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