1 // Copyright (C) 2011 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
21 #include "MeshJobManager_i.hxx"
23 #include <SALOMEconfig.h>
24 #include CORBA_SERVER_HEADER(SALOME_Exception)
26 #include "Basics_Utils.hxx" // For standard logging
27 #include "SALOME_KernelServices.hxx" // For CORBA logging
31 #include "testhelper.hxx"
34 // ====================================================================
35 // General purpose helper functions (to put elsewhere at least)
36 // ====================================================================
40 * This function must be used to associate a datetime tag to a job
42 static long timetag() {
45 long tag = tv.tv_usec + tv.tv_sec*1000000;
50 * This function returns true if the string text starts with the string
53 static bool myStartsWith(const std::string& text,const std::string& token){
54 if(text.length() < token.length())
56 return (text.compare(0, token.length(), token) == 0);
60 // ====================================================================
61 // Constructor/Destructor
62 // ====================================================================
64 MeshJobManager_i::MeshJobManager_i(CORBA::ORB_ptr orb,
65 PortableServer::POA_ptr poa,
66 PortableServer::ObjectId * contId,
67 const char *instanceName,
68 const char *interfaceName)
69 : Engines_Component_i(orb, poa, contId, instanceName, interfaceName)
71 LOG("Activating MESHJOB::MeshJobManager object");
73 _id = _poa->activate_object(_thisObj);
75 _salomeLauncher = KERNEL::getSalomeLauncher();
76 if(CORBA::is_nil(_salomeLauncher)){
77 LOG("The SALOME launcher can't be reached ==> STOP");
78 throw KERNEL::createSalomeException("SALOME launcher can't be reached");
81 _resourcesManager = KERNEL::getResourcesManager();
82 if(CORBA::is_nil(_resourcesManager)){
83 LOG("The SALOME resource manager can't be reached ==> STOP");
84 throw KERNEL::createSalomeException("The SALOME resource manager can't be reached");
88 MeshJobManager_i::~MeshJobManager_i() {
89 LOG("MeshJobManager_i::~MeshJobManager_i()");
93 // ====================================================================
94 // Helper functions to deals with the local and remote file systems
95 // ====================================================================
97 #include <fstream> // to get the file streams
98 #include <sys/stat.h> // to get mkdir
99 #include <sys/types.h> // to get mkdir options
100 #include <unistd.h> // to get basename
101 #include <stdlib.h> // to get system and getenv
103 static std::string OUTPUTFILE("output.med");
104 static std::string DATAFILE("data.txt");
105 static std::string SCRIPTFILE("padder.sh");
106 static std::string SEPARATOR(" ");
108 static std::string USER(getenv("USER"));
109 static std::string LOCAL_INPUTDIR("/tmp/spadder.local.inputdir."+USER);
110 static std::string LOCAL_RESULTDIR("/tmp/spadder.local.resultdir."+USER);
111 static std::string REMOTE_WORKDIR("/tmp/spadder.remote.workdir."+USER);
114 * This function creates the padder text input file containing the
115 * input data (list of filenames and groupnames) and returns the path
116 * of the created file. This function is the one that knows the format
117 * of the padder input file. If the input file format changes, then
118 * this function (and only this one) should be updated.
120 const char * MeshJobManager_i::_writeDataFile(std::vector<MESHJOB::MeshJobParameter> listConcreteMesh,
121 std::vector<MESHJOB::MeshJobParameter> listSteelBarMesh) {
123 mkdir(LOCAL_INPUTDIR.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
125 // Make it static so that it's allocated once (constant name)
126 static std::string * dataFilename = new std::string(LOCAL_INPUTDIR+"/"+DATAFILE);
127 std::ofstream dataFile(dataFilename->c_str());
129 // We first specify the concrete mesh data (filename and groupname)
131 line = std::string(basename(listConcreteMesh[0].file_name)) + " " + std::string(listConcreteMesh[0].group_name);
132 dataFile << line.c_str() << std::endl;
133 // Note that we use here the basename because the files are supposed
134 // to be copied in the REMOTE_WORKDIR for execution.
136 // The, we can specify the steelbar mesh data, starting by the
138 int nbSteelBarMesh=listSteelBarMesh.size();
139 line = std::string("nbSteelbarMesh") + SEPARATOR + ToString(nbSteelBarMesh);
140 dataFile << line.c_str() << std::endl;
141 for (int i=0; i<nbSteelBarMesh; i++) {
142 line = std::string(basename(listSteelBarMesh[i].file_name)) + " " + std::string(listSteelBarMesh[i].group_name);
143 dataFile << line.c_str() << std::endl;
146 // Finally, we conclude with the name of the output file
148 dataFile << line.c_str() << std::endl;
150 return dataFilename->c_str();
154 * This function creates a shell script that runs padder whith the
155 * specified data file, and returns the path of the created script
156 * file. The config id is used to retrieve the path to the binary file
157 * and other required files.
159 const char* MeshJobManager_i::_writeScriptFile(const char * dataFileName, const char * configId) {
160 mkdir(LOCAL_INPUTDIR.c_str(), S_IRWXU | S_IRWXG | S_IROTH | S_IXOTH);
162 // Make it static so that it's allocated once (constant name)
163 static std::string * scriptFilename = new std::string(LOCAL_INPUTDIR+"/"+SCRIPTFILE);
165 char * binpath = _configMap[configId].binpath;
166 char * envpath = _configMap[configId].envpath;
168 std::ofstream script(scriptFilename->c_str());
169 script << "#!/bin/sh" << std::endl;
170 script << "here=$(dirname $0)" << std::endl;
171 script << ". " << envpath << std::endl;
172 script << binpath << " $here/" << basename(dataFileName) << std::endl;
173 // Note that we use the basename of the datafile because all data
174 // files are supposed to have been copied in the REMOTE_WORKDIR.
176 return scriptFilename->c_str();
180 // ====================================================================
181 // Functions to initialize and supervise the mesh computation job
182 // ====================================================================
184 bool MeshJobManager_i::configure(const char *configId,
185 const MESHJOB::ConfigParameter & configParameter)
187 beginService("MeshJobManager_i::configure");
189 _configMap[configId] = configParameter;
191 LOG("Adding configuration for " << configId);
192 LOG("- binpath = " << _configMap[configId].binpath);
193 LOG("- envpath = " << _configMap[configId].envpath);
195 endService("MeshJobManager_i::configure");
199 long MeshJobManager_i::JOBID_UNDEFINED = -1;
201 /*! Initialize a smesh computation job and return the job identifier */
202 CORBA::Long MeshJobManager_i::initialize(const MESHJOB::MeshJobParameterList & meshJobParameterList,
203 const char * configId)
205 beginService("MeshJobManager_i::initialize");
208 // We first analyse the CORBA sequence to store data in C++ vectors
210 std::vector<MESHJOB::MeshJobParameter> listConcreteMesh;
211 std::vector<MESHJOB::MeshJobParameter> listSteelBarMesh;
212 for(CORBA::ULong i=0; i<meshJobParameterList.length(); i++) {
213 MESHJOB::MeshJobParameter currentMesh = meshJobParameterList[i];
214 switch ( currentMesh.file_type ) {
215 case MESHJOB::MED_CONCRETE:
216 listConcreteMesh.push_back(currentMesh);
218 case MESHJOB::MED_STEELBAR:
219 listSteelBarMesh.push_back(currentMesh);
222 LOG("The type of the file is not recognized");
223 return JOBID_UNDEFINED;
227 if ( listConcreteMesh.size() != 1 ) {
228 // Not consistent with the specification
229 LOG("You specify more than one concrete mesh");
230 return JOBID_UNDEFINED;
233 LOG("Nb. concrete mesh = " << listConcreteMesh.size());
234 LOG("Nb. steelbar mesh = " << listSteelBarMesh.size());
236 // We initiate here a datetime to tag the files and folder
237 // associated to this job.
238 long jobDatetimeTag = timetag();
239 // And a MESHJOB::MeshJobPaths structure to hold the directories
240 // where to find data
241 MESHJOB::MeshJobPaths * jobPaths = new MESHJOB::MeshJobPaths();
242 jobPaths->local_inputdir = LOCAL_INPUTDIR.c_str();
243 jobPaths->local_resultdir = (LOCAL_RESULTDIR + "." + ToString(jobDatetimeTag)).c_str();
244 jobPaths->remote_workdir = (REMOTE_WORKDIR + "." + ToString(jobDatetimeTag)).c_str();
247 // Then, we have to create the padder input data file. This input
248 // data is a text file containing the list of file names and group
251 const char * dataFilename = this->_writeDataFile(listConcreteMesh, listSteelBarMesh);
252 LOG("dataFilename = " << dataFilename);
253 const char * scriptFilename = this->_writeScriptFile(dataFilename, configId);
254 LOG("scriptFilename = " << scriptFilename);
257 // Then, the following instructions consists in preparing the job
258 // parameters to request the SALOME launcher for creating a new
261 Engines::JobParameters_var jobParameters = new Engines::JobParameters;
262 jobParameters->job_type = CORBA::string_dup("command");
263 // CAUTION: the job_file must be a single filename specifying a
264 // self-consistent script to be executed without any argument on the
266 jobParameters->job_file = CORBA::string_dup(scriptFilename);
269 // Specification of the working spaces:
271 // - local_directory: can be used to specify where to find the input
272 // files on the local resource. It's optionnal if you specify the
273 // absolute path name of input files.
275 // - result_directory: must be used to specify where to download the
276 // output files on the local resources
278 // - work_directory: must be used to specify the remote directory
279 // where to put all the stuff to run the job. Note that the job
280 // will be executed from within this directory, i.e. a change
281 // directory toward this working directory is done by the batch
282 // system before running the specified job script.
284 jobParameters->local_directory = CORBA::string_dup("");
285 jobParameters->result_directory = CORBA::string_dup(jobPaths->local_resultdir);
286 jobParameters->work_directory = CORBA::string_dup(jobPaths->remote_workdir);
288 // We specify the input files that are required to execute the
289 // job_file. If basenames are specified, then the files are supposed
290 // to be located in local_directory.
291 int nbFiles = listSteelBarMesh.size()+2;
292 // The number of input file is:
293 // (nb. of steelbar meshfile)
294 // + (1 concrete meshfile)
295 // + (1 padder input file)
296 // = nb steelbar meshfile + 2
297 jobParameters->in_files.length(nbFiles);
298 jobParameters->in_files[0] = CORBA::string_dup(listConcreteMesh[0].file_name);
299 for (int i=0; i<listSteelBarMesh.size(); i++) {
300 jobParameters->in_files[1+i] = CORBA::string_dup(listSteelBarMesh[i].file_name);
302 jobParameters->in_files[1+listSteelBarMesh.size()] = CORBA::string_dup(dataFilename);
303 // Note that all these input files will be copied in the
304 // REMOTE_WORKDIR on the remote host
306 // Then, we have to specify the existance of an output
307 // filenames. The path is supposed to be a path on the remote
308 // resource, i.e. where the job is executed.
309 jobParameters->out_files.length(1);
310 std::string outputfile_name = std::string(jobPaths->remote_workdir)+"/"+OUTPUTFILE;
311 jobParameters->out_files[0] = CORBA::string_dup(outputfile_name.c_str());
313 // CAUTION: the maximum duration has to be set with a format like "hh:mm"
314 jobParameters->maximum_duration = CORBA::string_dup("01:00");
315 jobParameters->queue = CORBA::string_dup("");
317 // Setting resource and additionnal properties (if needed)
318 // The resource parameters can be initiated from scratch, for
319 // example by specifying the values in hard coding:
321 //jobParameters->resource_required.name = CORBA::string_dup("localhost");
322 //jobParameters->resource_required.hostname = CORBA::string_dup("localhost");
323 //jobParameters->resource_required.mem_mb = 1024 * 10;
324 //jobParameters->resource_required.nb_proc = 1;
326 // But it's better to initiate these parameters from a resource
327 // definition known by the resource manager. This ensures that the
328 // resource will be available:
329 //const char * resourceName = "localhost";
330 //const char * resourceName = "boulant@claui2p1";
331 //const char * resourceName = "nepal@nepal";
332 const char * resourceName = _configMap[configId].resname;
333 Engines::ResourceDefinition * resourceDefinition = _resourcesManager->GetResourceDefinition(resourceName);
334 // CAUTION: This resource should have been defined in the
335 // CatalogResource.xml associated to the SALOME application.
337 // Then, the values can be used to initiate the resource parameters
339 jobParameters->resource_required.name = CORBA::string_dup(resourceDefinition->name.in());
340 // CAUTION: the additionnal two following parameters MUST be
341 // specified explicitly, because they are not provided by the
342 // resource definition:
343 jobParameters->resource_required.mem_mb = resourceDefinition->mem_mb;
344 jobParameters->resource_required.nb_proc = resourceDefinition->nb_proc_per_node;
345 // CAUTION: the parameter mem_mb specifies the maximum memory value
346 // that could be allocated for executing the job. This takes into
347 // account not only the data that could be loaded by the batch
348 // process but also the linked dynamic library.
350 // A possible problem, for exemple in the case where you use the ssh
351 // emulation of a batch system, is to get an error message as below
352 // when libBatch try to run the ssh command:
354 // ##Â /usr/bin/ssh: error while loading shared libraries: libcrypto.so.0.9.8: failed
355 // ## to map segment from shared object: Cannot allocate memory
357 // In this exemple, the mem_mb was set to 1MB, value that is not
358 // sufficient to load the dynamic libraries linked to the ssh
359 // executable (libcrypto.so in the error message).
361 // So, even in the case of a simple test shell script, you should
362 // set this value at least to a standard threshold as 500MB
364 int jobId = JOBID_UNDEFINED;
366 jobId = _salomeLauncher->createJob(jobParameters);
367 // We register the datetime tag of this job
368 _jobDateTimeMap[jobId]=jobDatetimeTag;
369 _jobPathsMap[jobId] = jobPaths;
371 catch (const SALOME::SALOME_Exception & ex) {
372 LOG("SALOME Exception in createJob !" <<ex.details.text.in());
373 //LOG(ex.details.text.in());
374 return JOBID_UNDEFINED;
376 catch (const CORBA::SystemException& ex) {
377 LOG("Receive SALOME System Exception: "<<ex);
378 LOG("Check SALOME servers...");
379 return JOBID_UNDEFINED;
382 endService("MeshJobManager_i::initialize");
386 /*! Submit the job execution and return true if submission is OK */
387 bool MeshJobManager_i::start(CORBA::Long jobId) {
388 beginService("MeshJobManager_i::start");
391 _salomeLauncher->launchJob(jobId);
393 catch (const SALOME::SALOME_Exception & ex) {
394 LOG("SALOME Exception in createJob !" <<ex.details.text.in());
395 //LOG(ex.details.text.in());
398 catch (const CORBA::SystemException& ex) {
399 LOG("Receive SALOME System Exception: "<<ex);
400 LOG("Check SALOME servers...");
404 endService("MeshJobManager_i::initialize");
408 /*! Request the launch manager for the state of the specified job */
409 char* MeshJobManager_i::getState(CORBA::Long jobId) {
410 beginService("MeshJobManager_i::getState");
415 state = _salomeLauncher->getJobState(jobId);
417 catch (const SALOME::SALOME_Exception & ex)
419 LOG("SALOME Exception in getJobState !");
420 state = ex.details.text;
422 catch (const CORBA::SystemException& ex)
424 LOG("Receive SALOME System Exception: " << ex);
425 state="SALOME System Exception - see logs";
427 LOG("jobId="<<ToString(jobId)<<" state="<<state);
428 endService("MeshJobManager_i::getState");
429 return CORBA::string_dup(state.c_str());
432 MESHJOB::MeshJobPaths * MeshJobManager_i::getPaths(CORBA::Long jobId) {
434 MESHJOB::MeshJobPaths * jobPaths = _jobPathsMap[jobId];
435 if ( jobPaths == NULL ) {
436 LOG("You request the working paths for an undefined job (jobId="<<ToString(jobId)<<")");
437 return NULL; // Maybe raise an exception?
443 MESHJOB::MeshJobResults * MeshJobManager_i::finalize(CORBA::Long jobId) {
444 beginService("MeshJobManager_i::getResults");
445 MESHJOB::MeshJobResults * result = new MESHJOB::MeshJobResults();
447 MESHJOB::MeshJobPaths * jobPaths = this->getPaths(jobId);
448 std::string local_resultdir(jobPaths->local_resultdir);
449 result->results_dirname = local_resultdir.c_str();
452 _salomeLauncher->getJobResults(jobId, local_resultdir.c_str());
454 // __BUG__: to prevent from a bug of the MED driver (SALOME
455 // 5.1.5), we change the basename of the output file to force the
456 // complete reloading of data by the med driver.
457 long jobDatetimeTag = _jobDateTimeMap[jobId];
458 std::string outputFileName = "output"+ToString(jobDatetimeTag)+".med";
459 rename((local_resultdir+"/"+OUTPUTFILE).c_str(), (local_resultdir+"/"+outputFileName).c_str());
461 result->outputmesh_filename = outputFileName.c_str();
462 result->status = "OK";
464 catch (const SALOME::SALOME_Exception & ex)
466 LOG("SALOME Exception in getResults !");
467 result->status = "SALOME Exception in getResults !";
469 catch (const CORBA::SystemException& ex)
471 LOG("Receive CORBA System Exception: " << ex);
472 result->status = "Receive CORBA System Exception: see log";
474 endService("MeshJobManager_i::getResults");
479 /*! Clean all data associated to this job and remove the job from the launch manager */
480 bool MeshJobManager_i::clean(CORBA::Long jobId) {
481 beginService("MeshJobManager_i::clean");
483 // __GBO__ WORK IN PROGRESS: we just clean the temporary local
484 // directories. The remote working directories are tag with the
485 // execution datetime and the we prevent the task from conflict
486 // with files of another task.
487 MESHJOB::MeshJobPaths * jobPaths = this->getPaths(jobId);
488 if ( jobPaths == NULL ) return false;
491 // For safety reason (and prevent from bug that could erase the
492 // filesystem), we cancel the operation in the case where the
493 // directories to delete are not in the /tmp folder.
494 std::string shell_command("rm -rf ");
495 std::string inputdir(jobPaths->local_inputdir);
496 std::string resultdir(jobPaths->local_resultdir);
497 if ( !myStartsWith(inputdir,"/tmp/") ) {
498 LOG("WRN: The directory "<<inputdir<<" is not in /tmp. NO DELETE is done");
500 shell_command+=inputdir+" ";
502 if ( !myStartsWith(resultdir,"/tmp/")) {
503 LOG("WRN: The directory "<<resultdir<<" is not in /tmp. NO DELETE is done");
505 shell_command+=resultdir;
508 LOG("DBG: clean shell command = "<<shell_command);
510 bool cleanOk = false;
511 int error = system(shell_command.c_str());
512 if (error == 0) cleanOk = true;
514 endService("MeshJobManager_i::clean");
519 std::vector<std::string> * MeshJobManager_i::_getResourceNames() {
522 // These part is just to control the available resources
524 Engines::ResourceParameters params;
525 KERNEL::getLifeCycleCORBA()->preSet(params);
527 Engines::ResourceList * resourceList = _resourcesManager->GetFittingResources(params);
528 Engines::ResourceDefinition * resourceDefinition = NULL;
529 LOG("### resource list:");
530 std::vector<std::string>* resourceNames = new std::vector<std::string>();
532 for (int i = 0; i < resourceList->length(); i++) {
533 const char* aResourceName = (*resourceList)[i];
534 resourceNames->push_back(std::string(aResourceName));
535 LOG("resource["<<i<<"] = "<<aResourceName);
536 resourceDefinition = _resourcesManager->GetResourceDefinition(aResourceName);
537 LOG("protocol["<<i<<"] = "<<resourceDefinition->protocol);
541 // Note: a ResourceDefinition is used to create a batch configuration
542 // in the Launcher. This operation is done at Launcher startup from
543 // the configuration file CatalogResources.xml provided by the
544 // SALOME application.
545 // In the code instructions, you just have to choose a resource
546 // configuration by its name and then define the ResourceParameters
547 // that specify additionnal properties for a specific job submission
548 // (use the attribute resource_required of the JobParameters).
550 return resourceNames;
555 // ==========================================================================
557 // ==========================================================================
561 PortableServer::ObjectId * MeshJobManagerEngine_factory( CORBA::ORB_ptr orb,
562 PortableServer::POA_ptr poa,
563 PortableServer::ObjectId * contId,
564 const char *instanceName,
565 const char *interfaceName)
567 MESSAGE("PortableServer::ObjectId * MeshJobManagerEngine_factory()");
568 MeshJobManager_i * myEngine = new MeshJobManager_i(orb, poa, contId, instanceName, interfaceName);
569 return myEngine->getId() ;