648 lines
28 KiB
XML
648 lines
28 KiB
XML
<workflow-app xmlns="uri:oozie:workflow:0.5" name="ranking-wf">
|
||
|
||
<!-- start using a decision node, so as to determine from which point onwards a job will continue -->
|
||
<!-- <start to="get-doi-synonyms" /> -->
|
||
<start to="entry-point-decision" />
|
||
|
||
<decision name="entry-point-decision">
|
||
<switch>
|
||
<!-- The default will be set as the normal start, a.k.a. get-doi-synonyms -->
|
||
<!-- If any different condition is set, go to the corresponding start -->
|
||
<case to="non-iterative-rankings">${resume eq "rankings-start"}</case>
|
||
<case to="spark-impulse">${resume eq "impulse"}</case>
|
||
<case to="iterative-rankings">${resume eq "rankings-iterative"}</case>
|
||
<case to="get-file-names">${resume eq "format-results"}</case>
|
||
<case to="map-openaire-to-doi">${resume eq "map-ids"}</case>
|
||
<case to="map-scores-to-dois">${resume eq "map-scores"}</case>
|
||
<case to="create-openaire-ranking-graph">${resume eq "start"}</case>
|
||
<!-- TODO: add action set creation here -->
|
||
<default to="create-openaire-ranking-graph" />
|
||
</switch>
|
||
</decision>
|
||
|
||
<!-- Script here written by Serafeim: maps openaire ids to their synonyms -->
|
||
<action name="create-openaire-ranking-graph">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
<!-- Delete previously created doi synonym folder -->
|
||
<prepare>
|
||
<delete path="${synonymFolder}"/>
|
||
</prepare>
|
||
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Openaire Ranking Graph Creation</name>
|
||
<!-- Script name goes here -->
|
||
<jar>create_openaire_ranking_graph.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 20G --executor-cores 4 --driver-memory 20G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<!-- The openaire graph data from which to read relations and objects -->
|
||
<arg>${openaireDataInput}</arg>
|
||
<!-- Year for filtering entries w/ larger values / empty -->
|
||
<arg>${currentYear}</arg>
|
||
<!-- number of partitions to be used on joins -->
|
||
<arg>7680</arg>
|
||
<!-- The output of the graph should be the openaire input graph for ranking-->
|
||
<arg>${openaireGraphInputPath}</arg>
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/create_openaire_ranking_graph.py#create_openaire_ranking_graph.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="non-iterative-rankings" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="openaire-graph-error" />
|
||
|
||
</action>
|
||
|
||
<!-- Citation Count and RAM are calculated in parallel-->
|
||
<!-- Impulse Requires resources and will be run after-->
|
||
<fork name="non-iterative-rankings">
|
||
<path start="spark-cc"/>
|
||
<!-- <path start="spark-impulse"/> -->
|
||
<path start="spark-ram"/>
|
||
</fork>
|
||
|
||
<!-- CC here -->
|
||
<action name="spark-cc">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Spark CC</name>
|
||
<!-- Script name goes here -->
|
||
<jar>CC.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 18G --executor-cores 4 --driver-memory 10G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<arg>${openaireGraphInputPath}</arg>
|
||
<!-- number of partitions to be used on joins -->
|
||
<arg>7680</arg>
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/CC.py#CC.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="join-non-iterative-rankings" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="cc-fail" />
|
||
|
||
</action>
|
||
|
||
<!-- IMPULSE here -->
|
||
<action name="spark-ram">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Spark RAM</name>
|
||
<!-- Script name goes here -->
|
||
<jar>TAR.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 18G --executor-cores 4 --driver-memory 10G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<arg>${openaireGraphInputPath}</arg>
|
||
<arg>${ramGamma}</arg>
|
||
<arg>${currentYear}</arg>
|
||
<arg>RAM</arg>
|
||
<!-- number of partitions to be used on joins -->
|
||
<arg>7680</arg>
|
||
<arg>${γιτ α}</arg>
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/TAR.py#TAR.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="join-non-iterative-rankings" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="ram-fail" />
|
||
|
||
</action>
|
||
|
||
<!-- JOIN NON-ITERATIVE METHODS AND THEN CONTINUE TO ITERATIVE ONES -->
|
||
<join name="join-non-iterative-rankings" to="spark-impulse"/>
|
||
|
||
<!-- IMPULSE here -->
|
||
<action name="spark-impulse">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Spark Impulse</name>
|
||
<!-- Script name goes here -->
|
||
<jar>CC.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 18G --executor-cores 4 --driver-memory 10G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<arg>${openaireGraphInputPath}</arg>
|
||
<!-- number of partitions to be used on joins -->
|
||
<arg>7680</arg>
|
||
<arg>3</arg>
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/CC.py#CC.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="iterative-rankings" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="impulse-fail" />
|
||
|
||
</action>
|
||
|
||
<fork name="iterative-rankings">
|
||
<path start="spark-pagerank"/>
|
||
<path start="spark-attrank"/>
|
||
</fork>
|
||
|
||
<!-- PAGERANK here -->
|
||
<action name="spark-pagerank">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
|
||
<!-- we could add map-reduce configs here, but I don't know if we need them -->
|
||
<!-- This is the type of master-client configuration for running spark -->
|
||
<!-- <master>yarn-client</master> -->
|
||
<!-- Reference says: The master element indicates the url of the Spark Master. Ex: spark://host:port, mesos://host:port, yarn-cluster, yarn-master, or local. -->
|
||
<!-- <master>local[*]</master> -->
|
||
<!-- Reference says: The mode element if present indicates the mode of spark, where to run spark driver program. Ex: client,cluster. | In my case I always have a client -->
|
||
<!-- <mode>client</mode> -->
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Spark Pagerank</name>
|
||
<!-- Script name goes here -->
|
||
<jar>PageRank.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 18G --executor-cores 4 --driver-memory 10G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<arg>${openaireGraphInputPath}</arg>
|
||
<arg>${pageRankAlpha}</arg>
|
||
<arg>${convergenceError}</arg>
|
||
<arg>${checkpointDir}</arg>
|
||
<!-- number of partitions to be used on joins -->
|
||
<arg>7680</arg>
|
||
<arg>dfs</arg>
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/PageRank.py#PageRank.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="join-iterative-rankings" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="pagerank-fail" />
|
||
|
||
</action>
|
||
|
||
<!-- ATTRANK here -->
|
||
<action name="spark-attrank">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Spark AttRank</name>
|
||
<!-- Script name goes here -->
|
||
<jar>AttRank.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 18G --executor-cores 4 --driver-memory 10G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<arg>${openaireGraphInputPath}</arg>
|
||
<arg>${attrankAlpha}</arg>
|
||
<arg>${attrankBeta}</arg>
|
||
<arg>${attrankGamma}</arg>
|
||
<arg>${attrankRho}</arg>
|
||
<arg>${currentYear}</arg>
|
||
<arg>${attrankStartYear}</arg>
|
||
<arg>${convergenceError}</arg>
|
||
<arg>${checkpointDir}</arg>
|
||
<!-- number of partitions to be used on joins -->
|
||
<arg>7680</arg>
|
||
<arg>dfs</arg>
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/AttRank.py#AttRank.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="join-iterative-rankings" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="attrank-fail" />
|
||
|
||
</action>
|
||
|
||
<!-- JOIN ITERATIVE METHODS AND THEN END -->
|
||
<join name="join-iterative-rankings" to="get-file-names"/>
|
||
|
||
|
||
<!-- This will be a shell action that will output key-value pairs for output files -->
|
||
<action name="get-file-names">
|
||
<!-- This is required as a tag for shell jobs -->
|
||
<shell xmlns="uri:oozie:shell-action:0.3">
|
||
<!-- Same for all -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs -->
|
||
<name-node>${nameNode}</name-node>
|
||
|
||
<!-- Exec is needed foor shell comands - points to type of shell command -->
|
||
<exec>/usr/bin/bash</exec>
|
||
<!-- name of script to run -->
|
||
<argument>get_ranking_files.sh</argument>
|
||
<!-- We only pass the directory where we expect to find the rankings -->
|
||
<argument>/${workflowDataDir}</argument>
|
||
|
||
<!-- the name of the file run -->
|
||
<file>${wfAppPath}/get_ranking_files.sh#get_ranking_files.sh</file>
|
||
<!-- Get the output in order to be usable by following actions -->
|
||
<capture-output/>
|
||
</shell>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="format-result-files" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="filename-getting-error" />
|
||
|
||
</action>
|
||
|
||
|
||
<!-- Now we will run in parallel the formatting of ranking files for BiP! DB and openaire (json files) -->
|
||
<fork name="format-result-files">
|
||
<path start="format-bip-files"/>
|
||
<path start="format-json-files"/>
|
||
</fork>
|
||
|
||
|
||
<!-- Format json files -->
|
||
<!-- Two parts: a) format files b) make the file endings .json.gz -->
|
||
<action name="format-json-files">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Format Ranking Results JSON</name>
|
||
<!-- Script name goes here -->
|
||
<jar>format_ranking_results.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 10G --executor-cores 4 --driver-memory 10G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<arg>json</arg>
|
||
<!-- Input files must be identified dynamically -->
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['pr_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['attrank_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['cc_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['impulse_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['ram_file']}</arg>
|
||
<!-- Num partitions -->
|
||
<arg>7680</arg>
|
||
<!-- Type of data to be produced [bip (dois) / openaire (openaire-ids) ] -->
|
||
<arg>openaire</arg>
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/format_ranking_results.py#format_ranking_results.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="join-file-formatting" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="json-formatting-fail" />
|
||
</action>
|
||
|
||
<!-- This is the second line of parallel workflow execution where we create the BiP! DB files -->
|
||
<action name="format-bip-files">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Format Ranking Results BiP! DB</name>
|
||
<!-- Script name goes here -->
|
||
<jar>format_ranking_results.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 10G --executor-cores 4 --driver-memory 10G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<arg>zenodo</arg>
|
||
<!-- Input files must be identified dynamically -->
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['pr_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['attrank_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['cc_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['impulse_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['ram_file']}</arg>
|
||
<!-- Num partitions -->
|
||
<arg>7680</arg>
|
||
<!-- Type of data to be produced [bip (dois) / openaire (openaire-ids) ] -->
|
||
<arg>openaire</arg>
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/format_ranking_results.py#format_ranking_results.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="join-file-formatting" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="bip-formatting-fail" />
|
||
</action>
|
||
|
||
<!-- Finish formatting data and end -->
|
||
<join name="join-file-formatting" to="map-openaire-to-doi"/>
|
||
|
||
<!-- Script here written by Serafeim: maps openaire ids to their synonyms -->
|
||
<action name="map-openaire-to-doi">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
<!-- Delete previously created doi synonym folder -->
|
||
<prepare>
|
||
<delete path="${synonymFolder}"/>
|
||
</prepare>
|
||
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Openaire-DOI synonym collection</name>
|
||
<!-- Script name goes here -->
|
||
<jar>map_openaire_ids_to_dois.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 18G --executor-cores 4 --driver-memory 15G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<arg>${openaireDataInput}</arg>
|
||
<!-- number of partitions to be used on joins -->
|
||
<arg>${synonymFolder}</arg>
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/map_openaire_ids_to_dois.py#map_openaire_ids_to_dois.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="map-scores-to-dois" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="synonym-collection-fail" />
|
||
|
||
</action>
|
||
|
||
|
||
<!-- Script here written by Serafeim: maps openaire ids to their synonyms -->
|
||
<action name="map-scores-to-dois">
|
||
<!-- This is required as a tag for spark jobs, regardless of programming language -->
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<!-- Is this yarn? Probably the answers are at the link serafeim sent me -->
|
||
<job-tracker>${jobTracker}</job-tracker>
|
||
<!-- This should give the machine/root of the hdfs, serafeim has provided a link with the required job properties -->
|
||
<name-node>${nameNode}</name-node>
|
||
|
||
|
||
<!-- using configs from an example on openaire -->
|
||
<master>yarn-cluster</master>
|
||
<mode>cluster</mode>
|
||
|
||
<!-- This is the name of our job -->
|
||
<name>Mapping Openaire Scores to DOIs</name>
|
||
<!-- Script name goes here -->
|
||
<jar>map_scores_to_dois.py</jar>
|
||
<!-- spark configuration options: I've taken most of them from an example from dhp workflows / Master value stolen from sandro -->
|
||
<spark-opts>--executor-memory 18G --executor-cores 4 --driver-memory 15G
|
||
--master yarn
|
||
--deploy-mode cluster
|
||
--conf spark.sql.shuffle.partitions=7680
|
||
--conf spark.extraListeners=${spark2ExtraListeners}
|
||
--conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
|
||
--conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
|
||
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}</spark-opts>
|
||
<!-- Script arguments here -->
|
||
<arg>${synonymFolder}</arg>
|
||
<!-- Number of partitions -->
|
||
<arg>7680</arg>
|
||
<!-- The remaining input are the ranking files fproduced for bip db-->
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['pr_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['attrank_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['cc_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['impulse_file']}</arg>
|
||
<arg>${nameNode}/${workflowDataDir}/${wf:actionData('get-file-names')['ram_file']}</arg>
|
||
|
||
<!-- This needs to point to the file on the hdfs i think -->
|
||
<file>${wfAppPath}/map_scores_to_dois.py#map_scores_to_dois.py</file>
|
||
</spark>
|
||
|
||
<!-- Do this after finishing okay -->
|
||
<ok to="deleteOutputPathForActionSet" />
|
||
<!-- Go there if we have an error -->
|
||
<error to="map-scores-fail" />
|
||
|
||
</action>
|
||
|
||
<action name="deleteOutputPathForActionSet">
|
||
<fs>
|
||
<delete path="${actionSetOutputPath}"/>
|
||
<mkdir path="${actionSetOutputPath}"/>
|
||
<!--
|
||
<delete path="${workingDir}"/>
|
||
<mkdir path="${workingDir}"/>
|
||
-->
|
||
</fs>
|
||
<ok to="createActionSet"/>
|
||
<error to="actionset-delete-fail"/>
|
||
</action>
|
||
|
||
|
||
<action name="createActionSet">
|
||
<spark xmlns="uri:oozie:spark-action:0.2">
|
||
<master>yarn</master>
|
||
<mode>cluster</mode>
|
||
<name>Produces the atomic action with the bip finder scores for publications</name>
|
||
<class>eu.dnetlib.dhp.actionmanager.bipfinder.SparkAtomicActionScoreJob</class>
|
||
<jar>dhp-aggregation-${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.warehouse.dir=${sparkSqlWarehouseDir}
|
||
</spark-opts>
|
||
<arg>--inputPath</arg><arg>${bipScorePath}</arg>
|
||
<arg>--outputPath</arg><arg>${actionSetOutputPath}</arg>
|
||
</spark>
|
||
<ok to="end"/>
|
||
<error to="actionset-creation-fail"/>
|
||
</action>
|
||
|
||
|
||
|
||
<!-- TODO: end the workflow-->
|
||
|
||
<!-- Define ending node -->
|
||
<end name="end" />
|
||
|
||
<!-- Definitions of failure messages -->
|
||
<kill name="pagerank-fail">
|
||
<message>PageRank failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
<kill name="attrank-fail">
|
||
<message>AttRank failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
<kill name="cc-fail">
|
||
<message>CC failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
<kill name="impulse-fail">
|
||
<message>Impulse failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
<kill name="ram-fail">
|
||
<message>RAM failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
<kill name="openaire-graph-error">
|
||
<message>Creation of openaire-graph failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
<kill name="synonym-collection-fail">
|
||
<message>Synonym collection failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
<kill name="map-scores-fail">
|
||
<message>Mapping scores to DOIs failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
<kill name="actionset-delete-fail">
|
||
<message>Deleting output path for actionsets failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
<kill name="actionset-creation-fail">
|
||
<message>ActionSet creation failed, error message[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
||
</kill>
|
||
|
||
</workflow-app>
|