2020-03-24 17:40:58 +01:00
|
|
|
<workflow-app name="Duplicate Scan" xmlns="uri:oozie:workflow:0.5">
|
2020-03-18 14:27:49 +01:00
|
|
|
<parameters>
|
|
|
|
<property>
|
2020-03-19 15:01:07 +01:00
|
|
|
<name>graphBasePath</name>
|
2020-03-18 14:27:49 +01:00
|
|
|
<description>the raw graph base path</description>
|
|
|
|
</property>
|
2020-03-18 17:41:56 +01:00
|
|
|
<property>
|
|
|
|
<name>isLookUpUrl</name>
|
|
|
|
<description>the address of the lookUp service</description>
|
|
|
|
</property>
|
|
|
|
<property>
|
|
|
|
<name>actionSetId</name>
|
|
|
|
<description>id of the actionSet</description>
|
|
|
|
</property>
|
2021-02-10 11:51:50 +01:00
|
|
|
<property>
|
|
|
|
<name>actionSetIdOpenorgs</name>
|
|
|
|
<description>id of the actionSet for OpenOrgs dedup</description>
|
|
|
|
</property>
|
2020-03-19 15:01:07 +01:00
|
|
|
<property>
|
|
|
|
<name>workingPath</name>
|
|
|
|
<description>path for the working directory</description>
|
|
|
|
</property>
|
2021-09-20 16:21:47 +02:00
|
|
|
<property>
|
|
|
|
<name>whiteListPath</name>
|
|
|
|
<description>path for the whitelist of similarity relations</description>
|
|
|
|
</property>
|
2020-04-17 13:12:44 +02:00
|
|
|
<property>
|
|
|
|
<name>dedupGraphPath</name>
|
|
|
|
<description>path for the output graph</description>
|
|
|
|
</property>
|
2020-07-13 16:07:07 +02:00
|
|
|
<property>
|
|
|
|
<name>cutConnectedComponent</name>
|
|
|
|
<description>max number of elements in a connected component</description>
|
|
|
|
</property>
|
2020-03-18 14:27:49 +01:00
|
|
|
<property>
|
|
|
|
<name>sparkDriverMemory</name>
|
|
|
|
<description>memory for driver process</description>
|
|
|
|
</property>
|
|
|
|
<property>
|
|
|
|
<name>sparkExecutorMemory</name>
|
|
|
|
<description>memory for individual executor</description>
|
|
|
|
</property>
|
|
|
|
<property>
|
|
|
|
<name>sparkExecutorCores</name>
|
|
|
|
<description>number of cores used by single executor</description>
|
|
|
|
</property>
|
2020-04-16 12:11:37 +02:00
|
|
|
<property>
|
|
|
|
<name>oozieActionShareLibForSpark2</name>
|
|
|
|
<description>oozie action sharelib for spark 2.*</description>
|
|
|
|
</property>
|
|
|
|
<property>
|
|
|
|
<name>spark2ExtraListeners</name>
|
|
|
|
<value>com.cloudera.spark.lineage.NavigatorAppListener</value>
|
|
|
|
<description>spark 2.* extra listeners classname</description>
|
|
|
|
</property>
|
|
|
|
<property>
|
|
|
|
<name>spark2SqlQueryExecutionListeners</name>
|
|
|
|
<value>com.cloudera.spark.lineage.NavigatorQueryListener</value>
|
|
|
|
<description>spark 2.* sql query execution listeners classname</description>
|
|
|
|
</property>
|
|
|
|
<property>
|
|
|
|
<name>spark2YarnHistoryServerAddress</name>
|
|
|
|
<description>spark 2.* yarn history server address</description>
|
|
|
|
</property>
|
|
|
|
<property>
|
|
|
|
<name>spark2EventLogDir</name>
|
|
|
|
<description>spark 2.* event log dir location</description>
|
|
|
|
</property>
|
2020-03-18 14:27:49 +01:00
|
|
|
</parameters>
|
|
|
|
|
2020-03-20 16:48:14 +01:00
|
|
|
<global>
|
|
|
|
<job-tracker>${jobTracker}</job-tracker>
|
|
|
|
<name-node>${nameNode}</name-node>
|
|
|
|
<configuration>
|
|
|
|
<property>
|
|
|
|
<name>mapreduce.job.queuename</name>
|
|
|
|
<value>${queueName}</value>
|
|
|
|
</property>
|
|
|
|
<property>
|
|
|
|
<name>oozie.launcher.mapred.job.queue.name</name>
|
|
|
|
<value>${oozieLauncherQueueName}</value>
|
|
|
|
</property>
|
2020-04-16 12:11:37 +02:00
|
|
|
<property>
|
|
|
|
<name>oozie.action.sharelib.for.spark</name>
|
|
|
|
<value>${oozieActionShareLibForSpark2}</value>
|
|
|
|
</property>
|
2020-03-20 16:48:14 +01:00
|
|
|
</configuration>
|
|
|
|
</global>
|
|
|
|
|
2021-04-13 10:15:43 +02:00
|
|
|
<start to="resetWorkingPath"/>
|
2020-03-18 14:27:49 +01:00
|
|
|
|
|
|
|
<kill name="Kill">
|
|
|
|
<message>Action failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
|
|
|
</kill>
|
|
|
|
|
2020-05-06 11:27:44 +02:00
|
|
|
<action name="resetWorkingPath">
|
|
|
|
<fs>
|
|
|
|
<delete path="${workingPath}"/>
|
2021-03-19 16:57:40 +01:00
|
|
|
<delete path="${dedupGraphPath}"/>
|
2020-05-06 11:27:44 +02:00
|
|
|
</fs>
|
|
|
|
<ok to="CreateSimRel"/>
|
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
|
|
|
|
2021-03-29 10:07:12 +02:00
|
|
|
<action name="testOpenorgs">
|
|
|
|
<fs>
|
|
|
|
<delete path="${workingPath}/${actionSetIdOpenorgs}"/>
|
|
|
|
<delete path="${dedupGraphPath}"/>
|
|
|
|
</fs>
|
|
|
|
<ok to="CopyOpenorgsMergeRels"/>
|
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
2021-03-19 16:57:40 +01:00
|
|
|
|
2020-03-24 17:40:58 +01:00
|
|
|
<action name="CreateSimRel">
|
2020-03-18 14:27:49 +01:00
|
|
|
<spark xmlns="uri:oozie:spark-action:0.2">
|
2020-03-20 16:48:14 +01:00
|
|
|
<master>yarn</master>
|
2020-03-18 14:27:49 +01:00
|
|
|
<mode>cluster</mode>
|
|
|
|
<name>Create Similarity Relations</name>
|
2020-03-27 10:42:17 +01:00
|
|
|
<class>eu.dnetlib.dhp.oa.dedup.SparkCreateSimRels</class>
|
2020-03-25 15:57:09 +01:00
|
|
|
<jar>dhp-dedup-openaire-${projectVersion}.jar</jar>
|
2020-03-20 16:48:14 +01:00
|
|
|
<spark-opts>
|
2020-04-16 12:11:37 +02:00
|
|
|
--executor-memory=${sparkExecutorMemory}
|
|
|
|
--executor-cores=${sparkExecutorCores}
|
2020-03-20 16:48:14 +01:00
|
|
|
--driver-memory=${sparkDriverMemory}
|
2020-04-16 12:11:37 +02:00
|
|
|
--conf spark.extraListeners=${spark2ExtraListeners}
|
|
|
|
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
|
|
|
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
|
|
|
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
|
2023-05-17 10:16:22 +02:00
|
|
|
--conf spark.sql.shuffle.partitions=15000
|
2020-03-18 14:27:49 +01:00
|
|
|
</spark-opts>
|
2020-07-13 16:07:07 +02:00
|
|
|
<arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
|
|
|
|
<arg>--isLookUpUrl</arg><arg>${isLookUpUrl}</arg>
|
|
|
|
<arg>--actionSetId</arg><arg>${actionSetId}</arg>
|
|
|
|
<arg>--workingPath</arg><arg>${workingPath}</arg>
|
2023-05-17 10:16:22 +02:00
|
|
|
<arg>--numPartitions</arg><arg>15000</arg>
|
2020-03-18 14:27:49 +01:00
|
|
|
</spark>
|
2023-06-28 11:38:16 +02:00
|
|
|
<!-- <ok to="WhitelistSimRels"/>-->
|
|
|
|
<ok to="End"/>
|
2021-09-20 16:21:47 +02:00
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
|
|
|
|
|
|
|
<action name="WhitelistSimRels">
|
|
|
|
<spark xmlns="uri:oozie:spark-action:0.2">
|
|
|
|
<master>yarn</master>
|
|
|
|
<mode>cluster</mode>
|
|
|
|
<name>Add Whitelist Similarity Relations</name>
|
|
|
|
<class>eu.dnetlib.dhp.oa.dedup.SparkWhitelistSimRels</class>
|
|
|
|
<jar>dhp-dedup-openaire-${projectVersion}.jar</jar>
|
|
|
|
<spark-opts>
|
|
|
|
--executor-memory=${sparkExecutorMemory}
|
|
|
|
--executor-cores=${sparkExecutorCores}
|
|
|
|
--driver-memory=${sparkDriverMemory}
|
|
|
|
--conf spark.extraListeners=${spark2ExtraListeners}
|
|
|
|
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
|
|
|
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
|
|
|
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
|
2023-05-17 10:16:22 +02:00
|
|
|
--conf spark.sql.shuffle.partitions=15000
|
2021-09-20 16:21:47 +02:00
|
|
|
</spark-opts>
|
|
|
|
<arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
|
|
|
|
<arg>--isLookUpUrl</arg><arg>${isLookUpUrl}</arg>
|
|
|
|
<arg>--actionSetId</arg><arg>${actionSetId}</arg>
|
|
|
|
<arg>--workingPath</arg><arg>${workingPath}</arg>
|
|
|
|
<arg>--whiteListPath</arg><arg>${whiteListPath}</arg>
|
2023-05-17 10:16:22 +02:00
|
|
|
<arg>--numPartitions</arg><arg>15000</arg>
|
2021-09-20 16:21:47 +02:00
|
|
|
</spark>
|
2020-03-24 17:40:58 +01:00
|
|
|
<ok to="CreateMergeRel"/>
|
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
|
|
|
|
|
|
|
<action name="CreateMergeRel">
|
|
|
|
<spark xmlns="uri:oozie:spark-action:0.2">
|
2020-03-25 11:15:35 +01:00
|
|
|
<master>yarn</master>
|
2020-03-24 17:40:58 +01:00
|
|
|
<mode>cluster</mode>
|
|
|
|
<name>Create Merge Relations</name>
|
2020-04-16 12:13:26 +02:00
|
|
|
<class>eu.dnetlib.dhp.oa.dedup.SparkCreateMergeRels</class>
|
2020-03-25 15:57:09 +01:00
|
|
|
<jar>dhp-dedup-openaire-${projectVersion}.jar</jar>
|
2020-03-24 17:40:58 +01:00
|
|
|
<spark-opts>
|
2020-04-16 12:11:37 +02:00
|
|
|
--executor-memory=${sparkExecutorMemory}
|
|
|
|
--executor-cores=${sparkExecutorCores}
|
2020-03-24 17:40:58 +01:00
|
|
|
--driver-memory=${sparkDriverMemory}
|
2020-04-16 12:11:37 +02:00
|
|
|
--conf spark.extraListeners=${spark2ExtraListeners}
|
|
|
|
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
|
|
|
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
|
|
|
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
|
2023-05-17 10:16:22 +02:00
|
|
|
--conf spark.sql.shuffle.partitions=15000
|
2020-03-24 17:40:58 +01:00
|
|
|
</spark-opts>
|
2020-07-13 16:07:07 +02:00
|
|
|
<arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
|
|
|
|
<arg>--workingPath</arg><arg>${workingPath}</arg>
|
|
|
|
<arg>--isLookUpUrl</arg><arg>${isLookUpUrl}</arg>
|
|
|
|
<arg>--actionSetId</arg><arg>${actionSetId}</arg>
|
|
|
|
<arg>--cutConnectedComponent</arg><arg>${cutConnectedComponent}</arg>
|
2020-03-24 17:40:58 +01:00
|
|
|
</spark>
|
|
|
|
<ok to="CreateDedupRecord"/>
|
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
|
|
|
|
|
|
|
<action name="CreateDedupRecord">
|
|
|
|
<spark xmlns="uri:oozie:spark-action:0.2">
|
2020-03-25 11:15:35 +01:00
|
|
|
<master>yarn</master>
|
2020-03-24 17:40:58 +01:00
|
|
|
<mode>cluster</mode>
|
|
|
|
<name>Create Dedup Record</name>
|
2020-03-27 10:42:17 +01:00
|
|
|
<class>eu.dnetlib.dhp.oa.dedup.SparkCreateDedupRecord</class>
|
2020-03-25 15:57:09 +01:00
|
|
|
<jar>dhp-dedup-openaire-${projectVersion}.jar</jar>
|
2020-03-24 17:40:58 +01:00
|
|
|
<spark-opts>
|
2020-04-16 12:11:37 +02:00
|
|
|
--executor-memory=${sparkExecutorMemory}
|
|
|
|
--executor-cores=${sparkExecutorCores}
|
2020-03-24 17:40:58 +01:00
|
|
|
--driver-memory=${sparkDriverMemory}
|
2020-04-16 12:11:37 +02:00
|
|
|
--conf spark.extraListeners=${spark2ExtraListeners}
|
|
|
|
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
|
|
|
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
|
|
|
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
|
2023-05-17 10:16:22 +02:00
|
|
|
--conf spark.sql.shuffle.partitions=15000
|
2021-01-18 16:48:08 +01:00
|
|
|
</spark-opts>
|
|
|
|
<arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
|
|
|
|
<arg>--workingPath</arg><arg>${workingPath}</arg>
|
|
|
|
<arg>--isLookUpUrl</arg><arg>${isLookUpUrl}</arg>
|
|
|
|
<arg>--actionSetId</arg><arg>${actionSetId}</arg>
|
|
|
|
</spark>
|
2021-02-10 11:51:50 +01:00
|
|
|
<ok to="CopyOpenorgsMergeRels"/>
|
2021-01-18 16:48:08 +01:00
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
|
|
|
|
2021-04-06 14:31:00 +02:00
|
|
|
<!-- copy organization merge relations in the working dir (in the organization_mergerel dir)-->
|
2021-02-10 11:51:50 +01:00
|
|
|
<action name="CopyOpenorgsMergeRels">
|
2021-01-18 16:48:08 +01:00
|
|
|
<spark xmlns="uri:oozie:spark-action:0.2">
|
|
|
|
<master>yarn</master>
|
|
|
|
<mode>cluster</mode>
|
2021-03-19 16:57:40 +01:00
|
|
|
<name>Copy Openorgs Merge Relations</name>
|
2021-02-10 11:51:50 +01:00
|
|
|
<class>eu.dnetlib.dhp.oa.dedup.SparkCopyOpenorgsMergeRels</class>
|
2021-01-18 16:48:08 +01:00
|
|
|
<jar>dhp-dedup-openaire-${projectVersion}.jar</jar>
|
|
|
|
<spark-opts>
|
|
|
|
--executor-memory=${sparkExecutorMemory}
|
|
|
|
--executor-cores=${sparkExecutorCores}
|
|
|
|
--driver-memory=${sparkDriverMemory}
|
|
|
|
--conf spark.extraListeners=${spark2ExtraListeners}
|
|
|
|
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
|
|
|
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
|
|
|
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
|
2023-05-17 10:16:22 +02:00
|
|
|
--conf spark.sql.shuffle.partitions=15000
|
2021-01-18 16:48:08 +01:00
|
|
|
</spark-opts>
|
|
|
|
<arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
|
|
|
|
<arg>--workingPath</arg><arg>${workingPath}</arg>
|
2021-02-10 11:51:50 +01:00
|
|
|
<arg>--isLookUpUrl</arg><arg>${isLookUpUrl}</arg>
|
|
|
|
<arg>--actionSetId</arg><arg>${actionSetIdOpenorgs}</arg>
|
2023-05-17 10:16:22 +02:00
|
|
|
<arg>--numPartitions</arg><arg>15000</arg>
|
2021-01-18 16:48:08 +01:00
|
|
|
</spark>
|
2021-03-29 10:07:12 +02:00
|
|
|
<ok to="CreateOrgsDedupRecord"/>
|
2021-01-18 16:48:08 +01:00
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
|
|
|
|
2021-03-29 10:07:12 +02:00
|
|
|
<action name="CreateOrgsDedupRecord">
|
2021-01-18 16:48:08 +01:00
|
|
|
<spark xmlns="uri:oozie:spark-action:0.2">
|
|
|
|
<master>yarn</master>
|
|
|
|
<mode>cluster</mode>
|
2021-03-29 10:07:12 +02:00
|
|
|
<name>Create Organizations Dedup Records</name>
|
2021-04-06 14:31:00 +02:00
|
|
|
<class>eu.dnetlib.dhp.oa.dedup.SparkCreateOrgsDedupRecord</class>
|
2021-01-18 16:48:08 +01:00
|
|
|
<jar>dhp-dedup-openaire-${projectVersion}.jar</jar>
|
|
|
|
<spark-opts>
|
|
|
|
--executor-memory=${sparkExecutorMemory}
|
|
|
|
--executor-cores=${sparkExecutorCores}
|
|
|
|
--driver-memory=${sparkDriverMemory}
|
|
|
|
--conf spark.extraListeners=${spark2ExtraListeners}
|
|
|
|
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
|
|
|
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
|
|
|
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
|
|
|
|
--conf spark.sql.shuffle.partitions=3840
|
2020-03-24 17:40:58 +01:00
|
|
|
</spark-opts>
|
2020-07-13 16:07:07 +02:00
|
|
|
<arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
|
|
|
|
<arg>--workingPath</arg><arg>${workingPath}</arg>
|
2021-03-29 10:07:12 +02:00
|
|
|
<arg>--isLookUpUrl</arg><arg>${isLookUpUrl}</arg>
|
2021-02-10 11:51:50 +01:00
|
|
|
<arg>--actionSetId</arg><arg>${actionSetIdOpenorgs}</arg>
|
2020-03-24 17:40:58 +01:00
|
|
|
</spark>
|
2020-04-17 13:12:44 +02:00
|
|
|
<ok to="UpdateEntity"/>
|
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
|
|
|
|
|
|
|
<action name="UpdateEntity">
|
|
|
|
<spark xmlns="uri:oozie:spark-action:0.2">
|
|
|
|
<master>yarn</master>
|
|
|
|
<mode>cluster</mode>
|
|
|
|
<name>Update Entity</name>
|
|
|
|
<class>eu.dnetlib.dhp.oa.dedup.SparkUpdateEntity</class>
|
|
|
|
<jar>dhp-dedup-openaire-${projectVersion}.jar</jar>
|
|
|
|
<spark-opts>
|
|
|
|
--executor-memory=${sparkExecutorMemory}
|
|
|
|
--executor-cores=${sparkExecutorCores}
|
|
|
|
--driver-memory=${sparkDriverMemory}
|
|
|
|
--conf spark.extraListeners=${spark2ExtraListeners}
|
|
|
|
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
|
|
|
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
|
|
|
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
|
2023-05-17 10:16:22 +02:00
|
|
|
--conf spark.sql.shuffle.partitions=10000
|
2020-04-17 13:12:44 +02:00
|
|
|
</spark-opts>
|
2020-07-13 16:07:07 +02:00
|
|
|
<arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
|
|
|
|
<arg>--workingPath</arg><arg>${workingPath}</arg>
|
|
|
|
<arg>--dedupGraphPath</arg><arg>${dedupGraphPath}</arg>
|
2020-04-17 13:12:44 +02:00
|
|
|
</spark>
|
|
|
|
<ok to="copyRelations"/>
|
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
|
|
|
|
2021-02-10 11:51:50 +01:00
|
|
|
<!-- copy all relations without openorgs relations to the dedupgraph -->
|
2020-04-17 13:12:44 +02:00
|
|
|
<action name="copyRelations">
|
2021-02-10 11:51:50 +01:00
|
|
|
<spark xmlns="uri:oozie:spark-action:0.2">
|
|
|
|
<master>yarn</master>
|
|
|
|
<mode>cluster</mode>
|
2021-03-19 16:57:40 +01:00
|
|
|
<name>Copy Non-Openorgs Relations</name>
|
2021-02-10 11:51:50 +01:00
|
|
|
<class>eu.dnetlib.dhp.oa.dedup.SparkCopyRelationsNoOpenorgs</class>
|
|
|
|
<jar>dhp-dedup-openaire-${projectVersion}.jar</jar>
|
|
|
|
<spark-opts>
|
|
|
|
--executor-memory=${sparkExecutorMemory}
|
|
|
|
--executor-cores=${sparkExecutorCores}
|
|
|
|
--driver-memory=${sparkDriverMemory}
|
|
|
|
--conf spark.extraListeners=${spark2ExtraListeners}
|
|
|
|
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
|
|
|
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
|
|
|
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
|
2023-05-17 10:16:22 +02:00
|
|
|
--conf spark.sql.shuffle.partitions=10000
|
2021-02-10 11:51:50 +01:00
|
|
|
</spark-opts>
|
|
|
|
<arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
|
|
|
|
<arg>--workingPath</arg><arg>${workingPath}</arg>
|
|
|
|
<arg>--dedupGraphPath</arg><arg>${dedupGraphPath}</arg>
|
|
|
|
</spark>
|
2020-03-18 14:27:49 +01:00
|
|
|
<ok to="End"/>
|
|
|
|
<error to="Kill"/>
|
|
|
|
</action>
|
|
|
|
|
|
|
|
<end name="End"/>
|
|
|
|
</workflow-app>
|