1 // Copyright (C) 2007-2010 CEA/DEN, EDF R&D, OPEN CASCADE
3 // Copyright (C) 2003-2007 OPEN CASCADE, EADS/CCR, LIP6, CEA/DEN,
4 // CEDRAT, EDF R&D, LEG, PRINCIPIA R&D, BUREAU VERITAS
6 // This library is free software; you can redistribute it and/or
7 // modify it under the terms of the GNU Lesser General Public
8 // License as published by the Free Software Foundation; either
9 // version 2.1 of the License.
11 // This library is distributed in the hope that it will be useful,
12 // but WITHOUT ANY WARRANTY; without even the implied warranty of
13 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 // Lesser General Public License for more details.
16 // You should have received a copy of the GNU Lesser General Public
17 // License along with this library; if not, write to the Free Software
18 // Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 // See http://www.salome-platform.org/ or email : webmaster.salome@opencascade.com
23 * BatchManager_ePBS.cxx : emulation of PBS client
25 * Auteur : Bernard SECHER - CEA DEN, André RIBES - EDF R&D
26 * Mail : mailto:bernard.secher@cea.fr
27 * Date : Thu Apr 24 10:17:22 2008
42 #include <Batch_config.h>
50 #include "Batch_BatchManager_ePBS.hxx"
51 #include "Batch_JobInfo_ePBS.hxx"
57 BatchManager_ePBS::BatchManager_ePBS(const FactBatchManager * parent, const char * host,
58 CommunicationProtocolType protocolType, const char * mpiImpl,
60 : BatchManager_eClient(parent, host, protocolType, mpiImpl),
61 BatchManager(parent, host)
64 _nb_proc_per_node = nb_proc_per_node;
68 BatchManager_ePBS::~BatchManager_ePBS()
73 // Methode pour le controle des jobs : soumet un job au gestionnaire
74 const JobId BatchManager_ePBS::submitJob(const Job & job)
77 Parametre params = job.getParametre();
78 const std::string workDir = params[WORKDIR];
79 const string fileToExecute = params[EXECUTABLE];
80 string::size_type p1 = fileToExecute.find_last_of("/");
81 string::size_type p2 = fileToExecute.find_last_of(".");
82 std::string fileNameToExecute = fileToExecute.substr(p1+1,p2-p1-1);
84 // export input files on cluster
85 exportInputFiles(job);
87 // build batch script for job
88 buildBatchScript(job);
90 // define name of log file (local)
91 string logFile = generateTemporaryFileName("PBS-submitlog");
93 // define command to submit batch
94 string subCommand = string("cd ") + workDir + "; qsub " + fileNameToExecute + "_Batch.sh";
95 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
99 cerr << command.c_str() << endl;
100 status = system(command.c_str());
103 ifstream error_message(logFile.c_str());
106 while(std::getline(error_message, temp))
108 error_message.close();
109 throw EmulationException("Error of connection on remote host, error was: " + mess);
112 // read id of submitted job in log file
113 ifstream idfile(logFile.c_str());
117 if (sline.size() == 0)
118 throw EmulationException("Error in the submission of the job on the remote host");
120 JobId id(this, sline);
124 // Methode pour le controle des jobs : retire un job du gestionnaire
125 void BatchManager_ePBS::deleteJob(const JobId & jobid)
129 istringstream iss(jobid.getReference());
132 // define command to delete batch
133 string subCommand = string("qdel ") + iss.str();
134 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
135 cerr << command.c_str() << endl;
136 status = system(command.c_str());
138 throw EmulationException("Error of connection on remote host");
140 cerr << "jobId = " << ref << "killed" << endl;
143 // Methode pour le controle des jobs : suspend un job en file d'attente
144 void BatchManager_ePBS::holdJob(const JobId & jobid)
146 throw EmulationException("Not yet implemented");
149 // Methode pour le controle des jobs : relache un job suspendu
150 void BatchManager_ePBS::releaseJob(const JobId & jobid)
152 throw EmulationException("Not yet implemented");
156 // Methode pour le controle des jobs : modifie un job en file d'attente
157 void BatchManager_ePBS::alterJob(const JobId & jobid, const Parametre & param, const Environnement & env)
159 throw EmulationException("Not yet implemented");
162 // Methode pour le controle des jobs : modifie un job en file d'attente
163 void BatchManager_ePBS::alterJob(const JobId & jobid, const Parametre & param)
165 alterJob(jobid, param, Environnement());
168 // Methode pour le controle des jobs : modifie un job en file d'attente
169 void BatchManager_ePBS::alterJob(const JobId & jobid, const Environnement & env)
171 alterJob(jobid, Parametre(), env);
174 // Methode pour le controle des jobs : renvoie l'etat du job
175 JobInfo BatchManager_ePBS::queryJob(const JobId & jobid)
178 istringstream iss(jobid.getReference());
181 // define name of log file (local)
182 string logFile = generateTemporaryFileName(string("PBS-querylog-id") + jobid.getReference());
184 // define command to query batch
185 string subCommand = string("qstat -f ") + iss.str();
186 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
189 cerr << command.c_str() << endl;
190 int status = system(command.c_str());
191 if(status && status != 153 && status != 256*153)
192 throw EmulationException("Error of connection on remote host");
194 JobInfo_ePBS ji = JobInfo_ePBS(id,logFile);
198 // Methode pour le controle des jobs : teste si un job est present en machine
199 bool BatchManager_ePBS::isRunning(const JobId & jobid)
201 throw EmulationException("Not yet implemented");
204 void BatchManager_ePBS::buildBatchScript(const Job & job)
206 std::cerr << "BuildBatchScript" << std::endl;
207 Parametre params = job.getParametre();
208 Environnement env = job.getEnvironnement();
212 string fileToExecute = "";
218 // Mandatory parameters
219 if (params.find(WORKDIR) != params.end())
220 workDir = params[WORKDIR].str();
222 throw EmulationException("params[WORKDIR] is not defined ! Please defined it, cannot submit this job");
223 if (params.find(EXECUTABLE) != params.end())
224 fileToExecute = params[EXECUTABLE].str();
226 throw EmulationException("params[EXECUTABLE] is not defined ! Please defined it, cannot submit this job");
228 // Optional parameters
229 if (params.find(NBPROC) != params.end())
230 nbproc = params[NBPROC];
231 if (params.find(MAXWALLTIME) != params.end())
232 edt = params[MAXWALLTIME];
233 if (params.find(MAXRAMSIZE) != params.end())
234 mem = params[MAXRAMSIZE];
235 if (params.find(QUEUE) != params.end())
236 queue = params[QUEUE].str();
238 string::size_type p1 = fileToExecute.find_last_of("/");
239 string::size_type p2 = fileToExecute.find_last_of(".");
240 string rootNameToExecute = fileToExecute.substr(p1+1,p2-p1-1);
241 string fileNameToExecute = fileToExecute.substr(p1+1);
243 // Create batch submit file
244 ofstream tempOutputFile;
245 std::string TmpFileName = createAndOpenTemporaryFile("PBS-script", tempOutputFile);
247 tempOutputFile << "#! /bin/sh -f" << endl;
250 // Division - arrondi supérieur
251 int nodes_requested = (nbproc + _nb_proc_per_node -1) / _nb_proc_per_node;
252 tempOutputFile << "#PBS -l nodes=" << nodes_requested << ":ppn=" << _nb_proc_per_node << endl;
255 tempOutputFile << "#PBS -q " << queue << endl;
257 tempOutputFile << "#PBS -l walltime=" << edt*60 << endl;
259 tempOutputFile << "#PBS -l mem=" << mem << "kb" << endl;
260 tempOutputFile << "#PBS -o " << workDir << "/logs/output.log." << rootNameToExecute << endl;
261 tempOutputFile << "#PBS -e " << workDir << "/logs/error.log." << rootNameToExecute << endl;
263 // Define environment for the job
265 tempOutputFile << "#PBS -v ";
266 Environnement::const_iterator iter;
267 for (iter = env.begin() ; iter != env.end() ; ++iter) {
268 tempOutputFile << iter->first << "=" << iter->second << ",";
270 tempOutputFile << endl;
273 // Abstraction of PBS_NODEFILE - TODO
274 tempOutputFile << "export LIBBATCH_NODEFILE=$PBS_NODEFILE" << endl;
276 // Launch the executable
277 tempOutputFile << "cd " << workDir << endl;
278 tempOutputFile << "./" + fileNameToExecute << endl;
279 tempOutputFile.flush();
280 tempOutputFile.close();
282 BATCH_CHMOD(TmpFileName.c_str(), 0x1ED);
283 cerr << "Batch script file generated is: " << TmpFileName.c_str() << endl;
285 int status = _protocol.copyFile(TmpFileName, "", "",
286 workDir + "/" + rootNameToExecute + "_Batch.sh",
287 _hostname, _username);
289 throw EmulationException("Error of connection on remote host, cannot copy batch submission file");