|
1
|
<workflow-app xmlns="uri:oozie:workflow:0.4" name="mainworkflows_metadataextraction_cached_by_checksum">
|
|
2
|
|
|
3
|
<parameters>
|
|
4
|
<property>
|
|
5
|
<name>input</name>
|
|
6
|
<description>input document content directory</description>
|
|
7
|
</property>
|
|
8
|
<property>
|
|
9
|
<name>output_root</name>
|
|
10
|
<description>metadata extraction output directory</description>
|
|
11
|
</property>
|
|
12
|
<property>
|
|
13
|
<name>excluded_ids</name>
|
|
14
|
<value>$UNDEFINED$</value>
|
|
15
|
<description>list of content identifiers excluded from metadataextraction processing</description>
|
|
16
|
</property>
|
|
17
|
<property>
|
|
18
|
<name>max_file_size_mb</name>
|
|
19
|
<value>500</value>
|
|
20
|
<description>maximum allowed file size in Megabytes</description>
|
|
21
|
</property>
|
|
22
|
<property>
|
|
23
|
<name>content_connection_timeout</name>
|
|
24
|
<value>60000</value>
|
|
25
|
<description>streaming content connection timeout</description>
|
|
26
|
</property>
|
|
27
|
<property>
|
|
28
|
<name>content_read_timeout</name>
|
|
29
|
<value>60000</value>
|
|
30
|
<description>streaming content read timeout</description>
|
|
31
|
</property>
|
|
32
|
<property>
|
|
33
|
<name>zk_session_timeout</name>
|
|
34
|
<value>60000</value>
|
|
35
|
<description>zookeeper session timeout when handling locks</description>
|
|
36
|
</property>
|
|
37
|
<property>
|
|
38
|
<name>default_cache_location</name>
|
|
39
|
<value>/cache/metadataextraction</value>
|
|
40
|
<description>default cache location stored in HDFS</description>
|
|
41
|
</property>
|
|
42
|
<property>
|
|
43
|
<name>mapred_max_split_size</name>
|
|
44
|
<value>50000</value>
|
|
45
|
<description>maximum input data split size, required by streaming version reading DocumentContentUrl to split input data into more chunks</description>
|
|
46
|
</property>
|
|
47
|
<property>
|
|
48
|
<name>output_name_meta</name>
|
|
49
|
<value>meta</value>
|
|
50
|
<description>metadata output subdirectory name</description>
|
|
51
|
</property>
|
|
52
|
<property>
|
|
53
|
<name>output_name_plaintext</name>
|
|
54
|
<value>plaintext</value>
|
|
55
|
<description>plaintext output subdirectory name</description>
|
|
56
|
</property>
|
|
57
|
<property>
|
|
58
|
<name>output_name_fault</name>
|
|
59
|
<value>fault</value>
|
|
60
|
<description>fault output subdirectory name</description>
|
|
61
|
</property>
|
|
62
|
</parameters>
|
|
63
|
|
|
64
|
<global>
|
|
65
|
<job-tracker>${jobTracker}</job-tracker>
|
|
66
|
<name-node>${nameNode}</name-node>
|
|
67
|
<configuration>
|
|
68
|
<property>
|
|
69
|
<name>mapred.job.queue.name</name>
|
|
70
|
<value>${queueName}</value>
|
|
71
|
</property>
|
|
72
|
</configuration>
|
|
73
|
</global>
|
|
74
|
|
|
75
|
<start to="preprocessing" />
|
|
76
|
|
|
77
|
<action name="preprocessing">
|
|
78
|
<sub-workflow>
|
|
79
|
<app-path>${wf:appPath()}/transformers_metadataextraction_checksum_preprocessing</app-path>
|
|
80
|
<propagate-configuration/>
|
|
81
|
<configuration>
|
|
82
|
<property>
|
|
83
|
<name>workingDir</name>
|
|
84
|
<value>${workingDir}/preprocessing/working_dir</value>
|
|
85
|
</property>
|
|
86
|
<property>
|
|
87
|
<name>input</name>
|
|
88
|
<value>${input}</value>
|
|
89
|
</property>
|
|
90
|
<property>
|
|
91
|
<name>output</name>
|
|
92
|
<value>${workingDir}/preprocessing/output</value>
|
|
93
|
</property>
|
|
94
|
</configuration>
|
|
95
|
</sub-workflow>
|
|
96
|
<ok to="mainworkflows_metadataextraction_cached"/>
|
|
97
|
<error to="fail" />
|
|
98
|
</action>
|
|
99
|
|
|
100
|
<action name="mainworkflows_metadataextraction_cached">
|
|
101
|
<sub-workflow>
|
|
102
|
<app-path>${wf:appPath()}/mainworkflows_metadataextraction_cached</app-path>
|
|
103
|
<propagate-configuration/>
|
|
104
|
<configuration>
|
|
105
|
<property>
|
|
106
|
<name>workingDir</name>
|
|
107
|
<value>${workingDir}/mainworkflows_metadataextraction_cached/working_dir</value>
|
|
108
|
</property>
|
|
109
|
<property>
|
|
110
|
<!-- checksum identified input -->
|
|
111
|
<name>input</name>
|
|
112
|
<value>${workingDir}/preprocessing/output</value>
|
|
113
|
</property>
|
|
114
|
<property>
|
|
115
|
<name>output_root</name>
|
|
116
|
<value>${workingDir}/mainworkflows_metadataextraction_cached/out</value>
|
|
117
|
</property>
|
|
118
|
<property>
|
|
119
|
<name>excluded_ids</name>
|
|
120
|
<value>${excluded_ids}</value>
|
|
121
|
</property>
|
|
122
|
<property>
|
|
123
|
<name>max_file_size_mb</name>
|
|
124
|
<value>${max_file_size_mb}</value>
|
|
125
|
</property>
|
|
126
|
<property>
|
|
127
|
<name>content_connection_timeout</name>
|
|
128
|
<value>${content_connection_timeout}</value>
|
|
129
|
</property>
|
|
130
|
<property>
|
|
131
|
<name>content_read_timeout</name>
|
|
132
|
<value>${content_read_timeout}</value>
|
|
133
|
</property>
|
|
134
|
<property>
|
|
135
|
<name>zk_session_timeout</name>
|
|
136
|
<value>${zk_session_timeout}</value>
|
|
137
|
</property>
|
|
138
|
<property>
|
|
139
|
<name>default_cache_location</name>
|
|
140
|
<value>${default_cache_location}</value>
|
|
141
|
</property>
|
|
142
|
<property>
|
|
143
|
<name>mapred_max_split_size</name>
|
|
144
|
<value>${mapred_max_split_size}</value>
|
|
145
|
</property>
|
|
146
|
<property>
|
|
147
|
<name>output_name_meta</name>
|
|
148
|
<value>${output_name_meta}</value>
|
|
149
|
</property>
|
|
150
|
<property>
|
|
151
|
<name>output_name_plaintext</name>
|
|
152
|
<value>${output_name_plaintext}</value>
|
|
153
|
</property>
|
|
154
|
<property>
|
|
155
|
<name>output_name_fault</name>
|
|
156
|
<value>${output_name_fault}</value>
|
|
157
|
</property>
|
|
158
|
</configuration>
|
|
159
|
</sub-workflow>
|
|
160
|
<ok to="postprocessing-forking"/>
|
|
161
|
<error to="fail" />
|
|
162
|
</action>
|
|
163
|
|
|
164
|
<fork name="postprocessing-forking">
|
|
165
|
<path start="postprocessing-meta"/>
|
|
166
|
<path start="postprocessing-text"/>
|
|
167
|
<path start="postprocessing-fault"/>
|
|
168
|
</fork>
|
|
169
|
|
|
170
|
<action name="postprocessing-meta">
|
|
171
|
<sub-workflow>
|
|
172
|
<app-path>${wf:appPath()}/transformers_metadataextraction_checksum_postprocessing_meta</app-path>
|
|
173
|
<propagate-configuration/>
|
|
174
|
<configuration>
|
|
175
|
<property>
|
|
176
|
<name>workingDir</name>
|
|
177
|
<value>${workingDir}/postprocessing_meta/working_dir</value>
|
|
178
|
</property>
|
|
179
|
<property>
|
|
180
|
<name>input_document_content_url</name>
|
|
181
|
<value>${input}</value>
|
|
182
|
</property>
|
|
183
|
<property>
|
|
184
|
<name>input_extracted_document_metadata</name>
|
|
185
|
<value>${workingDir}/mainworkflows_metadataextraction_cached/out/${output_name_meta}</value>
|
|
186
|
</property>
|
|
187
|
<property>
|
|
188
|
<name>output</name>
|
|
189
|
<value>${output_root}/${output_name_meta}</value>
|
|
190
|
</property>
|
|
191
|
</configuration>
|
|
192
|
</sub-workflow>
|
|
193
|
<ok to="postprocessing-joining"/>
|
|
194
|
<error to="fail" />
|
|
195
|
</action>
|
|
196
|
|
|
197
|
<action name="postprocessing-text">
|
|
198
|
<sub-workflow>
|
|
199
|
<app-path>${wf:appPath()}/transformers_metadataextraction_checksum_postprocessing_text</app-path>
|
|
200
|
<propagate-configuration/>
|
|
201
|
<configuration>
|
|
202
|
<property>
|
|
203
|
<name>workingDir</name>
|
|
204
|
<value>${workingDir}/postprocessing_text/working_dir</value>
|
|
205
|
</property>
|
|
206
|
<property>
|
|
207
|
<name>input_document_content_url</name>
|
|
208
|
<value>${input}</value>
|
|
209
|
</property>
|
|
210
|
<property>
|
|
211
|
<name>input_document_text</name>
|
|
212
|
<value>${workingDir}/mainworkflows_metadataextraction_cached/out/${output_name_plaintext}</value>
|
|
213
|
</property>
|
|
214
|
<property>
|
|
215
|
<name>output</name>
|
|
216
|
<value>${output_root}/${output_name_plaintext}</value>
|
|
217
|
</property>
|
|
218
|
</configuration>
|
|
219
|
</sub-workflow>
|
|
220
|
<ok to="postprocessing-joining"/>
|
|
221
|
<error to="fail" />
|
|
222
|
</action>
|
|
223
|
|
|
224
|
<action name="postprocessing-fault">
|
|
225
|
<distcp xmlns="uri:oozie:distcp-action:0.1">
|
|
226
|
<job-tracker>${jobTracker}</job-tracker>
|
|
227
|
<name-node>${nameNode}</name-node>
|
|
228
|
<prepare>
|
|
229
|
<delete path="${nameNode}${output_root}/${output_name_fault}" />
|
|
230
|
</prepare>
|
|
231
|
<arg>${nameNode}${workingDir}/mainworkflows_metadataextraction_cached/out/${output_name_fault}</arg>
|
|
232
|
<arg>${nameNode}${output_root}/${output_name_fault}</arg>
|
|
233
|
</distcp>
|
|
234
|
<ok to="postprocessing-joining"/>
|
|
235
|
<error to="fail"/>
|
|
236
|
</action>
|
|
237
|
|
|
238
|
<join name="postprocessing-joining" to="end"/>
|
|
239
|
|
|
240
|
<kill name="fail">
|
|
241
|
<message>Unfortunately, the process failed -- error message:
|
|
242
|
[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
|
243
|
</kill>
|
|
244
|
<end name="end" />
|
|
245
|
</workflow-app>
|