue on msi (MSI list of queues) as follows:
qsub -q short kinship.sh
You can then check the status of this job by running various commands.
To see when your job might start running use showstart <jobid>
like this:
[ln0006:hyoung] showstart 4560951
job 4560951 requires 10 procs for 10:00:00
Estimated Rsv based start in 2:10:44 on Wed Nov 29 23:58:00
Estimated Rsv based completion in 12:10:44 on Thu Nov 30 09:58:00
Best Partition: mesabipar
To check status use qstat
:
[ln0006:hyoung] qstat
Job ID Name User Time Use S Queue
------------------------- ---------------- --------------- -------- - -----
4560955.mesabim3.msi.umn.edu kinship.sh vrie0006 01:02:54 R small
Here's an example of an array of 22 jobs to perform rvtests, with 1 job per chromosome. Note that in the script ${PBS_ARRAYID}
denotes the job number, and can be used to differentiate jobs.
#!/bin/bash
#PBS -l walltime=24:00:00,nodes=1:ppn=1,mem=2gb
#PBS -m abe
#PBS -M datt0019@umn.edu
/home/vrie0006/datt0019/tools/rvtests/executable/rvtest
--inVcf /home/vrie0006/datt0019/mctfr/vcf_files/chr${PBS_ARRAYID}.withRS.filtered.PASS.beagled.MZadded.vcf.gz
--boltPlink /home/vrie0006/datt0019/mctfr/bed/cpd_merged
--pheno /home/vrie0006/datt0019/mctfr/phenotypes/residualized_phenotypes.ped
--pheno-name cpd
--meta bolt
--inverseNormal --qtl
--out /home/vrie0006/datt0019/mctfr/gwas/cpd/chr${PBS_ARRAYID}
--boltPlinkNoCheck
--siteMACMin 10
That script can then be submitted for scheduling to mesabi as follows:
qsub -t 1-22 rvtests.sh
Interactive job
qsub -I -l walltime=5:00:00,mem=10gb,nodes=1:ppn=1