Thursday Exercise 2.2: Using StashCache for Large Shared Data

This exercise will use a BLAST workflow to demonstrate the functionality of StashCache for transferring input files to jobs on OSG.

Because our individual blast jobs from Exercise 2.1 would take a bit longer with a larger database (too long for an workable exercise), we'll imagine for this exercise that our pdbaa_files.tar.gz file is too large for a web proxy (larger than ~1 GB). For this exercise, we will use the input from Exercise 2.1, but instead of using the web proxy for the pdbaa database, we will place it in StashCache via the OSG Connect server.

StashCache is a distributed set of caches spread across the U.S. They are connected with high bandwidth connections to each other, and to the data origin servers, where your data is originally placed.

StashCache Map

We will be using the command stashcp to copy files from Stash to the execute hosts. It has a cp like syntax.

Setup

Place the Database in StashCache

Copy to your public space on OSG Connect

StashCache provides a public space for you to store data which can be accessed through the caching servers. First, you need to move your BLAST database into this public directory. You have already placed files in the ~/stash/public directory in the previous exercise in order for it to be accessible to the HTTP proxies. The same directory is accessible to StashCache.

As the public directory name indicates, your files placed in the public directory will be accessible to anyone's jobs if they know how to use stashcp, though no one else will be able to edit the files, since only you can place or change files in your public space. For your own work in the future, make sure that you never put any sensitive data in such locations.

Check the file on OSG Connect

Next, you can check for the file and test the command that we'll use in jobs on the OSG Connect login node:

user@training $ ls ~/stash/public

Now, load the stashcache module, which will allow you to test a copy of the file from StashCache into your local scratch directory on training.osgconnect.net:

user@training $ module load stashcache
user@training $ stashcp /user/<USERNAME>/public/pdbaa_files.tar.gz /local-scratch2/<USERNAME>/

Replacing all instances of <USERNAME> with your username on training.osgconnect.net. You should now see the pdbaa_files.tar.gz file in your local scratch directory. Notice that we had to include the /user and your username in the file path for stashcp, which make sure you're copying from your public space.

Modify the Submit File and Wrapper

You will have to modify the wrapper and submit files to use StashCache:

  1. At the top of the wrapper script (after #!/bin/bash), add the lines that load the stashcache module and to copy the pdbaa_files.tar.gz file into the current directory of the job:

    module load stashcache
stashcp /user/<USERNAME>/public/pdbaa_files.tar.gz ./

    Replacing <USERNAME> with your own username.

  2. Since HTCondor will no longer transfer or download the file for you, make sure to add the following line (or modify your existing rm command, if you're confident) from your wrapper script to make sure the pdbaa_files.tar.gz file is also deleted and not copied back as perceived output.

    rm pdbaa_files.tar.gz

  3. Delete the wget and export http_proxy lines from the job wrapper script blast_wrapper.sh

  4. Add the following line to the submit file and update the "requirements" statement to require servers with OSG Connect modules (for accessing the stashcp module), somewhere before the word queue.

    +WantsStashCache = true
requirements = (OSGVO_OS_STRING == "RHEL 7") && (HAS_MODULES =?= true)

  5. Confirm that your queue statement is correct for the current directory. It should be something like:

    queue inputfile matching mouse_rna.fa.*

And that mouse_rna.fa.* files exist in the current directory (you should have copied them from the previous exercise directory).

Submit the Job

Now submit and monitor the job! If your 100 jobs from the previous exercise haven't started running yet, this job will not yet start. However, after it has been running for ~2 minutes, you're safe to continue to the next exercise!

Note: Keeping StashCache 'Clean'

Just as for data on a web proxy, it is VERY important to remove old files from StashCache when you no longer need them, especially so that you'll have plenty of space for such files in the future. For example, you would delete (rm) files from public on training.osgconnect.net when you don't need them there anymore, but only after all jobs have finished. The next time you use StashCache after the school, remember to first check for old files that you can delete.