Working with AliEn Kilian Schwarz ALICE Group Meeting April 19 2006

Download Report

Transcript Working with AliEn Kilian Schwarz ALICE Group Meeting April 19 2006

Working with AliEn
Kilian Schwarz
ALICE Group Meeting
April 19 2006
Initialisation and login




. /u/aliprod/bin/.alienlogin
alien proxy-init
alien-token-init
aliensh
Searching technique for simulated
data with certain properties





Not possible with PDC01-05-data 
 wait for PDC06 
Current method:
aliensh
alien: find /alice/cern.ch/demo/data %.root >
/tmp/collection.xml
 The XML collection file can be read directly by
ROOT
 DC data can be found, e.g. in
alien://alice/cern.ch/user/a/aliprod/production_200
5/
Large scale data simulation using
Grid
 Example at
alien://alice/cern.ch/demo/production/flow.jdl
 Needed configuration files can be found in
alien://alice/cern.ch/demo/production/configs (e.g.
Config.C, rec.C, sim.C, simrun.C,…)
 Start production with alien> submit flow.jdl 001
(run number, is used as SEED for random
generator to be able to repeat simulation with
same random generator initialisation)
 If you want to modify, create your own jdl and
configuration files, following the given examples.
The production JDL
 # our executable
 Executable = “aliroot”; (compulsary !!!)
 Has to be in alien://bin, /alice/bin or $HOME/bin. “aliroot” exists already
in /alice/bin and is a shell script executing AliRoot.
 # this are pseudo names to identify the production
 Jobtag={“comment:PDC05 flow events”, “round::2005”, “tag:v4-03-04”,
“type:Flow”};
 # the packages we need
 Packages={“AliRoot”::v4-03-04”,”GEANT3::v1-3”};
 Jobs runs only at sites where those packages are installed
 # the estimated maximal runtime in seconds
 TTL = “86400”; WN has to provide the requested run time
 # this script runs when the job is finished. If it returns 0, the job is
validated (SAVEDDONE), if it returns !=0, the job ends in status
ERROR_V
 Validationcommand =
“/alice/cern.ch/demo/production/scripts/validation.sh”;
The production JDL 2




# this uses only sites, which are member in this particular grid partition
Requirements = member(other.GridPartitions,”Production”);
# this are the input files/macros we need
InputFile={ “LF:/alice/cern.ch/demo/production/configs/Config.C”,
“LF:/alice/cern.ch/demo/production/configs/simrun.C”,
“LF:/alice/cern.ch/demo/production/configs/sim.C”,
“LF:/alice/cern.ch/demo/production/configs/rec.C”,
“LF:/alice/cern.ch/demo/production/configs/CheckESD.C”};
 These files will be transported to the node where the job will run and
can bei either lfn or pfn
 # we want to pack the output together into archive files. One with all the
root files, one with all the log files.
 OutputArchive={
“root_archive.zip:galice.root,Kinematics.root,AliESDs.root@Alice::CERN::Castor2”,
”log_archive:*.log,stdout, stderr@Alice::CERN::se01”};
The production JDL 3
 # we define the arguments for the simulation script as run number and
event number
 Splitargument=“simrun.C –run $1 –event #alien_counter#”;
 Arguments field for each job. Here the subjobs counter produced by
the Split tag below is given. If more than one value is defined, each
subjob will be submitted as many times as there are items in the array,
and the subjobs will have the element in the array as arguments.
 # we produce event number 1-10, the runnumber is given as argument
$1 during the submission
 Split = “production:1-10”;
 The job can be splitted in several sub jobs, according to file (1 job per
input data file), directory (all input files of 1 directory will be analysed in
1 job), event (all files with the same name of the last subdirectory in 1
job), production #start-&lt#end (no InputData required, it will submit the
same JDL several times from #start to #finish. This counter can be
referenced in Splitarguments using #alien_counter#
 Workdirectorysize={“1500 MB”}; The node has to provide the requested
size
Job status









Can be investigated by using
alien:// ps
e.g.: ps –f DONE (lists all finished jobs)
ps –u user (all jobs from the user)
ps –s site (all jobs on one site)
ps –trace <jobid> all (gives job trace information)
Output can be found in alien://proc/user
Copy output back to local via e.g.
alien://cp output.root file:/tmp/output.root