[Hallcsw] Farm resources for the commissioning runs

[Brad Sawatzky]

Now that some groups are moving away from 'online' analysis, I thought I
would post a reminder and some suggestions about how to migrate off the
cdaq cluster and use JLab compute resources effectively.

All Collaborators are encouraged to review the rules and tips on using these
shared systems:
  https://hallcweb.jlab.org/DocDB/0008/000868/001/Counting_House_TipsTricks-Jun2017.pdf


I'll try to hit a few of the highlights below:
------------------------------
Using /volatile:
------------------------------
Your offline workflow should generally involve running your analysis on
the Farm with your ROOTfiles/ output directory symlinked to a directory
path under /volatile that identifies your experiment.  For example:
  /volatile/hallc/comm2017/e1206107/

Often it is best to also subdivide that path per-user so people within
the collaboration do not step on each others files by accident.  For
example:
  /volatile/hallc/comm2017/e1206107/<username>/
You'll want to coordinate with your Spokespeople to come up with an
analysis plan and filesystem layout that works for you.

 ** Read and understand how /volatile works here: **
   https://scicomp.jlab.org/docs/volatile_disk_pool
 ** It is your responsibility to understand how the JLab **
 ** resources work.  Ignore this and you may lose files.  **


------------------------------
Using /work
------------------------------
/work should be used to store your analyzer software (hcana,
hallc_replay, etc).

NOTE: /work is on fault-tolerant hardware, but it is not backed up.
  If you 'rm -rf' a directory, your files are gone.  You should be using
  git and github to back up and version-control your software and/or
  establish an alternative backup plan.

  Never keep the only copy of your files on your laptop or personal
  computer.  9 times in 10, even well intentioned people forget to back
  them up properly.

  Students (and Spokespeople) should know what they would do if their
  favorite personal computer's hard disk died tomorrow and plan
  accordingly.


------------------------------
Accessing raw data files (ie. CODA files)
------------------------------
All coda files should only be accessed via this path:
  /cache/hallc/spring17/raw/
*Never* copy raw data files to any other shared JLab filesystem like
/volatile, /work, /home, /group, etc.  There is never a good reason to
do this.  If you think you want to do this, please talk to me first and
we will find a method that works better for you.

------------------------------
When to push files to tape
------------------------------
When you have determined that you have some ROOT files that are useful
(ie. they pass basic sanity checks and you know you want to work with
them for a while), you should push them to tape using 'jput' or
'jmirror'.

* Once they are on tape, you should delete them from disk
* and access them via /cache/hallc/.... just like the CODA files.

I've requested that the experiments each have their own tape volumes
established that are to be used for their analysis output (ROOT files,
etc).  Watch for your tape stub to show up under /mss/hallc/ with these
names:
  /mss/hallc/E12-06-107/
  /mss/hallc/E12-09-002/
  /mss/hallc/E12-10-002/
  /mss/hallc/E12-10-003/
  /mss/hallc/E12-10-008/
[ Hopefully these stubs will exist by the time you read this.  If not,
  check back tomorrow and they should be there.]

More general documentation on using the Farm and tape command can be found
here:
  https://scicomp.jlab.org/docs/FarmUsersGuide

As always, feel free to drop by and chat if you have any questions.

-- Brad

-- 
Brad Sawatzky, PhD <brads at jlab.org>  -<>-  Jefferson Lab / Hall C / C111
Ph: 757-269-5947  -<>-  Fax: 757-269-5235  -<>- Pager: brads-page at jlab.org
The most exciting phrase to hear in science, the one that heralds new
  discoveries, is not "Eureka!" but "That's funny..."   -- Isaac Asimov