test
CMS 3D CMS Logo

 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Properties Friends Macros Pages
MatrixInjector.py
Go to the documentation of this file.
1 import sys
2 import json
3 import os
4 import copy
5 import multiprocessing
6 
8  if opt.show:
9  print 'Not injecting to wmagent in --show mode. Need to run the worklfows.'
10  sys.exit(-1)
11  if opt.wmcontrol=='init':
12  #init means it'll be in test mode
13  opt.nProcs=0
14  if opt.wmcontrol=='test':
15  #means the wf were created already, and we just dryRun it.
16  opt.dryRun=True
17  if opt.wmcontrol=='submit' and opt.nProcs==0:
18  print 'Not injecting to wmagent in -j 0 mode. Need to run the worklfows.'
19  sys.exit(-1)
20  if opt.wmcontrol=='force':
21  print "This is an expert setting, you'd better know what you're doing"
22  opt.dryRun=True
23 
24 def upload_to_couch_oneArg(arguments):
25  from modules.wma import upload_to_couch
26  (filePath,labelInCouch,user,group,where) = arguments
27  cacheId=upload_to_couch(filePath,
28  labelInCouch,
29  user,
30  group,
31  test_mode=False,
32  url=where)
33  return cacheId
34 
35 
36 class MatrixInjector(object):
37 
38  def __init__(self,opt,mode='init',options=''):
39  self.count=1040
40 
41  self.dqmgui=None
42  self.wmagent=None
43  for k in options.split(','):
44  if k.startswith('dqm:'):
45  self.dqmgui=k.split(':',1)[-1]
46  elif k.startswith('wma:'):
47  self.wmagent=k.split(':',1)[-1]
48 
49  self.testMode=((mode!='submit') and (mode!='force'))
50  self.version =1
51  self.keep = opt.keep
52  self.memoryOffset = opt.memoryOffset
53  self.memPerCore = opt.memPerCore
54 
55  #wagemt stuff
56  if not self.wmagent:
57  self.wmagent=os.getenv('WMAGENT_REQMGR')
58  if not self.wmagent:
59  if not opt.testbed :
60  self.wmagent = 'cmsweb.cern.ch'
61  self.DbsUrl = "https://"+self.wmagent+"/dbs/prod/global/DBSReader"
62  else :
63  self.wmagent = 'cmsweb-testbed.cern.ch'
64  self.DbsUrl = "https://"+self.wmagent+"/dbs/int/global/DBSReader"
65 
66  if not self.dqmgui:
67  self.dqmgui="https://cmsweb.cern.ch/dqm/relval"
68  #couch stuff
69  self.couch = 'https://'+self.wmagent+'/couchdb'
70 # self.couchDB = 'reqmgr_config_cache'
71  self.couchCache={} # so that we do not upload like crazy, and recyle cfgs
72  self.user = os.getenv('USER')
73  self.group = 'ppd'
74  self.label = 'RelValSet_'+os.getenv('CMSSW_VERSION').replace('-','')+'_v'+str(self.version)
75  self.speciallabel=''
76  if opt.label:
77  self.speciallabel= '_'+opt.label
78 
79 
80  if not os.getenv('WMCORE_ROOT'):
81  print '\n\twmclient is not setup properly. Will not be able to upload or submit requests.\n'
82  if not self.testMode:
83  print '\n\t QUIT\n'
84  sys.exit(-18)
85  else:
86  print '\n\tFound wmclient\n'
87 
88  self.defaultChain={
89  "RequestType" : "TaskChain", #this is how we handle relvals
90  "SubRequestType" : "RelVal", #this is how we handle relvals, now that TaskChain is also used for central MC production
91  "RequestPriority": 500000,
92  "Requestor": self.user, #Person responsible
93  "Group": self.group, #group for the request
94  "CMSSWVersion": os.getenv('CMSSW_VERSION'), #CMSSW Version (used for all tasks in chain)
95  "Campaign": os.getenv('CMSSW_VERSION'), # only for wmstat purpose
96  "ScramArch": os.getenv('SCRAM_ARCH'), #Scram Arch (used for all tasks in chain)
97  "ProcessingVersion": self.version, #Processing Version (used for all tasks in chain)
98  "GlobalTag": None, #Global Tag (overridden per task)
99  "CouchURL": self.couch, #URL of CouchDB containing Config Cache
100  "ConfigCacheURL": self.couch, #URL of CouchDB containing Config Cache
101  "DbsUrl": self.DbsUrl,
102  #- Will contain all configs for all Tasks
103  #"SiteWhitelist" : ["T2_CH_CERN", "T1_US_FNAL"], #Site whitelist
104  "TaskChain" : None, #Define number of tasks in chain.
105  "nowmTasklist" : [], #a list of tasks as we put them in
106  "unmergedLFNBase" : "/store/unmerged",
107  "mergedLFNBase" : "/store/relval",
108  "dashboardActivity" : "relval",
109  "Multicore" : 1, # do not set multicore for the whole chain
110  "Memory" : 3000,
111  "SizePerEvent" : 1234,
112  "TimePerEvent" : 0.1
113  }
114 
116  "EnableHarvesting" : "True",
117  "DQMUploadUrl" : self.dqmgui,
118  "DQMConfigCacheID" : None,
119  "Multicore" : 1 # hardcode Multicore to be 1 for Harvest
120  }
121 
123  "TaskName" : None, #Task Name
124  "ConfigCacheID" : None, #Generator Config id
125  "GlobalTag": None,
126  "SplittingAlgo" : "EventBased", #Splitting Algorithm
127  "EventsPerJob" : None, #Size of jobs in terms of splitting algorithm
128  "RequestNumEvents" : None, #Total number of events to generate
129  "Seeding" : "AutomaticSeeding", #Random seeding method
130  "PrimaryDataset" : None, #Primary Dataset to be created
131  "nowmIO": {},
132  "Multicore" : opt.nThreads, # this is the per-taskchain Multicore; it's the default assigned to a task if it has no value specified
133  "KeepOutput" : False
134  }
136  "TaskName" : "DigiHLT", #Task Name
137  "ConfigCacheID" : None, #Processing Config id
138  "GlobalTag": None,
139  "InputDataset" : None, #Input Dataset to be processed
140  "SplittingAlgo" : "LumiBased", #Splitting Algorithm
141  "LumisPerJob" : 10, #Size of jobs in terms of splitting algorithm
142  "nowmIO": {},
143  "Multicore" : opt.nThreads, # this is the per-taskchain Multicore; it's the default assigned to a task if it has no value specified
144  "KeepOutput" : False
145  }
146  self.defaultTask={
147  "TaskName" : None, #Task Name
148  "InputTask" : None, #Input Task Name (Task Name field of a previous Task entry)
149  "InputFromOutputModule" : None, #OutputModule name in the input task that will provide files to process
150  "ConfigCacheID" : None, #Processing Config id
151  "GlobalTag": None,
152  "SplittingAlgo" : "LumiBased", #Splitting Algorithm
153  "LumisPerJob" : 10, #Size of jobs in terms of splitting algorithm
154  "nowmIO": {},
155  "Multicore" : opt.nThreads, # this is the per-taskchain Multicore; it's the default assigned to a task if it has no value specified
156  "KeepOutput" : False
157  }
158 
159  self.chainDicts={}
160 
161 
162  def prepare(self,mReader, directories, mode='init'):
163  try:
164  #from Configuration.PyReleaseValidation.relval_steps import wmsplit
165  wmsplit = {}
166  wmsplit['DIGIHI']=5
167  wmsplit['RECOHI']=5
168  wmsplit['HLTD']=5
169  wmsplit['RECODreHLT']=2
170  wmsplit['DIGIPU']=4
171  wmsplit['DIGIPU1']=4
172  wmsplit['RECOPU1']=1
173  wmsplit['DIGIUP15_PU50']=1
174  wmsplit['RECOUP15_PU50']=1
175  wmsplit['DIGIUP15_PU25']=1
176  wmsplit['RECOUP15_PU25']=1
177  wmsplit['DIGIUP15_PU25HS']=1
178  wmsplit['RECOUP15_PU25HS']=1
179  wmsplit['DIGIHIMIX']=5
180  wmsplit['RECOHIMIX']=5
181  wmsplit['RECODSplit']=1
182  wmsplit['SingleMuPt10_UP15_ID']=1
183  wmsplit['DIGIUP15_ID']=1
184  wmsplit['RECOUP15_ID']=1
185  wmsplit['TTbar_13_ID']=1
186  wmsplit['SingleMuPt10FS_ID']=1
187  wmsplit['TTbarFS_ID']=1
188  wmsplit['RECODR2_50nsreHLT']=1
189  wmsplit['RECODR2_25nsreHLT']=1
190  wmsplit['RECODR2_2016reHLT']=5
191  wmsplit['RECODR2_2016reHLT_skimSingleMu']=5
192  wmsplit['RECODR2_2016reHLT_skimDoubleEG']=5
193  wmsplit['RECODR2_2016reHLT_skimMuonEG']=5
194  wmsplit['RECODR2_2016reHLT_skimJetHT']=5
195  wmsplit['RECODR2_2016reHLT_skimMuOnia']=5
196  wmsplit['HLTDR2_50ns']=1
197  wmsplit['HLTDR2_25ns']=1
198  wmsplit['HLTDR2_2016']=1
199  wmsplit['Hadronizer']=1
200  wmsplit['DIGIUP15']=5
201  wmsplit['RECOUP15']=5
202  wmsplit['RECOAODUP15']=5
203  wmsplit['DBLMINIAODMCUP15NODQM']=5
204 
205 
206  #import pprint
207  #pprint.pprint(wmsplit)
208  except:
209  print "Not set up for step splitting"
210  wmsplit={}
211 
212  acqEra=False
213  for (n,dir) in directories.items():
214  chainDict=copy.deepcopy(self.defaultChain)
215  print "inspecting",dir
216  nextHasDSInput=None
217  for (x,s) in mReader.workFlowSteps.items():
218  #x has the format (num, prefix)
219  #s has the format (num, name, commands, stepList)
220  if x[0]==n:
221  #print "found",n,s[3]
222  #chainDict['RequestString']='RV'+chainDict['CMSSWVersion']+s[1].split('+')[0]
223  index=0
224  splitForThisWf=None
225  thisLabel=self.speciallabel
226  #if 'HARVESTGEN' in s[3]:
227  if len( [step for step in s[3] if "HARVESTGEN" in step] )>0:
228  chainDict['TimePerEvent']=0.01
229  thisLabel=thisLabel+"_gen"
230  # for double miniAOD test
231  if len( [step for step in s[3] if "DBLMINIAODMCUP15NODQM" in step] )>0:
232  thisLabel=thisLabel+"_dblMiniAOD"
233  processStrPrefix=''
234  setPrimaryDs=None
235  for step in s[3]:
236 
237  if 'INPUT' in step or (not isinstance(s[2][index],str)):
238  nextHasDSInput=s[2][index]
239 
240  else:
241 
242  if (index==0):
243  #first step and not input -> gen part
244  chainDict['nowmTasklist'].append(copy.deepcopy(self.defaultScratch))
245  try:
246  chainDict['nowmTasklist'][-1]['nowmIO']=json.loads(open('%s/%s.io'%(dir,step)).read())
247  except:
248  print "Failed to find",'%s/%s.io'%(dir,step),".The workflows were probably not run on cfg not created"
249  return -15
250 
251  chainDict['nowmTasklist'][-1]['PrimaryDataset']='RelVal'+s[1].split('+')[0]
252  if not '--relval' in s[2][index]:
253  print 'Impossible to create task from scratch without splitting information with --relval'
254  return -12
255  else:
256  arg=s[2][index].split()
257  ns=map(int,arg[arg.index('--relval')+1].split(','))
258  chainDict['nowmTasklist'][-1]['RequestNumEvents'] = ns[0]
259  chainDict['nowmTasklist'][-1]['EventsPerJob'] = ns[1]
260  if 'FASTSIM' in s[2][index] or '--fast' in s[2][index]:
261  thisLabel+='_FastSim'
262  if 'lhe' in s[2][index] in s[2][index]:
263  chainDict['nowmTasklist'][-1]['LheInputFiles'] =True
264 
265  elif nextHasDSInput:
266  chainDict['nowmTasklist'].append(copy.deepcopy(self.defaultInput))
267  try:
268  chainDict['nowmTasklist'][-1]['nowmIO']=json.loads(open('%s/%s.io'%(dir,step)).read())
269  except:
270  print "Failed to find",'%s/%s.io'%(dir,step),".The workflows were probably not run on cfg not created"
271  return -15
272  chainDict['nowmTasklist'][-1]['InputDataset']=nextHasDSInput.dataSet
273  splitForThisWf=nextHasDSInput.split
274  chainDict['nowmTasklist'][-1]['LumisPerJob']=splitForThisWf
275  if step in wmsplit:
276  chainDict['nowmTasklist'][-1]['LumisPerJob']=wmsplit[step]
277  # get the run numbers or #events
278  if len(nextHasDSInput.run):
279  chainDict['nowmTasklist'][-1]['RunWhitelist']=nextHasDSInput.run
280  if len(nextHasDSInput.ls):
281  chainDict['nowmTasklist'][-1]['LumiList']=nextHasDSInput.ls
282  #print "what is s",s[2][index]
283  if '--data' in s[2][index] and nextHasDSInput.label:
284  thisLabel+='_RelVal_%s'%nextHasDSInput.label
285  if 'filter' in chainDict['nowmTasklist'][-1]['nowmIO']:
286  print "This has an input DS and a filter sequence: very likely to be the PyQuen sample"
287  processStrPrefix='PU_'
288  setPrimaryDs = 'RelVal'+s[1].split('+')[0]
289  if setPrimaryDs:
290  chainDict['nowmTasklist'][-1]['PrimaryDataset']=setPrimaryDs
291  nextHasDSInput=None
292  else:
293  #not first step and no inputDS
294  chainDict['nowmTasklist'].append(copy.deepcopy(self.defaultTask))
295  try:
296  chainDict['nowmTasklist'][-1]['nowmIO']=json.loads(open('%s/%s.io'%(dir,step)).read())
297  except:
298  print "Failed to find",'%s/%s.io'%(dir,step),".The workflows were probably not run on cfg not created"
299  return -15
300  if splitForThisWf:
301  chainDict['nowmTasklist'][-1]['LumisPerJob']=splitForThisWf
302  if step in wmsplit:
303  chainDict['nowmTasklist'][-1]['LumisPerJob']=wmsplit[step]
304 
305  # change LumisPerJob for Hadronizer steps.
306  if 'Hadronizer' in step:
307  chainDict['nowmTasklist'][-1]['LumisPerJob']=wmsplit['Hadronizer']
308 
309  #print step
310  chainDict['nowmTasklist'][-1]['TaskName']=step
311  if setPrimaryDs:
312  chainDict['nowmTasklist'][-1]['PrimaryDataset']=setPrimaryDs
313  chainDict['nowmTasklist'][-1]['ConfigCacheID']='%s/%s.py'%(dir,step)
314  chainDict['nowmTasklist'][-1]['GlobalTag']=chainDict['nowmTasklist'][-1]['nowmIO']['GT'] # copy to the proper parameter name
315  chainDict['GlobalTag']=chainDict['nowmTasklist'][-1]['nowmIO']['GT'] #set in general to the last one of the chain
316  if 'pileup' in chainDict['nowmTasklist'][-1]['nowmIO']:
317  chainDict['nowmTasklist'][-1]['MCPileup']=chainDict['nowmTasklist'][-1]['nowmIO']['pileup']
318  if '--pileup ' in s[2][index]: # catch --pileup (scenarion) and not --pileup_ (dataset to be mixed) => works also making PRE-MIXed dataset
319  processStrPrefix='PU_' # take care of pu overlay done with GEN-SIM mixing
320  if ( s[2][index].split()[ s[2][index].split().index('--pileup')+1 ] ).find('25ns') > 0 :
321  processStrPrefix='PU25ns_'
322  elif ( s[2][index].split()[ s[2][index].split().index('--pileup')+1 ] ).find('50ns') > 0 :
323  processStrPrefix='PU50ns_'
324  if 'DIGIPREMIX_S2' in s[2][index] : # take care of pu overlay done with DIGI mixing of premixed events
325  if s[2][index].split()[ s[2][index].split().index('--pileup_input')+1 ].find('25ns') > 0 :
326  processStrPrefix='PUpmx25ns_'
327  elif s[2][index].split()[ s[2][index].split().index('--pileup_input')+1 ].find('50ns') > 0 :
328  processStrPrefix='PUpmx50ns_'
329 
330  if acqEra:
331  #chainDict['AcquisitionEra'][step]=(chainDict['CMSSWVersion']+'-PU_'+chainDict['nowmTasklist'][-1]['GlobalTag']).replace('::All','')+thisLabel
332  chainDict['AcquisitionEra'][step]=chainDict['CMSSWVersion']
333  chainDict['ProcessingString'][step]=processStrPrefix+chainDict['nowmTasklist'][-1]['GlobalTag'].replace('::All','')+thisLabel
334  else:
335  #chainDict['nowmTasklist'][-1]['AcquisitionEra']=(chainDict['CMSSWVersion']+'-PU_'+chainDict['nowmTasklist'][-1]['GlobalTag']).replace('::All','')+thisLabel
336  chainDict['nowmTasklist'][-1]['AcquisitionEra']=chainDict['CMSSWVersion']
337  chainDict['nowmTasklist'][-1]['ProcessingString']=processStrPrefix+chainDict['nowmTasklist'][-1]['GlobalTag'].replace('::All','')+thisLabel
338 
339  # specify different ProcessingString for double miniAOD dataset
340  if ('DBLMINIAODMCUP15NODQM' in step):
341  chainDict['nowmTasklist'][-1]['ProcessingString']=chainDict['nowmTasklist'][-1]['ProcessingString']+'_miniAOD'
342 
343  if( chainDict['nowmTasklist'][-1]['Multicore'] ):
344  # the scaling factor of 1.2GB / thread is empirical and measured on a SECOND round of tests with PU samples
345  # the number of threads is NO LONGER assumed to be the same for all tasks
346  # https://hypernews.cern.ch/HyperNews/CMS/get/edmFramework/3509/1/1/1.html
347  # now change to 1.5GB / additional thread according to discussion:
348  # https://hypernews.cern.ch/HyperNews/CMS/get/relval/4817/1/1.html
349 # chainDict['nowmTasklist'][-1]['Memory'] = 3000 + int( chainDict['nowmTasklist'][-1]['Multicore'] -1 )*1500
350  chainDict['nowmTasklist'][-1]['Memory'] = self.memoryOffset + int( chainDict['nowmTasklist'][-1]['Multicore'] -1 ) * self.memPerCore
351 
352  index+=1
353  #end of loop through steps
354  chainDict['RequestString']='RV'+chainDict['CMSSWVersion']+s[1].split('+')[0]
355  if processStrPrefix or thisLabel:
356  chainDict['RequestString']+='_'+processStrPrefix+thisLabel
357 
358 
359 
360  #wrap up for this one
361  import pprint
362  #print 'wrapping up'
363  #pprint.pprint(chainDict)
364  #loop on the task list
365  for i_second in reversed(range(len(chainDict['nowmTasklist']))):
366  t_second=chainDict['nowmTasklist'][i_second]
367  #print "t_second taskname", t_second['TaskName']
368  if 'primary' in t_second['nowmIO']:
369  #print t_second['nowmIO']['primary']
370  primary=t_second['nowmIO']['primary'][0].replace('file:','')
371  for i_input in reversed(range(0,i_second)):
372  t_input=chainDict['nowmTasklist'][i_input]
373  for (om,o) in t_input['nowmIO'].items():
374  if primary in o:
375  #print "found",primary,"procuced by",om,"of",t_input['TaskName']
376  t_second['InputTask'] = t_input['TaskName']
377  t_second['InputFromOutputModule'] = om
378  #print 't_second',pprint.pformat(t_second)
379  if t_second['TaskName'].startswith('HARVEST'):
380  chainDict.update(copy.deepcopy(self.defaultHarvest))
381  chainDict['DQMConfigCacheID']=t_second['ConfigCacheID']
382  ## the info are not in the task specific dict but in the general dict
383  #t_input.update(copy.deepcopy(self.defaultHarvest))
384  #t_input['DQMConfigCacheID']=t_second['ConfigCacheID']
385  break
386 
387  ## there is in fact only one acquisition era
388  #if len(set(chainDict['AcquisitionEra'].values()))==1:
389  # print "setting only one acq"
390  if acqEra:
391  chainDict['AcquisitionEra'] = chainDict['AcquisitionEra'].values()[0]
392 
393  ## clean things up now
394  itask=0
395  if self.keep:
396  for i in self.keep:
397  if type(i)==int and i < len(chainDict['nowmTasklist']):
398  chainDict['nowmTasklist'][i]['KeepOutput']=True
399  for (i,t) in enumerate(chainDict['nowmTasklist']):
400  if t['TaskName'].startswith('HARVEST'):
401  continue
402  if not self.keep:
403  t['KeepOutput']=True
404  elif t['TaskName'] in self.keep:
405  t['KeepOutput']=True
406  t.pop('nowmIO')
407  itask+=1
408  chainDict['Task%d'%(itask)]=t
409 
410 
411  ##
412 
413 
414  ## provide the number of tasks
415  chainDict['TaskChain']=itask#len(chainDict['nowmTasklist'])
416 
417  chainDict.pop('nowmTasklist')
418  self.chainDicts[n]=chainDict
419 
420 
421  return 0
422 
423  def uploadConf(self,filePath,label,where):
424  labelInCouch=self.label+'_'+label
425  cacheName=filePath.split('/')[-1]
426  if self.testMode:
427  self.count+=1
428  print '\tFake upload of',filePath,'to couch with label',labelInCouch
429  return self.count
430  else:
431  try:
432  from modules.wma import upload_to_couch,DATABASE_NAME
433  except:
434  print '\n\tUnable to find wmcontrol modules. Please include it in your python path\n'
435  print '\n\t QUIT\n'
436  sys.exit(-16)
437 
438  if cacheName in self.couchCache:
439  print "Not re-uploading",filePath,"to",where,"for",label
440  cacheId=self.couchCache[cacheName]
441  else:
442  print "Loading",filePath,"to",where,"for",label
443  ## totally fork the upload to couch to prevent cross loading of process configurations
444  pool = multiprocessing.Pool(1)
445  cacheIds = pool.map( upload_to_couch_oneArg, [(filePath,labelInCouch,self.user,self.group,where)] )
446  cacheId = cacheIds[0]
447  self.couchCache[cacheName]=cacheId
448  return cacheId
449 
450  def upload(self):
451  for (n,d) in self.chainDicts.items():
452  for it in d:
453  if it.startswith("Task") and it!='TaskChain':
454  #upload
455  couchID=self.uploadConf(d[it]['ConfigCacheID'],
456  str(n)+d[it]['TaskName'],
457  d['CouchURL']
458  )
459  print d[it]['ConfigCacheID']," uploaded to couchDB for",str(n),"with ID",couchID
460  d[it]['ConfigCacheID']=couchID
461  if it =='DQMConfigCacheID':
462  couchID=self.uploadConf(d['DQMConfigCacheID'],
463  str(n)+'harvesting',
464  d['CouchURL']
465  )
466  print d['DQMConfigCacheID'],"uploaded to couchDB for",str(n),"with ID",couchID
467  d['DQMConfigCacheID']=couchID
468 
469 
470  def submit(self):
471  try:
472  from modules.wma import makeRequest,approveRequest
473  from wmcontrol import random_sleep
474  print '\n\tFound wmcontrol\n'
475  except:
476  print '\n\tUnable to find wmcontrol modules. Please include it in your python path\n'
477  if not self.testMode:
478  print '\n\t QUIT\n'
479  sys.exit(-17)
480 
481  import pprint
482  for (n,d) in self.chainDicts.items():
483  if self.testMode:
484  print "Only viewing request",n
485  print pprint.pprint(d)
486  else:
487  #submit to wmagent each dict
488  print "For eyes before submitting",n
489  print pprint.pprint(d)
490  print "Submitting",n,"..........."
491  workFlow=makeRequest(self.wmagent,d,encodeDict=True)
492  approveRequest(self.wmagent,workFlow)
493  print "...........",n,"submitted"
494  random_sleep()
495 
496 
497 
boost::dynamic_bitset append(const boost::dynamic_bitset<> &bs1, const boost::dynamic_bitset<> &bs2)
this method takes two bitsets bs1 and bs2 and returns result of bs2 appended to the end of bs1 ...
void find(edm::Handle< EcalRecHitCollection > &hits, DetId thisDet, std::vector< EcalRecHitCollection::const_iterator > &hit, bool debug=false)
Definition: FindCaloHit.cc:7
def performInjectionOptionTest
if(dp >Float(M_PI)) dp-
def upload_to_couch_oneArg
double split
Definition: MVATrainer.cc:139