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");
98 _lastErrorMessage = "";
101 MeshJobManager_i::~MeshJobManager_i() {
102 LOG("MeshJobManager_i::~MeshJobManager_i()");
106 // ====================================================================
107 // Helper functions to deals with the local and remote file systems
108 // ====================================================================
110 #include <fstream> // to get the file streams
112 #include <stdlib.h> // to get _splitpath
113 #include <direct.h> // to get _mkdir
115 #include <unistd.h> // to get basename
116 #include <sys/stat.h> // to get mkdir
117 #include <sys/types.h> // to get mkdir options
120 #include <stdlib.h> // to get system and getenv
122 static std::string OUTPUTFILE("output.med");
123 static std::string DATAFILE("data.txt");
124 static std::string SCRIPTFILE("padder.sh");
125 static std::string SEPARATOR(" ");
127 static std::string USER(getenv("USER"));
128 static std::string LOCAL_INPUTDIR("/tmp/spadder.local.inputdir."+USER);
129 static std::string LOCAL_RESULTDIR("/tmp/spadder.local.resultdir."+USER);
130 static std::string REMOTE_WORKDIR("/tmp/spadder.remote.workdir."+USER);
133 * This function creates the padder text input file containing the
134 * input data (list of filenames and groupnames) and returns the path
135 * of the created file. This function is the one that knows the format
136 * of the padder input file. If the input file format changes, then
137 * this function (and only this one) should be updated.
139 const char * MeshJobManager_i::_writeDataFile(std::vector<MESHJOB::MeshJobParameter> listConcreteMesh,
140 std::vector<MESHJOB::MeshJobParameter> listSteelBarMesh) {
142 _mkdir(LOCAL_INPUTDIR.c_str());
144 mkdir(LOCAL_INPUTDIR.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
147 // Make it static so that it's allocated once (constant name)
148 static std::string * dataFilename = new std::string(LOCAL_INPUTDIR+"/"+DATAFILE);
149 std::ofstream dataFile(dataFilename->c_str());
151 // We first specify the concrete mesh data (filename and groupname)
154 char fname[ _MAX_FNAME ];
155 _splitpath( listConcreteMesh[0].file_name, NULL, NULL, fname, NULL );
156 char* bname = &fname[0];
158 char* bname = basename(listConcreteMesh[0].file_name);
160 line = std::string(bname) + " " + std::string(listConcreteMesh[0].group_name);
161 dataFile << line.c_str() << std::endl;
162 // Note that we use here the basename because the files are supposed
163 // to be copied in the REMOTE_WORKDIR for execution.
165 // The, we can specify the steelbar mesh data, starting by the
167 int nbSteelBarMesh=listSteelBarMesh.size();
168 line = std::string("nbSteelbarMesh") + SEPARATOR + ToString(nbSteelBarMesh);
169 dataFile << line.c_str() << std::endl;
170 for (int i=0; i<nbSteelBarMesh; i++) {
172 char fname[ _MAX_FNAME ];
173 _splitpath( listSteelBarMesh[i].file_name, NULL, NULL, fname, NULL );
174 char* bname = &fname[0];
176 char* bname = basename(listSteelBarMesh[i].file_name);
178 line = std::string(bname) + " " + std::string(listSteelBarMesh[i].group_name);
179 dataFile << line.c_str() << std::endl;
182 // Finally, we conclude with the name of the output file
184 dataFile << line.c_str() << std::endl;
186 return dataFilename->c_str();
190 * This function creates a shell script that runs padder whith the
191 * specified data file, and returns the path of the created script
192 * file. The config id is used to retrieve the path to the binary file
193 * and other required files.
195 const char* MeshJobManager_i::_writeScriptFile(const char * dataFileName, const char * configId) {
197 _mkdir(LOCAL_INPUTDIR.c_str());
199 mkdir(LOCAL_INPUTDIR.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
202 // Make it static so that it's allocated once (constant name)
203 static std::string * scriptFilename = new std::string(LOCAL_INPUTDIR+"/"+SCRIPTFILE);
205 char * binpath = _configMap[configId].binpath;
206 char * envpath = _configMap[configId].envpath;
209 char fname[ _MAX_FNAME ];
210 _splitpath( dataFileName, NULL, NULL, fname, NULL );
211 const char* bname = &fname[0];
213 const char* bname = basename(dataFileName);
217 std::ofstream script(scriptFilename->c_str());
218 script << "#!/bin/sh" << std::endl;
219 script << "here=$(dirname $0)" << std::endl;
220 script << ". " << envpath << std::endl;
221 script << binpath << " $here/" << bname << std::endl;
222 // Note that we use the basename of the datafile because all data
223 // files are supposed to have been copied in the REMOTE_WORKDIR.
225 return scriptFilename->c_str();
229 // ====================================================================
230 // Functions to initialize and supervise the mesh computation job
231 // ====================================================================
233 bool MeshJobManager_i::configure(const char *configId,
234 const MESHJOB::ConfigParameter & configParameter)
236 beginService("MeshJobManager_i::configure");
238 _configMap[configId] = configParameter;
240 LOG("Adding configuration for " << configId);
241 LOG("- binpath = " << _configMap[configId].binpath);
242 LOG("- envpath = " << _configMap[configId].envpath);
244 endService("MeshJobManager_i::configure");
248 long MeshJobManager_i::JOBID_UNDEFINED = -1;
250 /*! Initialize a smesh computation job and return the job identifier */
251 CORBA::Long MeshJobManager_i::initialize(const MESHJOB::MeshJobParameterList & meshJobParameterList,
252 const char * configId)
254 beginService("MeshJobManager_i::initialize");
255 std::cerr << "##################################### initialize" << std::endl;
256 std::cerr << "#####################################" << std::endl;
259 // We first analyse the CORBA sequence to store data in C++ vectors
261 std::vector<MESHJOB::MeshJobParameter> listConcreteMesh;
262 std::vector<MESHJOB::MeshJobParameter> listSteelBarMesh;
263 for(CORBA::ULong i=0; i<meshJobParameterList.length(); i++) {
264 MESHJOB::MeshJobParameter currentMesh = meshJobParameterList[i];
265 switch ( currentMesh.file_type ) {
266 case MESHJOB::MED_CONCRETE:
267 listConcreteMesh.push_back(currentMesh);
269 case MESHJOB::MED_STEELBAR:
270 listSteelBarMesh.push_back(currentMesh);
273 LOG("The type of the file is not recognized");
274 return JOBID_UNDEFINED;
278 if ( listConcreteMesh.size() != 1 ) {
279 // Not consistent with the specification
280 LOG("You specify more than one concrete mesh");
281 return JOBID_UNDEFINED;
284 LOG("Nb. concrete mesh = " << listConcreteMesh.size());
285 LOG("Nb. steelbar mesh = " << listSteelBarMesh.size());
287 // We initiate here a datetime to tag the files and folder
288 // associated to this job.
290 DWORD jobDatetimeTag = timeGetTime();
292 long jobDatetimeTag = timetag();
294 // And a MESHJOB::MeshJobPaths structure to hold the directories
295 // where to find data
296 MESHJOB::MeshJobPaths * jobPaths = new MESHJOB::MeshJobPaths();
297 jobPaths->local_inputdir = LOCAL_INPUTDIR.c_str();
298 jobPaths->local_resultdir = (LOCAL_RESULTDIR + "." + ToString(jobDatetimeTag)).c_str();
299 jobPaths->remote_workdir = (REMOTE_WORKDIR + "." + ToString(jobDatetimeTag)).c_str();
302 // Then, we have to create the padder input data file. This input
303 // data is a text file containing the list of file names and group
306 const char * dataFilename = this->_writeDataFile(listConcreteMesh, listSteelBarMesh);
307 LOG("dataFilename = " << dataFilename);
308 const char * scriptFilename = this->_writeScriptFile(dataFilename, configId);
309 LOG("scriptFilename = " << scriptFilename);
312 // Then, the following instructions consists in preparing the job
313 // parameters to request the SALOME launcher for creating a new
316 Engines::JobParameters_var jobParameters = new Engines::JobParameters;
317 jobParameters->job_type = CORBA::string_dup("command");
318 // CAUTION: the job_file must be a single filename specifying a
319 // self-consistent script to be executed without any argument on the
321 jobParameters->job_file = CORBA::string_dup(scriptFilename);
324 // Specification of the working spaces:
326 // - local_directory: can be used to specify where to find the input
327 // files on the local resource. It's optionnal if you specify the
328 // absolute path name of input files.
330 // - result_directory: must be used to specify where to download the
331 // output files on the local resources
333 // - work_directory: must be used to specify the remote directory
334 // where to put all the stuff to run the job. Note that the job
335 // will be executed from within this directory, i.e. a change
336 // directory toward this working directory is done by the batch
337 // system before running the specified job script.
339 jobParameters->local_directory = CORBA::string_dup("");
340 jobParameters->result_directory = CORBA::string_dup(jobPaths->local_resultdir);
341 jobParameters->work_directory = CORBA::string_dup(jobPaths->remote_workdir);
343 // We specify the input files that are required to execute the
344 // job_file. If basenames are specified, then the files are supposed
345 // to be located in local_directory.
346 int nbFiles = listSteelBarMesh.size()+2;
347 // The number of input file is:
348 // (nb. of steelbar meshfile)
349 // + (1 concrete meshfile)
350 // + (1 padder input file)
351 // = nb steelbar meshfile + 2
352 jobParameters->in_files.length(nbFiles);
353 jobParameters->in_files[0] = CORBA::string_dup(listConcreteMesh[0].file_name);
354 for (int i=0; i<listSteelBarMesh.size(); i++) {
355 jobParameters->in_files[1+i] = CORBA::string_dup(listSteelBarMesh[i].file_name);
357 jobParameters->in_files[1+listSteelBarMesh.size()] = CORBA::string_dup(dataFilename);
358 // Note that all these input files will be copied in the
359 // REMOTE_WORKDIR on the remote host
361 // Then, we have to specify the existance of an output
362 // filenames. The path is supposed to be a path on the remote
363 // resource, i.e. where the job is executed.
364 jobParameters->out_files.length(1);
365 std::string outputfile_name = std::string(jobPaths->remote_workdir)+"/"+OUTPUTFILE;
366 jobParameters->out_files[0] = CORBA::string_dup(outputfile_name.c_str());
368 // CAUTION: the maximum duration has to be set with a format like "hh:mm"
369 jobParameters->maximum_duration = CORBA::string_dup("01:00");
370 jobParameters->queue = CORBA::string_dup("");
372 // Setting resource and additionnal properties (if needed)
373 // The resource parameters can be initiated from scratch, for
374 // example by specifying the values in hard coding:
376 //jobParameters->resource_required.name = CORBA::string_dup("localhost");
377 //jobParameters->resource_required.hostname = CORBA::string_dup("localhost");
378 //jobParameters->resource_required.mem_mb = 1024 * 10;
379 //jobParameters->resource_required.nb_proc = 1;
381 // But it's better to initiate these parameters from a resource
382 // definition known by the resource manager. This ensures that the
383 // resource will be available:
384 //const char * resourceName = "localhost";
385 //const char * resourceName = "boulant@claui2p1";
386 //const char * resourceName = "nepal@nepal";
387 const char * resourceName = _configMap[configId].resname;
389 Engines::ResourceDefinition * resourceDefinition;
391 resourceDefinition = _resourcesManager->GetResourceDefinition(resourceName);
393 catch (const CORBA::SystemException& ex) {
394 _lastErrorMessage = std::string("We can not access to the ressource ") + std::string(resourceName);
395 _lastErrorMessage+= std::string("(check the file CatalogResource.xml)");
396 LOG(_lastErrorMessage);
397 return JOBID_UNDEFINED;
399 // CAUTION: This resource should have been defined in the
400 // CatalogResource.xml associated to the SALOME application.
402 // Then, the values can be used to initiate the resource parameters
404 jobParameters->resource_required.name = CORBA::string_dup(resourceDefinition->name.in());
405 // CAUTION: the additionnal two following parameters MUST be
406 // specified explicitly, because they are not provided by the
407 // resource definition:
408 jobParameters->resource_required.mem_mb = resourceDefinition->mem_mb;
409 jobParameters->resource_required.nb_proc = resourceDefinition->nb_proc_per_node;
410 // CAUTION: the parameter mem_mb specifies the maximum memory value
411 // that could be allocated for executing the job. This takes into
412 // account not only the data that could be loaded by the batch
413 // process but also the linked dynamic library.
415 // A possible problem, for exemple in the case where you use the ssh
416 // emulation of a batch system, is to get an error message as below
417 // when libBatch try to run the ssh command:
419 // ##Â /usr/bin/ssh: error while loading shared libraries: libcrypto.so.0.9.8: failed
420 // ## to map segment from shared object: Cannot allocate memory
422 // In this exemple, the mem_mb was set to 1MB, value that is not
423 // sufficient to load the dynamic libraries linked to the ssh
424 // executable (libcrypto.so in the error message).
426 // So, even in the case of a simple test shell script, you should
427 // set this value at least to a standard threshold as 500MB
428 int jobId = JOBID_UNDEFINED;
430 jobId = _salomeLauncher->createJob(jobParameters);
431 // We register the datetime tag of this job
432 _jobDateTimeMap[jobId]=jobDatetimeTag;
433 _jobPathsMap[jobId] = jobPaths;
435 catch (const SALOME::SALOME_Exception & ex) {
436 LOG("SALOME Exception at initialization step !" <<ex.details.text.in());
437 _lastErrorMessage = ex.details.text.in();
438 return JOBID_UNDEFINED;
440 catch (const CORBA::SystemException& ex) {
441 LOG("Receive SALOME System Exception: "<<ex);
442 LOG("Check SALOME servers...");
443 _lastErrorMessage = "Check the SALOME servers (or try to restart SALOME)";
444 return JOBID_UNDEFINED;
447 endService("MeshJobManager_i::initialize");
451 /*! Submit the job execution and return true if submission is OK */
452 bool MeshJobManager_i::start(CORBA::Long jobId) {
453 beginService("MeshJobManager_i::start");
456 _salomeLauncher->launchJob(jobId);
458 catch (const SALOME::SALOME_Exception & ex) {
459 LOG("SALOME Exception in launchjob !" <<ex.details.text.in());
460 _lastErrorMessage = ex.details.text.in();
463 catch (const CORBA::SystemException& ex) {
464 LOG("Receive SALOME System Exception: "<<ex);
465 LOG("Check SALOME servers...");
466 _lastErrorMessage = "Check the SALOME servers (or try to restart SALOME)";
470 endService("MeshJobManager_i::initialize");
474 /*! Request the launch manager for the state of the specified job */
475 char* MeshJobManager_i::getState(CORBA::Long jobId) {
476 beginService("MeshJobManager_i::getState");
481 state = _salomeLauncher->getJobState(jobId);
483 catch (const SALOME::SALOME_Exception & ex)
485 LOG("SALOME Exception in getJobState !");
486 _lastErrorMessage = ex.details.text.in();
487 state = ex.details.text;
489 catch (const CORBA::SystemException& ex)
491 LOG("Receive SALOME System Exception: " << ex);
492 state="SALOME System Exception - see logs";
494 LOG("jobId="<<ToString(jobId)<<" state="<<state);
495 endService("MeshJobManager_i::getState");
496 return CORBA::string_dup(state.c_str());
499 MESHJOB::MeshJobPaths * MeshJobManager_i::getPaths(CORBA::Long jobId) {
501 MESHJOB::MeshJobPaths * jobPaths = _jobPathsMap[jobId];
502 if ( jobPaths == NULL ) {
503 LOG("You request the working paths for an undefined job (jobId="<<ToString(jobId)<<")");
504 return NULL; // Maybe raise an exception?
510 MESHJOB::MeshJobResults * MeshJobManager_i::finalize(CORBA::Long jobId) {
511 beginService("MeshJobManager_i::getResults");
512 MESHJOB::MeshJobResults * result = new MESHJOB::MeshJobResults();
514 MESHJOB::MeshJobPaths * jobPaths = this->getPaths(jobId);
515 std::string local_resultdir(jobPaths->local_resultdir);
516 result->results_dirname = local_resultdir.c_str();
519 _salomeLauncher->getJobResults(jobId, local_resultdir.c_str());
521 // __BUG__: to prevent from a bug of the MED driver (SALOME
522 // 5.1.5), we change the basename of the output file to force the
523 // complete reloading of data by the med driver.
524 long jobDatetimeTag = _jobDateTimeMap[jobId];
525 std::string outputFileName = "output"+ToString(jobDatetimeTag)+".med";
526 rename((local_resultdir+"/"+OUTPUTFILE).c_str(), (local_resultdir+"/"+outputFileName).c_str());
528 result->outputmesh_filename = outputFileName.c_str();
529 result->status = "OK";
531 catch (const SALOME::SALOME_Exception & ex)
533 LOG("SALOME Exception in getResults !");
534 result->status = "SALOME Exception in getResults !";
535 _lastErrorMessage = ex.details.text.in();
537 catch (const CORBA::SystemException& ex)
539 LOG("Receive CORBA System Exception: " << ex);
540 result->status = "Receive CORBA System Exception: see log";
542 endService("MeshJobManager_i::getResults");
547 /*! Clean all data associated to this job and remove the job from the launch manager */
548 bool MeshJobManager_i::clean(CORBA::Long jobId) {
549 beginService("MeshJobManager_i::clean");
551 // __GBO__ WORK IN PROGRESS: we just clean the temporary local
552 // directories. The remote working directories are tag with the
553 // execution datetime and the we prevent the task from conflict
554 // with files of another task.
555 MESHJOB::MeshJobPaths * jobPaths = this->getPaths(jobId);
556 if ( jobPaths == NULL ) return false;
559 // For safety reason (and prevent from bug that could erase the
560 // filesystem), we cancel the operation in the case where the
561 // directories to delete are not in the /tmp folder.
562 std::string shell_command("rm -rf ");
563 std::string inputdir(jobPaths->local_inputdir);
564 std::string resultdir(jobPaths->local_resultdir);
565 if ( !myStartsWith(inputdir,"/tmp/") ) {
566 LOG("WRN: The directory "<<inputdir<<" is not in /tmp. NO DELETE is done");
568 shell_command+=inputdir+" ";
570 if ( !myStartsWith(resultdir,"/tmp/")) {
571 LOG("WRN: The directory "<<resultdir<<" is not in /tmp. NO DELETE is done");
573 shell_command+=resultdir;
576 LOG("DBG: clean shell command = "<<shell_command);
578 bool cleanOk = false;
579 int error = system(shell_command.c_str());
580 if (error == 0) cleanOk = true;
582 endService("MeshJobManager_i::clean");
587 std::vector<std::string> * MeshJobManager_i::_getResourceNames() {
590 // These part is just to control the available resources
592 Engines::ResourceParameters params;
593 KERNEL::getLifeCycleCORBA()->preSet(params);
595 Engines::ResourceList * resourceList = _resourcesManager->GetFittingResources(params);
596 Engines::ResourceDefinition * resourceDefinition = NULL;
597 LOG("### resource list:");
598 std::vector<std::string>* resourceNames = new std::vector<std::string>();
600 for (int i = 0; i < resourceList->length(); i++) {
601 const char* aResourceName = (*resourceList)[i];
602 resourceNames->push_back(std::string(aResourceName));
603 LOG("resource["<<i<<"] = "<<aResourceName);
604 resourceDefinition = _resourcesManager->GetResourceDefinition(aResourceName);
605 LOG("protocol["<<i<<"] = "<<resourceDefinition->protocol);
609 // Note: a ResourceDefinition is used to create a batch configuration
610 // in the Launcher. This operation is done at Launcher startup from
611 // the configuration file CatalogResources.xml provided by the
612 // SALOME application.
613 // In the code instructions, you just have to choose a resource
614 // configuration by its name and then define the ResourceParameters
615 // that specify additionnal properties for a specific job submission
616 // (use the attribute resource_required of the JobParameters).
618 return resourceNames;
621 char* MeshJobManager_i::getLastErrorMessage() {
622 beginService("MeshJobManager_i::getState");
623 endService("MeshJobManager_i::getState");
624 return CORBA::string_dup(_lastErrorMessage.c_str());
628 // ==========================================================================
630 // ==========================================================================
634 PortableServer::ObjectId * MeshJobManagerEngine_factory( CORBA::ORB_ptr orb,
635 PortableServer::POA_ptr poa,
636 PortableServer::ObjectId * contId,
637 const char *instanceName,
638 const char *interfaceName)
640 LOG("PortableServer::ObjectId * MeshJobManagerEngine_factory()");
641 MeshJobManager_i * myEngine = new MeshJobManager_i(orb, poa, contId, instanceName, interfaceName);
642 return myEngine->getId() ;