1 // Copyright (C) 2007-2015 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, or (at your option) any later version.
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 #include "SALOME_ContainerManager.hxx"
24 #include "SALOME_ResourcesManager.hxx"
25 #include "SALOME_LoadRateManager.hxx"
26 #include "SALOME_NamingService.hxx"
27 #include "SALOME_ResourcesManager_Client.hxx"
28 #include "SALOME_ModuleCatalog.hh"
29 #include "Basics_Utils.hxx"
30 #include "Basics_DirUtils.hxx"
31 #include <sys/types.h>
38 #include "Utils_CorbaException.hxx"
42 #include <SALOMEconfig.h>
43 #include CORBA_CLIENT_HEADER(SALOME_Session)
51 #define getpid _getpid
54 #ifdef WITH_PACO_PARALLEL
58 const int SALOME_ContainerManager::TIME_OUT_TO_LAUNCH_CONT=60;
60 const char *SALOME_ContainerManager::_ContainerManagerNameInNS =
63 omni_mutex SALOME_ContainerManager::_numInstanceMutex;
65 Utils_Mutex SALOME_ContainerManager::_getenvMutex;
67 Utils_Mutex SALOME_ContainerManager::_systemMutex;
69 //=============================================================================
73 * Define a CORBA single thread policy for the server, which avoid to deal
74 * with non thread-safe usage like Change_Directory in SALOME naming service
76 //=============================================================================
78 SALOME_ContainerManager::SALOME_ContainerManager(CORBA::ORB_ptr orb, PortableServer::POA_var poa, SALOME_NamingService *ns)
81 MESSAGE("constructor");
83 _resManager = new SALOME_ResourcesManager_Client(ns);
85 PortableServer::POAManager_var pman = poa->the_POAManager();
86 _orb = CORBA::ORB::_duplicate(orb) ;
87 CORBA::PolicyList policies;
89 PortableServer::ThreadPolicy_var threadPol(poa->create_thread_policy(PortableServer::ORB_CTRL_MODEL));
90 policies[0] = PortableServer::ThreadPolicy::_duplicate(threadPol);
92 _poa = poa->create_POA("MThreadPOA",pman,policies);
94 PortableServer::ObjectId_var id = _poa->activate_object(this);
95 CORBA::Object_var obj = _poa->id_to_reference(id);
96 Engines::ContainerManager_var refContMan =
97 Engines::ContainerManager::_narrow(obj);
99 _NS->Register(refContMan,_ContainerManagerNameInNS);
100 _isAppliSalomeDefined = (GetenvThreadSafe("APPLI") != 0);
105 // the urifile name depends on pid of the process
106 std::stringstream urifile;
107 urifile << GetenvThreadSafeAsString("HOME") << "/.urifile_" << getpid();
108 setenv("OMPI_URI_FILE",urifile.str().c_str(),1);
109 if( GetenvThreadSafe("OMPI_URI_FILE") != NULL ){
110 // get the pid of all ompi-server
111 std::set<pid_t> thepids1 = getpidofprogram("ompi-server");
112 // launch a new ompi-server
114 command = "ompi-server -r ";
115 command += GetenvThreadSafeAsString("OMPI_URI_FILE");
116 int status=SystemThreadSafe(command.c_str());
118 throw SALOME_Exception("Error when launching ompi-server");
119 // get the pid of all ompi-server
120 std::set<pid_t> thepids2 = getpidofprogram("ompi-server");
121 // my ompi-server is the new one
122 std::set<pid_t>::const_iterator it;
123 for(it=thepids2.begin();it!=thepids2.end();it++)
124 if(thepids1.find(*it) == thepids1.end())
125 _pid_mpiServer = *it;
126 if(_pid_mpiServer < 0)
127 throw SALOME_Exception("Error when getting ompi-server id");
131 // get the pid of all hydra_nameserver
132 std::set<pid_t> thepids1 = getpidofprogram("hydra_nameserver");
133 // launch a new hydra_nameserver
135 command = "hydra_nameserver &";
136 SystemThreadSafe(command.c_str());
137 // get the pid of all hydra_nameserver
138 std::set<pid_t> thepids2 = getpidofprogram("hydra_nameserver");
139 // my hydra_nameserver is the new one
140 std::set<pid_t>::const_iterator it;
141 for(it=thepids2.begin();it!=thepids2.end();it++)
142 if(thepids1.find(*it) == thepids1.end())
143 _pid_mpiServer = *it;
147 MESSAGE("constructor end");
150 //=============================================================================
154 //=============================================================================
156 SALOME_ContainerManager::~SALOME_ContainerManager()
158 MESSAGE("destructor");
162 if( GetenvThreadSafe("OMPI_URI_FILE") != NULL ){
163 // kill my ompi-server
164 if( kill(_pid_mpiServer,SIGTERM) != 0 )
165 throw SALOME_Exception("Error when killing ompi-server");
167 int status=SystemThreadSafe("rm -f ${OMPI_URI_FILE}");
169 throw SALOME_Exception("Error when removing urifile");
172 // kill my hydra_nameserver
173 if(_pid_mpiServer > -1)
174 if( kill(_pid_mpiServer,SIGTERM) != 0 )
175 throw SALOME_Exception("Error when killing hydra_nameserver");
180 //=============================================================================
181 //! shutdown all the containers, then the ContainerManager servant
184 //=============================================================================
186 void SALOME_ContainerManager::Shutdown()
189 ShutdownContainers();
190 _NS->Destroy_Name(_ContainerManagerNameInNS);
191 PortableServer::ObjectId_var oid = _poa->servant_to_id(this);
192 _poa->deactivate_object(oid);
195 //=============================================================================
196 //! Loop on all the containers listed in naming service, ask shutdown on each
199 //=============================================================================
201 void SALOME_ContainerManager::ShutdownContainers()
203 MESSAGE("ShutdownContainers");
205 SALOME::Session_var session = SALOME::Session::_nil();
207 CORBA::Object_var objS = _NS->Resolve("/Kernel/Session");
208 if (!CORBA::is_nil(objS))
210 session = SALOME::Session::_narrow(objS);
211 if (!CORBA::is_nil(session))
212 pid = session->getPID();
216 isOK = _NS->Change_Directory("/Containers");
218 std::vector<std::string> vec = _NS->list_directory_recurs();
219 std::list<std::string> lstCont;
220 for(std::vector<std::string>::iterator iter = vec.begin();iter!=vec.end();iter++)
223 CORBA::Object_var obj=_NS->Resolve((*iter).c_str());
226 Engines::Container_var cont=Engines::Container::_narrow(obj);
227 if(!CORBA::is_nil(cont) && pid != cont->getPID())
228 lstCont.push_back((*iter));
230 catch(const CORBA::Exception& e)
232 // ignore this entry and continue
235 MESSAGE("Container list: ");
236 for(std::list<std::string>::iterator iter=lstCont.begin();iter!=lstCont.end();iter++){
239 for(std::list<std::string>::iterator iter=lstCont.begin();iter!=lstCont.end();iter++)
244 CORBA::Object_var obj=_NS->Resolve((*iter).c_str());
245 Engines::Container_var cont=Engines::Container::_narrow(obj);
246 if(!CORBA::is_nil(cont))
248 MESSAGE("ShutdownContainers: " << (*iter));
252 MESSAGE("ShutdownContainers: no container ref for " << (*iter));
254 catch(CORBA::SystemException& e)
256 INFOS("CORBA::SystemException ignored : " << e);
258 catch(CORBA::Exception&)
260 INFOS("CORBA::Exception ignored.");
264 INFOS("Unknown exception ignored.");
270 //=============================================================================
271 //! Give a suitable Container given constraints
273 * \param params Container Parameters required for the container
274 * \return the container or nil
276 //=============================================================================
277 Engines::Container_ptr SALOME_ContainerManager::GiveContainer(const Engines::ContainerParameters& params)
279 std::string machFile;
280 Engines::Container_ptr ret(Engines::Container::_nil());
282 // Step 0: Default mode is start
283 Engines::ContainerParameters local_params(params);
284 if (std::string(local_params.mode.in()) == "")
285 local_params.mode = CORBA::string_dup("start");
286 std::string mode = local_params.mode.in();
287 MESSAGE("[GiveContainer] starting with mode: " << mode);
289 // Step 1: Find Container for find and findorstart mode
290 if (mode == "find" || mode == "findorstart")
292 ret = FindContainer(params, params.resource_params.resList);
293 if(!CORBA::is_nil(ret))
299 MESSAGE("[GiveContainer] no container found");
309 // Step 2: Get all possibleResources from the parameters
310 // Consider only resources that can run containers
311 resourceParams resource_params = resourceParameters_CORBAtoCPP(local_params.resource_params);
312 resource_params.can_run_containers = true;
313 std::vector<std::string> possibleResources = _resManager->GetFittingResources(resource_params);
314 MESSAGE("[GiveContainer] - length of possible resources " << possibleResources.size());
315 std::vector<std::string> local_resources;
317 // Step 3: if mode is "get" keep only machines with existing containers
320 for(unsigned int i=0; i < possibleResources.size(); i++)
322 Engines::Container_ptr cont = FindContainer(params, possibleResources[i]);
325 if(!cont->_non_existent())
326 local_resources.push_back(possibleResources[i]);
328 catch(CORBA::Exception&) {}
331 // if local_resources is empty, we cannot give a container
332 if (local_resources.size() == 0)
334 MESSAGE("[GiveContainer] cannot find a container for mode get");
339 local_resources = possibleResources;
341 // Step 4: select the resource where to get/start the container
342 bool resource_available = true;
343 std::string resource_selected;
344 std::vector<std::string> resources = local_resources;
345 while (resource_available)
347 if (resources.size() == 0)
348 resource_available = false;
353 resource_selected = _resManager->Find(params.resource_params.policy.in(), resources);
354 // Remove resource_selected from vector
355 std::vector<std::string>::iterator it;
356 for (it=resources.begin() ; it < resources.end(); it++ )
357 if (*it == resource_selected)
363 catch(const SALOME_Exception &ex)
365 MESSAGE("[GiveContainer] Exception in ResourceManager find !: " << ex.what());
368 MESSAGE("[GiveContainer] Resource selected is: " << resource_selected);
370 // Step 5: Create container name
371 ParserResourcesType resource_definition = _resManager->GetResourceDefinition(resource_selected);
372 std::string hostname(resource_definition.HostName);
373 std::string containerNameInNS;
376 if ( params.nb_proc <= 0 )
379 nbproc = params.nb_proc;
382 if( GetenvThreadSafe("LIBBATCH_NODEFILE") != NULL )
383 machFile = machinesFile(nbproc);
385 catch(const SALOME_Exception & ex)
387 std::string err_msg = ex.what();
388 err_msg += params.container_name;
389 INFOS(err_msg.c_str());
392 // A mpi parallel container register on zero node in NS
393 containerNameInNS = _NS->BuildContainerNameForNS(params, GetMPIZeroNode(hostname,machFile).c_str());
396 containerNameInNS = _NS->BuildContainerNameForNS(params, hostname.c_str());
397 MESSAGE("[GiveContainer] Container name in the naming service: " << containerNameInNS);
399 // Step 6: check if the name exists in naming service
400 //if params.mode == "getorstart" or "get" use the existing container
401 //if params.mode == "start" shutdown the existing container before launching a new one with that name
403 { // critical section
404 Utils_Locker lock (&_giveContainerMutex1);
405 CORBA::Object_var obj = _NS->Resolve(containerNameInNS.c_str());
406 if (!CORBA::is_nil(obj))
410 Engines::Container_var cont=Engines::Container::_narrow(obj);
411 if(!cont->_non_existent())
413 if(std::string(params.mode.in())=="getorstart" || std::string(params.mode.in())=="get"){
414 return cont._retn(); /* the container exists and params.mode is getorstart or get use it*/
418 INFOS("[GiveContainer] A container is already registered with the name: " << containerNameInNS << ", shutdown the existing container");
419 cont->Shutdown(); // shutdown the registered container if it exists
423 catch(CORBA::Exception&)
425 INFOS("[GiveContainer] CORBA::Exception ignored when trying to get the container - we start a new one");
428 } // end critical section
429 Engines::Container_var cont = LaunchContainer(params, resource_selected, hostname, machFile, containerNameInNS);
430 if (!CORBA::is_nil(cont))
432 INFOS("[GiveContainer] container " << containerNameInNS << " launched");
437 INFOS("[GiveContainer] Failed to launch container on resource " << resource_selected);
442 // We were not able to launch the container
443 INFOS("[GiveContainer] Cannot launch the container on the following selected resources:")
444 std::vector<std::string>::iterator it;
445 for (it=local_resources.begin() ; it < local_resources.end(); it++ )
446 INFOS("[GiveContainer] " << *it)
450 Engines::Container_ptr
451 SALOME_ContainerManager::LaunchContainer(const Engines::ContainerParameters& params,
452 const std::string & resource_selected,
453 const std::string & hostname,
454 const std::string & machFile,
455 const std::string & containerNameInNS)
457 std::string user,command,logFilename,tmpFileName;
459 Engines::Container_ptr ret(Engines::Container::_nil());
460 {//start of critical section
461 Utils_Locker lock (&_giveContainerMutex1);
462 // Step 1: type of container: PaCO, Exe, Mpi or Classic
463 // Mpi already tested in step 5, specific code on BuildCommandToLaunch Local/Remote Container methods
464 // TODO -> separates Mpi from Classic/Exe
466 std::string container_exe = "SALOME_Container"; // Classic container
467 Engines::ContainerParameters local_params(params);
471 CORBA::String_var container_exe_tmp;
472 CORBA::Object_var obj = _NS->Resolve("/Kernel/ModulCatalog");
473 SALOME_ModuleCatalog::ModuleCatalog_var Catalog = SALOME_ModuleCatalog::ModuleCatalog::_narrow(obj) ;
474 if (CORBA::is_nil (Catalog))
476 INFOS("[GiveContainer] Module Catalog is not found -> cannot launch a container");
479 // Loop through component list
480 for(unsigned int i=0; i < local_params.resource_params.componentList.length(); i++)
482 const char* compoi = local_params.resource_params.componentList[i];
483 SALOME_ModuleCatalog::Acomponent_var compoInfo = Catalog->GetComponent(compoi);
484 if (CORBA::is_nil (compoInfo))
488 SALOME_ModuleCatalog::ImplType impl=compoInfo->implementation_type();
489 container_exe_tmp=compoInfo->implementation_name();
490 if(impl==SALOME_ModuleCatalog::CEXE)
494 INFOS("ContainerManager Error: you can't have 2 CEXE component in the same container" );
495 return Engines::Container::_nil();
497 MESSAGE("[GiveContainer] Exe container found !: " << container_exe_tmp);
498 container_exe = container_exe_tmp.in();
503 catch (ServiceUnreachable&)
505 INFOS("Caught exception: Naming Service Unreachable");
510 INFOS("Caught unknown exception.");
514 // Step 2: test resource
515 // Only if an application directory is set
516 if(hostname != Kernel_Utils::GetHostname() && _isAppliSalomeDefined)
518 // Preparing remote command
519 std::string command = "";
520 const ParserResourcesType resInfo(_resManager->GetResourceDefinition(resource_selected));
521 command = getCommandToRunRemoteProcess(resInfo.Protocol, resInfo.HostName, resInfo.UserName);
522 if (resInfo.AppliPath != "")
523 command += resInfo.AppliPath;
526 ASSERT(GetenvThreadSafe("APPLI"));
527 command += GetenvThreadSafeAsString("APPLI");
529 command += "/runRemote.sh ";
530 ASSERT(GetenvThreadSafe("NSHOST"));
531 command += GetenvThreadSafeAsString("NSHOST"); // hostname of CORBA name server
533 ASSERT(GetenvThreadSafe("NSPORT"));
534 command += GetenvThreadSafeAsString("NSPORT"); // port of CORBA name server
535 command += " \"ls /tmp >/dev/null 2>&1\"";
537 // Launch remote command
538 int status = SystemThreadSafe(command.c_str());
541 // Error on resource - cannot launch commands
542 INFOS("[LaunchContainer] Cannot launch commands on machine " << hostname);
543 INFOS("[LaunchContainer] Command was " << command);
545 INFOS("[LaunchContainer] Command status is " << WEXITSTATUS(status));
547 return Engines::Container::_nil();
551 // Step 3: start a new container
552 // Check if a PaCO container
554 if (std::string(local_params.parallelLib.in()) != "")
556 ret = StartPaCOPPContainer(params, resource_selected);
559 // Other type of containers...
560 MESSAGE("[GiveContainer] Try to launch a new container on " << resource_selected);
561 // if a parallel container is launched in batch job, command is: "mpirun -np nbproc -machinefile nodesfile SALOME_MPIContainer"
562 if( GetenvThreadSafe("LIBBATCH_NODEFILE") != NULL && params.isMPI )
563 command = BuildCommandToLaunchLocalContainer(params, machFile, container_exe, tmpFileName);
564 // if a container is launched on localhost, command is "SALOME_Container" or "mpirun -np nbproc SALOME_MPIContainer"
565 else if(hostname == Kernel_Utils::GetHostname())
566 command = BuildCommandToLaunchLocalContainer(params, machFile, container_exe, tmpFileName);
567 // if a container is launched in remote mode, command is "ssh resource_selected SALOME_Container" or "ssh resource_selected mpirun -np nbproc SALOME_MPIContainer"
569 command = BuildCommandToLaunchRemoteContainer(resource_selected, params, container_exe);
571 //redirect stdout and stderr in a file
573 logFilename=GetenvThreadSafeAsString("TEMP");
575 user = GetenvThreadSafeAsString( "USERNAME" );
577 user = GetenvThreadSafeAsString( "USER" );
579 user = GetenvThreadSafeAsString( "LOGNAME" );
581 char* val = GetenvThreadSafe("SALOME_TMP_DIR");
584 struct stat file_info;
585 stat(val, &file_info);
586 bool is_dir = S_ISDIR(file_info.st_mode);
587 if (is_dir)logFilename=val;
588 else std::cerr << "SALOME_TMP_DIR environment variable is not a directory use /tmp instead" << std::endl;
592 logFilename += _NS->ContainerName(params)+"_"+ resource_selected +"_"+user;
593 std::ostringstream tmp;
594 tmp << "_" << getpid();
595 logFilename += tmp.str();
596 logFilename += ".log" ;
597 command += " > " + logFilename + " 2>&1";
598 MakeTheCommandToBeLaunchedASync(command);
600 // launch container with a system call
601 status=SystemThreadSafe(command.c_str());
602 }//end of critical of section
606 INFOS("[LaunchContainer] command failed (system command status -1): " << command);
607 RmTmpFile(tmpFileName); // command file can be removed here
608 return Engines::Container::_nil();
610 else if (status == 217)
612 INFOS("[LaunchContainer] command failed (system command status 217): " << command);
613 RmTmpFile(tmpFileName); // command file can be removed here
614 return Engines::Container::_nil();
618 // Step 4: Wait for the container
619 int count(GetTimeOutToLoaunchServer());
620 INFOS("[GiveContainer] waiting " << count << " second steps container " << containerNameInNS);
621 while (CORBA::is_nil(ret) && count)
625 MESSAGE("[GiveContainer] step " << count << " Waiting for container on " << resource_selected);
626 CORBA::Object_var obj(_NS->Resolve(containerNameInNS.c_str()));
627 ret=Engines::Container::_narrow(obj);
629 if (CORBA::is_nil(ret))
631 INFOS("[GiveContainer] was not able to launch container " << containerNameInNS);
635 // Setting log file name
636 logFilename=":"+logFilename;
637 logFilename="@"+Kernel_Utils::GetHostname()+logFilename;//threadsafe
638 logFilename=user+logFilename;
639 ret->logfilename(logFilename.c_str());
640 RmTmpFile(tmpFileName); // command file can be removed here
646 //=============================================================================
647 //! Find a container given constraints (params) on a list of machines (possibleComputers)
648 //! agy : this method is ThreadSafe
652 //=============================================================================
654 Engines::Container_ptr SALOME_ContainerManager::FindContainer(const Engines::ContainerParameters& params, const Engines::ResourceList& possibleResources)
656 MESSAGE("[FindContainer] FindContainer on " << possibleResources.length() << " resources");
657 for(unsigned int i=0; i < possibleResources.length();i++)
659 Engines::Container_ptr cont = FindContainer(params, possibleResources[i].in());
660 if(!CORBA::is_nil(cont))
663 MESSAGE("[FindContainer] no container found");
664 return Engines::Container::_nil();
667 //=============================================================================
668 //! Find a container given constraints (params) on a machine (theMachine)
669 //! agy : this method is ThreadSafe
673 //=============================================================================
675 Engines::Container_ptr
676 SALOME_ContainerManager::FindContainer(const Engines::ContainerParameters& params, const std::string& resource)
678 ParserResourcesType resource_definition = _resManager->GetResourceDefinition(resource);
679 std::string hostname(resource_definition.HostName);
680 std::string containerNameInNS(_NS->BuildContainerNameForNS(params, hostname.c_str()));
681 MESSAGE("[FindContainer] Try to find a container " << containerNameInNS << " on resource " << resource);
682 CORBA::Object_var obj = _NS->Resolve(containerNameInNS.c_str());
685 if(obj->_non_existent())
686 return Engines::Container::_nil();
688 return Engines::Container::_narrow(obj);
690 catch(const CORBA::Exception& e)
692 return Engines::Container::_nil();
697 bool isPythonContainer(const char* ContainerName);
699 //=============================================================================
701 * This is no longer valid (C++ container are also python containers)
703 //=============================================================================
704 bool isPythonContainer(const char* ContainerName)
706 return false; // VSR 02/08/2013: Python containers are no more supported
708 int len = strlen(ContainerName);
711 if (strcmp(ContainerName + len - 2, "Py") == 0)
717 //=============================================================================
719 * Builds the script to be launched
721 * If SALOME Application not defined ($APPLI),
722 * see BuildTempFileToLaunchRemoteContainer()
724 * Else rely on distant configuration. Command is under the form (example):
725 * ssh user@machine distantPath/runRemote.sh hostNS portNS WORKINGDIR workingdir \
726 * SALOME_Container containerName &"
728 * - where user is ommited if not specified in CatalogResources,
729 * - where distant path is always relative to user@machine $HOME, and
730 * equal to $APPLI if not specified in CatalogResources,
731 * - where hostNS is the hostname of CORBA naming server (set by scripts to
732 * use to launch SALOME and servers in $APPLI: runAppli.sh, runRemote.sh)
733 * - where portNS is the port used by CORBA naming server (set by scripts to
734 * use to launch SALOME and servers in $APPLI: runAppli.sh, runRemote.sh)
735 * - where workingdir is the requested working directory for the container.
736 * If WORKINGDIR (and workingdir) is not present the working dir will be $HOME
738 //=============================================================================
741 SALOME_ContainerManager::BuildCommandToLaunchRemoteContainer(const std::string& resource_name, const Engines::ContainerParameters& params, const std::string& container_exe) const
743 std::string command,tmpFileName;
744 if (!_isAppliSalomeDefined)
745 command = BuildTempFileToLaunchRemoteContainer(resource_name, params, tmpFileName);
749 const ParserResourcesType resInfo(_resManager->GetResourceDefinition(resource_name));
753 if ( params.nb_proc <= 0 )
756 nbproc = params.nb_proc;
759 // "ssh -l user machine distantPath/runRemote.sh hostNS portNS WORKINGDIR workingdir \
760 // SALOME_Container containerName &"
761 command = getCommandToRunRemoteProcess(resInfo.Protocol, resInfo.HostName, resInfo.UserName);
763 if (resInfo.AppliPath != "")
764 command += resInfo.AppliPath; // path relative to user@machine $HOME
767 ASSERT(GetenvThreadSafe("APPLI"));
768 command += GetenvThreadSafeAsString("APPLI"); // path relative to user@machine $HOME
771 command += "/runRemote.sh ";
773 ASSERT(GetenvThreadSafe("NSHOST"));
774 command += GetenvThreadSafeAsString("NSHOST"); // hostname of CORBA name server
777 ASSERT(GetenvThreadSafe("NSPORT"));
778 command += GetenvThreadSafeAsString("NSPORT"); // port of CORBA name server
780 std::string wdir = params.workingdir.in();
783 command += " WORKINGDIR ";
785 if(wdir == "$TEMPDIR")
787 command += wdir; // requested working directory
793 command += " mpirun -np ";
794 std::ostringstream o;
798 command += "-x PATH,LD_LIBRARY_PATH,OMNIORB_CONFIG,SALOME_trace ";
799 #elif defined(OPEN_MPI)
800 if( GetenvThreadSafe("OMPI_URI_FILE") == NULL )
801 command += "-x PATH -x LD_LIBRARY_PATH -x OMNIORB_CONFIG -x SALOME_trace";
803 command += "-x PATH -x LD_LIBRARY_PATH -x OMNIORB_CONFIG -x SALOME_trace -ompi-server file:";
804 command += GetenvThreadSafeAsString("OMPI_URI_FILE");
807 command += "-nameserver " + Kernel_Utils::GetHostname();
809 command += " SALOME_MPIContainer ";
812 command += " " +container_exe+ " ";
814 command += _NS->ContainerName(params);
816 AddOmninamesParams(command);
818 MESSAGE("command =" << command);
824 //=============================================================================
826 * builds the command to be launched.
828 //=============================================================================
829 std::string SALOME_ContainerManager::BuildCommandToLaunchLocalContainer(const Engines::ContainerParameters& params, const std::string& machinesFile, const std::string& container_exe, std::string& tmpFileName) const
831 tmpFileName = BuildTemporaryFileName();
835 std::ostringstream o;
841 if ( params.nb_proc <= 0 )
844 nbproc = params.nb_proc;
848 if( GetenvThreadSafe("LIBBATCH_NODEFILE") != NULL )
849 o << "-machinefile " << machinesFile << " ";
852 o << "-x PATH,LD_LIBRARY_PATH,OMNIORB_CONFIG,SALOME_trace ";
853 #elif defined(OPEN_MPI)
854 if( GetenvThreadSafe("OMPI_URI_FILE") == NULL )
855 o << "-x PATH -x LD_LIBRARY_PATH -x OMNIORB_CONFIG -x SALOME_trace";
858 o << "-x PATH -x LD_LIBRARY_PATH -x OMNIORB_CONFIG -x SALOME_trace -ompi-server file:";
859 o << GetenvThreadSafeAsString("OMPI_URI_FILE");
862 o << "-nameserver " + Kernel_Utils::GetHostname();
865 if (isPythonContainer(params.container_name))
866 o << " pyMPI SALOME_ContainerPy.py ";
868 o << " SALOME_MPIContainer ";
873 std::string wdir=params.workingdir.in();
876 // a working directory is requested
877 if(wdir == "$TEMPDIR")
879 // a new temporary directory is requested
880 std::string dir = Kernel_Utils::GetTmpDir();
882 o << "cd /d " << dir << std::endl;
884 o << "cd " << dir << ";";
890 // a permanent directory is requested use it or create it
892 o << "mkdir " + wdir << std::endl;
893 o << "cd /D " + wdir << std::endl;
895 o << "mkdir -p " << wdir << " && cd " << wdir + ";";
900 if (isPythonContainer(params.container_name))
901 o << "SALOME_ContainerPy.py ";
903 o << container_exe + " ";
907 o << _NS->ContainerName(params);
909 AddOmninamesParams(o);
911 std::ofstream command_file( tmpFileName.c_str() );
912 command_file << o.str();
913 command_file.close();
916 chmod(tmpFileName.c_str(), 0x1ED);
918 command = tmpFileName;
920 MESSAGE("Command is file ... " << command);
921 MESSAGE("Command is ... " << o.str());
926 //=============================================================================
928 * removes the generated temporary file in case of a remote launch.
929 * This method is thread safe
931 //=============================================================================
933 void SALOME_ContainerManager::RmTmpFile(std::string& tmpFileName)
935 int lenght = tmpFileName.size();
939 std::string command = "del /F ";
941 std::string command = "rm ";
944 command += tmpFileName.substr(0, lenght - 3 );
946 command += tmpFileName;
948 SystemThreadSafe(command.c_str());
949 //if dir is empty - remove it
950 std::string tmp_dir = Kernel_Utils::GetDirByPath( tmpFileName );
951 if ( Kernel_Utils::IsEmptyDir( tmp_dir ) )
954 command = "del /F " + tmp_dir;
956 command = "rmdir " + tmp_dir;
958 SystemThreadSafe(command.c_str());
963 //=============================================================================
965 * add to command all options relative to naming service.
967 //=============================================================================
969 void SALOME_ContainerManager::AddOmninamesParams(std::string& command) const
971 std::ostringstream oss;
972 AddOmninamesParams(oss);
976 //=============================================================================
978 * add to command all options relative to naming service.
980 //=============================================================================
982 void SALOME_ContainerManager::AddOmninamesParams(std::ostream& fileStream) const
984 AddOmninamesParams(fileStream,_NS);
987 //=============================================================================
989 * add to command all options relative to naming service.
991 //=============================================================================
993 void SALOME_ContainerManager::AddOmninamesParams(std::ostream& fileStream, SALOME_NamingService *ns)
995 CORBA::String_var iorstr(ns->getIORaddr());
996 fileStream << "ORBInitRef NameService=";
997 fileStream << iorstr;
1000 void SALOME_ContainerManager::MakeTheCommandToBeLaunchedASync(std::string& command)
1003 command = "%PYTHONBIN% -c \"import subprocess ; subprocess.Popen(command).pid\"";
1009 int SALOME_ContainerManager::GetTimeOutToLoaunchServer()
1011 int count(TIME_OUT_TO_LAUNCH_CONT);
1012 if (GetenvThreadSafe("TIMEOUT_TO_LAUNCH_CONTAINER") != 0)
1014 std::string new_count_str(GetenvThreadSafeAsString("TIMEOUT_TO_LAUNCH_CONTAINER"));
1016 std::istringstream ss(new_count_str);
1017 if (!(ss >> new_count))
1019 INFOS("[LaunchContainer] TIMEOUT_TO_LAUNCH_CONTAINER should be an int");
1027 void SALOME_ContainerManager::SleepInSecond(int ellapseTimeInSecond)
1030 sleep( ellapseTimeInSecond ) ;
1032 int timeInMS(1000*ellapseTimeInSecond);
1037 //=============================================================================
1039 * generate a file name in /tmp directory
1041 //=============================================================================
1043 std::string SALOME_ContainerManager::BuildTemporaryFileName()
1045 //build more complex file name to support multiple salome session
1046 std::string aFileName = Kernel_Utils::GetTmpFileName();
1050 aFileName += ".bat";
1055 //=============================================================================
1057 * Builds in a temporary file the script to be launched.
1059 * Used if SALOME Application ($APPLI) is not defined.
1060 * The command is build with data from CatalogResources, in which every path
1061 * used on remote computer must be defined.
1063 //=============================================================================
1065 std::string SALOME_ContainerManager::BuildTempFileToLaunchRemoteContainer (const std::string& resource_name, const Engines::ContainerParameters& params, std::string& tmpFileName) const
1069 tmpFileName = BuildTemporaryFileName();
1070 std::ofstream tempOutputFile;
1071 tempOutputFile.open(tmpFileName.c_str(), std::ofstream::out );
1072 const ParserResourcesType resInfo(_resManager->GetResourceDefinition(resource_name));
1073 tempOutputFile << "#! /bin/sh" << std::endl;
1077 tempOutputFile << "export SALOME_trace=local" << std::endl; // mkr : 27.11.2006 : PAL13967 - Distributed supervision graphs - Problem with "SALOME_trace"
1078 //tempOutputFile << "source " << resInfo.PreReqFilePath << endl;
1084 tempOutputFile << "mpirun -np ";
1087 if ( params.nb_proc <= 0 )
1090 nbproc = params.nb_proc;
1092 std::ostringstream o;
1094 tempOutputFile << nbproc << " ";
1096 tempOutputFile << "-x PATH,LD_LIBRARY_PATH,OMNIORB_CONFIG,SALOME_trace ";
1097 #elif defined(OPEN_MPI)
1098 if( GetenvThreadSafe("OMPI_URI_FILE") == NULL )
1099 tempOutputFile << "-x PATH -x LD_LIBRARY_PATH -x OMNIORB_CONFIG -x SALOME_trace";
1101 tempOutputFile << "-x PATH -x LD_LIBRARY_PATH -x OMNIORB_CONFIG -x SALOME_trace -ompi-server file:";
1102 tempOutputFile << GetenvThreadSafeAsString("OMPI_URI_FILE");
1104 #elif defined(MPICH)
1105 tempOutputFile << "-nameserver " + Kernel_Utils::GetHostname();
1109 tempOutputFile << GetenvThreadSafeAsString("KERNEL_ROOT_DIR") << "/bin/salome/";
1113 if (isPythonContainer(params.container_name))
1114 tempOutputFile << " pyMPI SALOME_ContainerPy.py ";
1116 tempOutputFile << " SALOME_MPIContainer ";
1121 if (isPythonContainer(params.container_name))
1122 tempOutputFile << "SALOME_ContainerPy.py ";
1124 tempOutputFile << "SALOME_Container ";
1127 tempOutputFile << _NS->ContainerName(params) << " -";
1128 AddOmninamesParams(tempOutputFile);
1129 tempOutputFile << " &" << std::endl;
1130 tempOutputFile.flush();
1131 tempOutputFile.close();
1133 chmod(tmpFileName.c_str(), 0x1ED);
1136 // --- Build command
1138 std::string command;
1140 if (resInfo.Protocol == rsh)
1143 std::string commandRcp = "rcp ";
1144 commandRcp += tmpFileName;
1146 commandRcp += resInfo.HostName;
1148 commandRcp += tmpFileName;
1149 status = SystemThreadSafe(commandRcp.c_str());
1152 else if (resInfo.Protocol == ssh)
1155 std::string commandRcp = "scp ";
1156 commandRcp += tmpFileName;
1158 commandRcp += resInfo.HostName;
1160 commandRcp += tmpFileName;
1161 status = SystemThreadSafe(commandRcp.c_str());
1164 else if (resInfo.Protocol == srun)
1166 command = "srun -n 1 -N 1 --share --nodelist=";
1167 std::string commandRcp = "rcp ";
1168 commandRcp += tmpFileName;
1170 commandRcp += resInfo.HostName;
1172 commandRcp += tmpFileName;
1173 status = SystemThreadSafe(commandRcp.c_str());
1176 throw SALOME_Exception("Unknown protocol");
1179 throw SALOME_Exception("Error of connection on remote host");
1181 command += resInfo.HostName;
1183 command += tmpFileName;
1191 std::string SALOME_ContainerManager::GetMPIZeroNode(const std::string machine, const std::string machinesFile) const
1194 std::string zeronode;
1195 std::string command;
1196 std::string tmpFile = BuildTemporaryFileName();
1198 if( GetenvThreadSafe("LIBBATCH_NODEFILE") == NULL )
1200 if (_isAppliSalomeDefined)
1202 const ParserResourcesType resInfo(_resManager->GetResourceDefinition(machine));
1204 if (resInfo.Protocol == rsh)
1206 else if (resInfo.Protocol == ssh)
1208 else if (resInfo.Protocol == srun)
1209 command = "srun -n 1 -N 1 --share --nodelist=";
1211 throw SALOME_Exception("Unknown protocol");
1213 if (resInfo.UserName != "")
1216 command += resInfo.UserName;
1220 command += resInfo.HostName;
1223 if (resInfo.AppliPath != "")
1224 command += resInfo.AppliPath; // path relative to user@machine $HOME
1227 ASSERT(GetenvThreadSafe("APPLI"));
1228 command += GetenvThreadSafeAsString("APPLI"); // path relative to user@machine $HOME
1231 command += "/runRemote.sh ";
1233 ASSERT(GetenvThreadSafe("NSHOST"));
1234 command += GetenvThreadSafeAsString("NSHOST"); // hostname of CORBA name server
1237 ASSERT(GetenvThreadSafe("NSPORT"));
1238 command += GetenvThreadSafeAsString("NSPORT"); // port of CORBA name server
1240 command += " mpirun -np 1 hostname -s > " + tmpFile;
1243 command = "mpirun -np 1 hostname -s > " + tmpFile;
1246 command = "mpirun -np 1 -machinefile " + machinesFile + " hostname -s > " + tmpFile;
1248 status = SystemThreadSafe(command.c_str());
1250 std::ifstream fp(tmpFile.c_str(),std::ios::in);
1251 while(fp >> zeronode);
1259 std::string SALOME_ContainerManager::machinesFile(const int nbproc)
1262 std::string nodesFile = GetenvThreadSafeAsString("LIBBATCH_NODEFILE");
1263 std::string machinesFile = Kernel_Utils::GetTmpFileName();
1264 std::ifstream fpi(nodesFile.c_str(),std::ios::in);
1265 std::ofstream fpo(machinesFile.c_str(),std::ios::out);
1267 _numInstanceMutex.lock();
1269 for(int i=0;i<_nbprocUsed;i++)
1272 for(int i=0;i<nbproc;i++)
1274 fpo << tmp << std::endl;
1276 throw SALOME_Exception("You need more processors than batch session have allocated for you! Unable to launch the mpi container: ");
1278 _nbprocUsed += nbproc;
1282 _numInstanceMutex.unlock();
1284 return machinesFile;
1288 std::set<pid_t> SALOME_ContainerManager::getpidofprogram(const std::string program)
1290 std::set<pid_t> thepids;
1291 std::string tmpFile = Kernel_Utils::GetTmpFileName();
1294 cmd = "pidof " + program + " > " + tmpFile;
1295 SystemThreadSafe(cmd.c_str());
1296 std::ifstream fpi(tmpFile.c_str(),std::ios::in);
1297 while(fpi >> thepid){
1298 thepids.insert(atoi(thepid.c_str()));
1303 std::string SALOME_ContainerManager::getCommandToRunRemoteProcess(AccessProtocolType protocol,
1304 const std::string & hostname,
1305 const std::string & username)
1307 std::ostringstream command;
1314 command << "-l " << username << " ";
1316 command << hostname << " ";
1322 command << "-l " << username << " ";
1324 command << hostname << " ";
1327 // no need to redefine the user with srun, the job user is taken by default
1328 // (note: for srun, user id can be specified with " --uid=<user>")
1329 command << "srun -n 1 -N 1 --share --nodelist=" << hostname << " ";
1332 command << "pbsdsh -o -h " << hostname << " ";
1335 command << "blaunch -no-shell " << hostname << " ";
1338 throw SALOME_Exception("Unknown protocol");
1341 return command.str();
1345 SALOME_ContainerManager::checkPaCOParameters(Engines::ContainerParameters & params, std::string resource_selected)
1349 // Step 1 : check ContainerParameters
1350 // Check container_name, has to be defined
1351 if (std::string(params.container_name.in()) == "")
1353 INFOS("[checkPaCOParameters] You must define a container_name to launch a PaCO++ container");
1356 // Check parallelLib
1357 std::string parallelLib = params.parallelLib.in();
1358 if (parallelLib != "Mpi" && parallelLib != "Dummy")
1360 INFOS("[checkPaCOParameters] parallelLib is not correctly defined");
1361 INFOS("[checkPaCOParameters] you can chosse between: Mpi and Dummy");
1362 INFOS("[checkPaCOParameters] you entered: " << parallelLib);
1366 if (params.nb_proc <= 0)
1368 INFOS("[checkPaCOParameters] You must define a nb_proc > 0");
1372 // Step 2 : check resource_selected
1373 const ParserResourcesType resource_definition = _resManager->GetResourceDefinition(resource_selected);
1374 //std::string protocol = resource_definition->protocol.in();
1375 std::string username = resource_definition.UserName;
1376 std::string applipath = resource_definition.AppliPath;
1378 //if (protocol == "" || username == "" || applipath == "")
1379 if (username == "" || applipath == "")
1381 INFOS("[checkPaCOParameters] resource selected is not well defined");
1382 INFOS("[checkPaCOParameters] resource name: " << resource_definition.Name);
1383 INFOS("[checkPaCOParameters] resource hostname: " << resource_definition.HostName);
1384 INFOS("[checkPaCOParameters] resource protocol: " << resource_definition.getAccessProtocolTypeStr());
1385 INFOS("[checkPaCOParameters] resource username: " << username);
1386 INFOS("[checkPaCOParameters] resource applipath: " << applipath);
1394 * :WARNING: Do not directly convert returned value to std::string
1395 * This function may return NULL if env variable is not defined.
1396 * And std::string(NULL) causes undefined behavior.
1397 * Use GetenvThreadSafeAsString to properly get a std::string.
1399 char *SALOME_ContainerManager::GetenvThreadSafe(const char *name)
1400 {// getenv is not thread safe. See man 7 pthread.
1401 Utils_Locker lock (&_getenvMutex);
1402 return getenv(name);
1406 * Return env variable as a std::string.
1407 * Return empty string if env variable is not set.
1409 std::string SALOME_ContainerManager::GetenvThreadSafeAsString(const char *name)
1411 char* var = GetenvThreadSafe(name);
1412 return var ? std::string(var) : std::string();
1415 int SALOME_ContainerManager::SystemThreadSafe(const char *command)
1417 Utils_Locker lock (&_systemMutex);
1418 return system(command);
1421 #ifdef WITH_PACO_PARALLEL
1423 //=============================================================================
1425 * Start a suitable PaCO++ Parallel Container in a list of machines.
1426 * \param params Container Parameters required for the container
1427 * \return CORBA container reference.
1429 //=============================================================================
1430 Engines::Container_ptr
1431 SALOME_ContainerManager::StartPaCOPPContainer(const Engines::ContainerParameters& params_const,
1432 std::string resource_selected)
1434 CORBA::Object_var obj;
1435 PaCO::InterfaceManager_var container_proxy;
1436 Engines::Container_ptr ret = Engines::Container::_nil();
1437 Engines::ContainerParameters params(params_const);
1438 params.resource_params.name = CORBA::string_dup(resource_selected.c_str());
1440 // Step 0 : Check parameters
1441 if (!checkPaCOParameters(params, resource_selected))
1443 INFOS("[StartPaCOPPContainer] check parameters failed ! see logs...");
1447 // Step 1 : Starting a new parallel container !
1448 INFOS("[StartPaCOPPContainer] Starting a PaCO++ parallel container");
1449 INFOS("[StartPaCOPPContainer] on resource : " << resource_selected);
1451 // Step 2 : Get a MachineFile for the parallel container
1452 std::string machine_file_name = _resManager->getMachineFile(resource_selected,
1454 params.parallelLib.in());
1456 if (machine_file_name == "")
1458 INFOS("[StartPaCOPPContainer] Machine file generation failed");
1462 // Step 3 : starting parallel container proxy
1463 std::string command_proxy("");
1464 std::string proxy_machine;
1467 command_proxy = BuildCommandToLaunchPaCOProxyContainer(params, machine_file_name, proxy_machine);
1469 catch(const SALOME_Exception & ex)
1471 INFOS("[StartPaCOPPContainer] Exception in BuildCommandToLaunchPaCOContainer");
1475 obj = LaunchPaCOProxyContainer(command_proxy, params, proxy_machine);
1476 if (CORBA::is_nil(obj))
1478 INFOS("[StartPaCOPPContainer] LaunchPaCOContainer for proxy returns NIL !");
1481 container_proxy = PaCO::InterfaceManager::_narrow(obj);
1482 MESSAGE("[StartPaCOPPContainer] PaCO container proxy is launched");
1484 // Step 4 : starting parallel container nodes
1485 std::string command_nodes("");
1486 SALOME_ContainerManager::actual_launch_machine_t nodes_machines;
1489 command_nodes = BuildCommandToLaunchPaCONodeContainer(params, machine_file_name, nodes_machines, proxy_machine);
1491 catch(const SALOME_Exception & ex)
1493 INFOS("[StarPaCOPPContainer] Exception in BuildCommandToLaunchPaCONodeContainer");
1498 std::string container_generic_node_name = std::string(params.container_name.in()) + std::string("Node");
1499 bool result = LaunchPaCONodeContainer(command_nodes, params, container_generic_node_name, nodes_machines);
1502 INFOS("[StarPaCOPPContainer] LaunchPaCONodeContainer failed !");
1503 // Il faut tuer le proxy
1506 Engines::Container_var proxy = Engines::Container::_narrow(container_proxy);
1511 INFOS("[StarPaCOPPContainer] Exception catched from proxy Shutdown...");
1516 // Step 4 : connecting nodes and the proxy to actually create a parallel container
1517 for (int i = 0; i < params.nb_proc; i++)
1519 std::ostringstream tmp;
1521 std::string proc_number = tmp.str();
1522 std::string container_node_name = container_generic_node_name + proc_number;
1524 std::string theNodeMachine(nodes_machines[i]);
1525 std::string containerNameInNS = _NS->BuildContainerNameForNS(container_node_name.c_str(), theNodeMachine.c_str());
1526 obj = _NS->Resolve(containerNameInNS.c_str());
1527 if (CORBA::is_nil(obj))
1529 INFOS("[StarPaCOPPContainer] CONNECTION FAILED From Naming Service !");
1530 INFOS("[StarPaCOPPContainer] Container name is " << containerNameInNS);
1535 MESSAGE("[StarPaCOPPContainer] Deploying node : " << container_node_name);
1536 PaCO::InterfaceParallel_var node = PaCO::InterfaceParallel::_narrow(obj);
1538 MESSAGE("[StarPaCOPPContainer] node " << container_node_name << " is deployed");
1540 catch(CORBA::SystemException& e)
1542 INFOS("[StarPaCOPPContainer] Exception in deploying node : " << containerNameInNS);
1543 INFOS("CORBA::SystemException : " << e);
1546 catch(CORBA::Exception& e)
1548 INFOS("[StarPaCOPPContainer] Exception in deploying node : " << containerNameInNS);
1549 INFOS("CORBA::Exception" << e);
1554 INFOS("[StarPaCOPPContainer] Exception in deploying node : " << containerNameInNS);
1555 INFOS("Unknown exception !");
1560 // Step 5 : starting parallel container
1563 MESSAGE ("[StarPaCOPPContainer] Starting parallel object");
1564 container_proxy->start();
1565 MESSAGE ("[StarPaCOPPContainer] Parallel object is started");
1566 ret = Engines::Container::_narrow(container_proxy);
1568 catch(CORBA::SystemException& e)
1570 INFOS("Caught CORBA::SystemException. : " << e);
1572 catch(PortableServer::POA::ServantAlreadyActive&)
1574 INFOS("Caught CORBA::ServantAlreadyActiveException");
1576 catch(CORBA::Exception&)
1578 INFOS("Caught CORBA::Exception.");
1580 catch(std::exception& exc)
1582 INFOS("Caught std::exception - "<<exc.what());
1586 INFOS("Caught unknown exception.");
1592 SALOME_ContainerManager::BuildCommandToLaunchPaCOProxyContainer(const Engines::ContainerParameters& params,
1593 std::string machine_file_name,
1594 std::string & proxy_hostname)
1596 // In the proxy case, we always launch a Dummy Proxy
1597 std::string exe_name = "SALOME_ParallelContainerProxyDummy";
1598 std::string container_name = params.container_name.in();
1600 // Convert nb_proc in string
1601 std::ostringstream tmp_string;
1602 tmp_string << params.nb_proc;
1603 std::string nb_proc_str = tmp_string.str();
1605 // Get resource definition
1606 ParserResourcesType resource_definition =
1607 _resManager->GetResourceDefinition(params.resource_params.name.in());
1610 std::string hostname;
1611 std::ifstream machine_file(machine_file_name.c_str());
1612 std::getline(machine_file, hostname, ' ');
1613 size_t found = hostname.find('\n');
1614 if (found!=std::string::npos)
1615 hostname.erase(found, 1); // Remove \n
1616 proxy_hostname = hostname;
1617 MESSAGE("[BuildCommandToLaunchPaCOProxyContainer] machine file name extracted is " << hostname);
1620 bool remote_execution = false;
1621 if (hostname != std::string(Kernel_Utils::GetHostname()))
1623 MESSAGE("[BuildCommandToLaunchPaCOProxyContainer] remote machine case detected !");
1624 remote_execution = true;
1627 // Log environnement
1628 std::string log_type("");
1629 char * get_val = GetenvThreadSafe("PARALLEL_LOG");
1633 // Generating the command
1634 std::string command_begin("");
1635 std::string command_end("");
1636 std::ostringstream command;
1638 LogConfiguration(log_type, "proxy", container_name, hostname, command_begin, command_end);
1639 command << command_begin;
1641 // Adding connection command
1642 // We can only have a remote execution with
1643 // a SALOME application
1644 if (remote_execution)
1646 ASSERT(GetenvThreadSafe("NSHOST"));
1647 ASSERT(GetenvThreadSafe("NSPORT"));
1649 command << resource_definition.getAccessProtocolTypeStr();
1651 command << resource_definition.UserName;
1652 command << " " << hostname;
1653 command << " " << resource_definition.AppliPath;
1654 command << "/runRemote.sh ";
1655 command << GetenvThreadSafeAsString("NSHOST") << " "; // hostname of CORBA name server
1656 command << GetenvThreadSafeAsString("NSPORT") << " "; // port of CORBA name server
1659 command << exe_name;
1660 command << " " << container_name;
1661 command << " Dummy";
1662 command << " " << hostname;
1663 command << " " << nb_proc_str;
1665 AddOmninamesParams(command);
1668 command << command_end;
1669 MESSAGE("[BuildCommandToLaunchPaCOProxyContainer] Command is: " << command.str());
1671 return command.str();
1675 SALOME_ContainerManager::BuildCommandToLaunchPaCONodeContainer(const Engines::ContainerParameters& params,
1676 const std::string & machine_file_name,
1677 SALOME_ContainerManager::actual_launch_machine_t & vect_machine,
1678 const std::string & proxy_hostname)
1681 std::string exe_name = "SALOME_ParallelContainerNode";
1682 exe_name += params.parallelLib.in();
1683 std::string container_name = params.container_name.in();
1685 // Convert nb_proc in string
1686 std::ostringstream nb_proc_stream;
1687 nb_proc_stream << params.nb_proc;
1689 // Get resource definition
1690 ParserResourcesType resource_definition =
1691 _resManager->GetResourceDefinition(params.resource_params.name.in());
1693 // Log environnement
1694 std::string log_type("");
1695 char * get_val = GetenvThreadSafe("PARALLEL_LOG");
1699 // Now the command is different according to paralleLib
1700 std::ostringstream command_nodes;
1701 std::ifstream machine_file(machine_file_name.c_str());
1702 if (std::string(params.parallelLib.in()) == "Dummy")
1704 for (int i= 0; i < params.nb_proc; i++)
1707 std::string hostname;
1708 std::getline(machine_file, hostname);
1709 MESSAGE("[BuildCommandToLaunchPaCONodeContainer] machine file name extracted is " << hostname);
1712 bool remote_execution = false;
1713 if (hostname != std::string(Kernel_Utils::GetHostname()))
1715 MESSAGE("[BuildCommandToLaunchPaCONodeContainer] remote machine case detected !");
1716 remote_execution = true;
1719 // For each node we have a new command
1720 // Generating the command
1721 std::ostringstream command_node_stream;
1722 std::string command_node_begin("");
1723 std::string command_node_end("");
1724 std::ostringstream node_number;
1726 std::string container_node_name = container_name + node_number.str();
1727 LogConfiguration(log_type, "node", container_node_name, hostname, command_node_begin, command_node_end);
1729 // Adding connection command
1730 // We can only have a remote execution with
1731 // a SALOME application
1732 if (remote_execution)
1734 ASSERT(GetenvThreadSafe("NSHOST"));
1735 ASSERT(GetenvThreadSafe("NSPORT"));
1737 command_node_stream << resource_definition.getAccessProtocolTypeStr();
1738 command_node_stream << " -l ";
1739 command_node_stream << resource_definition.UserName;
1740 command_node_stream << " " << hostname;
1741 command_node_stream << " " << resource_definition.AppliPath;
1742 command_node_stream << "/runRemote.sh ";
1743 command_node_stream << GetenvThreadSafeAsString("NSHOST") << " "; // hostname of CORBA name server
1744 command_node_stream << GetenvThreadSafeAsString("NSPORT") << " "; // port of CORBA name server
1747 command_node_stream << exe_name;
1748 command_node_stream << " " << container_name;
1749 command_node_stream << " " << params.parallelLib.in();
1750 command_node_stream << " " << proxy_hostname;
1751 command_node_stream << " " << node_number.str();
1752 command_node_stream << " -";
1753 AddOmninamesParams(command_node_stream);
1755 command_nodes << command_node_begin << command_node_stream.str() << command_node_end;
1756 vect_machine.push_back(hostname);
1760 else if (std::string(params.parallelLib.in()) == "Mpi")
1763 std::string hostname;
1764 std::getline(machine_file, hostname, ' ');
1765 MESSAGE("[BuildCommandToLaunchPaCONodeContainer] machine file name extracted is " << hostname);
1768 bool remote_execution = false;
1769 if (hostname != std::string(Kernel_Utils::GetHostname()))
1771 MESSAGE("[BuildCommandToLaunchPaCONodeContainer] remote machine case detected !");
1772 remote_execution = true;
1775 // In case of Mpi and Remote, we copy machine_file in the applipath
1776 // scp mpi_machine_file user@machine:Path
1777 std::ostringstream command_remote_stream;
1778 std::string::size_type last = machine_file_name.find_last_of("/");
1779 if (last == std::string::npos)
1782 if (resource_definition.Protocol == rsh)
1783 command_remote_stream << "rcp ";
1785 command_remote_stream << "scp ";
1786 command_remote_stream << machine_file_name << " ";
1787 command_remote_stream << resource_definition.UserName << "@";
1788 command_remote_stream << hostname << ":" << resource_definition.AppliPath;
1789 command_remote_stream << "/" << machine_file_name.substr(last+1);
1791 int status = SystemThreadSafe(command_remote_stream.str().c_str());
1794 INFOS("copy of the MPI machine file failed ! - sorry !");
1798 // Generating the command
1799 std::string command_begin("");
1800 std::string command_end("");
1802 LogConfiguration(log_type, "nodes", container_name, hostname, command_begin, command_end);
1803 command_nodes << command_begin;
1805 // Adding connection command
1806 // We can only have a remote execution with
1807 // a SALOME application
1808 if (remote_execution)
1810 ASSERT(GetenvThreadSafe("NSHOST"));
1811 ASSERT(GetenvThreadSafe("NSPORT"));
1813 command_nodes << resource_definition.getAccessProtocolTypeStr();
1814 command_nodes << " -l ";
1815 command_nodes << resource_definition.UserName;
1816 command_nodes << " " << hostname;
1817 command_nodes << " " << resource_definition.AppliPath;
1818 command_nodes << "/runRemote.sh ";
1819 command_nodes << GetenvThreadSafeAsString("NSHOST") << " "; // hostname of CORBA name server
1820 command_nodes << GetenvThreadSafeAsString("NSPORT") << " "; // port of CORBA name server
1823 if (resource_definition.mpi == lam)
1825 command_nodes << "mpiexec -ssi boot ";
1826 command_nodes << "-machinefile " << machine_file_name << " ";
1827 command_nodes << "-n " << params.nb_proc;
1831 command_nodes << "mpirun -np " << params.nb_proc;
1833 command_nodes << " " << exe_name;
1834 command_nodes << " " << container_name;
1835 command_nodes << " " << params.parallelLib.in();
1836 command_nodes << " " << proxy_hostname;
1837 command_nodes << " -";
1838 AddOmninamesParams(command_nodes);
1840 // We don't put hostname, because nodes are registered in the resource of the proxy
1841 for (int i= 0; i < params.nb_proc; i++)
1842 vect_machine.push_back(proxy_hostname);
1844 command_nodes << command_end;
1846 return command_nodes.str();
1850 SALOME_ContainerManager::LogConfiguration(const std::string & log_type,
1851 const std::string & exe_type,
1852 const std::string & container_name,
1853 const std::string & hostname,
1854 std::string & begin,
1857 if(log_type == "xterm")
1859 begin = "xterm -e \"export LD_LIBRARY_PATH=$LD_LIBRARY_PATH; export PATH=$PATH;";
1862 else if(log_type == "xterm_debug")
1864 begin = "xterm -e \"export LD_LIBRARY_PATH=$LD_LIBRARY_PATH; export PATH=$PATH;";
1869 // default into a file...
1870 std::string logFilename = "/tmp/" + container_name + "_" + hostname + "_" + exe_type + "_";
1871 std::string user = GetenvThreadSafeAsString("USER");
1873 user = GetenvThreadSafeAsString("LOGNAME");
1874 logFilename += user + ".log";
1875 end = " > " + logFilename + " 2>&1 & ";
1880 SALOME_ContainerManager::LaunchPaCOProxyContainer(const std::string& command,
1881 const Engines::ContainerParameters& params,
1882 const std::string & hostname)
1884 PaCO::InterfaceManager_ptr container_proxy = PaCO::InterfaceManager::_nil();
1886 MESSAGE("[LaunchPaCOProxyContainer] Launch command");
1887 int status = SystemThreadSafe(command.c_str());
1889 INFOS("[LaunchPaCOProxyContainer] failed : system command status -1");
1890 return container_proxy;
1892 else if (status == 217) {
1893 INFOS("[LaunchPaCOProxyContainer] failed : system command status 217");
1894 return container_proxy;
1897 int count(GetTimeOutToLoaunchServer());
1898 CORBA::Object_var obj = CORBA::Object::_nil();
1899 std::string containerNameInNS = _NS->BuildContainerNameForNS(params.container_name.in(),
1901 MESSAGE("[LaunchParallelContainer] Waiting for Parallel Container proxy : " << containerNameInNS);
1903 while (CORBA::is_nil(obj) && count)
1907 obj = _NS->Resolve(containerNameInNS.c_str());
1912 container_proxy = PaCO::InterfaceManager::_narrow(obj);
1914 catch(CORBA::SystemException& e)
1916 INFOS("[StarPaCOPPContainer] Exception in _narrow after LaunchParallelContainer for proxy !");
1917 INFOS("CORBA::SystemException : " << e);
1918 return container_proxy;
1920 catch(CORBA::Exception& e)
1922 INFOS("[StarPaCOPPContainer] Exception in _narrow after LaunchParallelContainer for proxy !");
1923 INFOS("CORBA::Exception" << e);
1924 return container_proxy;
1928 INFOS("[StarPaCOPPContainer] Exception in _narrow after LaunchParallelContainer for proxy !");
1929 INFOS("Unknown exception !");
1930 return container_proxy;
1932 if (CORBA::is_nil(container_proxy))
1934 INFOS("[StarPaCOPPContainer] PaCO::InterfaceManager::_narrow returns NIL !");
1935 return container_proxy;
1940 //=============================================================================
1941 /*! This method launches the parallel container.
1942 * It will may be placed on the ressources manager.
1944 * \param command to launch
1945 * \param container's parameters
1946 * \param name of the container
1948 * \return CORBA container reference
1950 //=============================================================================
1952 SALOME_ContainerManager::LaunchPaCONodeContainer(const std::string& command,
1953 const Engines::ContainerParameters& params,
1954 const std::string& name,
1955 SALOME_ContainerManager::actual_launch_machine_t & vect_machine)
1957 INFOS("[LaunchPaCONodeContainer] Launch command");
1958 int status = SystemThreadSafe(command.c_str());
1960 INFOS("[LaunchPaCONodeContainer] failed : system command status -1");
1963 else if (status == 217) {
1964 INFOS("[LaunchPaCONodeContainer] failed : system command status 217");
1968 INFOS("[LaunchPaCONodeContainer] Waiting for the nodes of the parallel container");
1969 // We are waiting all the nodes
1970 for (int i = 0; i < params.nb_proc; i++)
1972 CORBA::Object_var obj = CORBA::Object::_nil();
1973 std::string theMachine(vect_machine[i]);
1975 std::ostringstream tmp;
1977 std::string proc_number = tmp.str();
1978 std::string container_node_name = name + proc_number;
1979 std::string containerNameInNS = _NS->BuildContainerNameForNS((char*) container_node_name.c_str(), theMachine.c_str());
1980 INFOS("[LaunchPaCONodeContainer] Waiting for Parallel Container node " << containerNameInNS << " on " << theMachine);
1981 int count(GetTimeOutToLoaunchServer());
1982 while (CORBA::is_nil(obj) && count) {
1985 obj = _NS->Resolve(containerNameInNS.c_str());
1987 if (CORBA::is_nil(obj))
1989 INFOS("[LaunchPaCONodeContainer] Launch of node failed (or not found) !");
1998 Engines::Container_ptr
1999 SALOME_ContainerManager::StartPaCOPPContainer(const Engines::ContainerParameters& params,
2000 std::string resource_selected)
2002 Engines::Container_ptr ret = Engines::Container::_nil();
2003 INFOS("[StarPaCOPPContainer] is disabled !");
2004 INFOS("[StarPaCOPPContainer] recompile SALOME Kernel to enable PaCO++ parallel extension");
2009 SALOME_ContainerManager::BuildCommandToLaunchPaCOProxyContainer(const Engines::ContainerParameters& params,
2010 std::string machine_file_name,
2011 std::string & proxy_hostname)
2017 SALOME_ContainerManager::BuildCommandToLaunchPaCONodeContainer(const Engines::ContainerParameters& params,
2018 const std::string & machine_file_name,
2019 SALOME_ContainerManager::actual_launch_machine_t & vect_machine,
2020 const std::string & proxy_hostname)
2025 SALOME_ContainerManager::LogConfiguration(const std::string & log_type,
2026 const std::string & exe_type,
2027 const std::string & container_name,
2028 const std::string & hostname,
2029 std::string & begin,
2035 SALOME_ContainerManager::LaunchPaCOProxyContainer(const std::string& command,
2036 const Engines::ContainerParameters& params,
2037 const std::string& hostname)
2039 CORBA::Object_ptr ret = CORBA::Object::_nil();
2044 SALOME_ContainerManager::LaunchPaCONodeContainer(const std::string& command,
2045 const Engines::ContainerParameters& params,
2046 const std::string& name,
2047 SALOME_ContainerManager::actual_launch_machine_t & vect_machine)