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");
255 // We first analyse the CORBA sequence to store data in C++ vectors
257 std::vector<MESHJOB::MeshJobParameter> listConcreteMesh;
258 std::vector<MESHJOB::MeshJobParameter> listSteelBarMesh;
259 for(CORBA::ULong i=0; i<meshJobParameterList.length(); i++) {
260 MESHJOB::MeshJobParameter currentMesh = meshJobParameterList[i];
261 switch ( currentMesh.file_type ) {
262 case MESHJOB::MED_CONCRETE:
263 listConcreteMesh.push_back(currentMesh);
265 case MESHJOB::MED_STEELBAR:
266 listSteelBarMesh.push_back(currentMesh);
269 LOG("The type of the file is not recognized");
270 return JOBID_UNDEFINED;
274 if ( listConcreteMesh.size() != 1 ) {
275 // Not consistent with the specification
276 LOG("You specify more than one concrete mesh");
277 return JOBID_UNDEFINED;
280 LOG("Nb. concrete mesh = " << listConcreteMesh.size());
281 LOG("Nb. steelbar mesh = " << listSteelBarMesh.size());
283 // We initiate here a datetime to tag the files and folder
284 // associated to this job.
286 DWORD jobDatetimeTag = timeGetTime();
288 long jobDatetimeTag = timetag();
290 // And a MESHJOB::MeshJobPaths structure to hold the directories
291 // where to find data
292 MESHJOB::MeshJobPaths * jobPaths = new MESHJOB::MeshJobPaths();
293 jobPaths->local_inputdir = LOCAL_INPUTDIR.c_str();
294 jobPaths->local_resultdir = (LOCAL_RESULTDIR + "." + ToString(jobDatetimeTag)).c_str();
295 jobPaths->remote_workdir = (REMOTE_WORKDIR + "." + ToString(jobDatetimeTag)).c_str();
298 // Then, we have to create the padder input data file. This input
299 // data is a text file containing the list of file names and group
302 const char * dataFilename = this->_writeDataFile(listConcreteMesh, listSteelBarMesh);
303 LOG("dataFilename = " << dataFilename);
304 const char * scriptFilename = this->_writeScriptFile(dataFilename, configId);
305 LOG("scriptFilename = " << scriptFilename);
308 // Then, the following instructions consists in preparing the job
309 // parameters to request the SALOME launcher for creating a new
312 Engines::JobParameters_var jobParameters = new Engines::JobParameters;
313 jobParameters->job_type = CORBA::string_dup("command");
314 // CAUTION: the job_file must be a single filename specifying a
315 // self-consistent script to be executed without any argument on the
317 jobParameters->job_file = CORBA::string_dup(scriptFilename);
320 // Specification of the working spaces:
322 // - local_directory: can be used to specify where to find the input
323 // files on the local resource. It's optionnal if you specify the
324 // absolute path name of input files.
326 // - result_directory: must be used to specify where to download the
327 // output files on the local resources
329 // - work_directory: must be used to specify the remote directory
330 // where to put all the stuff to run the job. Note that the job
331 // will be executed from within this directory, i.e. a change
332 // directory toward this working directory is done by the batch
333 // system before running the specified job script.
335 jobParameters->local_directory = CORBA::string_dup("");
336 jobParameters->result_directory = CORBA::string_dup(jobPaths->local_resultdir);
337 jobParameters->work_directory = CORBA::string_dup(jobPaths->remote_workdir);
339 // We specify the input files that are required to execute the
340 // job_file. If basenames are specified, then the files are supposed
341 // to be located in local_directory.
342 int nbFiles = listSteelBarMesh.size()+2;
343 // The number of input file is:
344 // (nb. of steelbar meshfile)
345 // + (1 concrete meshfile)
346 // + (1 padder input file)
347 // = nb steelbar meshfile + 2
348 jobParameters->in_files.length(nbFiles);
349 jobParameters->in_files[0] = CORBA::string_dup(listConcreteMesh[0].file_name);
350 for (int i=0; i<listSteelBarMesh.size(); i++) {
351 jobParameters->in_files[1+i] = CORBA::string_dup(listSteelBarMesh[i].file_name);
353 jobParameters->in_files[1+listSteelBarMesh.size()] = CORBA::string_dup(dataFilename);
354 // Note that all these input files will be copied in the
355 // REMOTE_WORKDIR on the remote host
357 // Then, we have to specify the existance of an output
358 // filenames. The path is supposed to be a path on the remote
359 // resource, i.e. where the job is executed.
360 jobParameters->out_files.length(1);
361 std::string outputfile_name = std::string(jobPaths->remote_workdir)+"/"+OUTPUTFILE;
362 jobParameters->out_files[0] = CORBA::string_dup(outputfile_name.c_str());
364 // CAUTION: the maximum duration has to be set with a format like "hh:mm"
365 jobParameters->maximum_duration = CORBA::string_dup("01:00");
366 jobParameters->queue = CORBA::string_dup("");
368 // Setting resource and additionnal properties (if needed)
369 // The resource parameters can be initiated from scratch, for
370 // example by specifying the values in hard coding:
372 //jobParameters->resource_required.name = CORBA::string_dup("localhost");
373 //jobParameters->resource_required.hostname = CORBA::string_dup("localhost");
374 //jobParameters->resource_required.mem_mb = 1024 * 10;
375 //jobParameters->resource_required.nb_proc = 1;
377 // But it's better to initiate these parameters from a resource
378 // definition known by the resource manager. This ensures that the
379 // resource will be available:
380 //const char * resourceName = "localhost";
381 //const char * resourceName = "boulant@claui2p1";
382 //const char * resourceName = "nepal@nepal";
383 const char * resourceName = _configMap[configId].resname;
384 Engines::ResourceDefinition * resourceDefinition = _resourcesManager->GetResourceDefinition(resourceName);
385 // CAUTION: This resource should have been defined in the
386 // CatalogResource.xml associated to the SALOME application.
388 // Then, the values can be used to initiate the resource parameters
390 jobParameters->resource_required.name = CORBA::string_dup(resourceDefinition->name.in());
391 // CAUTION: the additionnal two following parameters MUST be
392 // specified explicitly, because they are not provided by the
393 // resource definition:
394 jobParameters->resource_required.mem_mb = resourceDefinition->mem_mb;
395 jobParameters->resource_required.nb_proc = resourceDefinition->nb_proc_per_node;
396 // CAUTION: the parameter mem_mb specifies the maximum memory value
397 // that could be allocated for executing the job. This takes into
398 // account not only the data that could be loaded by the batch
399 // process but also the linked dynamic library.
401 // A possible problem, for exemple in the case where you use the ssh
402 // emulation of a batch system, is to get an error message as below
403 // when libBatch try to run the ssh command:
405 // ##Â /usr/bin/ssh: error while loading shared libraries: libcrypto.so.0.9.8: failed
406 // ## to map segment from shared object: Cannot allocate memory
408 // In this exemple, the mem_mb was set to 1MB, value that is not
409 // sufficient to load the dynamic libraries linked to the ssh
410 // executable (libcrypto.so in the error message).
412 // So, even in the case of a simple test shell script, you should
413 // set this value at least to a standard threshold as 500MB
415 int jobId = JOBID_UNDEFINED;
417 jobId = _salomeLauncher->createJob(jobParameters);
418 // We register the datetime tag of this job
419 _jobDateTimeMap[jobId]=jobDatetimeTag;
420 _jobPathsMap[jobId] = jobPaths;
422 catch (const SALOME::SALOME_Exception & ex) {
423 LOG("SALOME Exception in createJob !" <<ex.details.text.in());
424 //LOG(ex.details.text.in());
425 return JOBID_UNDEFINED;
427 catch (const CORBA::SystemException& ex) {
428 LOG("Receive SALOME System Exception: "<<ex);
429 LOG("Check SALOME servers...");
430 return JOBID_UNDEFINED;
433 endService("MeshJobManager_i::initialize");
437 /*! Submit the job execution and return true if submission is OK */
438 bool MeshJobManager_i::start(CORBA::Long jobId) {
439 beginService("MeshJobManager_i::start");
442 _salomeLauncher->launchJob(jobId);
444 catch (const SALOME::SALOME_Exception & ex) {
445 LOG("SALOME Exception in createJob !" <<ex.details.text.in());
446 //LOG(ex.details.text.in());
449 catch (const CORBA::SystemException& ex) {
450 LOG("Receive SALOME System Exception: "<<ex);
451 LOG("Check SALOME servers...");
455 endService("MeshJobManager_i::initialize");
459 /*! Request the launch manager for the state of the specified job */
460 char* MeshJobManager_i::getState(CORBA::Long jobId) {
461 beginService("MeshJobManager_i::getState");
466 state = _salomeLauncher->getJobState(jobId);
468 catch (const SALOME::SALOME_Exception & ex)
470 LOG("SALOME Exception in getJobState !");
471 state = ex.details.text;
473 catch (const CORBA::SystemException& ex)
475 LOG("Receive SALOME System Exception: " << ex);
476 state="SALOME System Exception - see logs";
478 LOG("jobId="<<ToString(jobId)<<" state="<<state);
479 endService("MeshJobManager_i::getState");
480 return CORBA::string_dup(state.c_str());
483 MESHJOB::MeshJobPaths * MeshJobManager_i::getPaths(CORBA::Long jobId) {
485 MESHJOB::MeshJobPaths * jobPaths = _jobPathsMap[jobId];
486 if ( jobPaths == NULL ) {
487 LOG("You request the working paths for an undefined job (jobId="<<ToString(jobId)<<")");
488 return NULL; // Maybe raise an exception?
494 MESHJOB::MeshJobResults * MeshJobManager_i::finalize(CORBA::Long jobId) {
495 beginService("MeshJobManager_i::getResults");
496 MESHJOB::MeshJobResults * result = new MESHJOB::MeshJobResults();
498 MESHJOB::MeshJobPaths * jobPaths = this->getPaths(jobId);
499 std::string local_resultdir(jobPaths->local_resultdir);
500 result->results_dirname = local_resultdir.c_str();
503 _salomeLauncher->getJobResults(jobId, local_resultdir.c_str());
505 // __BUG__: to prevent from a bug of the MED driver (SALOME
506 // 5.1.5), we change the basename of the output file to force the
507 // complete reloading of data by the med driver.
508 long jobDatetimeTag = _jobDateTimeMap[jobId];
509 std::string outputFileName = "output"+ToString(jobDatetimeTag)+".med";
510 rename((local_resultdir+"/"+OUTPUTFILE).c_str(), (local_resultdir+"/"+outputFileName).c_str());
512 result->outputmesh_filename = outputFileName.c_str();
513 result->status = "OK";
515 catch (const SALOME::SALOME_Exception & ex)
517 LOG("SALOME Exception in getResults !");
518 result->status = "SALOME Exception in getResults !";
520 catch (const CORBA::SystemException& ex)
522 LOG("Receive CORBA System Exception: " << ex);
523 result->status = "Receive CORBA System Exception: see log";
525 endService("MeshJobManager_i::getResults");
530 /*! Clean all data associated to this job and remove the job from the launch manager */
531 bool MeshJobManager_i::clean(CORBA::Long jobId) {
532 beginService("MeshJobManager_i::clean");
534 // __GBO__ WORK IN PROGRESS: we just clean the temporary local
535 // directories. The remote working directories are tag with the
536 // execution datetime and the we prevent the task from conflict
537 // with files of another task.
538 MESHJOB::MeshJobPaths * jobPaths = this->getPaths(jobId);
539 if ( jobPaths == NULL ) return false;
542 // For safety reason (and prevent from bug that could erase the
543 // filesystem), we cancel the operation in the case where the
544 // directories to delete are not in the /tmp folder.
545 std::string shell_command("rm -rf ");
546 std::string inputdir(jobPaths->local_inputdir);
547 std::string resultdir(jobPaths->local_resultdir);
548 if ( !myStartsWith(inputdir,"/tmp/") ) {
549 LOG("WRN: The directory "<<inputdir<<" is not in /tmp. NO DELETE is done");
551 shell_command+=inputdir+" ";
553 if ( !myStartsWith(resultdir,"/tmp/")) {
554 LOG("WRN: The directory "<<resultdir<<" is not in /tmp. NO DELETE is done");
556 shell_command+=resultdir;
559 LOG("DBG: clean shell command = "<<shell_command);
561 bool cleanOk = false;
562 int error = system(shell_command.c_str());
563 if (error == 0) cleanOk = true;
565 endService("MeshJobManager_i::clean");
570 std::vector<std::string> * MeshJobManager_i::_getResourceNames() {
573 // These part is just to control the available resources
575 Engines::ResourceParameters params;
576 KERNEL::getLifeCycleCORBA()->preSet(params);
578 Engines::ResourceList * resourceList = _resourcesManager->GetFittingResources(params);
579 Engines::ResourceDefinition * resourceDefinition = NULL;
580 LOG("### resource list:");
581 std::vector<std::string>* resourceNames = new std::vector<std::string>();
583 for (int i = 0; i < resourceList->length(); i++) {
584 const char* aResourceName = (*resourceList)[i];
585 resourceNames->push_back(std::string(aResourceName));
586 LOG("resource["<<i<<"] = "<<aResourceName);
587 resourceDefinition = _resourcesManager->GetResourceDefinition(aResourceName);
588 LOG("protocol["<<i<<"] = "<<resourceDefinition->protocol);
592 // Note: a ResourceDefinition is used to create a batch configuration
593 // in the Launcher. This operation is done at Launcher startup from
594 // the configuration file CatalogResources.xml provided by the
595 // SALOME application.
596 // In the code instructions, you just have to choose a resource
597 // configuration by its name and then define the ResourceParameters
598 // that specify additionnal properties for a specific job submission
599 // (use the attribute resource_required of the JobParameters).
601 return resourceNames;
606 // ==========================================================================
608 // ==========================================================================
612 PortableServer::ObjectId * MeshJobManagerEngine_factory( CORBA::ORB_ptr orb,
613 PortableServer::POA_ptr poa,
614 PortableServer::ObjectId * contId,
615 const char *instanceName,
616 const char *interfaceName)
618 LOG("PortableServer::ObjectId * MeshJobManagerEngine_factory()");
619 MeshJobManager_i * myEngine = new MeshJobManager_i(orb, poa, contId, instanceName, interfaceName);
620 return myEngine->getId() ;