lanyuanxiaoyao/hudi
1
0
Fork 0
Code Activity

Adding support for custom scheduler configs with streaming sink (#4762)

Browse Source
This commit is contained in:
Sivabalan Narayanan
2022-02-08 04:14:10 -05:00
committed by GitHub
parent 1636876e8a
commit ab73047958
4 changed files with 38 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/hudi/DataSourceOptions.scala
View File

@@ -478,6 +478,32 @@ object DataSourceWriteOptions {
+ "Use this when you are in the process of migrating from "
+ "com.uber.hoodie to org.apache.hudi. Stop using this after you migrated the table definition to org.apache.hudi input format")
// spark data source write pool name. Incase of streaming sink, users might be interested to set custom scheduling configs
// for regular writes and async compaction. In such cases, this pool name will be used for spark datasource writes.
val SPARK_DATASOURCE_WRITER_POOL_NAME = "sparkdatasourcewrite"
/*
When async compaction is enabled (deltastreamer or streaming sink), users might be interested to set custom
scheduling configs for regular writes and async compaction. This is the property used to set custom scheduler config
file with spark. In Deltastreamer, the file is generated within hudi and set if necessary. Where as in case of streaming
sink, users have to set this property when they invoke spark shell.
Sample format of the file contents.
<?xml version="1.0"?>
<allocations>
<pool name="sparkdatasourcewrite">
<schedulingMode>FAIR</schedulingMode>
<weight>4</weight>
<minShare>2</minShare>
</pool>
<pool name="hoodiecompact">
<schedulingMode>FAIR</schedulingMode>
<weight>3</weight>
<minShare>1</minShare>
</pool>
</allocations>
*/
val SPARK_SCHEDULER_ALLOCATION_FILE_KEY = "spark.scheduler.allocation.file"
/** @deprecated Use {@link HIVE_SYNC_MODE} instead of this config from 0.9.0 */
@Deprecated
val HIVE_USE_JDBC: ConfigProperty[String] = ConfigProperty

5
hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/hudi/HoodieSparkSqlWriter.scala
View File

@@ -117,6 +117,11 @@ object HoodieSparkSqlWriter {
}
val jsc = new JavaSparkContext(sparkContext)
if (asyncCompactionTriggerFn.isDefined) {
if (jsc.getConf.getOption(DataSourceWriteOptions.SPARK_SCHEDULER_ALLOCATION_FILE_KEY).isDefined) {
jsc.setLocalProperty("spark.scheduler.pool", DataSourceWriteOptions.SPARK_DATASOURCE_WRITER_POOL_NAME)
}
}
val instantTime = HoodieActiveTimeline.createNewInstantTime()
val keyGenerator = HoodieSparkKeyGeneratorFactory.createKeyGenerator(new TypedProperties(hoodieConfig.getProps))