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_Constants.hxx"
51 #include "Batch_BatchManager_ePBS.hxx"
52 #include "Batch_JobInfo_ePBS.hxx"
58 BatchManager_ePBS::BatchManager_ePBS(const FactBatchManager * parent, const char * host,
59 const char * username,
60 CommunicationProtocolType protocolType, const char * mpiImpl,
62 : BatchManager(parent, host),
63 BatchManager_eClient(parent, host, username, protocolType, mpiImpl)
66 _nb_proc_per_node = nb_proc_per_node;
70 BatchManager_ePBS::~BatchManager_ePBS()
75 // Methode pour le controle des jobs : soumet un job au gestionnaire
76 const JobId BatchManager_ePBS::submitJob(const Job & job)
79 Parametre params = job.getParametre();
80 const std::string workDir = params[WORKDIR];
81 const string fileToExecute = params[EXECUTABLE];
82 string::size_type p1 = fileToExecute.find_last_of("/");
83 string::size_type p2 = fileToExecute.find_last_of(".");
84 std::string fileNameToExecute = fileToExecute.substr(p1+1,p2-p1-1);
86 // export input files on cluster
87 exportInputFiles(job);
89 // build batch script for job
90 buildBatchScript(job);
92 // define name of log file (local)
93 string logFile = generateTemporaryFileName("PBS-submitlog");
95 // define command to submit batch
96 string subCommand = string("cd ") + workDir + "; qsub " + fileNameToExecute + "_Batch.sh";
97 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
101 cerr << command.c_str() << endl;
102 status = system(command.c_str());
105 ifstream error_message(logFile.c_str());
108 while(std::getline(error_message, temp))
110 error_message.close();
111 throw EmulationException("Error of connection on remote host, error was: " + mess);
114 // read id of submitted job in log file
115 ifstream idfile(logFile.c_str());
119 if (sline.size() == 0)
120 throw EmulationException("Error in the submission of the job on the remote host");
122 JobId id(this, sline);
126 // Ce manager permet de faire de la reprise
128 BatchManager_ePBS::addJob(const Batch::Job & job, const std::string reference)
130 return JobId(this, reference);
133 // Methode pour le controle des jobs : retire un job du gestionnaire
134 void BatchManager_ePBS::deleteJob(const JobId & jobid)
138 istringstream iss(jobid.getReference());
141 // define command to delete batch
142 string subCommand = string("qdel ") + iss.str();
143 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
144 cerr << command.c_str() << endl;
145 status = system(command.c_str());
147 throw EmulationException("Error of connection on remote host");
149 cerr << "jobId = " << ref << "killed" << endl;
152 // Methode pour le controle des jobs : suspend un job en file d'attente
153 void BatchManager_ePBS::holdJob(const JobId & jobid)
155 throw EmulationException("Not yet implemented");
158 // Methode pour le controle des jobs : relache un job suspendu
159 void BatchManager_ePBS::releaseJob(const JobId & jobid)
161 throw EmulationException("Not yet implemented");
165 // Methode pour le controle des jobs : modifie un job en file d'attente
166 void BatchManager_ePBS::alterJob(const JobId & jobid, const Parametre & param, const Environnement & env)
168 throw EmulationException("Not yet implemented");
171 // Methode pour le controle des jobs : modifie un job en file d'attente
172 void BatchManager_ePBS::alterJob(const JobId & jobid, const Parametre & param)
174 alterJob(jobid, param, Environnement());
177 // Methode pour le controle des jobs : modifie un job en file d'attente
178 void BatchManager_ePBS::alterJob(const JobId & jobid, const Environnement & env)
180 alterJob(jobid, Parametre(), env);
183 // Methode pour le controle des jobs : renvoie l'etat du job
184 JobInfo BatchManager_ePBS::queryJob(const JobId & jobid)
187 istringstream iss(jobid.getReference());
190 // define name of log file (local)
191 string logFile = generateTemporaryFileName(string("PBS-querylog-id") + jobid.getReference());
193 // define command to query batch
194 string subCommand = string("qstat -f ") + iss.str();
195 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
198 cerr << command.c_str() << endl;
199 int status = system(command.c_str());
200 if(status && status != 153 && status != 256*153)
201 throw EmulationException("Error of connection on remote host");
203 JobInfo_ePBS ji = JobInfo_ePBS(id,logFile);
207 // Methode pour le controle des jobs : teste si un job est present en machine
208 bool BatchManager_ePBS::isRunning(const JobId & jobid)
210 throw EmulationException("Not yet implemented");
213 void BatchManager_ePBS::buildBatchScript(const Job & job)
215 std::cerr << "BuildBatchScript" << std::endl;
216 Parametre params = job.getParametre();
217 Environnement env = job.getEnvironnement();
221 string fileToExecute = "";
227 // Mandatory parameters
228 if (params.find(WORKDIR) != params.end())
229 workDir = params[WORKDIR].str();
231 throw EmulationException("params[WORKDIR] is not defined ! Please defined it, cannot submit this job");
232 if (params.find(EXECUTABLE) != params.end())
233 fileToExecute = params[EXECUTABLE].str();
235 throw EmulationException("params[EXECUTABLE] is not defined ! Please defined it, cannot submit this job");
237 // Optional parameters
238 if (params.find(NBPROC) != params.end())
239 nbproc = params[NBPROC];
240 if (params.find(MAXWALLTIME) != params.end())
241 edt = params[MAXWALLTIME];
242 if (params.find(MAXRAMSIZE) != params.end())
243 mem = params[MAXRAMSIZE];
244 if (params.find(QUEUE) != params.end())
245 queue = params[QUEUE].str();
247 string::size_type p1 = fileToExecute.find_last_of("/");
248 string::size_type p2 = fileToExecute.find_last_of(".");
249 string rootNameToExecute = fileToExecute.substr(p1+1,p2-p1-1);
250 string fileNameToExecute = fileToExecute.substr(p1+1);
252 // Create batch submit file
253 ofstream tempOutputFile;
254 std::string TmpFileName = createAndOpenTemporaryFile("PBS-script", tempOutputFile);
256 tempOutputFile << "#! /bin/sh -f" << endl;
259 // Division - arrondi supérieur
260 int nodes_requested = (nbproc + _nb_proc_per_node -1) / _nb_proc_per_node;
261 tempOutputFile << "#PBS -l nodes=" << nodes_requested << ":ppn=" << _nb_proc_per_node << endl;
264 tempOutputFile << "#PBS -q " << queue << endl;
266 tempOutputFile << "#PBS -l walltime=" << edt*60 << endl;
268 tempOutputFile << "#PBS -l mem=" << mem << "MB" << endl;
269 tempOutputFile << "#PBS -o " << workDir << "/logs/output.log." << rootNameToExecute << endl;
270 tempOutputFile << "#PBS -e " << workDir << "/logs/error.log." << rootNameToExecute << endl;
272 // Define environment for the job
274 tempOutputFile << "#PBS -v ";
275 Environnement::const_iterator iter;
276 for (iter = env.begin() ; iter != env.end() ; ++iter) {
277 tempOutputFile << iter->first << "=" << iter->second << ",";
279 tempOutputFile << endl;
282 // Abstraction of PBS_NODEFILE - TODO
283 tempOutputFile << "export LIBBATCH_NODEFILE=$PBS_NODEFILE" << endl;
285 // Launch the executable
286 tempOutputFile << "cd " << workDir << endl;
287 tempOutputFile << "./" + fileNameToExecute << endl;
288 tempOutputFile.flush();
289 tempOutputFile.close();
291 BATCH_CHMOD(TmpFileName.c_str(), 0x1ED);
292 cerr << "Batch script file generated is: " << TmpFileName.c_str() << endl;
294 int status = _protocol.copyFile(TmpFileName, "", "",
295 workDir + "/" + rootNameToExecute + "_Batch.sh",
296 _hostname, _username);
298 throw EmulationException("Error of connection on remote host, cannot copy batch submission file");