|
1
|
<workflow-app xmlns="uri:oozie:workflow:0.4" name="metadataextraction">
|
|
2
|
|
|
3
|
<parameters>
|
|
4
|
<property>
|
|
5
|
<name>input</name>
|
|
6
|
<description>metadata extraction input directory</description>
|
|
7
|
</property>
|
|
8
|
<property>
|
|
9
|
<name>output_root</name>
|
|
10
|
<description>metadata extraction output directory</description>
|
|
11
|
</property>
|
|
12
|
<property>
|
|
13
|
<name>output_name_meta</name>
|
|
14
|
<value>meta</value>
|
|
15
|
<description>metadata output subdirectory name</description>
|
|
16
|
</property>
|
|
17
|
<property>
|
|
18
|
<name>output_name_plaintext</name>
|
|
19
|
<value>plaintext</value>
|
|
20
|
<description>plaintext output subdirectory name</description>
|
|
21
|
</property>
|
|
22
|
<property>
|
|
23
|
<name>output_name_fault</name>
|
|
24
|
<value>fault</value>
|
|
25
|
<description>fault output subdirectory name</description>
|
|
26
|
</property>
|
|
27
|
<property>
|
|
28
|
<name>excluded_ids</name>
|
|
29
|
<value>$UNDEFINED$</value>
|
|
30
|
<description>excluded identifiers list</description>
|
|
31
|
</property>
|
|
32
|
<property>
|
|
33
|
<name>max_file_size_mb</name>
|
|
34
|
<value>$UNDEFINED$</value>
|
|
35
|
<description>maximum allowed file size in Megabytes</description>
|
|
36
|
</property>
|
|
37
|
<property>
|
|
38
|
<name>log_fault_processing_time_threshold_secs</name>
|
|
39
|
<value>300</value>
|
|
40
|
<description>processing time threshold expressed in seconds.
|
|
41
|
When exceeded apropriate object will be written to fault datastore</description>
|
|
42
|
</property>
|
|
43
|
<property>
|
|
44
|
<name>content_connection_timeout</name>
|
|
45
|
<value>60000</value>
|
|
46
|
<description>streaming content connection timeout</description>
|
|
47
|
</property>
|
|
48
|
<property>
|
|
49
|
<name>content_read_timeout</name>
|
|
50
|
<value>60000</value>
|
|
51
|
<description>streaming content read timeout</description>
|
|
52
|
</property>
|
|
53
|
<property>
|
|
54
|
<name>mapred_child_java_opts</name>
|
|
55
|
<value>-Xmx4096m</value>
|
|
56
|
<description>java-opts, e.g. maximum heap size for oozie</description>
|
|
57
|
</property>
|
|
58
|
<property>
|
|
59
|
<name>mapred_max_split_size</name>
|
|
60
|
<value>50000</value>
|
|
61
|
<description>maximum input data split size, required by streaming version reading DocumentContentUrl to split input data into more chunks</description>
|
|
62
|
</property>
|
|
63
|
|
|
64
|
<property>
|
|
65
|
<name>processing_mode</name>
|
|
66
|
<value>MetadataExtractorMapper</value>
|
|
67
|
<description>metadata extraction processing mode, supported values: MetadataExtractorMapper, StreamingMetadataExtractorMapper</description>
|
|
68
|
</property>
|
|
69
|
<property>
|
|
70
|
<name>inputport_classname</name>
|
|
71
|
<value>eu.dnetlib.iis.importer.schemas.DocumentContent</value>
|
|
72
|
<description>input classname, should be adjusted according to the processing_mode value. Supported values: eu.dnetlib.iis.importer.schemas.DocumentContent, eu.dnetlib.iis.importer.auxiliary.schemas.DocumentContentUrl</description>
|
|
73
|
</property>
|
|
74
|
</parameters>
|
|
75
|
|
|
76
|
<start to="metadata_extractor"/>
|
|
77
|
|
|
78
|
<action name="metadata_extractor">
|
|
79
|
<map-reduce>
|
|
80
|
<job-tracker>${jobTracker}</job-tracker>
|
|
81
|
<name-node>${nameNode}</name-node>
|
|
82
|
<!-- The data generated by this node is deleted in this section -->
|
|
83
|
<prepare>
|
|
84
|
<delete path="${nameNode}${output_root}" />
|
|
85
|
</prepare>
|
|
86
|
<configuration>
|
|
87
|
<property>
|
|
88
|
<name>mapred.task.timeout</name>
|
|
89
|
<value>1800000</value>
|
|
90
|
</property>
|
|
91
|
<property>
|
|
92
|
<name>mapreduce.inputformat.class</name>
|
|
93
|
<value>eu.dnetlib.iis.core.javamapreduce.hack.KeyInputFormat</value>
|
|
94
|
</property>
|
|
95
|
<property>
|
|
96
|
<name>mapred.mapoutput.key.class</name>
|
|
97
|
<value>org.apache.avro.mapred.AvroKey</value>
|
|
98
|
</property>
|
|
99
|
<property>
|
|
100
|
<name>mapred.mapoutput.value.class</name>
|
|
101
|
<value>org.apache.avro.mapred.AvroValue</value>
|
|
102
|
</property>
|
|
103
|
<property>
|
|
104
|
<name>mapred.output.key.class</name>
|
|
105
|
<value>org.apache.avro.mapred.AvroKey</value>
|
|
106
|
</property>
|
|
107
|
<property>
|
|
108
|
<name>mapred.output.value.class</name>
|
|
109
|
<value>org.apache.avro.mapred.AvroValue</value>
|
|
110
|
</property>
|
|
111
|
<property>
|
|
112
|
<name>mapred.output.key.comparator.class</name>
|
|
113
|
<value>eu.dnetlib.iis.core.javamapreduce.hack.KeyComparator</value>
|
|
114
|
</property>
|
|
115
|
<property>
|
|
116
|
<name>io.serializations</name>
|
|
117
|
<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
|
|
118
|
</value>
|
|
119
|
</property>
|
|
120
|
<property>
|
|
121
|
<name>mapred.output.value.groupfn.class</name>
|
|
122
|
<value>eu.dnetlib.iis.core.javamapreduce.hack.KeyComparator</value>
|
|
123
|
</property>
|
|
124
|
|
|
125
|
<property>
|
|
126
|
<name>rpc.engine.org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolPB
|
|
127
|
</name>
|
|
128
|
<value>org.apache.hadoop.ipc.ProtobufRpcEngine</value>
|
|
129
|
</property>
|
|
130
|
|
|
131
|
<!-- This is required for new api usage -->
|
|
132
|
<property>
|
|
133
|
<name>mapred.mapper.new-api</name>
|
|
134
|
<value>true</value>
|
|
135
|
</property>
|
|
136
|
<property>
|
|
137
|
<name>mapred.reducer.new-api</name>
|
|
138
|
<value>true</value>
|
|
139
|
</property>
|
|
140
|
|
|
141
|
<!-- Standard stuff for our framework -->
|
|
142
|
<property>
|
|
143
|
<name>mapred.job.queue.name</name>
|
|
144
|
<value>${queueName}</value>
|
|
145
|
</property>
|
|
146
|
|
|
147
|
<property>
|
|
148
|
<name>mapred.child.java.opts</name>
|
|
149
|
<value>${mapred_child_java_opts}</value>
|
|
150
|
</property>
|
|
151
|
|
|
152
|
<property>
|
|
153
|
<name>mapred.max.split.size</name>
|
|
154
|
<value>${mapred_max_split_size}</value>
|
|
155
|
</property>
|
|
156
|
|
|
157
|
<property>
|
|
158
|
<name>avro.mapreduce.multipleoutputs</name>
|
|
159
|
<value>${output_name_meta} ${output_name_plaintext} ${output_name_fault}</value>
|
|
160
|
</property>
|
|
161
|
|
|
162
|
<!-- ## Output classes for all output ports -->
|
|
163
|
<property>
|
|
164
|
<name>avro.mapreduce.multipleoutputs.namedOutput.${output_name_meta}.format
|
|
165
|
</name>
|
|
166
|
<value>org.apache.avro.mapreduce.AvroKeyOutputFormat</value>
|
|
167
|
</property>
|
|
168
|
<property>
|
|
169
|
<name>avro.mapreduce.multipleoutputs.namedOutput.${output_name_plaintext}.format
|
|
170
|
</name>
|
|
171
|
<value>org.apache.avro.mapreduce.AvroKeyOutputFormat</value>
|
|
172
|
</property>
|
|
173
|
<property>
|
|
174
|
<name>avro.mapreduce.multipleoutputs.namedOutput.${output_name_fault}.format
|
|
175
|
</name>
|
|
176
|
<value>org.apache.avro.mapreduce.AvroKeyOutputFormat</value>
|
|
177
|
</property>
|
|
178
|
<!-- ### Schema of the data ingested by the mapper. To be more precise,
|
|
179
|
it's the schema of Avro data passed as template parameter of the AvroKey
|
|
180
|
object passed to mapper. -->
|
|
181
|
<property>
|
|
182
|
<name>eu.dnetlib.iis.avro.input.class</name>
|
|
183
|
<value>${inputport_classname}</value>
|
|
184
|
</property>
|
|
185
|
<!-- As a convention, we're setting "null" values
|
|
186
|
since mapper does not produce any standard data in this example
|
|
187
|
(probably any other valid Avro schema would be OK as well).-->
|
|
188
|
<property>
|
|
189
|
<name>eu.dnetlib.iis.avro.map.output.key.class</name>
|
|
190
|
<value>org.apache.avro.Schema.Type.NULL</value>
|
|
191
|
</property>
|
|
192
|
<!-- As a convention, we're setting "null" values
|
|
193
|
since mapper does not produce any standard data in this example
|
|
194
|
(probably any other valid Avro schema would be OK as well).-->
|
|
195
|
|
|
196
|
<property>
|
|
197
|
<name>eu.dnetlib.iis.avro.map.output.value.class</name>
|
|
198
|
<value>org.apache.avro.Schema.Type.NULL</value>
|
|
199
|
</property>
|
|
200
|
|
|
201
|
|
|
202
|
<!-- ### Schema of multiple output ports. -->
|
|
203
|
<property>
|
|
204
|
<name>eu.dnetlib.iis.avro.multipleoutputs.class.${output_name_meta}</name>
|
|
205
|
<value>eu.dnetlib.iis.metadataextraction.schemas.ExtractedDocumentMetadata</value>
|
|
206
|
</property>
|
|
207
|
<property>
|
|
208
|
<name>eu.dnetlib.iis.avro.multipleoutputs.class.${output_name_plaintext}</name>
|
|
209
|
<value>eu.dnetlib.iis.metadataextraction.schemas.DocumentText</value>
|
|
210
|
</property>
|
|
211
|
<property>
|
|
212
|
<name>eu.dnetlib.iis.avro.multipleoutputs.class.${output_name_fault}</name>
|
|
213
|
<value>eu.dnetlib.iis.audit.schemas.Fault</value>
|
|
214
|
</property>
|
|
215
|
|
|
216
|
<property>
|
|
217
|
<name>mapred.input.dir</name>
|
|
218
|
<value>${input}</value>
|
|
219
|
</property>
|
|
220
|
<property>
|
|
221
|
<name>mapred.output.dir</name>
|
|
222
|
<value>${output_root}</value>
|
|
223
|
</property>
|
|
224
|
<property>
|
|
225
|
<name>output.meta</name>
|
|
226
|
<value>${output_name_meta}</value>
|
|
227
|
</property>
|
|
228
|
<property>
|
|
229
|
<name>output.plaintext</name>
|
|
230
|
<value>${output_name_plaintext}</value>
|
|
231
|
</property>
|
|
232
|
<property>
|
|
233
|
<name>output.fault</name>
|
|
234
|
<value>${output_name_fault}</value>
|
|
235
|
</property>
|
|
236
|
<property>
|
|
237
|
<name>excluded.ids</name>
|
|
238
|
<value>${excluded_ids}</value>
|
|
239
|
</property>
|
|
240
|
<property>
|
|
241
|
<name>import.content.max.file.size.mb</name>
|
|
242
|
<value>${max_file_size_mb}</value>
|
|
243
|
</property>
|
|
244
|
<property>
|
|
245
|
<name>log.fault.processing.time.threshold.secs</name>
|
|
246
|
<value>${log_fault_processing_time_threshold_secs}</value>
|
|
247
|
</property>
|
|
248
|
<property>
|
|
249
|
<name>import.content.connection.timeout</name>
|
|
250
|
<value>${content_connection_timeout}</value>
|
|
251
|
</property>
|
|
252
|
<property>
|
|
253
|
<name>import.content.read.timeout</name>
|
|
254
|
<value>${content_read_timeout}</value>
|
|
255
|
</property>
|
|
256
|
<property>
|
|
257
|
<name>mapreduce.map.class</name>
|
|
258
|
<value>eu.dnetlib.iis.metadataextraction.${processing_mode}</value>
|
|
259
|
</property>
|
|
260
|
<property>
|
|
261
|
<name>mapred.reduce.tasks</name>
|
|
262
|
<value>0</value>
|
|
263
|
</property>
|
|
264
|
</configuration>
|
|
265
|
</map-reduce>
|
|
266
|
<ok to="end"/>
|
|
267
|
<error to="fail"/>
|
|
268
|
</action>
|
|
269
|
|
|
270
|
<kill name="fail">
|
|
271
|
<message>Unfortunately, the process failed -- error message: [${wf:errorMessage(wf:lastErrorNode())}]</message>
|
|
272
|
</kill>
|
|
273
|
<end name="end"/>
|
|
274
|
</workflow-app>
|