1 // Copyright (C) 2007-2012 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 * Batch_BatchManager_eVishnu.cxx :
25 * Created on: 24 june 2011
26 * Author : Renaud BARATE - EDF R&D
38 #include <Batch_NotYetImplementedException.hxx>
39 #include <Batch_Constants.hxx>
40 #include <Batch_Utils.hxx>
42 #include "Batch_BatchManager_eVishnu.hxx"
43 #include "Batch_JobInfo_eVishnu.hxx"
49 BatchManager_eVishnu::BatchManager_eVishnu(const FactBatchManager * parent,
51 const char * username,
52 CommunicationProtocolType protocolType,
55 : BatchManager(parent, host),
56 // Force SH protocol for Vishnu
57 BatchManager_eClient(parent, host, username, SH, mpiImpl),
58 _nb_proc_per_node(nb_proc_per_node)
62 BatchManager_eVishnu::~BatchManager_eVishnu()
66 // Method to submit a job to the batch manager
67 const JobId BatchManager_eVishnu::submitJob(const Job & job)
69 // export input files on cluster
70 exportInputFiles(job);
72 // build command file to submit the job
73 string cmdFile = buildCommandFile(job);
75 // define extra parameters (that can not be defined in the command file)
76 Parametre params = job.getParametre();
77 ostringstream extraParams;
78 if (params.find(NBPROC) != params.end())
79 extraParams << "-P " << params[NBPROC] << " ";
80 if (params.find(MAXRAMSIZE) != params.end())
81 extraParams << "-m " << params[MAXRAMSIZE] << " ";
83 // define command to submit batch
84 string subCommand = string("export OMNIORB_CONFIG=$VISHNU_CONFIG_FILE; ");
85 subCommand += "vishnu_connect -p 2 && ";
86 subCommand += "vishnu_submit_job " + extraParams.str() + _hostname + " " + cmdFile;
87 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
89 cerr << command.c_str() << endl;
93 int status = Utils::getCommandOutput(command, output);
95 if (status != 0) throw EmulationException("Can't submit job, error was: " + output);
97 // find id of submitted job in output
98 string search = "Job Id : ";
99 string::size_type pos = output.find(search);
100 if (pos == string::npos)
101 throw EmulationException("Error in the submission of the job on the remote host");
102 pos += search.size();
103 string::size_type endl_pos = output.find('\n', pos);
104 string::size_type count = (endl_pos == string::npos)? string::npos : endl_pos - pos;
105 string jobref = output.substr(pos, count);
106 if (jobref.size() == 0)
107 throw EmulationException("Error in the submission of the job on the remote host");
109 JobId id(this, jobref);
114 void BatchManager_eVishnu::exportInputFiles(const Job& job)
116 Parametre params = job.getParametre();
117 string subCommand = string("export OMNIORB_CONFIG=$VISHNU_CONFIG_FILE; ");
118 subCommand += "vishnu_connect -p 2 && ";
120 // create remote directories
121 subCommand += "vishnu_create_dir -p " + _hostname + ":" + params[WORKDIR].str() + "/logs && ";
124 string executeFile = params[EXECUTABLE];
125 if (executeFile.size() != 0) {
126 subCommand += "vishnu_copy_file " + executeFile + " " +
127 _hostname + ":" + params[WORKDIR].str() + "/ && ";
130 // copy filesToExportList
131 const Versatile & V = params[INFILE];
132 Versatile::const_iterator Vit;
133 for (Vit=V.begin(); Vit!=V.end(); Vit++) {
134 CoupleType cpt = *static_cast< CoupleType * >(*Vit);
135 Couple inputFile = cpt;
137 // Get absolute paths
138 char * buf = getcwd(NULL, 0);
142 string absremote = (Utils::isAbsolutePath(inputFile.getRemote()))?
143 inputFile.getRemote() :
144 params[WORKDIR].str() + "/" + inputFile.getRemote();
145 string abslocal = (Utils::isAbsolutePath(inputFile.getLocal()))?
146 inputFile.getLocal() :
147 cwd + "/" + inputFile.getLocal();
149 if (Vit != V.begin())
150 subCommand += " && ";
151 subCommand += "vishnu_copy_file " + abslocal + " " + _hostname + ":" + absremote;
155 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
157 cerr << command.c_str() << endl;
159 int status = Utils::getCommandOutput(command, output);
162 throw EmulationException("Can't copy input files, error was: " + output);
166 * Create Vishnu command file and copy it on the server.
167 * Return the name of the remote file.
169 string BatchManager_eVishnu::buildCommandFile(const Job & job)
171 Parametre params = job.getParametre();
175 string fileToExecute = "";
178 // Mandatory parameters
179 if (params.find(WORKDIR) != params.end())
180 workDir = params[WORKDIR].str();
182 throw EmulationException("params[WORKDIR] is not defined. Please define it, cannot submit this job.");
183 if (params.find(EXECUTABLE) != params.end())
184 fileToExecute = params[EXECUTABLE].str();
186 throw EmulationException("params[EXECUTABLE] is not defined. Please define it, cannot submit this job.");
188 string::size_type p1 = fileToExecute.find_last_of("/");
189 string::size_type p2 = fileToExecute.find_last_of(".");
190 string rootNameToExecute = fileToExecute.substr(p1+1,p2-p1-1);
191 string fileNameToExecute = fileToExecute.substr(p1+1);
193 // Create batch submit file
194 ofstream tempOutputFile;
195 string tmpFileName = createAndOpenTemporaryFile("vishnu-script", tempOutputFile);
197 tempOutputFile << "#!/bin/sh" << endl;
198 tempOutputFile << "#% vishnu_output=" << workDir << "/logs/output.log." << rootNameToExecute << endl;
199 tempOutputFile << "#% vishnu_error=" << workDir << "/logs/error.log." << rootNameToExecute << endl;
201 if (params.find(NAME) != params.end())
202 tempOutputFile << "#% vishnu_job_name=\"" << params[NAME] << "\"" << endl;
204 // Optional parameters
205 if (params.find(MAXWALLTIME) != params.end()) {
206 long totalMinutes = params[MAXWALLTIME];
207 long h = totalMinutes / 60;
208 long m = totalMinutes - h * 60;
209 tempOutputFile << "#% vishnu_wallclocklimit=" << h << ":";
211 tempOutputFile << "0";
212 tempOutputFile << m << ":00" << endl;
214 if (params.find(QUEUE) != params.end())
215 tempOutputFile << "#% vishnu_queue=" << params[QUEUE] << endl;
217 // Define environment for the job
218 Environnement env = job.getEnvironnement();
219 for (Environnement::const_iterator iter = env.begin() ; iter != env.end() ; ++iter) {
220 tempOutputFile << "export " << iter->first << "=" << iter->second << endl;
224 tempOutputFile << "export LIBBATCH_NODEFILE=$VISHNU_BATCHJOB_NODEFILE" << endl;
226 // Launch the executable
227 tempOutputFile << "cd " << workDir << endl;
228 tempOutputFile << "./" + fileNameToExecute << endl;
230 tempOutputFile.flush();
231 tempOutputFile.close();
233 cerr << "Batch script file generated is: " << tmpFileName << endl;
237 void BatchManager_eVishnu::deleteJob(const JobId & jobid)
239 // define command to delete job
240 string subCommand = string("export OMNIORB_CONFIG=$VISHNU_CONFIG_FILE; ");
241 subCommand += "vishnu_connect -p 2 && ";
242 subCommand += "vishnu_cancel_job " + _hostname + " " + jobid.getReference();
243 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
244 cerr << command.c_str() << endl;
246 int status = system(command.c_str());
248 throw EmulationException("Can't delete job " + jobid.getReference());
250 cerr << "job " << jobid.getReference() << " killed" << endl;
253 void BatchManager_eVishnu::holdJob(const JobId & jobid)
255 throw NotYetImplementedException("BatchManager_eVishnu::holdJob");
258 void BatchManager_eVishnu::releaseJob(const JobId & jobid)
260 throw NotYetImplementedException("BatchManager_eVishnu::releaseJob");
263 void BatchManager_eVishnu::alterJob(const JobId & jobid, const Parametre & param, const Environnement & env)
265 throw NotYetImplementedException("BatchManager_eVishnu::alterJob");
268 void BatchManager_eVishnu::alterJob(const JobId & jobid, const Parametre & param)
270 throw NotYetImplementedException("BatchManager_eVishnu::alterJob");
273 void BatchManager_eVishnu::alterJob(const JobId & jobid, const Environnement & env)
275 throw NotYetImplementedException("BatchManager_eVishnu::alterJob");
278 JobInfo BatchManager_eVishnu::queryJob(const JobId & jobid)
280 // define command to query batch
281 string subCommand = string("export OMNIORB_CONFIG=$VISHNU_CONFIG_FILE; ");
282 subCommand += "vishnu_connect -p 2 && ";
283 subCommand += "vishnu_get_job_info " + _hostname + " " + jobid.getReference();
284 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
285 cerr << command.c_str() << endl;
288 int status = Utils::getCommandOutput(command, output);
290 throw EmulationException("Can't query job " + jobid.getReference());
291 JobInfo_eVishnu jobinfo = JobInfo_eVishnu(jobid.getReference(), output);
295 const JobId BatchManager_eVishnu::addJob(const Job & job, const string reference)
297 return JobId(this, reference);
300 void BatchManager_eVishnu::importOutputFiles(const Job & job, const std::string directory)
302 // Create local result directory
303 char * buf = getcwd(NULL, 0);
306 string absdir = (Utils::isAbsolutePath(directory))? directory : cwd + "/" + directory;
307 int status = CommunicationProtocol::getInstance(SH).makeDirectory(absdir, "", "");
309 throw EmulationException("Can't create result directory");
312 string subCommand = string("export OMNIORB_CONFIG=$VISHNU_CONFIG_FILE; ");
313 subCommand += "vishnu_connect -p 2 && ";
316 Parametre params = job.getParametre();
317 const Versatile & V = params[OUTFILE];
318 Versatile::const_iterator Vit;
319 for (Vit=V.begin(); Vit!=V.end(); Vit++) {
320 CoupleType cpt = *static_cast< CoupleType * >(*Vit);
321 Couple outputFile = cpt;
323 // Get absolute paths
324 string absremote = (Utils::isAbsolutePath(outputFile.getRemote()))?
325 outputFile.getRemote() :
326 params[WORKDIR].str() + "/" + outputFile.getRemote();
327 string abslocal = (Utils::isAbsolutePath(outputFile.getLocal()))?
328 outputFile.getLocal() :
329 absdir + "/" + outputFile.getLocal();
331 subCommand += "vishnu_copy_file " + _hostname + ":" + absremote + " " + abslocal + " && ";
335 subCommand += "vishnu_copy_file -r " +_hostname + ":" + params[WORKDIR].str() + "/logs" + " " + absdir;
338 string command = _protocol.getExecCommand(subCommand, _hostname, _username);
340 cerr << command.c_str() << endl;
342 status = Utils::getCommandOutput(command, output);
345 throw EmulationException("Can't import output files, error was: " + output);