1 // Copyright (C) 2007-2013 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_SGE.cxx : emulation of SGE client
25 * Auteur : Bernard SECHER - CEA DEN
26 * Mail : mailto:bernard.secher@cea.fr
27 * Date : Thu Apr 24 10:17:22 2008
48 #include <Constants.hxx>
50 #include <NotYetImplementedException.hxx>
52 #include "BatchManager_SGE.hxx"
53 #include "JobInfo_SGE.hxx"
60 BatchManager_SGE::BatchManager_SGE(const FactBatchManager * parent, const char * host,
61 const char * username,
62 CommunicationProtocolType protocolType, const char * mpiImpl)
63 : BatchManager(parent, host, username, protocolType, mpiImpl)
69 BatchManager_SGE::~BatchManager_SGE()
74 // Methode pour le controle des jobs : soumet un job au gestionnaire
75 const JobId BatchManager_SGE::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 command to submit batch
91 string subCommand = string("bash -l -c \\\"cd ") + workDir + "; qsub " + fileNameToExecute + "_Batch.sh\\\"";
92 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
98 int status = Utils::getCommandOutput(command, output);
100 if (status != 0) throw RunTimeException("Can't submit job, error was: " + output);
102 // find id of submitted job in output
104 istringstream iss(output);
105 iss >> strjob >> strjob >> strjob;
107 JobId id(this, strjob);
112 // Methode pour le controle des jobs : retire un job du gestionnaire
113 void BatchManager_SGE::deleteJob(const JobId & jobid)
117 istringstream iss(jobid.getReference());
120 // define command to delete batch
121 string subCommand = string("bash -l -c \\\"qdel ") + iss.str() + string("\\\"");
122 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
124 status = system(command.c_str());
126 throw RunTimeException("Error of connection on remote host");
128 LOG("jobId = " << ref << "killed");
131 // Methode pour le controle des jobs : renvoie l'etat du job
132 JobInfo BatchManager_SGE::queryJob(const JobId & jobid)
135 istringstream iss(jobid.getReference());
138 // define command to query batch
139 string subCommand = string("bash -l -c \\\"qstat | grep ") + iss.str() + string("\\\"");
140 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
144 int status = Utils::getCommandOutput(command, output);
145 if (status && status != 256)
146 throw RunTimeException("Error of connection on remote host");
148 JobInfo_SGE ji = JobInfo_SGE(id, output);
152 // Methode pour le controle des jobs : teste si un job est present en machine
153 bool BatchManager_SGE::isRunning(const JobId & jobid)
155 throw NotYetImplementedException("BatchManager_SGE::isRunning");
158 void BatchManager_SGE::buildBatchScript(const Job & job)
161 //TODO porting on Win32 platform
162 LOG("BuildBatchScript");
163 Parametre params = job.getParametre();
167 string fileToExecute = "";
173 // Mandatory parameters
174 if (params.find(WORKDIR) != params.end())
175 workDir = params[WORKDIR].str();
177 throw RunTimeException("params[WORKDIR] is not defined ! Please defined it, cannot submit this job");
178 if (params.find(EXECUTABLE) != params.end())
179 fileToExecute = params[EXECUTABLE].str();
181 throw RunTimeException("params[EXECUTABLE] is not defined ! Please defined it, cannot submit this job");
183 // Optional parameters
184 if (params.find(NBPROC) != params.end())
185 nbproc = params[NBPROC];
186 if (params.find(MAXWALLTIME) != params.end())
187 edt = params[MAXWALLTIME];
188 if (params.find(MAXRAMSIZE) != params.end())
189 mem = params[MAXRAMSIZE];
190 if (params.find(QUEUE) != params.end())
191 queue = params[QUEUE].str();
193 string::size_type p1 = fileToExecute.find_last_of("/");
194 string::size_type p2 = fileToExecute.find_last_of(".");
195 string rootNameToExecute = fileToExecute.substr(p1+1,p2-p1-1);
196 string fileNameToExecute = fileToExecute.substr(p1+1);
198 // Create batch submit file
199 ofstream tempOutputFile;
200 std::string TmpFileName = Utils::createAndOpenTemporaryFile("SGE-script", tempOutputFile);
202 tempOutputFile << "#! /bin/sh -f" << endl;
204 tempOutputFile << "#$ -q " << queue << endl;
205 tempOutputFile << "#$ -pe " << _mpiImpl->name() << " " << nbproc << endl;
207 tempOutputFile << "#$ -l h_rt=" << getWallTime(edt) << endl ;
209 tempOutputFile << "#$ -l h_vmem=" << mem << "M" << endl ;
210 tempOutputFile << "#$ -o " << workDir << "/logs/output.log." << rootNameToExecute << endl ;
211 tempOutputFile << "#$ -e " << workDir << "/logs/error.log." << rootNameToExecute << endl ;
213 // Abstraction of PBS_NODEFILE - TODO
214 tempOutputFile << "export LIBBATCH_NODEFILE=$TMPDIR/machines" << endl;
216 // Launch the executable
217 tempOutputFile << "cd " << workDir << endl ;
218 tempOutputFile << "./" + fileNameToExecute << endl;
219 tempOutputFile.flush();
220 tempOutputFile.close();
222 Utils::chmod(TmpFileName.c_str(), 0x1ED);
223 LOG("Batch script file generated is: " << TmpFileName.c_str());
225 int status = _protocol.copyFile(TmpFileName, "", "",
226 workDir + "/" + rootNameToExecute + "_Batch.sh",
227 _hostname, _username);
229 throw RunTimeException("Error of connection on remote host");
234 std::string BatchManager_SGE::getWallTime(const long edt)
241 oss << h << ":" << m;
243 oss << h << ":0" << m;
244 oss << ":00"; // the seconds