1 // Copyright (C) 2007-2022 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_ResourcesManager.hxx"
24 #include "Utils_ExceptHandlers.hxx"
25 #include "Utils_CorbaException.hxx"
43 #include <sys/types.h>
45 #include "utilities.h"
47 #define MAX_SIZE_FOR_HOSTNAME 256;
51 const char *SALOME_ResourcesManager::_ResourcesManagerNameInNS = "/ResourcesManager";
53 //=============================================================================
57 //=============================================================================
59 SALOME_ResourcesManager::SALOME_ResourcesManager(CORBA::ORB_ptr orb,
60 PortableServer::POA_var poa,
61 SALOME_NamingService_Abstract *ns,
62 const char *xmlFilePath)
63 : _rm(new ResourcesManager_cpp(xmlFilePath))
65 MESSAGE("SALOME_ResourcesManager constructor");
67 _orb = CORBA::ORB::_duplicate(orb) ;
69 PortableServer::POAManager_var pman = poa->the_POAManager();
70 CORBA::PolicyList policies;
72 PortableServer::ThreadPolicy_var threadPol(poa->create_thread_policy(PortableServer::SINGLE_THREAD_MODEL));
73 policies[0] = PortableServer::ThreadPolicy::_duplicate(threadPol);
74 _poa = poa->create_POA("SingleThreadPOA",pman,policies);
77 PortableServer::ObjectId_var id = _poa->activate_object(this);
78 CORBA::Object_var obj = _poa->id_to_reference(id);
79 Engines::ResourcesManager_var refContMan = Engines::ResourcesManager::_narrow(obj);
80 _NS->Register(refContMan,_ResourcesManagerNameInNS);
81 MESSAGE("SALOME_ResourcesManager constructor end");
84 //=============================================================================
86 * Standard constructor, parse resource file.
87 * - if ${APPLI} exists in environment,
88 * look for ${HOME}/${APPLI}/CatalogResources.xml
89 * - else look for default:
90 * ${KERNEL_ROOT_DIR}/share/salome/resources/kernel/CatalogResources.xml
91 * - parse XML resource file.
93 //=============================================================================
95 SALOME_ResourcesManager::SALOME_ResourcesManager(CORBA::ORB_ptr orb,
96 PortableServer::POA_var poa,
97 SALOME_NamingService_Abstract *ns) : _rm(new ResourcesManager_cpp())
99 MESSAGE("SALOME_ResourcesManager constructor");
101 _orb = CORBA::ORB::_duplicate(orb) ;
103 PortableServer::POAManager_var pman = poa->the_POAManager();
104 CORBA::PolicyList policies;
106 PortableServer::ThreadPolicy_var threadPol(poa->create_thread_policy(PortableServer::SINGLE_THREAD_MODEL));
107 policies[0] = PortableServer::ThreadPolicy::_duplicate(threadPol);
108 _poa = poa->create_POA("SingleThreadPOA",pman,policies);
109 threadPol->destroy();
111 PortableServer::ObjectId_var id = _poa->activate_object(this);
112 CORBA::Object_var obj = _poa->id_to_reference(id);
113 Engines::ResourcesManager_var refContMan = Engines::ResourcesManager::_narrow(obj);
115 _NS->Register(refContMan,_ResourcesManagerNameInNS);
117 MESSAGE("SALOME_ResourcesManager constructor end");
120 //=============================================================================
122 * Standard Destructor
124 //=============================================================================
126 SALOME_ResourcesManager::~SALOME_ResourcesManager()
128 MESSAGE("SALOME_ResourcesManager destructor");
132 //=============================================================================
134 * shutdown all the containers, then the ContainerManager servant
136 //=============================================================================
138 void SALOME_ResourcesManager::Shutdown()
143 _NS->Destroy_Name(_ResourcesManagerNameInNS);
144 PortableServer::ObjectId_var oid = _poa->servant_to_id(this);
145 _poa->deactivate_object(oid);
149 * Return list of resources available (regarding content of CatalogResources.xml) but select only those with canRunContainers attribute set to true.
150 * And for each resource the number of proc available of it.
152 * \sa SALOME_ResourcesManager::ListAllResourcesInCatalog
154 void SALOME_ResourcesManager::ListAllAvailableResources(Engines::ResourceList_out machines, Engines::IntegerList_out nbProcsOfMachines)
156 const MapOfParserResourcesType& zeList(_rm->GetList());
157 std::vector<std::string> ret0;
158 std::vector<int> ret1;
159 for(MapOfParserResourcesType::const_iterator it=zeList.begin();it!=zeList.end();it++)
161 const ParserResourcesType& elt((*it).second);
162 if(elt.can_run_containers)
164 ret0.push_back(elt.HostName);
165 ret1.push_back(elt.DataForSort._nbOfNodes*elt.DataForSort._nbOfProcPerNode);
168 machines=new Engines::ResourceList;
169 nbProcsOfMachines=new Engines::IntegerList;
170 std::size_t sz(ret0.size());
171 machines->length((CORBA::ULong)sz); nbProcsOfMachines->length((CORBA::ULong)sz);
172 for(std::size_t j=0;j<sz;j++)
174 (*machines)[(CORBA::ULong)j]=CORBA::string_dup(ret0[j].c_str());
175 (*nbProcsOfMachines)[(CORBA::ULong)j]=ret1[j];
180 * Return list of resources available (regarding content of CatalogResources.xml) whatever canRunContainers attribute value.
182 * \sa SALOME_ResourcesManager::ListAllAvailableResources
184 Engines::ResourceList *SALOME_ResourcesManager::ListAllResourcesInCatalog()
186 const MapOfParserResourcesType& zeList(_rm->GetList());
187 auto sz = zeList.size();
188 Engines::ResourceList *ret(new Engines::ResourceList);
191 for(auto it : zeList)
193 (*ret)[i++] = CORBA::string_dup( it.second.HostName.c_str() );
198 //=============================================================================
199 //! get the name of resources fitting the specified constraints (params)
201 * If hostname specified, check it is local or known in resources catalog.
204 * - select first machines with corresponding OS (all machines if
205 * parameter OS empty),
206 * - then select the sublist of machines on which the component is known
207 * (if the result is empty, that probably means that the inventory of
208 * components is probably not done, so give complete list from previous step)
210 //=============================================================================
212 Engines::ResourceList *
213 SALOME_ResourcesManager::GetFittingResources(const Engines::ResourceParameters& params)
215 //MESSAGE("ResourcesManager::GetFittingResources");
216 Engines::ResourceList_var ret;
219 resourceParams p = resourceParameters_CORBAtoCPP(params);
223 // Call C++ ResourceManager
224 std::vector <std::string> vec = _rm->GetFittingResources(p);
227 ret = resourceList_CPPtoCORBA(vec);
229 catch(const ResourcesException &ex)
231 INFOS("Caught exception in GetFittingResources C++: " << ex.msg);
232 THROW_SALOME_CORBA_EXCEPTION(ex.msg.c_str(),SALOME::BAD_PARAM);
238 //=============================================================================
240 * dynamically obtains the first machines
242 //=============================================================================
245 SALOME_ResourcesManager::FindFirst(const Engines::ResourceList& listOfResources)
248 std::vector<std::string> rl = resourceList_CORBAtoCPP(listOfResources);
250 return CORBA::string_dup(_rm->Find("first", rl).c_str());
254 SALOME_ResourcesManager::Find(const char* policy, const Engines::ResourceList& listOfResources)
257 std::vector<std::string> rl = resourceList_CORBAtoCPP(listOfResources);
259 return CORBA::string_dup(_rm->Find(policy, rl).c_str());
262 Engines::ResourceDefinition*
263 SALOME_ResourcesManager::GetResourceDefinition(const char * name)
265 Engines::ResourceDefinition_var resDef;
267 ParserResourcesType resource = _rm->GetResourcesDescr(name);
268 resDef = resourceDefinition_CPPtoCORBA(resource);
269 } catch (const exception & ex) {
270 INFOS("Caught exception in GetResourceDefinition: " << ex.what());
271 THROW_SALOME_CORBA_EXCEPTION(ex.what(), SALOME::BAD_PARAM);
274 return resDef._retn();
278 SALOME_ResourcesManager::AddResource(const Engines::ResourceDefinition& new_resource,
279 CORBA::Boolean write,
280 const char * xml_file)
284 ParserResourcesType resource = resourceDefinition_CORBAtoCPP(new_resource);
285 _rm->AddResourceInCatalog(resource);
289 _rm->WriteInXmlFile(std::string(xml_file));
290 _rm->ParseXmlFiles();
293 catch (const SALOME_Exception & e)
295 INFOS("Error in AddResourceInCatalog: " << e);
296 THROW_SALOME_CORBA_EXCEPTION(e.what(), SALOME::BAD_PARAM);
298 catch (const ResourcesException & e)
300 INFOS("Error in AddResourceInCatalog: " << e.msg);
301 THROW_SALOME_CORBA_EXCEPTION(e.msg.c_str(), SALOME::BAD_PARAM);
306 SALOME_ResourcesManager::RemoveResource(const char * resource_name,
307 CORBA::Boolean write,
308 const char * xml_file)
312 _rm->DeleteResourceInCatalog(resource_name);
314 catch (const SALOME_Exception & e)
316 INFOS("Error in DeleteResourceInCatalog: " << e);
317 THROW_SALOME_CORBA_EXCEPTION(e.what(), SALOME::BAD_PARAM);
322 _rm->WriteInXmlFile(std::string(xml_file));
323 _rm->ParseXmlFiles();
328 SALOME_ResourcesManager::getMachineFile(const char * resource_name,
329 CORBA::Long nb_procs,
330 const char * parallelLib)
332 std::string machine_file_name("");
334 if (std::string(parallelLib) == "Dummy")
336 MESSAGE("[getMachineFile] parallelLib is Dummy");
337 MapOfParserResourcesType resourcesList = _rm->GetList();
338 if (resourcesList.find(std::string(resource_name)) != resourcesList.end())
340 ParserResourcesType resource = resourcesList[std::string(resource_name)];
342 // Check if resource is cluster or not
343 if (resource.ClusterMembersList.empty())
345 //It is not a cluster so we create a cluster with one machine
346 ParserResourcesType fake_node;
347 fake_node.HostName = resource.HostName;
348 fake_node.Protocol = resource.Protocol;
349 fake_node.ClusterInternalProtocol = resource.ClusterInternalProtocol;
350 fake_node.UserName = resource.UserName;
351 fake_node.AppliPath = resource.AppliPath;
352 fake_node.DataForSort = resource.DataForSort;
354 resource.ClusterMembersList.push_front(fake_node);
357 // Creating list of machines for creating the machine file
358 std::list<std::string> list_of_machines;
359 std::list<ParserResourcesType>::iterator cluster_it =
360 resource.ClusterMembersList.begin();
361 while (cluster_it != resource.ClusterMembersList.end())
363 // For each member of the cluster we add a nbOfNodes * nbOfProcPerNode in the list
364 unsigned int number_of_proc = (*cluster_it).DataForSort._nbOfNodes *
365 (*cluster_it).DataForSort._nbOfProcPerNode;
366 for (unsigned int i = 0; i < number_of_proc; i++)
367 list_of_machines.push_back((*cluster_it).HostName);
371 // Creating machine file
372 machine_file_name = tmpnam(NULL);
373 std::ofstream machine_file(machine_file_name.c_str(), std::ios_base::out);
375 CORBA::Long machine_number = 0;
376 std::list<std::string>::iterator it = list_of_machines.begin();
377 while (machine_number != nb_procs)
379 // Adding a new node to the machine file
380 machine_file << *it << std::endl;
384 if (it == list_of_machines.end())
385 it = list_of_machines.begin();
390 INFOS("[getMachineFile] Error resource_name not found in resourcesList -> " << resource_name);
392 else if (std::string(parallelLib) == "Mpi")
394 MESSAGE("[getMachineFile] parallelLib is Mpi");
396 MapOfParserResourcesType resourcesList = _rm->GetList();
397 if (resourcesList.find(std::string(resource_name)) != resourcesList.end())
399 ParserResourcesType resource = resourcesList[std::string(resource_name)];
400 // Check if resource is cluster or not
401 if (resource.ClusterMembersList.empty())
403 //It is not a cluster so we create a cluster with one machine
404 ParserResourcesType fake_node;
405 fake_node.HostName = resource.HostName;
406 fake_node.Protocol = resource.Protocol;
407 fake_node.ClusterInternalProtocol = resource.ClusterInternalProtocol;
408 fake_node.UserName = resource.UserName;
409 fake_node.AppliPath = resource.AppliPath;
410 fake_node.DataForSort = resource.DataForSort;
412 resource.ClusterMembersList.push_front(fake_node);
415 // Choose mpi implementation -> each MPI implementation has is own machinefile...
416 if (resource.mpi == lam)
418 // Creating machine file
419 machine_file_name = tmpnam(NULL);
420 std::ofstream machine_file(machine_file_name.c_str(), std::ios_base::out);
422 // We add all cluster machines to the file
423 std::list<ParserResourcesType>::iterator cluster_it =
424 resource.ClusterMembersList.begin();
425 while (cluster_it != resource.ClusterMembersList.end())
427 unsigned int number_of_proc = (*cluster_it).DataForSort._nbOfNodes *
428 (*cluster_it).DataForSort._nbOfProcPerNode;
429 machine_file << (*cluster_it).HostName << " cpu=" << number_of_proc << std::endl;
433 else if ((resource.mpi == openmpi) || (resource.mpi == ompi))
435 // Creating machine file
436 machine_file_name = tmpnam(NULL);
437 std::ofstream machine_file(machine_file_name.c_str(), std::ios_base::out);
439 // We add all cluster machines to the file
440 std::list<ParserResourcesType>::iterator cluster_it =
441 resource.ClusterMembersList.begin();
442 while (cluster_it != resource.ClusterMembersList.end())
444 unsigned int number_of_proc = (*cluster_it).DataForSort._nbOfNodes *
445 (*cluster_it).DataForSort._nbOfProcPerNode;
446 machine_file << (*cluster_it).HostName << " slots=" << number_of_proc << std::endl;
450 else if (resource.mpi == nompi)
452 INFOS("[getMachineFile] Error resource_name MPI implementation was defined for " << resource_name);
455 INFOS("[getMachineFile] Error resource_name MPI implementation not currently handled for " << resource_name);
458 INFOS("[getMachineFile] Error resource_name not found in resourcesList -> " << resource_name);
461 INFOS("[getMachineFile] Error parallelLib is not handled -> " << parallelLib);
463 return CORBA::string_dup(machine_file_name.c_str());