INFN - Ferrara BaBar Meeting SPGrid: status in Ferrara Enrica Antonioli - Paolo Veronesi Ferrara, 12/02/2003
Ferrara - 12/02/032 Topics The DataGrid project Ferrara Farm Configuration First SP submissions through the Grid Work in Progress Future Plans Conclusions
Ferrara - 12/02/033 INFN-GRID To manage and use computing resources distributed on Garr-b sites Deployment of Testbed sites, in order to validate EDG software release and to adapt them to High Energy Physics requests Special project of INFN European DataGrid and INFN-GRID Funded by European Union Computing Grids permit: High Throughput Computing Analysis of large dimension data Sharing resources and data Applications involved: Biomedical Sciences Earth Observation High Energy Physics EDG CT BO PD TO CA ROMA MI To USA To Russia/Japan Current prototipe of INFN DataGrid testbed connected to EDG testbed – US and ASIA Cern FE R.A.L Manchester
Ferrara - 12/02/034 EDG Architecture and Services OS & Net services APPLICATION Layer ALICEATLASCMSLHCbBaBar High level GRID middleware Basics Services GLOBUS toolkit DataGRID Architecture
Ferrara - 12/02/035 CE / WN SE u The DataGrid Testbed consists of different types of machines (Grid Elements). u In Ferrara the farm is composed by one Computing Element (CE), three Worker Nodes (WN), one User Interface (UI) and one Storage Element (SE). u All these machines are managed by a LCFGng (Local ConFiGuration system new generation) server and they are automatically configured. UI LCFGngServer Grid Elements in Ferrara
Ferrara - 12/02/036 UI (User Interface): component for accessing to the workload management system. Users can submit a job and retrieve the output, they sholud have an account and a personal certificate installed in their home directory. UI Certificate Authorities To access the GRID you have to request a certificate to a certification authority. INFN-GRID users can obtain a certificate from the INFN CA ( User Interface To use the BaBar Grid, you must register that certificate with the BaBar Virtual Organisation (BaBar VO).
Ferrara - 12/02/037 Job Submission Storage Element UI JDL Logging & Book-keeping (LB) Resource Broker (RB) Job Submission Service (JSS) ComputingElement Information Service (IS) Replica Catalogue Job Submit Event Input Sandbox Job Status submitted waitingreadyscheduled Input Sandbox running Job Status done outputready Job Status Output Sandbox cleared
Ferrara - 12/02/038 SPGrid Farm in Ferrara LCFGng Server Management Lock server Data server SCSI 250 GB IR0AD SE CE-WN UI Ferrara - EDG RB CNAF - BolognaCERN
Ferrara - 12/02/039 Configuration INFN Grid Testbed Status: EDG (RedHat 6.2). A BaBar software special release (12.3.2y) has been built and installed to: Write Kanga files Run Moose on RH 6.2 A special tag of ProdTools has been installed to perform tests. A pool of BaBar accounts (babar000, babar001,…) has been created in the EDG farm of Ferrara. Each member of BaBar VO is able to submit jobs to the farm of Ferrara through the RB located at CNAF (grid009g.cnaf.infn.it).
Ferrara - 12/02/0310 Current Status Created a JDL file to run Moose on Grid resources. Created scripts containing EDG commands to submit jobs, to check their status and retrieve output files. An user can submit a range of runs. For each run a job is created and submitted to the Resource Broker, then it is sent to the Ferrara CE (grid0.fe.infn.it). The output file is then transferred to the closest SE (grid2.fe.infn.it).
Ferrara - 12/02/0311 Moose.jdl grid1> more Moose.jdl Executable ="Moose.csh"; InputSandbox ={"Moose.csh",".cshrc","config.csh"}; StdOutput ="Moose.txt"; StdError ="Moose.log"; OutputSandbox ={"Moose.txt","Moose.log"}; General environment configurations Config file for BaBar. Similar to SP standard scripts Similar to SP standard scripts (Job.Xsh) […] tar -czvf run${RUNNUM}.tar.gz *.root globus-url-copy -vb file://`pwd`/run${RUNNUM}.tar.gz \ gsiftp://grid2.fe.infn.it/flatfiles/SE00/paolo/run${RUNNUM}.tar.gz Globus command: To copy output files from WN to SE
Ferrara - 12/02/0312 grid1> more launch #!/bin/tcsh num_f = fin = $2 while ( $num_f <= $fin ) ####build the run directories […] ####build a config.csh with the appropriate environment variables echo "#\!/bin/tcsh -v" > config.csh […] #### now run the job dg-job-submit -o run$num_f.jobid -r \ grid0.fe.infn.it:2119/jobmanager-pbs-long Moose.jdl num_f++ end The launch script A config file is created for each run EDG job submission command Range of runs to submit For each run a job is created runtime
Ferrara - 12/02/0313 Job Submission grid1> ls Moose.csh Moose.jdl config.csh launch monitor retrieve grid1>./launch […] dg-job-submit -o run$num_f.jobid -r grid0.fe.infn.it:2119/jobmanager-pbs-long Moose.jdl Connecting to host grid009g.cnaf.infn.it, port 7771 Logging to host grid009g.cnaf.infn.it, port ================== dg-job-submit Success ================ The job has been successfully submitted to the Resource Broker. Use dg-job-status command to check job current status. Your job identifier (dg_jobId) is: The dg_jobId has been saved in the following file: /home/enrica/stress/ /run jobid […] grid1> ls / Moose.csh Moose.jdl config.csh run jobid grid1> ls / Moose.csh Moose.jdl config.csh run jobid Range of runs to submit Job ID CNAF RB
Ferrara - 12/02/0314 grid1>./monitor Run is Status = Scheduled Status Reason = initial Run is Status = Scheduled Status Reason = initial grid1>./monitor Run is Status = Running Status Run is Status = Running Status The monitor script grid1>./monitor Run is Status = Ready Status Reason = job accepted Run is Status = Ready Status Reason = job accepted grid1> more monitor num_f = fin = $2 while ( $num_f <= $fin ) echo Run $num_f is `dg-job-status -i \ $num_f/run$num_f.jobid | grep num_f++ end EDG command grid1>./monitor Run is Status = OutputReady Status Reason = terminated Run is Status = OutputReady Status Reason = terminated
Ferrara - 12/02/0315 The retrieve script grid1> more retrieve #!/bin/tcsh num_f = fin = $2 while ( $num_f <= $fin ) cd $num_f #### get logfiles dg-job-get-output -i run$num_f.jobid --dir $PWD #### get rootfiles globus-url-copy \ gsiftp://grid2.fe.infn.it/flatfiles/SE00/paolo/run$num_f.tar.gz \ file://`pwd`/run$num_f.tar.gz tar -xzvf run$num_f.tar.gz rm -f run$num_f.tar.gz #### delete rootfiles form SE globus-job-run grid2.fe.infn.it /bin/rm \ /flatfiles/SE00/paolo/run$num_f.tar.gz num_f++ end EDG command Globus command: Direct copy of file from SE to UI Globus command: delete file fromSE
Ferrara - 12/02/0316 Retrieving Output grid1> ls Moose.csh Moose.jdl config.csh launch monitor retrieve grid1> ls / Moose.jdl rootdef-tru.root Moose.csh config.csh run jobid rootdef-tag.root rootdef-aod.root grid1> ls / Moose.jdl Moose.csh config.csh run jobid rootdef-tru.root rootdef-tag.root rootdef-aod.root grid1> ls / / Moose.log Moose.txt grid1> ls / / Moose.log Moose.txt
Ferrara - 12/02/0317 Future Plans Data server Lock server Objectivity DB RB(UK) Ferrara MOOSE RPM 3) Install Objy DB on the SE 2) MOOSE in RPM format 1) Use of IC RB and others Integration of Moose Application with Application with EDG software releases SPGrid Farm Management LCFGng Server SE CE-WN UI
Ferrara - 12/02/0318 Documentation The DataGrid Project: EDG tutorials Archive Web Site: INFN-Grid Testbed: BaBar-Grid: Status of the Farm in Ferrara: