[HUDI-1690] Fix StackOverflowError while running clustering with large number of partitions

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.9.0
• Fix Version/s: 0.9.0
• Component/s: spark
• Labels: pull-request-available sev:high user-support-issues
• Assignee: satish
• Reporter: Rong Ma
• Created: 15/Mar/21 04:09
• Updated: 30/Apr/21 03:40
• Resolved: 30/Apr/21 03:40

Description

We are testing clustering on a hudi table with about 3000 partitions. The spark driver throws StackOverflowError before all the partitions sorted:

21/03/11 19:51:20 ERROR [main] UtilHelpers: Cluster failed
java.lang.StackOverflowError
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at java.io.ObjectStreamClass.invokeWriteReplace(ObjectStreamClass.java:1118)
at java.io.ObjectOutputStream.writeObject0(ObjectOutputStream.java:1136)
at java.io.ObjectOutputStream.writeObject(ObjectOutputStream.java:348)
at org.apache.spark.RangePartitioner.$anonfun$writeObject$1(Partitioner.scala:261)
at scala.runtime.java8.JFunction0$mcV$sp.apply(JFunction0$mcV$sp.java:23)
at org.apache.spark.util.Utils$.tryOrIOException(Utils.scala:1343)
at org.apache.spark.RangePartitioner.writeObject(Partitioner.scala:254)
at sun.reflect.GeneratedMethodAccessor201.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at java.io.ObjectStreamClass.invokeWriteObject(ObjectStreamClass.java:1028)
at java.io.ObjectOutputStream.writeSerialData(ObjectOutputStream.java:1496)
at java.io.ObjectOutputStream.writeOrdinaryObject(ObjectOutputStream.java:1432)
at java.io.ObjectOutputStream.writeObject0(ObjectOutputStream.java:1178)
at java.io.ObjectOutputStream.defaultWriteFields(ObjectOutputStream.java:1548)
at java.io.ObjectOutputStream.writeSerialData(ObjectOutputStream.java:1509)
at java.io.ObjectOutputStream.writeOrdinaryObject(ObjectOutputStream.java:1432)
at java.io.ObjectOutputStream.writeObject0(ObjectOutputStream.java:1178)
at java.io.ObjectOutputStream.defaultWriteFields(ObjectOutputStream.java:1548)
at java.io.ObjectOutputStream.writeSerialData(ObjectOutputStream.java:1509)
at java.io.ObjectOutputStream.writeOrdinaryObject(ObjectOutputStream.java:1432)
at java.io.ObjectOutputStream.writeObject0(ObjectOutputStream.java:1178)
at java.io.ObjectOutputStream.writeObject(ObjectOutputStream.java:348)
at scala.collection.immutable.List$SerializationProxy.writeObject(List.scala:477)
at sun.reflect.GeneratedMethodAccessor51.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at java.io.ObjectStreamClass.invokeWriteObject(ObjectStreamClass.java:1028)
at java.io.ObjectOutputStream.writeSerialData(ObjectOutputStream.java:1496)
at java.io.ObjectOutputStream.writeOrdinaryObject(ObjectOutputStream.java:1432)
at java.io.ObjectOutputStream.writeObject0(ObjectOutputStream.java:1178)
at java.io.ObjectOutputStream.defaultWriteFields(ObjectOutputStream.java:1548)

...

 

I see similar issue here:

https://stackoverflow.com/questions/30522564/spark-when-union-a-lot-of-rdd-throws-stack-overflow-error

Setting the driver's stack size to 100M still has this error. So this is probably because the rdd.union has been called too many times and the result of rdd lineage is too large. I think we should use JavaSparkContext.union instead RDD.union here https://github.com/apache/hudi/blob/e93c6a569310ce55c5a0fc0655328e7fd32a9da2/hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/table/action/cluster/SparkExecuteClusteringCommitActionExecutor.java#L96

Issue Links

• links to: GitHub Pull Request #2872