[SPARK-31422] Fix NPE when BlockManagerSource is used after BlockManagerMaster stops

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.3.4, 2.4.5, 3.0.0
• Fix Version/s: 2.4.6, 3.0.0
• Component/s: Spark Core
• Assignee: Dongjoon Hyun
• Reporter: Dongjoon Hyun
• Created: 11/Apr/20 06:24
• Updated: 11/Apr/20 15:28
• Resolved: 11/Apr/20 15:28

Description

In `SparkEnv.stop`, the following stop sequence is used.

blockManager.stop()
blockManager.master.stop()
metricsSystem.stop()

During `metricsSystem.stop`, some sink can invoke `BlockManagerSource` and ends up with NPE.

sinks.foreach(_.stop)
registry.removeMatching((_: String, _: Metric) => true)

java.lang.NullPointerException
	at org.apache.spark.storage.BlockManagerMaster.getStorageStatus(BlockManagerMaster.scala:170)
	at org.apache.spark.storage.BlockManagerSource$$anonfun$10.apply(BlockManagerSource.scala:63)
	at org.apache.spark.storage.BlockManagerSource$$anonfun$10.apply(BlockManagerSource.scala:63)
	at org.apache.spark.storage.BlockManagerSource$$anon$1.getValue(BlockManagerSource.scala:31)
	at org.apache.spark.storage.BlockManagerSource$$anon$1.getValue(BlockManagerSource.scala:30)

Since SparkContext registers and forgets `BlockManagerSource` without deregistering, we had better avoid NPE inside `BlockManagerMaster`.

    _env.metricsSystem.registerSource(new BlockManagerSource(_env.blockManager))

Issue Links

• links to: GitHub Pull Request #28187