1
|
<RESOURCE_PROFILE>
|
2
|
<HEADER>
|
3
|
<RESOURCE_IDENTIFIER value="80f76a00-0eb3-4df0-a256-99f1df7b5fd8_V29ya2Zsb3dEU1Jlc291cmNlcy9Xb3JrZmxvd0RTUmVzb3VyY2VUeXBl"/>
|
4
|
<RESOURCE_TYPE value="WorkflowDSResourceType"/>
|
5
|
<RESOURCE_KIND value="WorkflowDSResources"/>
|
6
|
<RESOURCE_URI value=""/>
|
7
|
<DATE_OF_CREATION value="2006-05-04T18:13:51.0Z"/>
|
8
|
</HEADER>
|
9
|
<BODY>
|
10
|
<WORKFLOW_NAME menuSection="InfoSpace Deduplication">InfoSpace Deduplication</WORKFLOW_NAME>
|
11
|
<WORKFLOW_DESCRIPTION>OpenAIRE Deduplication</WORKFLOW_DESCRIPTION>
|
12
|
<WORKFLOW_INFO/>
|
13
|
<WORKFLOW_FAMILY>InfoSpace Deduplication</WORKFLOW_FAMILY>
|
14
|
<WORKFLOW_PRIORITY>35</WORKFLOW_PRIORITY>
|
15
|
<!-- <ADMIN_EMAIL>alessia.bardi@isti.cnr.it,claudio.atzori@isti.cnr.it</ADMIN_EMAIL> -->
|
16
|
<CONFIGURATION status="EXECUTABLE" start="MANUAL">
|
17
|
|
18
|
<PARAMETERS>
|
19
|
<PARAM name="table" description="HBase table to be used in this workflow" required="true" managedBy="user" type="string"></PARAM>
|
20
|
<PARAM name="cluster" description="Hadoop cluster logical name" required="true" managedBy="user" type="string" function="validValues(['DM','IIS'])"></PARAM>
|
21
|
<PARAM name="reuseRegionInfo" description="Reuse table splits" required="true" managedBy="user" type="boolean">true</PARAM>
|
22
|
|
23
|
<PARAM name="tableConf" description="HBase table configuration param name" required="true" managedBy="system" type="string">hbase.table.conf</PARAM>
|
24
|
<PARAM name="tableCols" description="HBase table schema param name" required="true" managedBy="system" type="string">hbase.table.cols</PARAM>
|
25
|
|
26
|
<PARAM name="db" description="relational db name" required="true" type="string" managedBy="user"/>
|
27
|
|
28
|
<PARAM name="sqlQueryORG" description="the SQL query" required="true" managedBy="system" type="string">/eu/dnetlib/msro/workflows/hbase/queryOrganizations.sql</PARAM>
|
29
|
<PARAM name="mappingORG" description="xslt mapping function" required="true" type="string" managedBy="user" function="obtainValues('dbmf2hbaseMappings', {})"/>
|
30
|
|
31
|
<PARAM name="webcrawlActionSet" description="action set name for WebCrawl" required="true" type="string" managedBy="system">iis-wos-entities</PARAM>
|
32
|
|
33
|
<PARAM name="mappingPublication" description="xslt mapping for publications" required="true" type="string" managedBy="user" function="obtainValues('oaf2hbaseMappings', {})"/>
|
34
|
<PARAM name="reusePublication" description="reuse publications on HDFS?" required="true" type="boolean" managedBy="user"/>
|
35
|
<PARAM name="hdfsPathPublication" description="hdfs path for publications" required="true" type="string" managedBy="user"/>
|
36
|
|
37
|
<PARAM name="dedupConfigSequence" description="dedup configuration orchestration name" required="true" type="string" function="obtainValues('dedupOrchestrations', {})" managedBy="user"/>
|
38
|
<PARAM name="minDistWorkDir" description="work directory for the minDist algorithm" required="true" type="string" managedBy="user">/tmp/dedup/prod/mindist</PARAM>
|
39
|
|
40
|
<PARAM name="mappingSimilarities" description="xslt mapping function for similarities" required="true" type="string" managedBy="user" function="obtainValues('dbmf2hbaseMappings', {})"/>
|
41
|
<PARAM name="mappingDissimilarities" description="xslt mapping function for dissimilarities" required="true" type="string" managedBy="user" function="obtainValues('dbmf2hbaseMappings', {})"/>
|
42
|
</PARAMETERS>
|
43
|
<WORKFLOW>
|
44
|
|
45
|
<NODE name="resetHbase" type="LaunchWorkflowTemplate" isStart="true">
|
46
|
<DESCRIPTION>Reset HBase table</DESCRIPTION>
|
47
|
<PARAMETERS>
|
48
|
<PARAM name="wfTemplateId" value="75345aba-c069-43f4-90aa-e13688d9845e_V29ya2Zsb3dUZW1wbGF0ZURTUmVzb3VyY2VzL1dvcmtmbG93VGVtcGxhdGVEU1Jlc291cmNlVHlwZQ=="/>
|
49
|
<PARAM name="wfTemplateParams">
|
50
|
<MAP>
|
51
|
<ENTRY key="table" ref="table"/>
|
52
|
<ENTRY key="cluster" ref="cluster"/>
|
53
|
<ENTRY key="reuseRegionInfo" ref="reuseRegionInfo"/>
|
54
|
</MAP>
|
55
|
</PARAM>
|
56
|
</PARAMETERS>
|
57
|
|
58
|
<ARCS>
|
59
|
<ARC to="select"/>
|
60
|
</ARCS>
|
61
|
</NODE>
|
62
|
|
63
|
<NODE name="select" type="SelectWfPathByDedupEntity">
|
64
|
<DESCRIPTION>Select the workflow path</DESCRIPTION>
|
65
|
<PARAMETERS>
|
66
|
<PARAM name="dedupConfigSequence" ref="dedupConfigSequence"/>
|
67
|
</PARAMETERS>
|
68
|
<ARCS>
|
69
|
<ARC name="organization" to="orgs2hbase"/>
|
70
|
<ARC name="result" to="webcrawl2hbase"/>
|
71
|
</ARCS>
|
72
|
</NODE>
|
73
|
|
74
|
<NODE name="orgs2hbase" type="LaunchWorkflowTemplate">
|
75
|
<DESCRIPTION>Organizations to HBase</DESCRIPTION>
|
76
|
<PARAMETERS>
|
77
|
<PARAM name="wfTemplateId" value="9bb04513-be1d-4d13-8e86-785cc0375635_V29ya2Zsb3dUZW1wbGF0ZURTUmVzb3VyY2VzL1dvcmtmbG93VGVtcGxhdGVEU1Jlc291cmNlVHlwZQ=="/>
|
78
|
<PARAM name="wfTemplateParams">
|
79
|
<MAP>
|
80
|
<ENTRY key="sqlQuery" ref="sqlQueryORG"/>
|
81
|
<ENTRY key="mapping" ref="mappingORG"/>
|
82
|
<ENTRY key="sqlPrepareQuery" value="/eu/dnetlib/msro/workflows/hbase/prepareQueryDefault.sql"/>
|
83
|
<ENTRY key="db" ref="db"/>
|
84
|
<ENTRY key="hbaseTable" ref="table"/>
|
85
|
<ENTRY key="cluster" ref="cluster"/>
|
86
|
</MAP>
|
87
|
</PARAM>
|
88
|
</PARAMETERS>
|
89
|
<ARCS>
|
90
|
<ARC to="deduplication"/>
|
91
|
</ARCS>
|
92
|
</NODE>
|
93
|
|
94
|
<NODE name="webcrawl2hbase" type="PromoteActionsHDFS">
|
95
|
<DESCRIPTION>Promote Actions</DESCRIPTION>
|
96
|
<PARAMETERS>
|
97
|
<!-- dedup-similarity-organization-simple,dedup-similarity-result-levenstein,iis-dataset-entities-main,iis-referenced-datasets-preprocessing,iis-dataset-entities-preprocessing,iis-document-citations,iis-document-classes,iis-referenced-projects-main,iis-document-similarities,iis-document-statistics,iis-extracted-metadata,iis-referenced-datasets-main,iis-referenced-projects-preprocessing,iis-researchinitiative,iis-wos-entities,iis-referenceextraction-pdb -->
|
98
|
<PARAM name="set" ref="webcrawlActionSet"/>
|
99
|
<PARAM name="tableName" ref="table"/>
|
100
|
</PARAMETERS>
|
101
|
<ARCS>
|
102
|
<ARC to="oaf2hbase"/>
|
103
|
</ARCS>
|
104
|
</NODE>
|
105
|
|
106
|
|
107
|
<NODE name="oaf2hbase" type="LaunchWorkflowTemplate">
|
108
|
<DESCRIPTION>OAF Publications to HBase</DESCRIPTION>
|
109
|
<PARAMETERS>
|
110
|
<PARAM name="wfTemplateId" value="5b05a65a-4eeb-4862-bc55-b35c7ec3baf0_V29ya2Zsb3dUZW1wbGF0ZURTUmVzb3VyY2VzL1dvcmtmbG93VGVtcGxhdGVEU1Jlc291cmNlVHlwZQ=="/>
|
111
|
<PARAM name="wfTemplateParams">
|
112
|
<MAP>
|
113
|
<ENTRY key="hbaseTable" ref="table"/>
|
114
|
<ENTRY key="cluster" ref="cluster"/>
|
115
|
<ENTRY key="mdFormat" value="OAF"/>
|
116
|
<ENTRY key="mapping" ref="mappingPublication"/>
|
117
|
<ENTRY key="reuseMdRecords" ref="reusePublication"/>
|
118
|
<ENTRY key="hdfsPath" ref="hdfsPathPublication"/>
|
119
|
</MAP>
|
120
|
</PARAM>
|
121
|
</PARAMETERS>
|
122
|
<ARCS>
|
123
|
<ARC to="deduplication"/>
|
124
|
</ARCS>
|
125
|
</NODE>
|
126
|
|
127
|
|
128
|
<NODE name="deduplication" type="LaunchWorkflowTemplate">
|
129
|
<DESCRIPTION>Deduplication</DESCRIPTION>
|
130
|
<PARAMETERS>
|
131
|
<PARAM name="wfTemplateId" value="01ed11e8-e874-4478-a8ac-83e63e9699e4_V29ya2Zsb3dUZW1wbGF0ZURTUmVzb3VyY2VzL1dvcmtmbG93VGVtcGxhdGVEU1Jlc291cmNlVHlwZQ=="/>
|
132
|
<PARAM name="wfTemplateParams">
|
133
|
<MAP>
|
134
|
<ENTRY key="hbaseTable" ref="table"/>
|
135
|
<ENTRY key="cluster" ref="cluster"/>
|
136
|
<ENTRY key="dedupConfigSequence" ref="dedupConfigSequence"/>
|
137
|
<ENTRY key="minDistWorkDir" ref="minDistWorkDir"/>
|
138
|
<ENTRY key="mappingSimilarities" ref="mappingSimilarities"/>
|
139
|
<ENTRY key="mappingDissimilarities" ref="mappingDissimilarities"/>
|
140
|
</MAP>
|
141
|
</PARAM>
|
142
|
</PARAMETERS>
|
143
|
<ARCS>
|
144
|
<ARC to="index"/>
|
145
|
</ARCS>
|
146
|
</NODE>
|
147
|
|
148
|
|
149
|
<NODE name="index" type="LaunchWorkflowTemplate">
|
150
|
<DESCRIPTION>Update dedup index</DESCRIPTION>
|
151
|
<PARAMETERS>
|
152
|
<PARAM name="wfTemplateId" value="a1b0b30c-7e19-46e2-888c-67c0b880b346_V29ya2Zsb3dUZW1wbGF0ZURTUmVzb3VyY2VzL1dvcmtmbG93VGVtcGxhdGVEU1Jlc291cmNlVHlwZQ=="/>
|
153
|
<PARAM name="wfTemplateParams">
|
154
|
<MAP>
|
155
|
<ENTRY key="hbaseTable" ref="table"/>
|
156
|
<ENTRY key="cluster" ref="cluster"/>
|
157
|
<ENTRY key="dedupConfigSequence" ref="dedupConfigSequence"/>
|
158
|
</MAP>
|
159
|
</PARAM>
|
160
|
</PARAMETERS>
|
161
|
<ARCS>
|
162
|
<ARC to="success"/>
|
163
|
</ARCS>
|
164
|
</NODE>
|
165
|
|
166
|
|
167
|
</WORKFLOW>
|
168
|
</CONFIGURATION>
|
169
|
<NOTIFICATIONS/>
|
170
|
<SCHEDULING enabled="false">
|
171
|
<CRON>29 5 22 ? * *</CRON>
|
172
|
<MININTERVAL>10080</MININTERVAL>
|
173
|
</SCHEDULING>
|
174
|
<STATUS/>
|
175
|
</BODY>
|
176
|
</RESOURCE_PROFILE>
|