Merge pull request 'Use SparkSQL in place of Hive for executing step16-createIndicatorsTables.sql of stats update wf' (#386) from stats_with_spark_sql into beta

Reviewed-on: D-Net/dnet-hadoop#386

dhp-workflows/dhp-stats-update/pom.xml

@@ -8,6 +8,11 @@
     <modelVersion>4.0.0</modelVersion>
     <artifactId>dhp-stats-update</artifactId>
     <dependencies>
+        <dependency>
+            <groupId>eu.dnetlib.dhp</groupId>
+            <artifactId>dhp-common</artifactId>
+            <version>${project.version}</version>
+        </dependency>
         <dependency>
             <groupId>org.apache.spark</groupId>
             <artifactId>spark-core_${scala.binary.version}</artifactId>

dhp-workflows/dhp-stats-update/src/main/resources/eu/dnetlib/dhp/oa/graph/stats/oozie_app/workflow.xml

@@ -64,6 +64,26 @@
             <name>hadoop_user_name</name>
             <description>user name of the wf owner</description>
         </property>
+
+        <property>
+            <name>sparkSqlWarehouseDir</name>
+        </property>
+        <!-- General oozie workflow properties -->
+        <property>
+            <name>sparkClusterOpts</name>
+            <value>--conf spark.network.timeout=600 --conf spark.extraListeners= --conf spark.sql.queryExecutionListeners= --conf spark.yarn.historyServer.address=http://iis-cdh5-test-m3.ocean.icm.edu.pl:18088 --conf spark.eventLog.dir=hdfs://nameservice1/user/spark/applicationHistory</value>
+            <description>spark cluster-wide options</description>
+        </property>
+        <property>
+            <name>sparkResourceOpts</name>
+            <value>--executor-memory=6G --conf spark.executor.memoryOverhead=4G --executor-cores=6 --driver-memory=8G --driver-cores=4</value>
+            <description>spark resource options</description>
+        </property>
+        <property>
+            <name>sparkApplicationOpts</name>
+            <value>--conf spark.sql.shuffle.partitions=3840</value>
+            <description>spark resource options</description>
+        </property>
     </parameters>
 
     <global>
@@ -82,6 +102,10 @@
                 <name>hive.mapjoin.followby.gby.localtask.max.memory.usage</name>
                 <value>0.80</value>
             </property>
+            <property>
+                <name>oozie.action.sharelib.for.spark</name>
+                <value>${oozieActionShareLibForSpark2}</value>
+            </property>
             <property>
                 <name>mapred.job.queue.name</name>
                 <value>analytics</value>
@@ -322,12 +346,23 @@
     </action>
 
     <action name="Step16-createIndicatorsTables">
-        <hive2 xmlns="uri:oozie:hive2-action:0.1">
+        <spark xmlns="uri:oozie:spark-action:0.2">
-            <jdbc-url>${hive_jdbc_url}</jdbc-url>
+            <master>yarn</master>
-            <script>scripts/step16-createIndicatorsTables.sql</script>
+            <mode>cluster</mode>
-            <param>stats_db_name=${stats_db_name}</param>
+            <name>Step16-createIndicatorsTables</name>
-            <param>external_stats_db_name=${external_stats_db_name}</param>
+            <class>eu.dnetlib.dhp.oozie.RunSQLSparkJob</class>
-        </hive2>
+            <jar>dhp-stats-update-${projectVersion}.jar</jar>
+            <spark-opts>
+                --conf spark.sql.warehouse.dir=${sparkSqlWarehouseDir}
+                ${sparkClusterOpts}
+                ${sparkResourceOpts}
+                ${sparkApplicationOpts}
+            </spark-opts>
+            <arg>--hiveMetastoreUris</arg><arg>${hive_metastore_uris}</arg>
+            <arg>--sql</arg><arg>eu/dnetlib/dhp/oa/graph/stats/oozie_app/scripts/step16-createIndicatorsTables.sql</arg>
+            <arg>--stats_db_name</arg><arg>${stats_db_name}</arg>
+            <arg>--external_stats_db_name</arg><arg>${external_stats_db_name}</arg>
+        </spark>
         <ok to="Step16_1-definitions"/>
         <error to="Kill"/>
     </action>