1
|
<?xml version="1.0" encoding="UTF-8"?>
|
2
|
<RESOURCE_PROFILE>
|
3
|
<HEADER>
|
4
|
<RESOURCE_IDENTIFIER value="251db67f-24ac-4d1d-9e7b-d38fe8898234_V29ya2Zsb3dEU1Jlc291cmNlcy9Xb3JrZmxvd0RTUmVzb3VyY2VUeXBl"/>
|
5
|
<RESOURCE_TYPE value="WorkflowDSResourceType"/>
|
6
|
<RESOURCE_KIND value="WorkflowDSResources"/>
|
7
|
<RESOURCE_URI value=""/>
|
8
|
<DATE_OF_CREATION value="2006-05-04T18:13:51.0Z"/>
|
9
|
</HEADER>
|
10
|
<BODY>
|
11
|
<WORKFLOW_NAME>Offline Deduplication v2</WORKFLOW_NAME>
|
12
|
<WORKFLOW_TYPE>Deduplication</WORKFLOW_TYPE>
|
13
|
<WORKFLOW_PRIORITY>30</WORKFLOW_PRIORITY>
|
14
|
<CONFIGURATION start="manual">
|
15
|
<NODE name="fetchRelClasses" type="FetchRelClasses" isStart="true">
|
16
|
<DESCRIPTION />
|
17
|
<PARAMETERS>
|
18
|
<PARAM name="relClassesProperty" type="string" managedBy="system" required="true">dnet.openaire.model.relclasses.xquery</PARAM>
|
19
|
<PARAM name="relClassesName" type="string" managedBy="system" required="true">relClasses</PARAM>
|
20
|
</PARAMETERS>
|
21
|
<ARCS>
|
22
|
<ARC to="checkConf" />
|
23
|
</ARCS>
|
24
|
</NODE>
|
25
|
<NODE name="setDedupConfigs" type="SetDedupConfiguration" isStart="true">
|
26
|
<DESCRIPTION>Set Dedup conf</DESCRIPTION>
|
27
|
<PARAMETERS>
|
28
|
<PARAM function="obtainValues('dedupOrchestrations', {})" required="true" type="string" name="dedupConfigSequence" managedBy="user"></PARAM>
|
29
|
<PARAM required="true" type="string" name="dedupConfigSequenceParam" managedBy="system">dedup.conf.queue</PARAM>
|
30
|
</PARAMETERS>
|
31
|
<ARCS>
|
32
|
<ARC to="checkConf" />
|
33
|
</ARCS>
|
34
|
</NODE>
|
35
|
<NODE name="hadoopConfig" type="SetClusterAndTable" isStart="true">
|
36
|
<DESCRIPTION>Set table name</DESCRIPTION>
|
37
|
<PARAMETERS>
|
38
|
<PARAM required="true" type="string" name="cluster" managedBy="system">DM</PARAM>
|
39
|
<PARAM required="true" type="string" name="tableParam" managedBy="system">hbaseTable</PARAM>
|
40
|
<PARAM managedBy="user" name="table" required="true" type="string"></PARAM>
|
41
|
</PARAMETERS>
|
42
|
<ARCS>
|
43
|
<ARC to="checkConf" />
|
44
|
</ARCS>
|
45
|
</NODE>
|
46
|
|
47
|
<NODE name="setWorkingPath" type="SetEnvParameter" isStart="true">
|
48
|
<DESCRIPTION>Set the working dir on HDFS for the MinDist Algo</DESCRIPTION>
|
49
|
<PARAMETERS>
|
50
|
<PARAM managedBy="system" name="parameterName" required="true" type="string">workDir</PARAM>
|
51
|
<PARAM managedBy="user" name="parameterValue" required="true" type="string">/tmp/dedup/mindist</PARAM>
|
52
|
</PARAMETERS>
|
53
|
<ARCS>
|
54
|
<ARC to="checkConf"/>
|
55
|
</ARCS>
|
56
|
</NODE>
|
57
|
|
58
|
|
59
|
<NODE name="checkConf" type="DedupCheckConfiguration" isJoin="true">
|
60
|
<DESCRIPTION/>
|
61
|
<PARAMETERS>
|
62
|
<PARAM required="true" type="string" name="dedupConfigSequenceParam" managedBy="system">dedup.conf.queue</PARAM>
|
63
|
</PARAMETERS>
|
64
|
<ARCS>
|
65
|
<ARC to="resetPath" />
|
66
|
</ARCS>
|
67
|
</NODE>
|
68
|
|
69
|
<NODE name="resetPath" type="CreateHdfsDirJob">
|
70
|
<DESCRIPTION>input files cleanup</DESCRIPTION>
|
71
|
<PARAMETERS>
|
72
|
<PARAM required="true" type="boolean" name="force" managedBy="system">true</PARAM>
|
73
|
<PARAM required="true" type="string" name="envParams" managedBy="system">
|
74
|
{
|
75
|
'path' : 'workDir',
|
76
|
'cluster' : 'cluster'
|
77
|
}
|
78
|
</PARAM>
|
79
|
</PARAMETERS>
|
80
|
<ARCS>
|
81
|
<ARC to="deduplicateScan"/>
|
82
|
</ARCS>
|
83
|
</NODE>
|
84
|
|
85
|
<NODE name="deduplicateScan" type="DuplicateScanJob">
|
86
|
<DESCRIPTION>Dup Scan</DESCRIPTION>
|
87
|
<PARAMETERS>
|
88
|
<PARAM required="true" type="string" name="hadoopJob" managedBy="system">dedupCandidateScanJob</PARAM>
|
89
|
<PARAM required="true" type="string" name="dedupConfigSequenceParam" managedBy="system">dedup.conf.queue</PARAM>
|
90
|
<PARAM required="true" type="string" name="envParams" managedBy="system">
|
91
|
{
|
92
|
'entityTypeId' : 'entityTypeId',
|
93
|
'entityType' : 'entityType',
|
94
|
'cluster' : 'cluster',
|
95
|
'hbase.mapred.inputtable' : 'hbaseTable',
|
96
|
'hbase.mapred.outputtable' : 'hbaseTable',
|
97
|
'hbase.mapreduce.inputtable' : 'hbaseTable',
|
98
|
'hbase.mapreduce.outputtable' : 'hbaseTable'
|
99
|
}
|
100
|
</PARAM>
|
101
|
</PARAMETERS>
|
102
|
<ARCS>
|
103
|
<ARC to="deduplicateScan" />
|
104
|
<ARC name="done" to="prepareActionSet" />
|
105
|
<ARC name="done" to="queryUserSimilarities" />
|
106
|
</ARCS>
|
107
|
</NODE>
|
108
|
|
109
|
|
110
|
|
111
|
|
112
|
<NODE name="queryUserSimilarities" type="QueryUserActionDbJob">
|
113
|
<DESCRIPTION>query user similarity</DESCRIPTION>
|
114
|
<PARAMETERS>
|
115
|
<PARAM name="dedupConfigSequenceParam" type="string" required="true" managedBy="system">dedup.conf.queue</PARAM>
|
116
|
<PARAM name="dbProperty" type="string" managedBy="system" required="true">dnet.dedup.db.name</PARAM>
|
117
|
<PARAM name="sql" type="string" managedBy="system" required="true">/eu/dnetlib/msro/workflows/dedup/querySimilaritiesBySet.sql.st</PARAM>
|
118
|
<PARAM name="outputEprParam" type="string" managedBy="system" required="true">simEpr</PARAM>
|
119
|
</PARAMETERS>
|
120
|
<ARCS>
|
121
|
<ARC to="buildSimilarityMesh"/>
|
122
|
</ARCS>
|
123
|
</NODE>
|
124
|
<NODE name="buildSimilarityMesh" type="BuildSimilarityMeshJob">
|
125
|
<DESCRIPTION>build mesh</DESCRIPTION>
|
126
|
<PARAMETERS>
|
127
|
<PARAM name="inputEprParam" type="string" managedBy="system" required="true">simEpr</PARAM>
|
128
|
<PARAM name="outputEprParam" type="string" managedBy="system" required="true">simMeshEpr</PARAM>
|
129
|
</PARAMETERS>
|
130
|
<ARCS>
|
131
|
<ARC to="storeSimilarities"/>
|
132
|
</ARCS>
|
133
|
</NODE>
|
134
|
|
135
|
|
136
|
<NODE name="storeSimilarities" type="StoreHBase">
|
137
|
<DESCRIPTION>store similarity</DESCRIPTION>
|
138
|
<PARAMETERS>
|
139
|
<PARAM name="inputEprParam" type="string" managedBy="system" required="true">simMeshEpr</PARAM>
|
140
|
<PARAM name="cluster" type="string" managedBy="system" required="true">DM</PARAM>
|
141
|
<PARAM name="mapping" type="string" managedBy="user" required="true" function="obtainValues('dbmf2hbaseMappings', {})"></PARAM>
|
142
|
<PARAM name="simulation" type="boolean" managedBy="user" required="false">false</PARAM>
|
143
|
</PARAMETERS>
|
144
|
<ARCS>
|
145
|
<ARC to="queryUserDissimilarities"/>
|
146
|
</ARCS>
|
147
|
</NODE>
|
148
|
|
149
|
<NODE name="queryUserDissimilarities" type="QueryUserActionDbJob">
|
150
|
<DESCRIPTION>query user dissimilarity</DESCRIPTION>
|
151
|
<PARAMETERS>
|
152
|
<PARAM name="dedupConfigSequenceParam" type="string" required="true" managedBy="system">dedup.conf.queue</PARAM>
|
153
|
<PARAM name="dbProperty" type="string" managedBy="system" required="true">dnet.dedup.db.name</PARAM>
|
154
|
<PARAM name="sql" type="string" managedBy="system" required="true">/eu/dnetlib/msro/workflows/dedup/queryDissimilaritiesBySet.sql.st</PARAM>
|
155
|
<PARAM name="outputEprParam" type="string" managedBy="system" required="true">dissimEpr</PARAM>
|
156
|
</PARAMETERS>
|
157
|
<ARCS>
|
158
|
<ARC to="storeDissimilarities"/>
|
159
|
</ARCS>
|
160
|
</NODE>
|
161
|
|
162
|
<NODE name="storeDissimilarities" type="DeleteFromHBase">
|
163
|
<DESCRIPTION>store dissimilarity</DESCRIPTION>
|
164
|
<PARAMETERS>
|
165
|
<PARAM name="inputEprParam" type="string" managedBy="system" required="true">dissimEpr</PARAM>
|
166
|
<PARAM name="cluster" type="string" managedBy="system" required="true">DM</PARAM>
|
167
|
<PARAM name="mapping" type="string" managedBy="user" required="true" function="obtainValues('dbmf2hbaseMappings', {})"></PARAM>
|
168
|
<PARAM name="simulation" type="boolean" managedBy="user" required="false">false</PARAM>
|
169
|
</PARAMETERS>
|
170
|
<ARCS>
|
171
|
<ARC to="doneActions"/>
|
172
|
</ARCS>
|
173
|
</NODE>
|
174
|
|
175
|
|
176
|
<NODE name="prepareActionSet" type="PrepareConfiguredActionSet">
|
177
|
<DESCRIPTION>prepare action sets</DESCRIPTION>
|
178
|
<PARAMETERS>
|
179
|
<PARAM required="true" type="string" name="dedupConfigSequenceParam" managedBy="system">dedup.conf.queue</PARAM>
|
180
|
<PARAM required="true" type="string" name="jobProperty" managedBy="system">rawSetId</PARAM>
|
181
|
<PARAM required="true" type="string" name="actionSetPathParam" managedBy="system">actionSetPath</PARAM>
|
182
|
</PARAMETERS>
|
183
|
<ARCS>
|
184
|
<ARC to="similarity2actions" />
|
185
|
</ARCS>
|
186
|
</NODE>
|
187
|
|
188
|
<NODE name="similarity2actions" type="DedupSimilarityToActionsJobNode">
|
189
|
<DESCRIPTION>export the similarity rels as Actions</DESCRIPTION>
|
190
|
<PARAMETERS>
|
191
|
<PARAM required="true" type="string" name="hadoopJob" managedBy="system">dedupSimilarity2HdfsActionsJob</PARAM>
|
192
|
<PARAM required="true" type="string" name="dedupConfigSequenceParam" managedBy="system">dedup.conf.queue</PARAM>
|
193
|
<PARAM required="true" type="string" name="envParams" managedBy="system">
|
194
|
{
|
195
|
'dedup.conf' : 'dedup.conf',
|
196
|
'entityTypeId' : 'entityTypeId',
|
197
|
'entityType' : 'entityType',
|
198
|
'cluster' : 'cluster',
|
199
|
'rawSetId' : 'rawSetId',
|
200
|
'hbase.mapred.inputtable' : 'hbaseTable',
|
201
|
'hbase.mapreduce.inputtable' : 'hbaseTable',
|
202
|
'mapred.output.dir' : 'actionSetPath'
|
203
|
}
|
204
|
</PARAM>
|
205
|
</PARAMETERS>
|
206
|
<ARCS>
|
207
|
<ARC to="updateActionSets" />
|
208
|
</ARCS>
|
209
|
</NODE>
|
210
|
|
211
|
<NODE name="updateActionSets" type="UpdateActionSets">
|
212
|
<DESCRIPTION>update action sets</DESCRIPTION>
|
213
|
<PARAMETERS/>
|
214
|
<ARCS>
|
215
|
<ARC to="doneActions" />
|
216
|
</ARCS>
|
217
|
</NODE>
|
218
|
|
219
|
<NODE name="doneActions" isJoin="true">
|
220
|
<DESCRIPTION>done actions</DESCRIPTION>
|
221
|
<PARAMETERS/>
|
222
|
<ARCS>
|
223
|
<ARC to="mindist"/>
|
224
|
</ARCS>
|
225
|
</NODE>
|
226
|
|
227
|
|
228
|
<NODE name="mindist" type="MinDistSearchHadoopJob">
|
229
|
<DESCRIPTION>find the minimum verted in each adjacency lists</DESCRIPTION>
|
230
|
<PARAMETERS>
|
231
|
<PARAM required="true" type="boolean" name="debug" managedBy="user">false</PARAM>
|
232
|
<PARAM required="true" type="boolean" name="outPathParam" managedBy="system">outputPath</PARAM>
|
233
|
<PARAM required="true" type="string" name="dedupConfigSequenceParam" managedBy="system">dedup.conf.queue</PARAM>
|
234
|
<PARAM required="true" type="string" name="envParams" managedBy="system">
|
235
|
{
|
236
|
'cluster' : 'cluster',
|
237
|
'entityTypeId' : 'entityTypeId',
|
238
|
'hbase.mapred.inputtable' : 'hbaseTable',
|
239
|
'hbase.mapreduce.inputtable' : 'hbaseTable',
|
240
|
'workDir' : 'workDir'
|
241
|
}
|
242
|
</PARAM>
|
243
|
</PARAMETERS>
|
244
|
<ARCS>
|
245
|
<ARC name="depth_n" to="mindist"/>
|
246
|
<ARC to="components"/>
|
247
|
</ARCS>
|
248
|
</NODE>
|
249
|
|
250
|
<NODE name="components" type="DedupConfigurationAwareJob">
|
251
|
<DESCRIPTION>joins all the vertex ids to build the connected components in the graph</DESCRIPTION>
|
252
|
<PARAMETERS>
|
253
|
<PARAM required="true" type="string" name="hadoopJob" managedBy="system">connectedComponentsJob</PARAM>
|
254
|
<PARAM required="true" type="string" name="dedupConfigSequenceParam" managedBy="system">dedup.conf.queue</PARAM>
|
255
|
<PARAM required="true" type="string" name="envParams" managedBy="system">
|
256
|
{
|
257
|
'cluster' : 'cluster',
|
258
|
'entityType' : 'entityType',
|
259
|
'entityTypeId' : 'entityTypeId',
|
260
|
'mapred.input.dir' : 'outputPath',
|
261
|
'hbase.mapred.outputtable' : 'hbaseTable',
|
262
|
'hbase.mapreduce.outputtable' : 'hbaseTable'
|
263
|
}
|
264
|
</PARAM>
|
265
|
</PARAMETERS>
|
266
|
<ARCS>
|
267
|
<ARC to="markDeleted"/>
|
268
|
</ARCS>
|
269
|
</NODE>
|
270
|
|
271
|
<NODE name="markDeleted" type="DedupConfigurationAwareJob">
|
272
|
<DESCRIPTION>mark duplicates as deleted by inference</DESCRIPTION>
|
273
|
<PARAMETERS>
|
274
|
<PARAM required="true" type="string" name="hadoopJob" managedBy="system">dedupMarkDeletedEntityJob</PARAM>
|
275
|
<PARAM required="true" type="string" name="dedupConfigSequenceParam" managedBy="system">dedup.conf.queue</PARAM>
|
276
|
<PARAM required="true" type="string" name="envParams" managedBy="system">
|
277
|
{
|
278
|
'cluster' : 'cluster',
|
279
|
'entityType' : 'entityType',
|
280
|
'entityTypeId' : 'entityTypeId',
|
281
|
'hbase.mapred.inputtable' : 'hbaseTable',
|
282
|
'hbase.mapreduce.inputtable' : 'hbaseTable',
|
283
|
'hbase.mapred.outputtable' : 'hbaseTable',
|
284
|
'hbase.mapreduce.outputtable' : 'hbaseTable'
|
285
|
}
|
286
|
</PARAM>
|
287
|
</PARAMETERS>
|
288
|
<ARCS>
|
289
|
<ARC to="buildRoots"/>
|
290
|
</ARCS>
|
291
|
</NODE>
|
292
|
|
293
|
<NODE name="buildRoots" type="SubmitHadoopJob">
|
294
|
<DESCRIPTION>redirect rels</DESCRIPTION>
|
295
|
<PARAMETERS>
|
296
|
<PARAM required="true" type="string" name="hadoopJob" managedBy="system">dedupBuildRootsJob</PARAM>
|
297
|
<PARAM required="true" type="string" name="envParams" managedBy="system">
|
298
|
{
|
299
|
'dedup.conf' : 'dedup.conf',
|
300
|
'relClasses' : 'relClasses',
|
301
|
'entityTypeId' : 'entityTypeId',
|
302
|
'entityType' : 'entityType',
|
303
|
'cluster' : 'cluster',
|
304
|
'hbase.mapred.inputtable' : 'hbaseTable',
|
305
|
'hbase.mapreduce.inputtable' : 'hbaseTable',
|
306
|
'hbase.mapred.outputtable' : 'hbaseTable',
|
307
|
'hbase.mapreduce.outputtable' : 'hbaseTable'
|
308
|
}
|
309
|
</PARAM>
|
310
|
</PARAMETERS>
|
311
|
<ARCS>
|
312
|
<ARC to="findIndex" />
|
313
|
</ARCS>
|
314
|
</NODE>
|
315
|
|
316
|
<NODE name="findIndex" type="FindIndex">
|
317
|
<DESCRIPTION />
|
318
|
<PARAMETERS>
|
319
|
<PARAM name="mdFormat" type="string" managedBy="system" required="true">OPENAIRE</PARAM>
|
320
|
<PARAM name="layout" type="string" managedBy="system" required="true">index</PARAM>
|
321
|
<PARAM name="interpretation" type="string" managedBy="system" required="true">dedup</PARAM>
|
322
|
</PARAMETERS>
|
323
|
<ARCS>
|
324
|
<ARC name="found" to="prepareIndex" />
|
325
|
<ARC name="notFound" to="createIndex" />
|
326
|
</ARCS>
|
327
|
</NODE>
|
328
|
|
329
|
<NODE name="createIndex" type="CreateIndex">
|
330
|
<DESCRIPTION />
|
331
|
<PARAMETERS />
|
332
|
<ARCS>
|
333
|
<ARC to="prepareIndex" />
|
334
|
</ARCS>
|
335
|
</NODE>
|
336
|
|
337
|
<NODE name="prepareIndex" type="PrepareDedupIndexJob">
|
338
|
<DESCRIPTION>Prepare indexing</DESCRIPTION>
|
339
|
<PARAMETERS>
|
340
|
<PARAM name="rottenRecordsPathParam" type="string" required="true" managedBy="system">rottenRecordsPath</PARAM>
|
341
|
<PARAM required="true" type="string" name="dedupConfig" managedBy="system">
|
342
|
{
|
343
|
'dedupConfig' : 'dedup.conf'
|
344
|
}
|
345
|
</PARAM>
|
346
|
</PARAMETERS>
|
347
|
<ARCS>
|
348
|
<ARC to="cleanupRotten" />
|
349
|
</ARCS>
|
350
|
</NODE>
|
351
|
|
352
|
<NODE name="cleanupRotten" type="DeleteHdfsPathJob">
|
353
|
<DESCRIPTION>hdfs cleanup (rotten)</DESCRIPTION>
|
354
|
<PARAMETERS>
|
355
|
<PARAM required="true" type="string" name="cluster" managedBy="system">DM</PARAM>
|
356
|
<PARAM required="true" type="string" name="envParams" managedBy="system">
|
357
|
{
|
358
|
'path' : 'rottenRecordsPath'
|
359
|
}
|
360
|
</PARAM>
|
361
|
</PARAMETERS>
|
362
|
<ARCS>
|
363
|
<ARC to="updateIndex" />
|
364
|
</ARCS>
|
365
|
</NODE>
|
366
|
|
367
|
<NODE name="updateIndex" type="SubmitHadoopJob">
|
368
|
<DESCRIPTION>M/O index records</DESCRIPTION>
|
369
|
<PARAMETERS>
|
370
|
<PARAM required="true" type="string" name="hadoopJob" managedBy="system">dedupIndexFeedJob</PARAM>
|
371
|
<PARAM required="true" type="string" name="envParams" managedBy="system">
|
372
|
{
|
373
|
'mapred.output.dir' : 'rottenRecordsPath',
|
374
|
'index.fields' : 'index.fields',
|
375
|
'index.solr.url' : 'index.solr.url',
|
376
|
'index.solr.collection' : 'index.solr.collection',
|
377
|
'index.buffer.flush.threshold' : 'index.buffer.flush.threshold',
|
378
|
'index.shutdown.wait.time' : 'index.shutdown.wait.time',
|
379
|
'index.solr.sim.mode' : 'index.solr.sim.mode',
|
380
|
'index.feed.timestamp' : 'index.feed.timestamp',
|
381
|
'entityTypeId' : 'entityTypeId',
|
382
|
'entityType' : 'entityType',
|
383
|
'actionset' : 'actionset',
|
384
|
'cluster' : 'cluster'
|
385
|
}
|
386
|
</PARAM>
|
387
|
<PARAM required="true" type="string" name="sysParams" managedBy="system">
|
388
|
{
|
389
|
'hbase.mapred.inputtable' : 'hbase.mapred.datatable',
|
390
|
'hbase.mapreduce.inputtable' : 'hbase.mapred.datatable'
|
391
|
}
|
392
|
</PARAM>
|
393
|
</PARAMETERS>
|
394
|
<ARCS>
|
395
|
<ARC to="finalize" />
|
396
|
</ARCS>
|
397
|
</NODE>
|
398
|
|
399
|
<NODE name="finalize" type="FinalizeDedupIndexFeeding">
|
400
|
<DESCRIPTION>commit changes</DESCRIPTION>
|
401
|
<PARAMETERS />
|
402
|
<ARCS>
|
403
|
<ARC to="updateDs" />
|
404
|
</ARCS>
|
405
|
</NODE>
|
406
|
|
407
|
<NODE name="updateDs" type="IndexDsUpdateJob">
|
408
|
<DESCRIPTION>update DS</DESCRIPTION>
|
409
|
<PARAMETERS />
|
410
|
<ARCS>
|
411
|
<ARC to="success" />
|
412
|
</ARCS>
|
413
|
</NODE>
|
414
|
|
415
|
</CONFIGURATION>
|
416
|
<STATUS />
|
417
|
</BODY>
|
418
|
</RESOURCE_PROFILE>
|