1 // Copyright (C) 2011-2012 EDF R&D
3 // This library is free software; you can redistribute it and/or
4 // modify it under the terms of the GNU Lesser General Public
5 // License as published by the Free Software Foundation; either
6 // version 2.1 of the License.
8 // This library is distributed in the hope that it will be useful,
9 // but WITHOUT ANY WARRANTY; without even the implied warranty of
10 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 // Lesser General Public License for more details.
13 // You should have received a copy of the GNU Lesser General Public
14 // License along with this library; if not, write to the Free Software
15 // Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17 // See http://www.salome-platform.org/ or email : webmaster.salome@opencascade.com
19 // Authors : Guillaume Boulant (EDF) - 01/03/2011
28 #include "MeshJobManager_i.hxx"
30 #include <SALOMEconfig.h>
31 #include CORBA_SERVER_HEADER(SALOME_Exception)
34 #include "Basics_Utils.hxx" // For standard logging
36 #include "SALOME_KernelServices.hxx" // For CORBA logging
42 // ====================================================================
43 // General purpose helper functions (to put elsewhere at least)
44 // ====================================================================
48 * This function must be used to associate a datetime tag to a job
52 static long timetag() {
55 long tag = tv.tv_usec + tv.tv_sec*1000000;
61 * This function returns true if the string text starts with the string
64 static bool myStartsWith(const std::string& text,const std::string& token){
65 if(text.length() < token.length())
67 return (text.compare(0, token.length(), token) == 0);
71 // ====================================================================
72 // Constructor/Destructor
73 // ====================================================================
75 MeshJobManager_i::MeshJobManager_i(CORBA::ORB_ptr orb,
76 PortableServer::POA_ptr poa,
77 PortableServer::ObjectId * contId,
78 const char *instanceName,
79 const char *interfaceName)
80 : Engines_Component_i(orb, poa, contId, instanceName, interfaceName)
82 LOG("Activating MESHJOB::MeshJobManager object");
84 _id = _poa->activate_object(_thisObj);
86 _salomeLauncher = KERNEL::getSalomeLauncher();
87 if(CORBA::is_nil(_salomeLauncher)){
88 LOG("The SALOME launcher can't be reached ==> STOP");
89 throw KERNEL::createSalomeException("SALOME launcher can't be reached");
92 _resourcesManager = KERNEL::getResourcesManager();
93 if(CORBA::is_nil(_resourcesManager)){
94 LOG("The SALOME resource manager can't be reached ==> STOP");
95 throw KERNEL::createSalomeException("The SALOME resource manager can't be reached");
99 MeshJobManager_i::~MeshJobManager_i() {
100 LOG("MeshJobManager_i::~MeshJobManager_i()");
104 // ====================================================================
105 // Helper functions to deals with the local and remote file systems
106 // ====================================================================
108 #include <fstream> // to get the file streams
110 #include <stdlib.h> // to get _splitpath
111 #include <direct.h> // to get _mkdir
113 #include <unistd.h> // to get basename
114 #include <sys/stat.h> // to get mkdir
115 #include <sys/types.h> // to get mkdir options
118 #include <stdlib.h> // to get system and getenv
120 static std::string OUTPUTFILE("output.med");
121 static std::string DATAFILE("data.txt");
122 static std::string SCRIPTFILE("padder.sh");
123 static std::string SEPARATOR(" ");
125 static std::string USER(getenv("USER"));
126 static std::string LOCAL_INPUTDIR("/tmp/spadder.local.inputdir."+USER);
127 static std::string LOCAL_RESULTDIR("/tmp/spadder.local.resultdir."+USER);
128 static std::string REMOTE_WORKDIR("/tmp/spadder.remote.workdir."+USER);
131 * This function creates the padder text input file containing the
132 * input data (list of filenames and groupnames) and returns the path
133 * of the created file. This function is the one that knows the format
134 * of the padder input file. If the input file format changes, then
135 * this function (and only this one) should be updated.
137 const char * MeshJobManager_i::_writeDataFile(std::vector<MESHJOB::MeshJobParameter> listConcreteMesh,
138 std::vector<MESHJOB::MeshJobParameter> listSteelBarMesh) {
140 _mkdir(LOCAL_INPUTDIR.c_str());
142 mkdir(LOCAL_INPUTDIR.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
145 // Make it static so that it's allocated once (constant name)
146 static std::string * dataFilename = new std::string(LOCAL_INPUTDIR+"/"+DATAFILE);
147 std::ofstream dataFile(dataFilename->c_str());
149 // We first specify the concrete mesh data (filename and groupname)
152 char fname[ _MAX_FNAME ];
153 _splitpath( listConcreteMesh[0].file_name, NULL, NULL, fname, NULL );
154 char* bname = &fname[0];
156 char* bname = basename(listConcreteMesh[0].file_name);
158 line = std::string(bname) + " " + std::string(listConcreteMesh[0].group_name);
159 dataFile << line.c_str() << std::endl;
160 // Note that we use here the basename because the files are supposed
161 // to be copied in the REMOTE_WORKDIR for execution.
163 // The, we can specify the steelbar mesh data, starting by the
165 int nbSteelBarMesh=listSteelBarMesh.size();
166 line = std::string("nbSteelbarMesh") + SEPARATOR + ToString(nbSteelBarMesh);
167 dataFile << line.c_str() << std::endl;
168 for (int i=0; i<nbSteelBarMesh; i++) {
170 char fname[ _MAX_FNAME ];
171 _splitpath( listSteelBarMesh[i].file_name, NULL, NULL, fname, NULL );
172 char* bname = &fname[0];
174 char* bname = basename(listSteelBarMesh[i].file_name);
176 line = std::string(bname) + " " + std::string(listSteelBarMesh[i].group_name);
177 dataFile << line.c_str() << std::endl;
180 // Finally, we conclude with the name of the output file
182 dataFile << line.c_str() << std::endl;
184 return dataFilename->c_str();
188 * This function creates a shell script that runs padder whith the
189 * specified data file, and returns the path of the created script
190 * file. The config id is used to retrieve the path to the binary file
191 * and other required files.
193 const char* MeshJobManager_i::_writeScriptFile(const char * dataFileName, const char * configId) {
195 _mkdir(LOCAL_INPUTDIR.c_str());
197 mkdir(LOCAL_INPUTDIR.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
200 // Make it static so that it's allocated once (constant name)
201 static std::string * scriptFilename = new std::string(LOCAL_INPUTDIR+"/"+SCRIPTFILE);
203 char * binpath = _configMap[configId].binpath;
204 char * envpath = _configMap[configId].envpath;
207 char fname[ _MAX_FNAME ];
208 _splitpath( dataFileName, NULL, NULL, fname, NULL );
209 const char* bname = &fname[0];
211 const char* bname = basename(dataFileName);
215 std::ofstream script(scriptFilename->c_str());
216 script << "#!/bin/sh" << std::endl;
217 script << "here=$(dirname $0)" << std::endl;
218 script << ". " << envpath << std::endl;
219 script << binpath << " $here/" << bname << std::endl;
220 // Note that we use the basename of the datafile because all data
221 // files are supposed to have been copied in the REMOTE_WORKDIR.
223 return scriptFilename->c_str();
227 // ====================================================================
228 // Functions to initialize and supervise the mesh computation job
229 // ====================================================================
231 bool MeshJobManager_i::configure(const char *configId,
232 const MESHJOB::ConfigParameter & configParameter)
234 beginService("MeshJobManager_i::configure");
236 _configMap[configId] = configParameter;
238 LOG("Adding configuration for " << configId);
239 LOG("- binpath = " << _configMap[configId].binpath);
240 LOG("- envpath = " << _configMap[configId].envpath);
242 endService("MeshJobManager_i::configure");
246 long MeshJobManager_i::JOBID_UNDEFINED = -1;
248 /*! Initialize a smesh computation job and return the job identifier */
249 CORBA::Long MeshJobManager_i::initialize(const MESHJOB::MeshJobParameterList & meshJobParameterList,
250 const char * configId)
252 beginService("MeshJobManager_i::initialize");
253 std::cerr << "##################################### initialize" << std::endl;
254 std::cerr << "#####################################" << std::endl;
257 // We first analyse the CORBA sequence to store data in C++ vectors
259 std::vector<MESHJOB::MeshJobParameter> listConcreteMesh;
260 std::vector<MESHJOB::MeshJobParameter> listSteelBarMesh;
261 for(CORBA::ULong i=0; i<meshJobParameterList.length(); i++) {
262 MESHJOB::MeshJobParameter currentMesh = meshJobParameterList[i];
263 switch ( currentMesh.file_type ) {
264 case MESHJOB::MED_CONCRETE:
265 listConcreteMesh.push_back(currentMesh);
267 case MESHJOB::MED_STEELBAR:
268 listSteelBarMesh.push_back(currentMesh);
271 LOG("The type of the file is not recognized");
272 return JOBID_UNDEFINED;
276 if ( listConcreteMesh.size() != 1 ) {
277 // Not consistent with the specification
278 LOG("You specify more than one concrete mesh");
279 return JOBID_UNDEFINED;
282 LOG("Nb. concrete mesh = " << listConcreteMesh.size());
283 LOG("Nb. steelbar mesh = " << listSteelBarMesh.size());
285 // We initiate here a datetime to tag the files and folder
286 // associated to this job.
288 DWORD jobDatetimeTag = timeGetTime();
290 long jobDatetimeTag = timetag();
292 // And a MESHJOB::MeshJobPaths structure to hold the directories
293 // where to find data
294 MESHJOB::MeshJobPaths * jobPaths = new MESHJOB::MeshJobPaths();
295 jobPaths->local_inputdir = LOCAL_INPUTDIR.c_str();
296 jobPaths->local_resultdir = (LOCAL_RESULTDIR + "." + ToString(jobDatetimeTag)).c_str();
297 jobPaths->remote_workdir = (REMOTE_WORKDIR + "." + ToString(jobDatetimeTag)).c_str();
300 // Then, we have to create the padder input data file. This input
301 // data is a text file containing the list of file names and group
304 const char * dataFilename = this->_writeDataFile(listConcreteMesh, listSteelBarMesh);
305 LOG("dataFilename = " << dataFilename);
306 const char * scriptFilename = this->_writeScriptFile(dataFilename, configId);
307 LOG("scriptFilename = " << scriptFilename);
310 // Then, the following instructions consists in preparing the job
311 // parameters to request the SALOME launcher for creating a new
314 Engines::JobParameters_var jobParameters = new Engines::JobParameters;
315 jobParameters->job_type = CORBA::string_dup("command");
316 // CAUTION: the job_file must be a single filename specifying a
317 // self-consistent script to be executed without any argument on the
319 jobParameters->job_file = CORBA::string_dup(scriptFilename);
322 // Specification of the working spaces:
324 // - local_directory: can be used to specify where to find the input
325 // files on the local resource. It's optionnal if you specify the
326 // absolute path name of input files.
328 // - result_directory: must be used to specify where to download the
329 // output files on the local resources
331 // - work_directory: must be used to specify the remote directory
332 // where to put all the stuff to run the job. Note that the job
333 // will be executed from within this directory, i.e. a change
334 // directory toward this working directory is done by the batch
335 // system before running the specified job script.
337 jobParameters->local_directory = CORBA::string_dup("");
338 jobParameters->result_directory = CORBA::string_dup(jobPaths->local_resultdir);
339 jobParameters->work_directory = CORBA::string_dup(jobPaths->remote_workdir);
341 // We specify the input files that are required to execute the
342 // job_file. If basenames are specified, then the files are supposed
343 // to be located in local_directory.
344 int nbFiles = listSteelBarMesh.size()+2;
345 // The number of input file is:
346 // (nb. of steelbar meshfile)
347 // + (1 concrete meshfile)
348 // + (1 padder input file)
349 // = nb steelbar meshfile + 2
350 jobParameters->in_files.length(nbFiles);
351 jobParameters->in_files[0] = CORBA::string_dup(listConcreteMesh[0].file_name);
352 for (int i=0; i<listSteelBarMesh.size(); i++) {
353 jobParameters->in_files[1+i] = CORBA::string_dup(listSteelBarMesh[i].file_name);
355 jobParameters->in_files[1+listSteelBarMesh.size()] = CORBA::string_dup(dataFilename);
356 // Note that all these input files will be copied in the
357 // REMOTE_WORKDIR on the remote host
359 // Then, we have to specify the existance of an output
360 // filenames. The path is supposed to be a path on the remote
361 // resource, i.e. where the job is executed.
362 jobParameters->out_files.length(1);
363 std::string outputfile_name = std::string(jobPaths->remote_workdir)+"/"+OUTPUTFILE;
364 jobParameters->out_files[0] = CORBA::string_dup(outputfile_name.c_str());
366 // CAUTION: the maximum duration has to be set with a format like "hh:mm"
367 jobParameters->maximum_duration = CORBA::string_dup("01:00");
368 jobParameters->queue = CORBA::string_dup("");
370 // Setting resource and additionnal properties (if needed)
371 // The resource parameters can be initiated from scratch, for
372 // example by specifying the values in hard coding:
374 //jobParameters->resource_required.name = CORBA::string_dup("localhost");
375 //jobParameters->resource_required.hostname = CORBA::string_dup("localhost");
376 //jobParameters->resource_required.mem_mb = 1024 * 10;
377 //jobParameters->resource_required.nb_proc = 1;
379 // But it's better to initiate these parameters from a resource
380 // definition known by the resource manager. This ensures that the
381 // resource will be available:
382 //const char * resourceName = "localhost";
383 //const char * resourceName = "boulant@claui2p1";
384 //const char * resourceName = "nepal@nepal";
385 const char * resourceName = _configMap[configId].resname;
386 Engines::ResourceDefinition * resourceDefinition = _resourcesManager->GetResourceDefinition(resourceName);
387 // CAUTION: This resource should have been defined in the
388 // CatalogResource.xml associated to the SALOME application.
390 // Then, the values can be used to initiate the resource parameters
392 jobParameters->resource_required.name = CORBA::string_dup(resourceDefinition->name.in());
393 // CAUTION: the additionnal two following parameters MUST be
394 // specified explicitly, because they are not provided by the
395 // resource definition:
396 jobParameters->resource_required.mem_mb = resourceDefinition->mem_mb;
397 jobParameters->resource_required.nb_proc = resourceDefinition->nb_proc_per_node;
398 // CAUTION: the parameter mem_mb specifies the maximum memory value
399 // that could be allocated for executing the job. This takes into
400 // account not only the data that could be loaded by the batch
401 // process but also the linked dynamic library.
403 // A possible problem, for exemple in the case where you use the ssh
404 // emulation of a batch system, is to get an error message as below
405 // when libBatch try to run the ssh command:
407 // ## /usr/bin/ssh: error while loading shared libraries: libcrypto.so.0.9.8: failed
408 // ## to map segment from shared object: Cannot allocate memory
410 // In this exemple, the mem_mb was set to 1MB, value that is not
411 // sufficient to load the dynamic libraries linked to the ssh
412 // executable (libcrypto.so in the error message).
414 // So, even in the case of a simple test shell script, you should
415 // set this value at least to a standard threshold as 500MB
417 int jobId = JOBID_UNDEFINED;
419 std::cerr << "#####################################" << std::endl;
420 std::cerr << "#####################################" << std::endl;
421 std::cerr << "jobUndef = " << JOBID_UNDEFINED << std::endl;
422 jobId = _salomeLauncher->createJob(jobParameters);
423 std::cerr << "#####################################" << std::endl;
424 std::cerr << "#####################################" << std::endl;
425 std::cerr << "#####################################" << std::endl;
426 std::cerr << "jobId = " << jobId << std::endl;
427 // We register the datetime tag of this job
428 _jobDateTimeMap[jobId]=jobDatetimeTag;
429 _jobPathsMap[jobId] = jobPaths;
431 catch (const SALOME::SALOME_Exception & ex) {
432 LOG("SALOME Exception in createJob !" <<ex.details.text.in());
433 //LOG(ex.details.text.in());
434 return JOBID_UNDEFINED;
436 catch (const CORBA::SystemException& ex) {
437 LOG("Receive SALOME System Exception: "<<ex);
438 LOG("Check SALOME servers...");
439 return JOBID_UNDEFINED;
442 endService("MeshJobManager_i::initialize");
446 /*! Submit the job execution and return true if submission is OK */
447 bool MeshJobManager_i::start(CORBA::Long jobId) {
448 beginService("MeshJobManager_i::start");
451 _salomeLauncher->launchJob(jobId);
453 catch (const SALOME::SALOME_Exception & ex) {
454 LOG("SALOME Exception in launchjob !" <<ex.details.text.in());
455 //LOG(ex.details.text.in());
458 catch (const CORBA::SystemException& ex) {
459 LOG("Receive SALOME System Exception: "<<ex);
460 LOG("Check SALOME servers...");
464 endService("MeshJobManager_i::initialize");
468 /*! Request the launch manager for the state of the specified job */
469 char* MeshJobManager_i::getState(CORBA::Long jobId) {
470 beginService("MeshJobManager_i::getState");
475 state = _salomeLauncher->getJobState(jobId);
477 catch (const SALOME::SALOME_Exception & ex)
479 LOG("SALOME Exception in getJobState !");
480 state = ex.details.text;
482 catch (const CORBA::SystemException& ex)
484 LOG("Receive SALOME System Exception: " << ex);
485 state="SALOME System Exception - see logs";
487 LOG("jobId="<<ToString(jobId)<<" state="<<state);
488 endService("MeshJobManager_i::getState");
489 return CORBA::string_dup(state.c_str());
492 MESHJOB::MeshJobPaths * MeshJobManager_i::getPaths(CORBA::Long jobId) {
494 MESHJOB::MeshJobPaths * jobPaths = _jobPathsMap[jobId];
495 if ( jobPaths == NULL ) {
496 LOG("You request the working paths for an undefined job (jobId="<<ToString(jobId)<<")");
497 return NULL; // Maybe raise an exception?
503 MESHJOB::MeshJobResults * MeshJobManager_i::finalize(CORBA::Long jobId) {
504 beginService("MeshJobManager_i::getResults");
505 MESHJOB::MeshJobResults * result = new MESHJOB::MeshJobResults();
507 MESHJOB::MeshJobPaths * jobPaths = this->getPaths(jobId);
508 std::string local_resultdir(jobPaths->local_resultdir);
509 result->results_dirname = local_resultdir.c_str();
512 _salomeLauncher->getJobResults(jobId, local_resultdir.c_str());
514 // __BUG__: to prevent from a bug of the MED driver (SALOME
515 // 5.1.5), we change the basename of the output file to force the
516 // complete reloading of data by the med driver.
517 long jobDatetimeTag = _jobDateTimeMap[jobId];
518 std::string outputFileName = "output"+ToString(jobDatetimeTag)+".med";
519 rename((local_resultdir+"/"+OUTPUTFILE).c_str(), (local_resultdir+"/"+outputFileName).c_str());
521 result->outputmesh_filename = outputFileName.c_str();
522 result->status = "OK";
524 catch (const SALOME::SALOME_Exception & ex)
526 LOG("SALOME Exception in getResults !");
527 result->status = "SALOME Exception in getResults !";
529 catch (const CORBA::SystemException& ex)
531 LOG("Receive CORBA System Exception: " << ex);
532 result->status = "Receive CORBA System Exception: see log";
534 endService("MeshJobManager_i::getResults");
539 /*! Clean all data associated to this job and remove the job from the launch manager */
540 bool MeshJobManager_i::clean(CORBA::Long jobId) {
541 beginService("MeshJobManager_i::clean");
543 // __GBO__ WORK IN PROGRESS: we just clean the temporary local
544 // directories. The remote working directories are tag with the
545 // execution datetime and the we prevent the task from conflict
546 // with files of another task.
547 MESHJOB::MeshJobPaths * jobPaths = this->getPaths(jobId);
548 if ( jobPaths == NULL ) return false;
551 // For safety reason (and prevent from bug that could erase the
552 // filesystem), we cancel the operation in the case where the
553 // directories to delete are not in the /tmp folder.
554 std::string shell_command("rm -rf ");
555 std::string inputdir(jobPaths->local_inputdir);
556 std::string resultdir(jobPaths->local_resultdir);
557 if ( !myStartsWith(inputdir,"/tmp/") ) {
558 LOG("WRN: The directory "<<inputdir<<" is not in /tmp. NO DELETE is done");
560 shell_command+=inputdir+" ";
562 if ( !myStartsWith(resultdir,"/tmp/")) {
563 LOG("WRN: The directory "<<resultdir<<" is not in /tmp. NO DELETE is done");
565 shell_command+=resultdir;
568 LOG("DBG: clean shell command = "<<shell_command);
570 bool cleanOk = false;
571 int error = system(shell_command.c_str());
572 if (error == 0) cleanOk = true;
574 endService("MeshJobManager_i::clean");
579 std::vector<std::string> * MeshJobManager_i::_getResourceNames() {
582 // These part is just to control the available resources
584 Engines::ResourceParameters params;
585 KERNEL::getLifeCycleCORBA()->preSet(params);
587 Engines::ResourceList * resourceList = _resourcesManager->GetFittingResources(params);
588 Engines::ResourceDefinition * resourceDefinition = NULL;
589 LOG("### resource list:");
590 std::vector<std::string>* resourceNames = new std::vector<std::string>();
592 for (int i = 0; i < resourceList->length(); i++) {
593 const char* aResourceName = (*resourceList)[i];
594 resourceNames->push_back(std::string(aResourceName));
595 LOG("resource["<<i<<"] = "<<aResourceName);
596 resourceDefinition = _resourcesManager->GetResourceDefinition(aResourceName);
597 LOG("protocol["<<i<<"] = "<<resourceDefinition->protocol);
601 // Note: a ResourceDefinition is used to create a batch configuration
602 // in the Launcher. This operation is done at Launcher startup from
603 // the configuration file CatalogResources.xml provided by the
604 // SALOME application.
605 // In the code instructions, you just have to choose a resource
606 // configuration by its name and then define the ResourceParameters
607 // that specify additionnal properties for a specific job submission
608 // (use the attribute resource_required of the JobParameters).
610 return resourceNames;
615 // ==========================================================================
617 // ==========================================================================
621 PortableServer::ObjectId * MeshJobManagerEngine_factory( CORBA::ORB_ptr orb,
622 PortableServer::POA_ptr poa,
623 PortableServer::ObjectId * contId,
624 const char *instanceName,
625 const char *interfaceName)
627 LOG("PortableServer::ObjectId * MeshJobManagerEngine_factory()");
628 MeshJobManager_i * myEngine = new MeshJobManager_i(orb, poa, contId, instanceName, interfaceName);
629 return myEngine->getId() ;