1
|
<workflow-app xmlns="uri:oozie:workflow:0.4" name="mainworkflows_importer_content_url">
|
2
|
|
3
|
<parameters>
|
4
|
<!-- input -->
|
5
|
<property>
|
6
|
<name>objectstore_service_location</name>
|
7
|
<description>object store service location to retrieve PDF/text contents from</description>
|
8
|
</property>
|
9
|
<property>
|
10
|
<name>approved_objectstores_csv</name>
|
11
|
<value>$UNDEFINED$</value>
|
12
|
<description>CSV of approved object stores</description>
|
13
|
</property>
|
14
|
<property>
|
15
|
<name>blacklisted_objectstores_csv</name>
|
16
|
<value>$UNDEFINED$</value>
|
17
|
<description>CSV of blacklisted object stores</description>
|
18
|
</property>
|
19
|
<property>
|
20
|
<name>approved_datasources_csv</name>
|
21
|
<value>$UNDEFINED$</value>
|
22
|
<description>CSV of approved datasources, used interchangeably with approved_objectstores_csv</description>
|
23
|
</property>
|
24
|
<property>
|
25
|
<name>lookup_service_location</name>
|
26
|
<value>$UNDEFINED$</value>
|
27
|
<description>ISLookup service location, required when processing approved_datasources_csv in order to find datasource and objectstore relation</description>
|
28
|
</property>
|
29
|
<property>
|
30
|
<name>mimetypes_pdf</name>
|
31
|
<description>pdf mime types</description>
|
32
|
</property>
|
33
|
<property>
|
34
|
<name>mimetypes_text</name>
|
35
|
<description>text mime types</description>
|
36
|
</property>
|
37
|
<property>
|
38
|
<name>mimetypes_html</name>
|
39
|
<description>html mime types</description>
|
40
|
</property>
|
41
|
<property>
|
42
|
<name>mimetypes_xml_pmc</name>
|
43
|
<description>EuropePMC xml mime types</description>
|
44
|
</property>
|
45
|
<property>
|
46
|
<name>mimetypes_wos</name>
|
47
|
<description>WoS mime types</description>
|
48
|
</property>
|
49
|
<property>
|
50
|
<name>resultset_client_read_timeout</name>
|
51
|
<value>60000</value>
|
52
|
<description>result set client reading timeout</description>
|
53
|
</property>
|
54
|
<!-- input datastores -->
|
55
|
<property>
|
56
|
<name>input_id_mapping</name>
|
57
|
<value>$UNDEFINED$</value>
|
58
|
<description>input containing id mappings used by id-replacer module to translate identifiers of deduplicated objects</description>
|
59
|
</property>
|
60
|
<property>
|
61
|
<name>input_id</name>
|
62
|
<value>$UNDEFINED$</value>
|
63
|
<description>input containing list of identifiers used by existence filter module in order to omit contents which do not have metadata representatives</description>
|
64
|
</property>
|
65
|
<!-- output -->
|
66
|
<property>
|
67
|
<name>output_root</name>
|
68
|
<description>output root directory</description>
|
69
|
</property>
|
70
|
<property>
|
71
|
<name>output_name_pdf</name>
|
72
|
<value>pdf</value>
|
73
|
<description>pdf output subdirectory name</description>
|
74
|
</property>
|
75
|
<property>
|
76
|
<name>output_name_text</name>
|
77
|
<value>text</value>
|
78
|
<description>text output subdirectory name</description>
|
79
|
</property>
|
80
|
<property>
|
81
|
<name>output_name_html</name>
|
82
|
<value>html</value>
|
83
|
<description>html output subdirectory name</description>
|
84
|
</property>
|
85
|
<property>
|
86
|
<name>output_name_xml_pmc</name>
|
87
|
<value>xmlpmc</value>
|
88
|
<description>XML PMC output subdirectory name</description>
|
89
|
</property>
|
90
|
<property>
|
91
|
<name>output_name_wos</name>
|
92
|
<value>wos</value>
|
93
|
<description>WoS output subdirectory name</description>
|
94
|
</property>
|
95
|
</parameters>
|
96
|
|
97
|
<global>
|
98
|
<job-tracker>${jobTracker}</job-tracker>
|
99
|
<name-node>${nameNode}</name-node>
|
100
|
<configuration>
|
101
|
<property>
|
102
|
<name>mapred.job.queue.name</name>
|
103
|
<value>${queueName}</value>
|
104
|
</property>
|
105
|
</configuration>
|
106
|
</global>
|
107
|
|
108
|
<start to="objectstore-content-url-importer" />
|
109
|
|
110
|
<action name="objectstore-content-url-importer">
|
111
|
<java>
|
112
|
<!-- The data generated by this node is deleted in this section -->
|
113
|
<prepare>
|
114
|
<delete path="${nameNode}${workingDir}/imported-urls" />
|
115
|
</prepare>
|
116
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
117
|
<arg>eu.dnetlib.iis.importer.content.ObjectStoreDocumentContentUrlImporterProcess</arg>
|
118
|
<arg>-SworkingDir=${workingDir}</arg>
|
119
|
<arg>-Pimport.content.object.store.location=${objectstore_service_location}</arg>
|
120
|
<arg>-Pimport.content.approved.objectstores.csv=${approved_objectstores_csv}</arg>
|
121
|
<arg>-Pimport.content.blacklisted.objectstores.csv=${blacklisted_objectstores_csv}</arg>
|
122
|
<arg>-Pimport.approved.datasources.csv=${approved_datasources_csv}</arg>
|
123
|
<arg>-Pimport.content.lookup.service.location=${lookup_service_location}</arg>
|
124
|
<arg>-Pimport.resultset.client.read.timeout=${resultset_client_read_timeout}</arg>
|
125
|
<arg>-Ocontent_url=${workingDir}/imported-urls</arg>
|
126
|
</java>
|
127
|
<ok to="input_id_mapping-path-setter" />
|
128
|
<error to="fail" />
|
129
|
</action>
|
130
|
|
131
|
<action name='input_id_mapping-path-setter'>
|
132
|
<java>
|
133
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
134
|
<arg>eu.dnetlib.iis.common.oozie.property.ConditionalPropertySetter</arg>
|
135
|
<arg>-SworkingDir=${workingDir}</arg>
|
136
|
<arg>-Pcondition=${input_id_mapping eq "$UNDEFINED$"}</arg>
|
137
|
<arg>-PinCaseOfTrue=${workingDir}/imported-urls</arg>
|
138
|
<arg>-PelseCase=${workingDir}/transformers_idreplacer/output</arg>
|
139
|
<capture-output />
|
140
|
</java>
|
141
|
<ok to="decision-id-replacer" />
|
142
|
<error to="fail" />
|
143
|
</action>
|
144
|
|
145
|
<decision name="decision-id-replacer">
|
146
|
<switch>
|
147
|
<case to="input_id-path-setter">${input_id_mapping eq "$UNDEFINED$"}</case>
|
148
|
<default to="transformers_idreplacer"/>
|
149
|
</switch>
|
150
|
</decision>
|
151
|
|
152
|
<action name="transformers_idreplacer">
|
153
|
<sub-workflow>
|
154
|
<app-path>${wf:appPath()}/transformers_idreplacer</app-path>
|
155
|
<propagate-configuration/>
|
156
|
<configuration>
|
157
|
<property>
|
158
|
<name>workingDir</name>
|
159
|
<value>${workingDir}/transformers_idreplacer/working_dir</value>
|
160
|
</property>
|
161
|
<property>
|
162
|
<name>id_field_to_replace1</name>
|
163
|
<value>id</value>
|
164
|
</property>
|
165
|
<property>
|
166
|
<name>schema</name>
|
167
|
<value>eu.dnetlib.iis.importer.auxiliary.schemas.DocumentContentUrl</value>
|
168
|
</property>
|
169
|
<property>
|
170
|
<name>input</name>
|
171
|
<value>${workingDir}/imported-urls</value>
|
172
|
</property>
|
173
|
<property>
|
174
|
<name>input_id_mapping</name>
|
175
|
<value>${input_id_mapping}</value>
|
176
|
</property>
|
177
|
<property>
|
178
|
<name>output</name>
|
179
|
<value>${workingDir}/transformers_idreplacer/output</value>
|
180
|
</property>
|
181
|
</configuration>
|
182
|
</sub-workflow>
|
183
|
<ok to="input_id-path-setter" />
|
184
|
<error to="fail" />
|
185
|
</action>
|
186
|
|
187
|
<action name='input_id-path-setter'>
|
188
|
<java>
|
189
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
190
|
<arg>eu.dnetlib.iis.common.oozie.property.ConditionalPropertySetter</arg>
|
191
|
<arg>-SworkingDir=${workingDir}</arg>
|
192
|
<arg>-Pcondition=${input_id eq "$UNDEFINED$"}</arg>
|
193
|
<arg>-PinCaseOfTrue=${wf:actionData('input_id_mapping-path-setter')['result']}</arg>
|
194
|
<arg>-PelseCase=${workingDir}/transformers_common_existencefilter/output</arg>
|
195
|
<capture-output />
|
196
|
</java>
|
197
|
<ok to="decision-existence-filter" />
|
198
|
<error to="fail" />
|
199
|
</action>
|
200
|
|
201
|
<decision name="decision-existence-filter">
|
202
|
<switch>
|
203
|
<case to="content-url-dispatcher">${input_id eq "$UNDEFINED$"}</case>
|
204
|
<default to="transformers_common_existencefilter"/>
|
205
|
</switch>
|
206
|
</decision>
|
207
|
|
208
|
<action name="transformers_common_existencefilter">
|
209
|
<sub-workflow>
|
210
|
<app-path>${wf:appPath()}/transformers_common_existencefilter</app-path>
|
211
|
<propagate-configuration/>
|
212
|
<configuration>
|
213
|
<property>
|
214
|
<name>workingDir</name>
|
215
|
<value>${workingDir}/transformers_common_existencefilter/working_dir</value>
|
216
|
</property>
|
217
|
<property>
|
218
|
<name>input_data</name>
|
219
|
<value>${wf:actionData('input_id_mapping-path-setter')['result']}</value>
|
220
|
</property>
|
221
|
<property>
|
222
|
<name>input_existent_id</name>
|
223
|
<value>${input_id}</value>
|
224
|
</property>
|
225
|
<property>
|
226
|
<name>output_filtered</name>
|
227
|
<value>${workingDir}/transformers_common_existencefilter/output</value>
|
228
|
</property>
|
229
|
</configuration>
|
230
|
</sub-workflow>
|
231
|
<ok to="content-url-dispatcher" />
|
232
|
<error to="fail" />
|
233
|
</action>
|
234
|
|
235
|
<action name="content-url-dispatcher">
|
236
|
<map-reduce>
|
237
|
<prepare>
|
238
|
<delete path="${nameNode}${output_root}" />
|
239
|
</prepare>
|
240
|
<configuration>
|
241
|
<property>
|
242
|
<name>mapreduce.inputformat.class</name>
|
243
|
<value>eu.dnetlib.iis.core.javamapreduce.hack.KeyInputFormat</value>
|
244
|
</property>
|
245
|
<!-- The output format seems to be required, JSON is written when missing -->
|
246
|
<property>
|
247
|
<name>mapreduce.outputformat.class</name>
|
248
|
<value>eu.dnetlib.iis.core.javamapreduce.hack.KeyOutputFormat</value>
|
249
|
</property>
|
250
|
<property>
|
251
|
<name>mapred.mapoutput.key.class</name>
|
252
|
<value>org.apache.avro.mapred.AvroKey</value>
|
253
|
</property>
|
254
|
<property>
|
255
|
<name>mapred.mapoutput.value.class</name>
|
256
|
<value>org.apache.avro.mapred.AvroValue</value>
|
257
|
</property>
|
258
|
<property>
|
259
|
<name>mapred.output.key.class</name>
|
260
|
<value>org.apache.avro.mapred.AvroKey</value>
|
261
|
</property>
|
262
|
<property>
|
263
|
<name>mapred.output.value.class</name>
|
264
|
<value>org.apache.avro.mapred.AvroValue</value>
|
265
|
</property>
|
266
|
<property>
|
267
|
<name>mapred.output.key.comparator.class</name>
|
268
|
<value>eu.dnetlib.iis.core.javamapreduce.hack.KeyComparator</value>
|
269
|
</property>
|
270
|
<property>
|
271
|
<name>eu.dnetlib.iis.avro.output.class</name>
|
272
|
<value>org.apache.avro.Schema.Type.NULL</value>
|
273
|
</property>
|
274
|
<property>
|
275
|
<name>io.serializations</name>
|
276
|
<value>org.apache.hadoop.io.serializer.WritableSerialization,org.apache.hadoop.io.serializer.avro.AvroSpecificSerialization,org.apache.hadoop.io.serializer.avro.AvroReflectSerialization,org.apache.avro.hadoop.io.AvroSerialization
|
277
|
</value>
|
278
|
</property>
|
279
|
<property>
|
280
|
<name>mapred.output.value.groupfn.class</name>
|
281
|
<value>eu.dnetlib.iis.core.javamapreduce.hack.KeyComparator</value>
|
282
|
</property>
|
283
|
<property>
|
284
|
<name>rpc.engine.org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolPB
|
285
|
</name>
|
286
|
<value>org.apache.hadoop.ipc.ProtobufRpcEngine</value>
|
287
|
</property>
|
288
|
<property>
|
289
|
<name>mapred.mapper.new-api</name>
|
290
|
<value>true</value>
|
291
|
</property>
|
292
|
<property>
|
293
|
<name>mapred.reducer.new-api</name>
|
294
|
<value>true</value>
|
295
|
</property>
|
296
|
<!-- required for multiple outputs only -->
|
297
|
<property>
|
298
|
<name>avro.mapreduce.multipleoutputs</name>
|
299
|
<value>${output_name_pdf} ${output_name_text} ${output_name_html} ${output_name_xml_pmc} ${output_name_wos}</value>
|
300
|
</property>
|
301
|
<property>
|
302
|
<name>avro.mapreduce.multipleoutputs.namedOutput.${output_name_pdf}.format
|
303
|
</name>
|
304
|
<value>org.apache.avro.mapreduce.AvroKeyOutputFormat</value>
|
305
|
</property>
|
306
|
<property>
|
307
|
<name>avro.mapreduce.multipleoutputs.namedOutput.${output_name_text}.format
|
308
|
</name>
|
309
|
<value>org.apache.avro.mapreduce.AvroKeyOutputFormat</value>
|
310
|
</property>
|
311
|
<property>
|
312
|
<name>avro.mapreduce.multipleoutputs.namedOutput.${output_name_html}.format
|
313
|
</name>
|
314
|
<value>org.apache.avro.mapreduce.AvroKeyOutputFormat</value>
|
315
|
</property>
|
316
|
<property>
|
317
|
<name>avro.mapreduce.multipleoutputs.namedOutput.${output_name_xml_pmc}.format
|
318
|
</name>
|
319
|
<value>org.apache.avro.mapreduce.AvroKeyOutputFormat</value>
|
320
|
</property>
|
321
|
<property>
|
322
|
<name>avro.mapreduce.multipleoutputs.namedOutput.${output_name_wos}.format
|
323
|
</name>
|
324
|
<value>org.apache.avro.mapreduce.AvroKeyOutputFormat</value>
|
325
|
</property>
|
326
|
<property>
|
327
|
<name>eu.dnetlib.iis.avro.multipleoutputs.class.${output_name_pdf}</name>
|
328
|
<value>eu.dnetlib.iis.importer.auxiliary.schemas.DocumentContentUrl</value>
|
329
|
</property>
|
330
|
<property>
|
331
|
<name>eu.dnetlib.iis.avro.multipleoutputs.class.${output_name_text}</name>
|
332
|
<value>eu.dnetlib.iis.importer.auxiliary.schemas.DocumentContentUrl</value>
|
333
|
</property>
|
334
|
<property>
|
335
|
<name>eu.dnetlib.iis.avro.multipleoutputs.class.${output_name_html}</name>
|
336
|
<value>eu.dnetlib.iis.importer.auxiliary.schemas.DocumentContentUrl</value>
|
337
|
</property>
|
338
|
<property>
|
339
|
<name>eu.dnetlib.iis.avro.multipleoutputs.class.${output_name_xml_pmc}</name>
|
340
|
<value>eu.dnetlib.iis.importer.auxiliary.schemas.DocumentContentUrl</value>
|
341
|
</property>
|
342
|
<property>
|
343
|
<name>eu.dnetlib.iis.avro.multipleoutputs.class.${output_name_wos}</name>
|
344
|
<value>eu.dnetlib.iis.importer.auxiliary.schemas.DocumentContentUrl</value>
|
345
|
</property>
|
346
|
<!-- mimetypes related -->
|
347
|
<property>
|
348
|
<name>mimetypes.csv.${output_name_pdf}</name>
|
349
|
<value>${mimetypes_pdf}</value>
|
350
|
</property>
|
351
|
<property>
|
352
|
<name>mimetypes.csv.${output_name_text}</name>
|
353
|
<value>${mimetypes_text}</value>
|
354
|
</property>
|
355
|
<property>
|
356
|
<name>mimetypes.csv.${output_name_html}</name>
|
357
|
<value>${mimetypes_html}</value>
|
358
|
</property>
|
359
|
<property>
|
360
|
<name>mimetypes.csv.${output_name_xml_pmc}</name>
|
361
|
<value>${mimetypes_xml_pmc}</value>
|
362
|
</property>
|
363
|
<property>
|
364
|
<name>mimetypes.csv.${output_name_wos}</name>
|
365
|
<value>${mimetypes_wos}</value>
|
366
|
</property>
|
367
|
<property>
|
368
|
<name>mapreduce.map.class</name>
|
369
|
<value>eu.dnetlib.iis.importer.content.DocumentContentUrlDispatcher</value>
|
370
|
</property>
|
371
|
<property>
|
372
|
<name>eu.dnetlib.iis.avro.input.class</name>
|
373
|
<value>eu.dnetlib.iis.importer.auxiliary.schemas.DocumentContentUrl</value>
|
374
|
</property>
|
375
|
<property>
|
376
|
<name>eu.dnetlib.iis.avro.map.output.key.class</name>
|
377
|
<value>org.apache.avro.Schema.Type.STRING</value>
|
378
|
</property>
|
379
|
<property>
|
380
|
<name>eu.dnetlib.iis.avro.map.output.value.class</name>
|
381
|
<value>org.apache.avro.Schema.Type.NULL</value>
|
382
|
</property>
|
383
|
<property>
|
384
|
<name>mapred.input.dir</name>
|
385
|
<value>${wf:actionData('input_id-path-setter')['result']}</value>
|
386
|
</property>
|
387
|
<property>
|
388
|
<name>mapred.output.dir</name>
|
389
|
<value>${output_root}</value>
|
390
|
</property>
|
391
|
<property>
|
392
|
<name>mapred.reduce.tasks</name>
|
393
|
<value>0</value>
|
394
|
</property>
|
395
|
</configuration>
|
396
|
</map-reduce>
|
397
|
<ok to="end" />
|
398
|
<error to="fail" />
|
399
|
</action>
|
400
|
|
401
|
<kill name="fail">
|
402
|
<message>Unfortunately, the process failed -- error message:
|
403
|
[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
404
|
</kill>
|
405
|
<end name="end" />
|
406
|
</workflow-app>
|