[YARN-9016] DocumentStore as a backend for ATSv2

• Type: New Feature
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.3.0
• Component/s: ATSv2
• Hadoop Flags: Reviewed
• Assignee: Sushil Ks
• Reporter: Sushil Ks
• Created: 13/Nov/18 15:44
• Updated: 20/Jul/20 07:53
• Resolved: 13/Mar/19 23:46

Description

Document Store for ATSv2

               The Document Store for ATSv2 is a framework for plugging in any Document Store Vendor as a backend for ATSv2 i.e Azure CosmosDB , MongoDB, ElasticSearch etc.

• Supports multiple Document Store Vendors like CosmosDB, ElasticSearch, MongoDB etc by just adding new configurations properties and writing Document Store reader and writer clients.
• Currently has support for CosmosDB.
• All writes are Async and buffered, latest document would be flushed to the store either if the document buffer gets full or periodically at every flush interval in background without adding any additional latency to the running jobs..
• All the REST API's of Timeline Reader Server are supported.

How to enable?

Add the flowing properties under yarn-site.xml

<!-- config required for ATSv2 to use DocumentStore-->
<property>
 <name>yarn.timeline-service.writer.class </name>
 <value>org.apache.hadoop.yarn.server.timelineservice.documentstore.DocumentStoreTimelineWriterImpl</value>
</property>

<property>
   <name>yarn.timeline-service.reader.class </name>             <value>org.apache.hadoop.yarn.server.timelineservice.documentstore.DocumentStoreTimelineReaderImpl</value>
</property>

<property> 
   <name>yarn.timeline-service.document-store.db-name</name>       
   <value>YOUR_DATABASE_NAME</value> <!-- default is timeline_service --> 
</property>

Creating DB and Collections for storing documents

                      The following config needs to be set inside yarn-site.xml for creating the database and collections for storing documents.

<!-- Using schema creator class for DocumentStore-->
<property>
 <name>yarn.timeline-service.schema-creator.class </name>
 <value>org.apache.hadoop.yarn.server.timelineservice.documentstore.DocumentStoreCollectionCreator</value>
</property>

            Running the schema creator tool to create the necessary collections.

bin/hadoop org.apache.hadoop.yarn.server.timelineservice.storage.TimelineSchemaCreator

 Azure CosmosDB 

       To use Azure CosmosDB as a DocumentStore for ATSv2, the additional properties under yarn-site.xml is required..

<!-- config required for using Azure CosmosDB as a DocumentStore for ATSv2 -->
<property>  
   <name>yarn.timeline-service.document-store-type</name>  
   <value>COSMOS_DB</value>
</property>

<property>
   <name>yarn.timeline-service.document-store.cosmos-db.endpoint</name>
   <value>http://YOUR_AZURE_COSMOS_DB_URL:443/</value>
</property>

<property>
   <name>yarn.timeline-service.document-store.cosmos-db.masterkey</name>
   <value>YOUR_AZURE_COSMOS_DB_MASTER_KEY_CREDENTIAL</value>
</property>

 
  Testing locally
              In order to test the Azure CosmosDB as a DocumentStore locally, install the emulator from here and start it locally. Set the endpoint and master key under yarn-site.xml as mentioned above and run any example job like DistributedShell etc. Later you can check the data explorer UI of Azure CosmosDB locally to query the documents or even launch the TimelineReader locally to fetch/query the data from REST API's.   
                  

      

Issue Links

• breaks: HADOOP-16381 The JSON License is included in binary tarball via azure-documentdb:1.16.2
• relates to: YARN-9150 Making TimelineSchemaCreator support different backends for Timeline Schema Creation in ATSv2