|
1
|
<workflow-app xmlns="uri:oozie:workflow:0.4" name="mainworkflows_importer_plaintext_cached">
|
|
2
|
|
|
3
|
<parameters>
|
|
4
|
<property>
|
|
5
|
<name>input</name>
|
|
6
|
<description>input document content directory</description>
|
|
7
|
</property>
|
|
8
|
<property>
|
|
9
|
<name>output</name>
|
|
10
|
<description>plaintext importer output directory</description>
|
|
11
|
</property>
|
|
12
|
<property>
|
|
13
|
<name>content_connection_timeout</name>
|
|
14
|
<value>60000</value>
|
|
15
|
<description>streaming content connection timeout</description>
|
|
16
|
</property>
|
|
17
|
<property>
|
|
18
|
<name>content_read_timeout</name>
|
|
19
|
<value>60000</value>
|
|
20
|
<description>streaming content read timeout</description>
|
|
21
|
</property>
|
|
22
|
<property>
|
|
23
|
<name>zk_session_timeout</name>
|
|
24
|
<value>60000</value>
|
|
25
|
<description>zookeeper session timeout when handling locks</description>
|
|
26
|
</property>
|
|
27
|
<property>
|
|
28
|
<name>default_cache_location</name>
|
|
29
|
<description>default cache location stored in HDFS</description>
|
|
30
|
</property>
|
|
31
|
<property>
|
|
32
|
<name>mapred_max_split_size</name>
|
|
33
|
<value>50000</value>
|
|
34
|
<description>maximum input data split size, required by streaming version reading DocumentContentUrl to split input data into more chunks</description>
|
|
35
|
</property>
|
|
36
|
</parameters>
|
|
37
|
|
|
38
|
<global>
|
|
39
|
<job-tracker>${jobTracker}</job-tracker>
|
|
40
|
<name-node>${nameNode}</name-node>
|
|
41
|
<configuration>
|
|
42
|
<property>
|
|
43
|
<name>mapred.job.queue.name</name>
|
|
44
|
<value>${queueName}</value>
|
|
45
|
</property>
|
|
46
|
</configuration>
|
|
47
|
</global>
|
|
48
|
|
|
49
|
<start to="check_input_isempty" />
|
|
50
|
|
|
51
|
<action name="check_input_isempty">
|
|
52
|
<java>
|
|
53
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
54
|
<arg>eu.dnetlib.iis.common.utils.EmptyDatastoreVerifierProcess</arg>
|
|
55
|
<arg>-Iinput=${input}</arg>
|
|
56
|
<capture-output />
|
|
57
|
</java>
|
|
58
|
<ok to="decision-is-input-empty" />
|
|
59
|
<error to="fail" />
|
|
60
|
</action>
|
|
61
|
|
|
62
|
<decision name="decision-is-input-empty">
|
|
63
|
<switch>
|
|
64
|
<!-- skipping metadataextraction merging process -->
|
|
65
|
<case to="get-existing-cache-id">${wf:actionData('check_input_isempty')['isEmpty'] eq "false"}</case>
|
|
66
|
<default to="generate-empty-output"/>
|
|
67
|
</switch>
|
|
68
|
</decision>
|
|
69
|
|
|
70
|
<action name="generate-empty-output">
|
|
71
|
<java>
|
|
72
|
<prepare>
|
|
73
|
<!-- notice: directory have to aligned with skipped action output -->
|
|
74
|
<delete path="${nameNode}${output}" />
|
|
75
|
</prepare>
|
|
76
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
77
|
<arg>eu.dnetlib.iis.core.java.jsonworkflownodes.Producer</arg>
|
|
78
|
<arg>-C{plaintext,
|
|
79
|
eu.dnetlib.iis.metadataextraction.schemas.DocumentText,
|
|
80
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
81
|
<arg>-Oplaintext=${output}</arg>
|
|
82
|
</java>
|
|
83
|
<ok to="end"/>
|
|
84
|
<error to="fail"/>
|
|
85
|
</action>
|
|
86
|
|
|
87
|
<action name='get-existing-cache-id'>
|
|
88
|
<java>
|
|
89
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
90
|
<arg>eu.dnetlib.iis.common.cache.CacheMetadataManagingProcess</arg>
|
|
91
|
<arg>-Pdefault_cache_location=${default_cache_location}</arg>
|
|
92
|
<arg>-Pmode=read_current_id</arg>
|
|
93
|
<capture-output />
|
|
94
|
</java>
|
|
95
|
<ok to="decision-is-cache-empty" />
|
|
96
|
<error to="fail" />
|
|
97
|
</action>
|
|
98
|
|
|
99
|
<decision name="decision-is-cache-empty">
|
|
100
|
<switch>
|
|
101
|
<case to="importer_plaintext_on_full_input">${wf:actionData('get-existing-cache-id')['cache_id'] eq "$UNDEFINED$"}</case>
|
|
102
|
<default to="transformers_importer_plaintext_skip_extracted"/>
|
|
103
|
</switch>
|
|
104
|
</decision>
|
|
105
|
|
|
106
|
<!-- end of cache based processing block, cache was provided as an input -->
|
|
107
|
<action name="transformers_importer_plaintext_skip_extracted">
|
|
108
|
<sub-workflow>
|
|
109
|
<app-path>${wf:appPath()}/transformers_importer_plaintext_skip_extracted</app-path>
|
|
110
|
<propagate-configuration/>
|
|
111
|
<configuration>
|
|
112
|
<property>
|
|
113
|
<name>workingDir</name>
|
|
114
|
<value>${workingDir}/transformers_importer_plaintext_skip_extracted/working_dir</value>
|
|
115
|
</property>
|
|
116
|
<property>
|
|
117
|
<name>input_document_content</name>
|
|
118
|
<value>${input}</value>
|
|
119
|
</property>
|
|
120
|
<property>
|
|
121
|
<name>input_document_text</name>
|
|
122
|
<value>${default_cache_location}/${wf:actionData('get-existing-cache-id')['cache_id']}</value>
|
|
123
|
</property>
|
|
124
|
<property>
|
|
125
|
<name>output_document_content</name>
|
|
126
|
<value>${workingDir}/transformers_importer_plaintext_skip_extracted/tobeprocessed_content</value>
|
|
127
|
</property>
|
|
128
|
<property>
|
|
129
|
<name>output_document_text</name>
|
|
130
|
<value>${workingDir}/transformers_importer_plaintext_skip_extracted/tobereturned_text</value>
|
|
131
|
</property>
|
|
132
|
</configuration>
|
|
133
|
</sub-workflow>
|
|
134
|
<ok to="importer_plaintext_on_filtered_input"/>
|
|
135
|
<error to="fail" />
|
|
136
|
</action>
|
|
137
|
|
|
138
|
<action name="importer_plaintext_on_filtered_input">
|
|
139
|
<sub-workflow>
|
|
140
|
<app-path>${wf:appPath()}/importer_plaintext</app-path>
|
|
141
|
<propagate-configuration/>
|
|
142
|
<configuration>
|
|
143
|
<property>
|
|
144
|
<name>workingDir</name>
|
|
145
|
<value>${workingDir}/importer_plaintext/working_dir</value>
|
|
146
|
</property>
|
|
147
|
<property>
|
|
148
|
<name>input</name>
|
|
149
|
<value>${workingDir}/transformers_importer_plaintext_skip_extracted/tobeprocessed_content</value>
|
|
150
|
</property>
|
|
151
|
<property>
|
|
152
|
<name>output</name>
|
|
153
|
<value>${workingDir}/importer_plaintext/output</value>
|
|
154
|
</property>
|
|
155
|
<!-- all the other properties are autmatically propagated-->
|
|
156
|
</configuration>
|
|
157
|
</sub-workflow>
|
|
158
|
<ok to="transformers_common_union_plaintext_merge_outputs"/>
|
|
159
|
<error to="fail" />
|
|
160
|
</action>
|
|
161
|
|
|
162
|
<action name="transformers_common_union_plaintext_merge_outputs">
|
|
163
|
<sub-workflow>
|
|
164
|
<app-path>${wf:appPath()}/transformers_common_union</app-path>
|
|
165
|
<propagate-configuration/>
|
|
166
|
<configuration>
|
|
167
|
<property>
|
|
168
|
<name>input_a</name>
|
|
169
|
<value>${workingDir}/transformers_importer_plaintext_skip_extracted/tobereturned_text</value>
|
|
170
|
</property>
|
|
171
|
<property>
|
|
172
|
<name>input_b</name>
|
|
173
|
<value>${workingDir}/importer_plaintext/output</value>
|
|
174
|
</property>
|
|
175
|
<property>
|
|
176
|
<name>output</name>
|
|
177
|
<value>${output}</value>
|
|
178
|
</property>
|
|
179
|
<property>
|
|
180
|
<name>schema</name>
|
|
181
|
<value>eu.dnetlib.iis.metadataextraction.schemas.DocumentText</value>
|
|
182
|
</property>
|
|
183
|
</configuration>
|
|
184
|
</sub-workflow>
|
|
185
|
<ok to="check_importer_plaintext_output_isempty"/>
|
|
186
|
<error to="fail" />
|
|
187
|
</action>
|
|
188
|
|
|
189
|
<action name='check_importer_plaintext_output_isempty'>
|
|
190
|
<java>
|
|
191
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
192
|
<arg>eu.dnetlib.iis.common.utils.EmptyDatastoreVerifierProcess</arg>
|
|
193
|
<arg>-Iinput=${workingDir}/importer_plaintext/output</arg>
|
|
194
|
<capture-output />
|
|
195
|
</java>
|
|
196
|
<ok to="decision-is-importer_plaintext-output-empty" />
|
|
197
|
<error to="fail" />
|
|
198
|
</action>
|
|
199
|
|
|
200
|
<decision name="decision-is-importer_plaintext-output-empty">
|
|
201
|
<switch>
|
|
202
|
<!-- skipping metadataextraction merging process -->
|
|
203
|
<case to="obtain-lock_for_merging">${wf:actionData('check_importer_plaintext_output_isempty')['isEmpty'] eq "false"}</case>
|
|
204
|
<default to="end"/>
|
|
205
|
</switch>
|
|
206
|
</decision>
|
|
207
|
|
|
208
|
<action name="obtain-lock_for_merging">
|
|
209
|
<java>
|
|
210
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
211
|
<arg>eu.dnetlib.iis.common.lock.LockManagingProcess</arg>
|
|
212
|
<arg>-Pzk_session_timeout=${zk_session_timeout}</arg>
|
|
213
|
<arg>-Pnode_id=${default_cache_location}</arg>
|
|
214
|
<arg>-Pmode=obtain</arg>
|
|
215
|
</java>
|
|
216
|
<ok to="get-new-cache-id_for_merging"/>
|
|
217
|
<error to="release-lock-and-fail"/>
|
|
218
|
</action>
|
|
219
|
|
|
220
|
<action name='get-new-cache-id_for_merging'>
|
|
221
|
<java>
|
|
222
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
223
|
<arg>eu.dnetlib.iis.common.cache.CacheMetadataManagingProcess</arg>
|
|
224
|
<arg>-Pdefault_cache_location=${default_cache_location}</arg>
|
|
225
|
<arg>-Pmode=generate_new_id</arg>
|
|
226
|
<capture-output />
|
|
227
|
</java>
|
|
228
|
<ok to="transformers_common_union_plaintext_merge_cache" />
|
|
229
|
<error to="release-lock-and-fail" />
|
|
230
|
</action>
|
|
231
|
|
|
232
|
<action name="transformers_common_union_plaintext_merge_cache">
|
|
233
|
<sub-workflow>
|
|
234
|
<app-path>${wf:appPath()}/transformers_common_union</app-path>
|
|
235
|
<propagate-configuration/>
|
|
236
|
<configuration>
|
|
237
|
<property>
|
|
238
|
<name>input_a</name>
|
|
239
|
<value>${default_cache_location}/${wf:actionData('get-existing-cache-id')['cache_id']}</value>
|
|
240
|
</property>
|
|
241
|
<property>
|
|
242
|
<name>input_b</name>
|
|
243
|
<value>${workingDir}/importer_plaintext/output</value>
|
|
244
|
</property>
|
|
245
|
<property>
|
|
246
|
<name>output</name>
|
|
247
|
<value>${default_cache_location}/${wf:actionData('get-new-cache-id_for_merging')['cache_id']}</value>
|
|
248
|
</property>
|
|
249
|
<property>
|
|
250
|
<name>schema</name>
|
|
251
|
<value>eu.dnetlib.iis.metadataextraction.schemas.DocumentText</value>
|
|
252
|
</property>
|
|
253
|
</configuration>
|
|
254
|
</sub-workflow>
|
|
255
|
<ok to="write-new-cache-id_for_merging"/>
|
|
256
|
<error to="fail-merge_cache-temp_files_cleanup" />
|
|
257
|
</action>
|
|
258
|
|
|
259
|
<action name="fail-merge_cache-temp_files_cleanup">
|
|
260
|
<fs>
|
|
261
|
<delete path="${nameNode}${default_cache_location}/${wf:actionData('get-new-cache-id_for_merging')['cache_id']}" />
|
|
262
|
</fs>
|
|
263
|
<ok to="release-lock-and-fail"/>
|
|
264
|
<error to="release-lock-and-fail"/>
|
|
265
|
</action>
|
|
266
|
|
|
267
|
<action name='write-new-cache-id_for_merging'>
|
|
268
|
<java>
|
|
269
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
270
|
<arg>eu.dnetlib.iis.common.cache.CacheMetadataManagingProcess</arg>
|
|
271
|
<arg>-Pdefault_cache_location=${default_cache_location}</arg>
|
|
272
|
<arg>-Pmode=write_id</arg>
|
|
273
|
<arg>-Pid=${wf:actionData('get-new-cache-id_for_merging')['cache_id']}</arg>
|
|
274
|
<capture-output />
|
|
275
|
</java>
|
|
276
|
<ok to="release-lock-and-end" />
|
|
277
|
<error to="fail-merge_cache-temp_files_cleanup" />
|
|
278
|
</action>
|
|
279
|
|
|
280
|
<!-- end of cache based processing block, cache was provided as an input -->
|
|
281
|
|
|
282
|
<!-- full input processing block, no cache was provided as an input -->
|
|
283
|
<action name="importer_plaintext_on_full_input">
|
|
284
|
<sub-workflow>
|
|
285
|
<app-path>${wf:appPath()}/importer_plaintext</app-path>
|
|
286
|
<propagate-configuration/>
|
|
287
|
<configuration>
|
|
288
|
<property>
|
|
289
|
<name>workingDir</name>
|
|
290
|
<value>${workingDir}/importer_plaintext/working_dir</value>
|
|
291
|
</property>
|
|
292
|
<!-- all the other properties are autmatically propagated-->
|
|
293
|
</configuration>
|
|
294
|
</sub-workflow>
|
|
295
|
<ok to="obtain-lock_for_initializing"/>
|
|
296
|
<error to="fail" />
|
|
297
|
</action>
|
|
298
|
|
|
299
|
<action name="obtain-lock_for_initializing">
|
|
300
|
<java>
|
|
301
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
302
|
<arg>eu.dnetlib.iis.common.lock.LockManagingProcess</arg>
|
|
303
|
<arg>-Pzk_session_timeout=${zk_session_timeout}</arg>
|
|
304
|
<arg>-Pnode_id=${default_cache_location}</arg>
|
|
305
|
<arg>-Pmode=obtain</arg>
|
|
306
|
</java>
|
|
307
|
<ok to="get-new-cache-id_for_initializing"/>
|
|
308
|
<error to="release-lock-and-fail"/>
|
|
309
|
</action>
|
|
310
|
|
|
311
|
<action name='get-new-cache-id_for_initializing'>
|
|
312
|
<java>
|
|
313
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
314
|
<arg>eu.dnetlib.iis.common.cache.CacheMetadataManagingProcess</arg>
|
|
315
|
<arg>-Pdefault_cache_location=${default_cache_location}</arg>
|
|
316
|
<arg>-Pmode=generate_new_id</arg>
|
|
317
|
<capture-output />
|
|
318
|
</java>
|
|
319
|
<ok to="initialize_plaintext_cache" />
|
|
320
|
<error to="release-lock-and-fail" />
|
|
321
|
</action>
|
|
322
|
|
|
323
|
<action name="initialize_plaintext_cache">
|
|
324
|
<distcp xmlns="uri:oozie:distcp-action:0.1">
|
|
325
|
<job-tracker>${jobTracker}</job-tracker>
|
|
326
|
<name-node>${nameNode}</name-node>
|
|
327
|
<arg>${nameNode}${output}</arg>
|
|
328
|
<arg>${nameNode}${default_cache_location}/${wf:actionData('get-new-cache-id_for_initializing')['cache_id']}</arg>
|
|
329
|
</distcp>
|
|
330
|
<ok to="write-new-cache-id_for_initializing"/>
|
|
331
|
<error to="fail-initialize_cache-temp_files_cleanup"/>
|
|
332
|
</action>
|
|
333
|
|
|
334
|
<action name="fail-initialize_cache-temp_files_cleanup">
|
|
335
|
<fs>
|
|
336
|
<delete path="${nameNode}${default_cache_location}/${wf:actionData('get-new-cache-id_for_initializing')['cache_id']}" />
|
|
337
|
</fs>
|
|
338
|
<ok to="release-lock-and-fail"/>
|
|
339
|
<error to="release-lock-and-fail"/>
|
|
340
|
</action>
|
|
341
|
|
|
342
|
<action name='write-new-cache-id_for_initializing'>
|
|
343
|
<java>
|
|
344
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
345
|
<arg>eu.dnetlib.iis.common.cache.CacheMetadataManagingProcess</arg>
|
|
346
|
<arg>-Pdefault_cache_location=${default_cache_location}</arg>
|
|
347
|
<arg>-Pmode=write_id</arg>
|
|
348
|
<arg>-Pid=${wf:actionData('get-new-cache-id_for_initializing')['cache_id']}</arg>
|
|
349
|
<capture-output />
|
|
350
|
</java>
|
|
351
|
<ok to="release-lock-and-end" />
|
|
352
|
<error to="fail-initialize_cache-temp_files_cleanup" />
|
|
353
|
</action>
|
|
354
|
|
|
355
|
<!-- end of full input processing block, no cache was provided as an input -->
|
|
356
|
|
|
357
|
<!-- lock releasing actions -->
|
|
358
|
<action name="release-lock-and-fail">
|
|
359
|
<java>
|
|
360
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
361
|
<arg>eu.dnetlib.iis.common.lock.LockManagingProcess</arg>
|
|
362
|
<arg>-Pzk_session_timeout=${zk_session_timeout}</arg>
|
|
363
|
<arg>-Pnode_id=${default_cache_location}</arg>
|
|
364
|
<arg>-Pmode=release</arg>
|
|
365
|
</java>
|
|
366
|
<ok to="fail"/>
|
|
367
|
<error to="fail"/>
|
|
368
|
</action>
|
|
369
|
|
|
370
|
<action name="release-lock-and-end">
|
|
371
|
<java>
|
|
372
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
373
|
<arg>eu.dnetlib.iis.common.lock.LockManagingProcess</arg>
|
|
374
|
<arg>-Pzk_session_timeout=${zk_session_timeout}</arg>
|
|
375
|
<arg>-Pnode_id=${default_cache_location}</arg>
|
|
376
|
<arg>-Pmode=release</arg>
|
|
377
|
</java>
|
|
378
|
<ok to="end"/>
|
|
379
|
<error to="fail"/>
|
|
380
|
</action>
|
|
381
|
<!-- end of lock releasing actions -->
|
|
382
|
|
|
383
|
<kill name="fail">
|
|
384
|
<message>Unfortunately, the process failed -- error message:
|
|
385
|
[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
|
386
|
</kill>
|
|
387
|
<end name="end" />
|
|
388
|
</workflow-app>
|