1 // Copyright (C) 2007-2014 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, PortableServer::POA_var poa, SALOME_NamingService *ns, const char *xmlFilePath) : _rm(xmlFilePath)
61 MESSAGE("SALOME_ResourcesManager constructor");
63 _orb = CORBA::ORB::_duplicate(orb) ;
65 PortableServer::POAManager_var pman = poa->the_POAManager();
66 CORBA::PolicyList policies;
68 PortableServer::ThreadPolicy_var threadPol(poa->create_thread_policy(PortableServer::SINGLE_THREAD_MODEL));
69 policies[0] = PortableServer::ThreadPolicy::_duplicate(threadPol);
70 _poa = poa->create_POA("SingleThreadPOA",pman,policies);
73 PortableServer::ObjectId_var id = _poa->activate_object(this);
74 CORBA::Object_var obj = _poa->id_to_reference(id);
75 Engines::ResourcesManager_var refContMan = Engines::ResourcesManager::_narrow(obj);
76 _NS->Register(refContMan,_ResourcesManagerNameInNS);
77 MESSAGE("SALOME_ResourcesManager constructor end");
80 //=============================================================================
82 * Standard constructor, parse resource file.
83 * - if ${APPLI} exists in environment,
84 * look for ${HOME}/${APPLI}/CatalogResources.xml
85 * - else look for default:
86 * ${KERNEL_ROOT_DIR}/share/salome/resources/kernel/CatalogResources.xml
87 * - parse XML resource file.
89 //=============================================================================
91 SALOME_ResourcesManager::SALOME_ResourcesManager(CORBA::ORB_ptr orb,
92 PortableServer::POA_var poa,
93 SALOME_NamingService *ns) : _rm()
95 MESSAGE("SALOME_ResourcesManager constructor");
97 _orb = CORBA::ORB::_duplicate(orb) ;
99 PortableServer::POAManager_var pman = poa->the_POAManager();
100 CORBA::PolicyList policies;
102 PortableServer::ThreadPolicy_var threadPol(poa->create_thread_policy(PortableServer::SINGLE_THREAD_MODEL));
103 policies[0] = PortableServer::ThreadPolicy::_duplicate(threadPol);
104 _poa = poa->create_POA("SingleThreadPOA",pman,policies);
105 threadPol->destroy();
107 PortableServer::ObjectId_var id = _poa->activate_object(this);
108 CORBA::Object_var obj = _poa->id_to_reference(id);
109 Engines::ResourcesManager_var refContMan = Engines::ResourcesManager::_narrow(obj);
110 _NS->Register(refContMan,_ResourcesManagerNameInNS);
112 MESSAGE("SALOME_ResourcesManager constructor end");
115 //=============================================================================
117 * Standard Destructor
119 //=============================================================================
121 SALOME_ResourcesManager::~SALOME_ResourcesManager()
123 MESSAGE("SALOME_ResourcesManager destructor");
127 //=============================================================================
129 * shutdown all the containers, then the ContainerManager servant
131 //=============================================================================
133 void SALOME_ResourcesManager::Shutdown()
136 _NS->Destroy_Name(_ResourcesManagerNameInNS);
137 PortableServer::ObjectId_var oid = _poa->servant_to_id(this);
138 _poa->deactivate_object(oid);
141 //=============================================================================
142 //! get the name of resources fitting the specified constraints (params)
144 * If hostname specified, check it is local or known in resources catalog.
147 * - select first machines with corresponding OS (all machines if
148 * parameter OS empty),
149 * - then select the sublist of machines on which the component is known
150 * (if the result is empty, that probably means that the inventory of
151 * components is probably not done, so give complete list from previous step)
153 //=============================================================================
155 Engines::ResourceList *
156 SALOME_ResourcesManager::GetFittingResources(const Engines::ResourceParameters& params)
158 MESSAGE("ResourcesManager::GetFittingResources");
159 Engines::ResourceList_var ret;
162 resourceParams p = resourceParameters_CORBAtoCPP(params);
166 // Call C++ ResourceManager
167 std::vector <std::string> vec = _rm.GetFittingResources(p);
170 ret = resourceList_CPPtoCORBA(vec);
172 catch(const ResourcesException &ex)
174 INFOS("Caught exception in GetFittingResources C++: " << ex.msg);
175 THROW_SALOME_CORBA_EXCEPTION(ex.msg.c_str(),SALOME::BAD_PARAM);
181 //=============================================================================
183 * dynamically obtains the first machines
185 //=============================================================================
188 SALOME_ResourcesManager::FindFirst(const Engines::ResourceList& listOfResources)
191 std::vector<std::string> rl = resourceList_CORBAtoCPP(listOfResources);
193 return CORBA::string_dup(_rm.Find("first", rl).c_str());
197 SALOME_ResourcesManager::Find(const char* policy, const Engines::ResourceList& listOfResources)
200 std::vector<std::string> rl = resourceList_CORBAtoCPP(listOfResources);
202 return CORBA::string_dup(_rm.Find(policy, rl).c_str());
205 Engines::ResourceDefinition*
206 SALOME_ResourcesManager::GetResourceDefinition(const char * name)
208 Engines::ResourceDefinition_var resDef;
210 ParserResourcesType resource = _rm.GetResourcesDescr(name);
211 resDef = resourceDefinition_CPPtoCORBA(resource);
212 } catch (const exception & ex) {
213 INFOS("Caught exception in GetResourceDefinition: " << ex.what());
214 THROW_SALOME_CORBA_EXCEPTION(ex.what(), SALOME::BAD_PARAM);
217 return resDef._retn();
221 SALOME_ResourcesManager::AddResource(const Engines::ResourceDefinition& new_resource,
222 CORBA::Boolean write,
223 const char * xml_file)
227 ParserResourcesType resource = resourceDefinition_CORBAtoCPP(new_resource);
228 _rm.AddResourceInCatalog(resource);
232 _rm.WriteInXmlFile(std::string(xml_file));
236 catch (const SALOME_Exception & e)
238 INFOS("Error in AddResourceInCatalog: " << e);
239 THROW_SALOME_CORBA_EXCEPTION(e.what(), SALOME::BAD_PARAM);
244 SALOME_ResourcesManager::RemoveResource(const char * resource_name,
245 CORBA::Boolean write,
246 const char * xml_file)
250 _rm.DeleteResourceInCatalog(resource_name);
252 catch (const SALOME_Exception & e)
254 INFOS("Error in DeleteResourceInCatalog: " << e);
255 THROW_SALOME_CORBA_EXCEPTION(e.what(), SALOME::BAD_PARAM);
260 _rm.WriteInXmlFile(std::string(xml_file));
266 SALOME_ResourcesManager::getMachineFile(std::string resource_name,
267 CORBA::Long nb_procs,
268 std::string parallelLib)
270 std::string machine_file_name("");
272 if (parallelLib == "Dummy")
274 MESSAGE("[getMachineFile] parallelLib is Dummy");
275 MapOfParserResourcesType resourcesList = _rm.GetList();
276 if (resourcesList.find(resource_name) != resourcesList.end())
278 ParserResourcesType resource = resourcesList[resource_name];
280 // Check if resource is cluster or not
281 if (resource.ClusterMembersList.empty())
283 //It is not a cluster so we create a cluster with one machine
284 ParserResourcesType fake_node;
285 fake_node.HostName = resource.HostName;
286 fake_node.Protocol = resource.Protocol;
287 fake_node.ClusterInternalProtocol = resource.ClusterInternalProtocol;
288 fake_node.UserName = resource.UserName;
289 fake_node.AppliPath = resource.AppliPath;
290 fake_node.DataForSort = resource.DataForSort;
292 resource.ClusterMembersList.push_front(fake_node);
295 // Creating list of machines for creating the machine file
296 std::list<std::string> list_of_machines;
297 std::list<ParserResourcesType>::iterator cluster_it =
298 resource.ClusterMembersList.begin();
299 while (cluster_it != resource.ClusterMembersList.end())
301 // For each member of the cluster we add a nbOfNodes * nbOfProcPerNode in the list
302 unsigned int number_of_proc = (*cluster_it).DataForSort._nbOfNodes *
303 (*cluster_it).DataForSort._nbOfProcPerNode;
304 for (unsigned int i = 0; i < number_of_proc; i++)
305 list_of_machines.push_back((*cluster_it).HostName);
309 // Creating machine file
310 machine_file_name = tmpnam(NULL);
311 std::ofstream machine_file(machine_file_name.c_str(), std::ios_base::out);
313 CORBA::Long machine_number = 0;
314 std::list<std::string>::iterator it = list_of_machines.begin();
315 while (machine_number != nb_procs)
317 // Adding a new node to the machine file
318 machine_file << *it << std::endl;
322 if (it == list_of_machines.end())
323 it = list_of_machines.begin();
328 INFOS("[getMachineFile] Error resource_name not found in resourcesList -> " << resource_name);
330 else if (parallelLib == "Mpi")
332 MESSAGE("[getMachineFile] parallelLib is Mpi");
334 MapOfParserResourcesType resourcesList = _rm.GetList();
335 if (resourcesList.find(resource_name) != resourcesList.end())
337 ParserResourcesType resource = resourcesList[resource_name];
338 // Check if resource is cluster or not
339 if (resource.ClusterMembersList.empty())
341 //It is not a cluster so we create a cluster with one machine
342 ParserResourcesType fake_node;
343 fake_node.HostName = resource.HostName;
344 fake_node.Protocol = resource.Protocol;
345 fake_node.ClusterInternalProtocol = resource.ClusterInternalProtocol;
346 fake_node.UserName = resource.UserName;
347 fake_node.AppliPath = resource.AppliPath;
348 fake_node.DataForSort = resource.DataForSort;
350 resource.ClusterMembersList.push_front(fake_node);
353 // Choose mpi implementation -> each MPI implementation has is own machinefile...
354 if (resource.mpi == lam)
356 // Creating machine file
357 machine_file_name = tmpnam(NULL);
358 std::ofstream machine_file(machine_file_name.c_str(), std::ios_base::out);
360 // We add all cluster machines to the file
361 std::list<ParserResourcesType>::iterator cluster_it =
362 resource.ClusterMembersList.begin();
363 while (cluster_it != resource.ClusterMembersList.end())
365 unsigned int number_of_proc = (*cluster_it).DataForSort._nbOfNodes *
366 (*cluster_it).DataForSort._nbOfProcPerNode;
367 machine_file << (*cluster_it).HostName << " cpu=" << number_of_proc << std::endl;
371 else if ((resource.mpi == openmpi) || (resource.mpi == ompi))
373 // Creating machine file
374 machine_file_name = tmpnam(NULL);
375 std::ofstream machine_file(machine_file_name.c_str(), std::ios_base::out);
377 // We add all cluster machines to the file
378 std::list<ParserResourcesType>::iterator cluster_it =
379 resource.ClusterMembersList.begin();
380 while (cluster_it != resource.ClusterMembersList.end())
382 unsigned int number_of_proc = (*cluster_it).DataForSort._nbOfNodes *
383 (*cluster_it).DataForSort._nbOfProcPerNode;
384 machine_file << (*cluster_it).HostName << " slots=" << number_of_proc << std::endl;
388 else if (resource.mpi == nompi)
390 INFOS("[getMachineFile] Error resource_name MPI implementation was defined for " << resource_name);
393 INFOS("[getMachineFile] Error resource_name MPI implementation not currenly handled for " << resource_name);
396 INFOS("[getMachineFile] Error resource_name not found in resourcesList -> " << resource_name);
399 INFOS("[getMachineFile] Error parallelLib is not handled -> " << parallelLib);
401 return machine_file_name;