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 * BatchManager_Local.cxx :
25 * Auteur : Ivan DUTKA-MALEN - EDF R&D
26 * Mail : mailto:ivan.dutka-malen@der.edf.fr
27 * Date : Thu Nov 6 10:17:22 2003
30 * Refactored by Renaud Barate (EDF R&D) in September 2009 to use
31 * CommunicationProtocol classes and merge Local_SH, Local_RSH and Local_SSH batch
41 #include <sys/types.h>
54 #include "Constants.hxx"
55 #include "IOMutex.hxx"
56 #include "BatchManager_Local.hxx"
57 #include "RunTimeException.hxx"
67 BatchManager_Local::BatchManager_Local(const Batch::FactBatchManager * parent, const char * host,
68 const char * username,
69 CommunicationProtocolType protocolType, const char * mpiImpl)
70 : BatchManager(parent, host, username, protocolType, mpiImpl), _connect(0),
73 pthread_mutex_init(&_threads_mutex, NULL);
74 pthread_cond_init(&_threadSyncCondition, NULL);
78 BatchManager_Local::~BatchManager_Local()
80 for (map<Id, Child>::iterator iter = _threads.begin() ; iter != _threads.end() ; iter++) {
81 pthread_mutex_lock(&_threads_mutex);
82 string state = iter->second.param[STATE];
83 if (state != FINISHED && state != FAILED) {
84 UNDER_LOCK( LOG("Warning: Job " << iter->first <<
85 " is not finished, it will now be canceled."));
86 pthread_cancel(iter->second.thread_id);
87 pthread_cond_wait(&_threadSyncCondition, &_threads_mutex);
89 pthread_mutex_unlock(&_threads_mutex);
91 pthread_mutex_destroy(&_threads_mutex);
92 pthread_cond_destroy(&_threadSyncCondition);
95 // Methode pour le controle des jobs : soumet un job au gestionnaire
96 const JobId BatchManager_Local::submitJob(const Job & job)
98 // export input files in the working directory of the execution host
99 exportInputFiles(job);
101 Job_Local jobLocal = job;
102 Id id = _idCounter++;
103 ThreadAdapter * p_ta = new ThreadAdapter(*this, job, id);
105 // Les attributs du thread a sa creation
106 pthread_attr_t thread_attr;
107 pthread_attr_init(&thread_attr);
108 pthread_attr_setdetachstate(&thread_attr, PTHREAD_CREATE_DETACHED);
110 // Creation du thread qui va executer la commande systeme qu'on lui passe
112 pthread_mutex_lock(&_threads_mutex);
113 int rc = pthread_create(&thread_id,
116 static_cast<void *>(p_ta));
118 // Liberation des zones memoire maintenant inutiles occupees par les attributs du thread
119 pthread_attr_destroy(&thread_attr);
122 pthread_mutex_unlock(&_threads_mutex);
123 throw RunTimeException("Can't create new thread in BatchManager_Local");
126 pthread_cond_wait(&_threadSyncCondition, &_threads_mutex);
127 pthread_mutex_unlock(&_threads_mutex);
129 ostringstream id_sst;
131 return JobId(this, id_sst.str());
134 // Methode pour le controle des jobs : retire un job du gestionnaire
135 void BatchManager_Local::deleteJob(const JobId & jobid)
139 istringstream iss(jobid.getReference());
142 // @@@ --------> SECTION CRITIQUE <-------- @@@
143 pthread_mutex_lock(&_threads_mutex);
144 bool idFound = (_threads.find(id) != _threads.end());
146 string state = _threads[id].param[STATE];
147 if (state != FINISHED && state != FAILED) {
148 pthread_cancel(_threads[id].thread_id);
149 pthread_cond_wait(&_threadSyncCondition, &_threads_mutex);
151 LOG("Cannot delete job " << jobid.getReference() << ". Job is already finished.");
154 pthread_mutex_unlock(&_threads_mutex);
155 // @@@ --------> SECTION CRITIQUE <-------- @@@
158 throw RunTimeException(string("Job with id ") + jobid.getReference() + " does not exist");
161 // Methode pour le controle des jobs : suspend un job en file d'attente
162 void BatchManager_Local::holdJob(const JobId & jobid)
165 istringstream iss(jobid.getReference());
168 UNDER_LOCK( LOG("BatchManager is sending HOLD command to the thread " << id) );
170 // On introduit une commande dans la queue du thread
171 // @@@ --------> SECTION CRITIQUE <-------- @@@
172 pthread_mutex_lock(&_threads_mutex);
173 if (_threads.find(id) != _threads.end())
174 _threads[id].command_queue.push(HOLD);
175 pthread_mutex_unlock(&_threads_mutex);
176 // @@@ --------> SECTION CRITIQUE <-------- @@@
179 // Methode pour le controle des jobs : relache un job suspendu
180 void BatchManager_Local::releaseJob(const JobId & jobid)
183 istringstream iss(jobid.getReference());
186 UNDER_LOCK( LOG("BatchManager is sending RELEASE command to the thread " << id) );
188 // On introduit une commande dans la queue du thread
189 // @@@ --------> SECTION CRITIQUE <-------- @@@
190 pthread_mutex_lock(&_threads_mutex);
191 if (_threads.find(id) != _threads.end())
192 _threads[id].command_queue.push(RELEASE);
193 pthread_mutex_unlock(&_threads_mutex);
194 // @@@ --------> SECTION CRITIQUE <-------- @@@
197 // Methode pour le controle des jobs : renvoie l'etat du job
198 JobInfo BatchManager_Local::queryJob(const JobId & jobid)
201 istringstream iss(jobid.getReference());
207 //UNDER_LOCK( cout << "JobInfo BatchManager_Local::queryJob(const JobId & jobid) : AVANT section critique" << endl );
208 // @@@ --------> SECTION CRITIQUE <-------- @@@
209 pthread_mutex_lock(&_threads_mutex);
210 std::map<Id, Child >::iterator pos = _threads.find(id);
211 bool found = (pos != _threads.end());
213 param = pos->second.param;
214 env = pos->second.env;
216 pthread_mutex_unlock(&_threads_mutex);
217 // @@@ --------> SECTION CRITIQUE <-------- @@@
218 //UNDER_LOCK( cout << "JobInfo BatchManager_Local::queryJob(const JobId & jobid) : APRES section critique" << endl );
220 if (!found) throw InvalidArgumentException("Invalid JobId argument for queryJob");
222 JobInfo_Local ji(param, env);
227 // Ce manager ne peut pas reprendre un job
228 // On force donc l'état du job à erreur - pour cela on ne donne pas d'Id
231 BatchManager_Local::addJob(const Batch::Job & job, const std::string & reference)
233 return JobId(this, "undefined");
236 // Methode pour le controle des jobs : teste si un job est present en machine
237 bool BatchManager_Local::isRunning(const JobId & jobid)
240 istringstream iss(jobid.getReference());
243 //UNDER_LOCK( cout << "JobInfo BatchManager_Local::queryJob(const JobId & jobid) : AVANT section critique" << endl );
244 // @@@ --------> SECTION CRITIQUE <-------- @@@
245 pthread_mutex_lock(&_threads_mutex);
246 bool running = (_threads[id].param[STATE].str() == RUNNING);
247 pthread_mutex_unlock(&_threads_mutex);
248 // @@@ --------> SECTION CRITIQUE <-------- @@@
249 //UNDER_LOCK( cout << "JobInfo BatchManager_Local::queryJob(const JobId & jobid) : APRES section critique" << endl );
254 string BatchManager_Local::ThreadAdapter::buildCommandFile(const Job_Local & job)
256 Parametre param = job.getParametre();
258 // Mandatory parameters
260 if (param.find(WORKDIR) != param.end())
261 workDir = param[WORKDIR].str();
263 throw RunTimeException("param[WORKDIR] is not defined. Please define it, cannot submit this job.");
264 string fileToExecute = "";
265 if (param.find(EXECUTABLE) != param.end())
266 fileToExecute = param[EXECUTABLE].str();
268 throw RunTimeException("param[EXECUTABLE] is not defined. Please define it, cannot submit this job.");
270 string::size_type p1 = fileToExecute.find_last_of("/");
271 string::size_type p2 = fileToExecute.find_last_of(".");
272 string rootNameToExecute = fileToExecute.substr(p1+1,p2-p1-1);
273 string fileNameToExecute = fileToExecute.substr(p1+1);
274 string remotePath = workDir + "/" + rootNameToExecute + "_launch_job";
276 // Create batch submit file
277 ofstream tempOutputFile;
278 string tmpFileName = Utils::createAndOpenTemporaryFile("launch-job", tempOutputFile);
281 if (_bm._protocol.getType() == SH) {
282 char drive[_MAX_DRIVE];
283 _splitpath(workDir.c_str(), drive, NULL, NULL, NULL);
284 if (strlen(drive) > 0) tempOutputFile << drive << endl;
285 tempOutputFile << "cd " << Utils::fixPath(workDir) << endl;
286 // Define environment for the job
287 Environnement env = job.getEnvironnement();
288 for (Environnement::const_iterator iter = env.begin() ; iter != env.end() ; ++iter) {
289 tempOutputFile << "set " << iter->first << "=" << iter->second << endl;
291 // Launch the executable
292 tempOutputFile << fileNameToExecute;
293 if (param.find(ARGUMENTS) != param.end()) {
294 Versatile V = param[ARGUMENTS];
295 for(Versatile::const_iterator it=V.begin(); it!=V.end(); it++) {
296 StringType argt = * static_cast<StringType *>(*it);
298 tempOutputFile << " " << arg;
301 remotePath += ".bat";
305 tempOutputFile << "#!/bin/sh" << endl;
306 tempOutputFile << "cd " << workDir << endl;
308 // Optional parameters (system limits on the job process)
309 if (param.find(MAXCPUTIME) != param.end()) {
310 long maxcputime = (long)param[MAXCPUTIME] * 60;
311 tempOutputFile << "ulimit -H -t " << maxcputime << endl;
314 if (param.find(MAXDISKSIZE) != param.end()) {
315 long maxdisksize = (long)param[MAXDISKSIZE] * 1024;
316 tempOutputFile << "ulimit -H -f " << maxdisksize << endl;
319 if (param.find(MAXRAMSIZE) != param.end()) {
320 long maxramsize = (long)param[MAXRAMSIZE] * 1024;
321 tempOutputFile << "ulimit -H -v " << maxramsize << endl;
324 // Define environment for the job
325 Environnement env = job.getEnvironnement();
326 for (Environnement::const_iterator iter = env.begin() ; iter != env.end() ; ++iter) {
327 tempOutputFile << "export " << iter->first << "=" << iter->second << endl;
330 // generate nodes file
331 tempOutputFile << "LIBBATCH_NODEFILE=`mktemp nodefile-XXXXXXXXXX`" << endl;
332 tempOutputFile << "echo `hostname` > $LIBBATCH_NODEFILE" << endl;
333 tempOutputFile << "export LIBBATCH_NODEFILE" << endl;
335 // Launch the executable
336 tempOutputFile << "./" + fileNameToExecute;
337 if (param.find(ARGUMENTS) != param.end()) {
338 Versatile V = param[ARGUMENTS];
339 for(Versatile::const_iterator it=V.begin(); it!=V.end(); it++) {
340 StringType argt = * static_cast<StringType *>(*it);
342 tempOutputFile << " " << arg;
346 // Standard input and output
347 if (param.find(INFILE) != param.end()) {
348 Versatile V = param[INFILE];
349 for(Versatile::const_iterator it=V.begin(); it!=V.end(); it++) {
350 Couple cpl = * static_cast<CoupleType*>(*it);
351 string remote = cpl.getRemote();
352 if (remote == "stdin")
353 tempOutputFile << " <stdin";
357 string stdoutFile = workDir + "/logs/output.log." + rootNameToExecute;
358 string stderrFile = workDir + "/logs/error.log." + rootNameToExecute;
359 if (param.find(OUTFILE) != param.end()) {
360 Versatile V = param[OUTFILE];
361 for(Versatile::const_iterator it=V.begin(); it!=V.end(); it++) {
362 Couple cpl = * static_cast<CoupleType*>(*it);
363 string remote = cpl.getRemote();
364 if (remote == "stdout") stdoutFile = "stdout";
365 if (remote == "stderr") stderrFile = "stderr";
368 tempOutputFile << " 1>" << stdoutFile << " 2>" << stderrFile << endl;
370 // Remove the node file
371 tempOutputFile << "rm $LIBBATCH_NODEFILE" << endl;
377 tempOutputFile.flush();
378 tempOutputFile.close();
380 LOG("Batch script file generated is: " << tmpFileName);
382 Utils::chmod(tmpFileName.c_str(), 0x1ED);
383 int status = _bm._protocol.copyFile(tmpFileName, "", "",
384 remotePath, _bm._hostname, _bm._username);
386 throw RunTimeException("Cannot copy command file on host " + _bm._hostname);
389 if (_bm._protocol.getType() != SH) {
390 // On Windows, we make the remote file executable afterward because
391 // pscp does not preserve access permissions on files
392 string subCommand = string("chmod u+x ") + remotePath;
393 string command = _bm._protocol.getExecCommand(subCommand, _bm._hostname, _bm._username);
395 status = system(command.c_str());
397 std::ostringstream oss;
398 oss << "Cannot change permissions of file " << remotePath << " on host " << _bm._hostname;
399 oss << ". Return status is " << status;
400 throw RunTimeException(oss.str());
410 // Constructeur de la classe ThreadAdapter
411 BatchManager_Local::ThreadAdapter::ThreadAdapter(BatchManager_Local & bm, const Job_Local & job, Id id) :
412 _bm(bm), _job(job), _id(id)
419 // Methode d'execution du thread
420 void * BatchManager_Local::ThreadAdapter::run(void * arg)
422 ThreadAdapter * p_ta = static_cast<ThreadAdapter *>(arg);
425 // On bloque tous les signaux pour ce thread
427 sigfillset(&setmask);
428 pthread_sigmask(SIG_BLOCK, &setmask, NULL);
431 // On autorise la terminaison differee du thread
432 // (ces valeurs sont les valeurs par defaut mais on les force par precaution)
433 pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL);
434 pthread_setcanceltype(PTHREAD_CANCEL_DEFERRED, NULL);
436 // On enregistre la fonction de suppression du fils en cas d'arret du thread
437 // Cette fontion sera automatiquement appelee lorsqu'une demande d'annulation
438 // sera prise en compte par pthread_testcancel()
440 pthread_cleanup_push(BatchManager_Local::delete_on_exit, arg);
441 pthread_cleanup_push(BatchManager_Local::setFailedOnCancel, arg);
442 pthread_cleanup_push(BatchManager_Local::kill_child_on_exit, static_cast<void *> (&child));
444 // On forke/exec un nouveau process pour pouvoir controler le fils
445 // (plus finement qu'avec un appel system)
446 // int rc = system(commande.c_str());
447 //char *const parmList[] = {"/usr/bin/ssh", "localhost", "-l", "aribes", "sleep 10 && echo end", NULL};
448 //execv("/usr/bin/ssh", parmList);
450 child = p_ta->launchWin32ChildProcess();
454 if (child < 0) { // erreur
455 UNDER_LOCK( LOG("Fork impossible (rc=" << child << ")") );
457 } else if (child > 0) { // pere
465 pthread_mutex_lock(&p_ta->_bm._threads_mutex);
467 // Set the job state to FINISHED or FAILED
468 p_ta->_bm._threads[p_ta->_id].param[STATE] = (p_ta->_bm._threads[p_ta->_id].hasFailed) ? FAILED : FINISHED;
470 // On retire la fonction de nettoyage de la memoire
471 pthread_cleanup_pop(0);
473 // On retire la fonction de suppression du fils
474 pthread_cleanup_pop(0);
476 // remove setFailedOnCancel function from cancel stack
477 pthread_cleanup_pop(0);
479 pthread_mutex_unlock(&p_ta->_bm._threads_mutex);
481 // On invoque la fonction de nettoyage de la memoire
484 UNDER_LOCK( LOG("Father is leaving") );
492 void BatchManager_Local::ThreadAdapter::pere(Process child)
494 time_t child_starttime = time(NULL);
496 // On enregistre le fils dans la table des threads
497 pthread_t thread_id = pthread_self();
499 Parametre param = _job.getParametre();
500 Environnement env = _job.getEnvironnement();
502 ostringstream id_sst;
504 param[ID] = id_sst.str();
505 param[STATE] = Batch::RUNNING;
507 _bm._threads[_id].thread_id = thread_id;
509 _bm._threads[_id].pid = child;
511 _bm._threads[_id].hasFailed = false;
512 _bm._threads[_id].param = param;
513 _bm._threads[_id].env = env;
514 _bm._threads[_id].command_queue.push(NOP);
516 // Unlock the master thread. From here, all shared variables must be protected
517 // from concurrent access
518 pthread_cond_signal(&_bm._threadSyncCondition);
521 // on boucle en attendant que le fils ait termine
525 GetExitCodeProcess(child, &exitCode);
526 if (exitCode != STILL_ACTIVE) {
527 UNDER_LOCK( LOG("Father sees his child is DONE: exit code = " << exitCode) );
532 pid_t child_wait_rc = waitpid(child, &child_rc, WNOHANG /* | WUNTRACED */);
533 if (child_wait_rc > 0) {
534 UNDER_LOCK( LOG("Status is: " << WIFEXITED( child_rc)) );
535 UNDER_LOCK( LOG("Status is: " << WEXITSTATUS( child_rc)) );
536 UNDER_LOCK( LOG("Status is: " << WIFSIGNALED( child_rc)) );
537 UNDER_LOCK( LOG("Status is: " << WTERMSIG( child_rc)) );
538 UNDER_LOCK( LOG("Status is: " << WCOREDUMP( child_rc)) );
539 UNDER_LOCK( LOG("Status is: " << WIFSTOPPED( child_rc)) );
540 UNDER_LOCK( LOG("Status is: " << WSTOPSIG( child_rc)) );
542 UNDER_LOCK( LOG("Status is: " << WIFCONTINUED( child_rc)) ); // not compilable on sarge
544 if (WIFSTOPPED(child_rc)) {
545 // NOTA : pour rentrer dans cette section, il faut que le flag WUNTRACED
546 // soit positionne dans l'appel a waitpid ci-dessus. Ce flag est couramment
547 // desactive car s'il est possible de detecter l'arret d'un process, il est
548 // plus difficile de detecter sa reprise.
550 // Le fils est simplement stoppe
551 // @@@ --------> SECTION CRITIQUE <-------- @@@
552 pthread_mutex_lock(&_bm._threads_mutex);
553 _bm._threads[_id].param[STATE] = Batch::PAUSED;
554 pthread_mutex_unlock(&_bm._threads_mutex);
555 // @@@ --------> SECTION CRITIQUE <-------- @@@
556 UNDER_LOCK( LOG("Father sees his child is STOPPED : " << child_wait_rc) );
560 // Le fils est termine, on sort de la boucle et du if englobant
561 UNDER_LOCK( LOG("Father sees his child is DONE : " << child_wait_rc << " (child_rc=" << (WIFEXITED(child_rc) ? WEXITSTATUS(child_rc) : -1) << ")") );
565 else if (child_wait_rc == -1) {
566 // Le fils a disparu ...
567 // @@@ --------> SECTION CRITIQUE <-------- @@@
568 pthread_mutex_lock(&_bm._threads_mutex);
569 _bm._threads[_id].hasFailed = true;
570 pthread_mutex_unlock(&_bm._threads_mutex);
571 // @@@ --------> SECTION CRITIQUE <-------- @@@
572 UNDER_LOCK( LOG("Father sees his child is DEAD : " << child_wait_rc << " (Reason : " << strerror(errno) << ")") );
577 // On teste si le thread doit etre detruit
578 pthread_testcancel();
582 // On regarde si le fils n'a pas depasse son temps (wallclock time)
583 time_t child_currenttime = time(NULL);
584 long child_elapsedtime_minutes = (child_currenttime - child_starttime) / 60L;
585 if (param.find(MAXWALLTIME) != param.end()) {
586 long maxwalltime = param[MAXWALLTIME];
587 // cout << "child_starttime = " << child_starttime << endl
588 // << "child_currenttime = " << child_currenttime << endl
589 // << "child_elapsedtime = " << child_elapsedtime << endl
590 // << "maxwalltime = " << maxwalltime << endl
591 // << "int(maxwalltime * 1.1) = " << int(maxwalltime * 1.1) << endl;
592 if (child_elapsedtime_minutes > long((float)maxwalltime * 1.1) ) { // On se donne 10% de marge avant le KILL
593 UNDER_LOCK( LOG("Father is sending KILL command to the thread " << _id) );
594 // On introduit une commande dans la queue du thread
595 // @@@ --------> SECTION CRITIQUE <-------- @@@
596 pthread_mutex_lock(&_bm._threads_mutex);
597 if (_bm._threads.find(_id) != _bm._threads.end())
598 _bm._threads[_id].command_queue.push(KILL);
599 pthread_mutex_unlock(&_bm._threads_mutex);
600 // @@@ --------> SECTION CRITIQUE <-------- @@@
603 } else if (child_elapsedtime_minutes > maxwalltime ) {
604 UNDER_LOCK( LOG("Father is sending TERM command to the thread " << _id) );
605 // On introduit une commande dans la queue du thread
606 // @@@ --------> SECTION CRITIQUE <-------- @@@
607 pthread_mutex_lock(&_bm._threads_mutex);
608 if (_bm._threads.find(_id) != _bm._threads.end())
609 _bm._threads[_id].command_queue.push(TERM);
610 pthread_mutex_unlock(&_bm._threads_mutex);
611 // @@@ --------> SECTION CRITIQUE <-------- @@@
617 // On regarde s'il y a quelque chose a faire dans la queue de commande
618 // @@@ --------> SECTION CRITIQUE <-------- @@@
619 pthread_mutex_lock(&_bm._threads_mutex);
620 if (_bm._threads.find(_id) != _bm._threads.end()) {
621 while (_bm._threads[_id].command_queue.size() > 0) {
622 Commande cmd = _bm._threads[_id].command_queue.front();
623 _bm._threads[_id].command_queue.pop();
627 UNDER_LOCK( LOG("Father does nothing to his child") );
631 UNDER_LOCK( LOG("Father is sending SIGSTOP signal to his child") );
632 kill(child, SIGSTOP);
636 UNDER_LOCK( LOG("Father is sending SIGCONT signal to his child") );
637 kill(child, SIGCONT);
641 UNDER_LOCK( LOG("Father is sending SIGTERM signal to his child") );
642 kill(child, SIGTERM);
646 UNDER_LOCK( LOG("Father is sending SIGKILL signal to his child") );
647 kill(child, SIGKILL);
659 pthread_mutex_unlock(&_bm._threads_mutex);
660 // @@@ --------> SECTION CRITIQUE <-------- @@@
662 // On fait une petite pause pour ne pas surcharger inutilement le processeur
677 void BatchManager_Local::ThreadAdapter::fils()
679 Parametre param = _job.getParametre();
680 Parametre::iterator it;
684 // build script file to launch the job and copy it on the server
685 string cmdFilePath = buildCommandFile(_job);
687 // define command to submit the job
688 vector<string> command = _bm._protocol.getExecCommandArgs(cmdFilePath, _bm._hostname, _bm._username);
690 // Build the argument array argv from the command
691 char ** argv = new char * [command.size() + 1];
693 for (string::size_type i=0 ; i<command.size() ; i++) {
694 argv[i] = new char[command[i].size() + 1];
695 strncpy(argv[i], command[i].c_str(), command[i].size() + 1);
696 comstr += command[i] + " ";
698 argv[command.size()] = NULL;
699 UNDER_LOCK( LOG("*** debug_command = " << comstr) );
701 // On cree une session pour le fils de facon a ce qu'il ne soit pas
702 // detruit lorsque le shell se termine (le shell ouvre une session et
703 // tue tous les process appartenant a la session en quittant)
706 // On ferme les descripteurs de fichiers standards
707 //close(STDIN_FILENO);
708 //close(STDOUT_FILENO);
709 //close(STDERR_FILENO);
711 // On execute la commande du fils
712 execv(argv[0], argv);
713 UNDER_LOCK( LOG("*** debug_command = " << strerror(errno)) );
714 // No need to deallocate since nothing happens after a successful exec
716 // Normalement on ne devrait jamais arriver ici
717 ofstream file_err("error.log");
718 UNDER_LOCK( file_err << "Echec de l'appel a execve" << endl );
720 } catch (GenericException & e) {
722 LOG("Caught exception : " << e.type << " : " << e.message);
730 BatchManager_Local::Process BatchManager_Local::ThreadAdapter::launchWin32ChildProcess()
732 Parametre param = _job.getParametre();
733 Parametre::iterator it;
734 PROCESS_INFORMATION pi;
738 // build script file to launch the job and copy it on the server
739 string cmdFilePath = buildCommandFile(_job);
741 // define command to submit the job
742 vector<string> command = _bm._protocol.getExecCommandArgs(cmdFilePath, _bm._hostname, _bm._username);
744 // Build the command string from the command argument vector
746 for (unsigned int i=0 ; i<command.size() ; i++) {
747 if (i>0) comstr += " ";
748 comstr += command[i];
751 UNDER_LOCK( LOG("*** debug_command = " << comstr) );
754 ZeroMemory( &si, sizeof(si) );
756 ZeroMemory( &pi, sizeof(pi) );
758 // Copy the command to a non-const buffer
759 char * buffer = strdup(comstr.c_str());
761 // launch the new process
762 bool res = CreateProcess(NULL, buffer, NULL, NULL, FALSE,
763 CREATE_NO_WINDOW, NULL, NULL, &si, &pi);
765 if (buffer) free(buffer);
766 if (!res) throw RunTimeException("Error while creating new process");
768 CloseHandle(pi.hThread);
770 } catch (GenericException & e) {
772 LOG("Caught exception : " << e.type << " : " << e.message);
781 void BatchManager_Local::kill_child_on_exit(void * p_pid)
784 //TODO: porting of following functionality
785 pid_t child = * static_cast<pid_t *>(p_pid);
788 kill(child, SIGTERM);
790 // Nota : on pourrait aussi faire a la suite un kill(child, SIGKILL)
791 // mais cette option n'est pas implementee pour le moment, car il est
792 // preferable de laisser le process fils se terminer normalement et seul.
796 void BatchManager_Local::delete_on_exit(void * arg)
798 ThreadAdapter * p_ta = static_cast<ThreadAdapter *>(arg);
802 void BatchManager_Local::setFailedOnCancel(void * arg)
804 ThreadAdapter * p_ta = static_cast<ThreadAdapter *>(arg);
805 pthread_mutex_lock(&p_ta->getBatchManager()._threads_mutex);
806 p_ta->getBatchManager()._threads[p_ta->getId()].param[STATE] = FAILED;
807 pthread_mutex_unlock(&p_ta->getBatchManager()._threads_mutex);
809 // Unlock the master thread. From here, the batch manager instance should not be used.
810 pthread_cond_signal(&p_ta->getBatchManager()._threadSyncCondition);