Week 10 Exercises - First Steps on Talapas
Short, ungraded practice tasks - about 40 minutes, all on Talapas. Companion reading: Ch. 20 and Ch. 21; reference: App. I. You need your Talapas account working (the account check at the end of Ex 7) - if it is not, that is task 0: contact me. HW2 is due today; these tasks are the cluster leg of your term project.
- Log in to Open OnDemand at https://ondemand.talapas.uoregon.edu with your Duck ID and open the file explorer. What is the path of your home directory?
- Open a shell from OnDemand (or
ssh yourduckid@login.talapas.uoregon.edu) and runhostnameandpwd. - Run
groups- which PIRG(s) are you in? - Recreate your course skeleton on the cluster:
mkdir -p ~/comptools/{data,scripts,results}. - Transfer
Stickle_RNAseq.tsvto~/comptools/data(OnDemand’s file explorer upload button, orscpfrom your laptop). Confirm withwc -l. - Run
module avail R(pressqto exit the list), thenmodule loadan R version and start R on the login node just long enough to runR.version.stringand quit. - Write a minimal Slurm script
scripts/hello.sbatchthat just printshostnameanddate: start from the batch-script template in the book (Ch. 20), keep the--job-name,--partition,--accountand--timelines, and drop the rest (including--output, so task 8’s log gets the defaultslurm-*.outname). - Submit it with
sbatch scripts/hello.sbatch, watch the queue withsqueue -u $USER, and when it finishes,cattheslurm-*.outfile. Which node did it run on? - Run
sinfoand identify one regular partition and the GPU partition. - From the shell, run your Week 3 grep pipeline on the dataset - identical commands, bigger computer. That is the point of the shell.
- Bring your own software:
git cloneyour project repository onto the cluster (that is why it lives on GitHub), or transfer theenvironment.ymlyou exported in Week 9. Thenmodule spider miniconda,module loadit, andconda env create -f environment.yml. conda activate rnaseqand checkwhich python3- the same environment as your laptop, built from one text file. Then prove it travels into batch jobs: editscripts/hello.sbatchtomodule loadminiconda,conda activate rnaseq, and runpython3 -c "import pandas as pd; print(pd.__version__)", and resubmit. The pandas version in the log is your environment running on a compute node - the full loop, closed.