Difference between revisions of "Off Line Analysis for transversity"

From Hall A Wiki
Jump to: navigation, search
(disk1, 800GB)
(disk2, 800GB)
Line 114: Line 114:
 
It's shared with GEN, Pi0.
 
It's shared with GEN, Pi0.
  
== disk2, 800GB ==  
+
=== disk2, 800GB ===  
  
 
Shared with E08007, mostly full. Could be used to buffer test files.
 
Shared with E08007, mostly full. Could be used to buffer test files.

Revision as of 07:21, 14 February 2009

This is temporary site of Off Line Analysis for transversity.

Farm Replay

Structure

On farm file system, there are 3 disk:

  • disk1 : Our main script disk.
  • disk2 : Shared with E08007, mostly full
  • disk3 : 2TB disk for large root files

THe main analysis directory is /work/halla/transversity/disk1/analysis/, which includes various of components:

  • replay.sh = SourceMeFirst : script setting up environmental pathes. Configured for CSH
  • replay : replay directory holding replay scripts/configure files
  • farmreplay : specialized folder generating batch farm replay scripts
  • DB : database
  • analyzer : analyzer
  • bigbitelib : Bigbite library
  • RICHlib : rich lib
  • BBNormAna : Normalization analysis lib
  • onlineGUI : Online GUI

analysis directory could be copied to your own location to be customized.


HowTo do a quick test replay

One can directly use the common replay directory to do quick tests.

ssh ifarml5    # or any computer seeing working disk
cd /work/halla/transversity/disk1/analysis
source replay.sh
cd replay
analyzer
replay_phys(4029,100000);   //just like online replay

HowTo build your own batch farm replay directory

Build a directory for your own replay. You can change replay code and control the rootfile storage. And it will be capable of replaying on both ifarm and batch farm.

ssh ifarml5    # or any computer seeing working disk

cd /work/halla/transversity/disk1/<your name>

mkdir analysis    # this is the main directory for replay codes
cd analysis

cp -rv /work/halla/transversity/disk1/analysis/replay  ./
cp -rv /work/halla/transversity/disk1/analysis/farmreplay ./
cp -rv /work/halla/transversity/disk1/analysis/replay.sh ./

mkdir /work/halla/transversity/disk3/<your root file folder>
ln -svf /work/halla/transversity/disk3/<your root file folder> ROOTfiles

Now you have the basic structure:

  • To test replay on ifarm, follow Off Line Analysis for transversity#HowTo build your own replay directory
  • To do batch farm replay, follow example below:
    1. enter directory farmreplay
    2. edit construct.pl, modify following sections:
      • $neve : the event number to replay
      • @runnumber : list of run to replay
      • $ReplayType : name of replay script
      • $current : path of your analysis directory
    3. run construct.pl; it will generate
      1. in_run# : the script for batch running for each run; double check it carefully
      2. sub_run# : Batch Job Command File for each run; double check it carefully
    4. (Optional), test in_run# scripts by copying it to a tmp directory and run it.
    5. jsub sub_run# will run the replay in batch farm

HowTO Customize Analyzer / Libs / Database

There are more folders in /work/halla/transversity/disk1/analysis/, which are various of components:

To use your own version of components, first copy its direcory to your own analysis folder, then

  • for database and analyzer, modify your replay.sh to set its path
  • for other libraries, just modify the *.so link in your replay directory to your own location

General Rule for Farm Disk Usage

All transversity farm disks are shared with other hall A experiment. The total space is limited. In the case that the disk is full, old file will be deleted by Hall A manager. Therefore, we need to avoid heavy disk usage by following proper rules:

On farm file system, there are 3 disks mounted at /work/halla/transversity/:

disk1, 800GB

Disk1 is our main script disk. Please make a directory with your user name.

It could be used to store small files. Please avoid large data. Please always keep a decent percentage of free space.

It's shared with GEN, Pi0.

disk2, 800GB

Shared with E08007, mostly full. Could be used to buffer test files.

disk3, 1-2TB

TB disk for large root files buffering. Quota is controlled by Dr. Robot Michaels. If the disk is full, old files will be deleted!

  • Please make a directory with your user name to separate your storage space
  • Please keep your root file as small as possible
  • Please delete unused root files as soon as possible
  • TmpROOTfiles which is linked to common replay directory will be periodically cleaned
  • ProductionROOTfiles is for production level of replay performed and managed by collaboration.