How to setup your directory and launch your first job

How to setup your directory and launch your first job

In order to run you first LArSoft SBND job, you will need to execute the following steps. Some explanation is included, but is not yet at the level to be a definitive guide.

You have here two main options:

  1. you want just to run existing software: this is a [production setup]
  2. you want to change existing code or add your own, and run it: this is a [development setup]

Both start with a [preliminary set up], and then diverge.

Set up the general SBND environment

The most important task this step performs is to tell UPS where to find the programs you need (like UPS itself, git, MRB, ROOT, and LArSoft!).

Two options:

  1. using CVMFS

    source /cvmfs/sbnd.opensciencegrid.org/products/sbnd/setup_sbnd.sh
    
  2. using a local UPS database /path/to/products filled with pullProducts:

    source /path/to/products/setup
    export MRB_PROJECT=larsoft
    setup mrb
    

In doubt, go for CVMFS, as it is the one that works more widely and therefore a good habit(TM).

Set up for running existing code (production like)

After the [general setup above], you just need to set up the version of SBND code you need.

There are different versions of LArSoft that you can set up against, and we have a list of them.
Once you pick your version and qualifiers, you set up both with a single command:

setup sbndcode v06_53_00 -q e14:prof

If this list is not up to date or you want to know what qualifiers there are run:

ups list -aK+ sbndcode 

The qualifiers are a list of strings that are separated by colons, and can be in any order. We have:

optimisation level (mandatory)

prof

optimisations turned all the way on, with stack pointers left to facilitate a bit of debugging

debug

optimisations turned off, NDEBUG not set, assertions are active: much slower, but cooperates well with debuggers

compiler (mandatory)

e10, e14 etc.

which compiler was used; we usually have no choice

Don't know/care which to pick? as a special case, you can run e.g. setup sbndcode v06_68_00, without qualifiers (this will pick the most optimised build available, with the recommended compiler; supported since v06_63_00).

Set up for changing or adding code (development like)

When you need to change existing code or writing new one, you need to have a development area and an environment (that is, an interactive shell) with a [development environment].
While the development environment allows running of code on almost all the cases, there are some corner cases where that might fail. For that reason, or to keep the development environment clean from additional setup required when running, setting up a [run-only environment] in a separate interactive shell is a good idea (as it is to learn to use a terminal multiplexer, especially when developing on a remote server).

If you will be developing code and intent to add it to the repo, you need to be listed as a developer here on the Redmine page. Look at the Overview tab above and see if your name is listed under Developer. If you're not there, ask someone under Manager to add you.

Development setup

First you need to go through the [general setup above].
Then, the first three steps create your working area for the first time. To reuse that area later on (that is, after you log out and in again), only the per-login setup part is needed.
We assume you work in a SBND GPVM, which is painfully slow but safe: ssh ${USER}@sbndgpvm01.fnal.gov.

  1. first time only (or, new working area) setup:
    1. choose the latest SBND code release

    2. go to your user directory in CephFS, /exp/sbnd/app - don't have one? feel free to create it! And create a directory for your new LArSoft release.

      cd /exp/sbnd/app/users/${USER}
      mkdir larsoft_VERSION # or whatever path you like
      cd larsoft_VERSION
      
    3. create a new larsoft development area:

      mrb newDev -v VERSION -q QUAL
      

      The qualifiers are the same described in the previous section

  2. this step belongs to the per-login setup: tell MRB that you will be running from this directory (the output of mrb newDev tells you exactly the command to run):

    source localProducts_larsoft_VERSION_QUAL/setup
    

    (note that this QUAL is with underscores in stead of colons: e.g., e14_prof)

  3. go to the srcs directory and pull down your own copy of the sbndcode repository:

    cd "$MRB_SOURCE" 
    mrb gitCheckout sbndcode
    

    See the important note below!! Particularly check to see if you haven been granted write access.

  4. build the code in your area:

    mrbsetenv
    mrb install -j4
    
  5. you are now ready to run your first job. Do that in a data area (create your directory as above). This will generate your first 10 muons and simulate the detector response to them:

    mkdir -p "/exp/sbnd/data/users/${USER}/larsoft_VERSION" 
    cd "/exp/sbnd/data/users/${USER}/larsoft_VERSION" 
    lar -c prodsingle_sbnd.fcl -n 10
    

Important note: mrb gitCheckout (mrb g in short) prepares the development version of every repository, regardless of which version of LArSoft you have setup! This can cause nasty version dependency problems, so if you want to pull other LArSoft packages (like larsim, larreco etc… ), it is prudent to add a -t LARSOFT_SUITE-VERSION to your mrb g command, where VERSION is the LArSoft version sbndcode current development branch (develop) builds against (i.e. mrb g -t v06_34_00 larsim ). You can find dependancy for the LArsoft version in the LArSoft release list page.
Or you can do that immediately after, going into the repository directory and running git checkout LARSOFT_SUITE-VERSION.

Write access to the repo note: go to the repository directory and type git remote -v if the url doesn't start with ssh: you won't be able to push your code. Check if you're listed as developer, and change the remote with git remote set-url origin ssh://p-sbncode@cdcvs.fnal.gov/cvs/projects/sbncode

While this set of instructions will bring you to the muons, it's a good idea to design a strategy before you run your second job:

  • where do you keep the code?
  • where do you run?
  • how do you set up the area for developing? (write your own setup script)
  • how do you set up the area for running your code? (write your own setup script)
  • do you use the same shell for compiling and for running? (it's in fact a good idea to have two separate shells and two separate directories)

There is a big jump from "making it work" to "making it work well".

A better explanation than this was given in the art/LArSoft course: it's a good idea to go through that material (but keep in mind that names of packages, setup scripts and versions may need to be replaced by SBND specific ones).

Run-only setup

While the development environment set up above is good for almost everything, there are quirks that may be hard to understand.
A setup exclusively for running with the code you are developing is as follows:

  1. setup the general SBND environment [in whatever way you did before]:

    source /cvmfs/sbnd.opensciencegrid.org/products/sbnd/setup_sbnd.sh
    
  2. tell UPS to pick and prefer the data products from your development area in /exp/sbnd/app/users:

    source /exp/sbnd/app/users/${USER}/larsoft_VERSION/localProducts_larsoft_VERSION_QUAL/setup
    
  3. set up SBND code (and all the highest level UPS products you need in order to run):

    setup sbndcode v06_53_00 -q e14:prof
    

You are now ready to run your job. You can move to your run directory and run jobs.

To make sure the setup is correct, run ups active and verify that the repositories from your development area have actually been picked up, e.g.:

ups active | grep localProducts_larsoft