[SUPPORT]Possible memory leak issue for org.apache.hadoop.hive.conf.HiveConf while using Flink into Hudi

[numberlabs-developers]

Tips before filing an issue

Have you gone through our FAQs?

Join the mailing list to engage in conversations and get faster support at dev-subscribe@hudi.apache.org.

If you have triaged this as a bug, then file an issue directly.

Describe the problem you faced

I am using Flink SQL to ingest data from AWS Kinesis to Hudi on S3. I used AWS Glue catalog as Hive metastore. hive_sync.enable is set to true in SQL. The ingestion works well. However after running a few hours or days, the jobmanager failed with OutOfMemory. I checked the hdump and found org.apache.hadoop.hive.conf.HiveConf took 80.77% memory. It seems to be related to HiveSyncContext.

The suspect leak from Eclipse Memory Analyzer:

12 instances of "org.apache.hadoop.hive.conf.HiveConf", loaded by "sun.misc.Launcher$AppClassLoader @ 0xe400bdf8" occupy 338,544,384 (80.77%) bytes.

Biggest instances: •org.apache.hadoop.hive.conf.HiveConf @ 0xe71197b0 - 33,702,712 (8.04%) bytes. •org.apache.hadoop.hive.conf.HiveConf @ 0xe72d9e30 - 33,702,712 (8.04%) bytes. •org.apache.hadoop.hive.conf.HiveConf @ 0xe77c62c0 - 33,702,712 (8.04%) bytes. •org.apache.hadoop.hive.conf.HiveConf @ 0xe787f640 - 33,702,712 (8.04%) bytes. •org.apache.hadoop.hive.conf.HiveConf @ 0xe798fd00 - 33,702,712 (8.04%) bytes. •org.apache.hadoop.hive.conf.HiveConf @ 0xe7a9b0f0 - 33,702,712 (8.04%) bytes. •org.apache.hadoop.hive.conf.HiveConf @ 0xe812a8c8 - 33,702,712 (8.04%) bytes. •org.apache.hadoop.hive.conf.HiveConf @ 0xe82d0af0 - 33,702,712 (8.04%) bytes. •org.apache.hadoop.hive.conf.HiveConf @ 0xe84c10c8 - 33,702,712 (8.04%) bytes. •org.apache.hadoop.hive.conf.HiveConf @ 0xe8736300 - 33,702,712 (8.04%) bytes.

Keywords sun.misc.Launcher$AppClassLoader @ 0xe400bdf8 org.apache.hadoop.hive.conf.HiveConf

To Reproduce

Steps to reproduce the behavior:

Set up an AWS EMR 6.10.0 with Flink 1.16.0 +Hive 3.1 + Hudi 0.13.0 Set up an AWS Kinesis and ingest data into it. Run a Flink SQL job to ingest to Hudi on S3 from Kinesis Run for a few hours or days, could get OOM. Expected behavior No OOM issue.

Environment Description

Hudi version : 0.13.0

Spark version : 3.3.1

Hive version : 3.1

Hadoop version : 3.3.3

Storage (HDFS/S3/GCS..) : S3

Running on Docker? (yes/no) : no

Additional context

Add any other context about the problem here.

Stacktrace

Add the stacktrace of the error.

2023-11-09 06:59:55,475 ERROR org.apache.hudi.sink.StreamWriteOperatorCoordinator [] - Executor executes action [commits the instant 20231109065505712] error java.lang.OutOfMemoryError: GC overhead limit exceeded at java.util.stream.StreamSupport.stream(StreamSupport.java:69) ~[?:1.8.0_392] at java.util.Collection.stream(Collection.java:581) ~[?:1.8.0_392] at org.apache.hudi.common.table.timeline.TimelineLayout$TimelineLayoutV1.lambda$filterHoodieInstants$2(TimelineLayout.java:68) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.timeline.TimelineLayout$TimelineLayoutV1$$Lambda$1187/1033743503.apply(Unknown Source) ~[?:?] at java.util.stream.ReferencePipeline$3$1.accept(ReferencePipeline.java:193) ~[?:1.8.0_392] at java.util.HashMap$ValueSpliterator.forEachRemaining(HashMap.java:1652) ~[?:1.8.0_392] at java.util.stream.AbstractPipeline.copyInto(AbstractPipeline.java:482) ~[?:1.8.0_392] at java.util.stream.AbstractPipeline.wrapAndCopyInto(AbstractPipeline.java:472) ~[?:1.8.0_392] at java.util.stream.ReduceOps$ReduceOp.evaluateSequential(ReduceOps.java:708) ~[?:1.8.0_392] at java.util.stream.AbstractPipeline.evaluate(AbstractPipeline.java:234) ~[?:1.8.0_392] at java.util.stream.ReferencePipeline.collect(ReferencePipeline.java:566) ~[?:1.8.0_392] at org.apache.hudi.common.table.HoodieTableMetaClient.scanHoodieInstantsFromFileSystem(HoodieTableMetaClient.java:651) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.HoodieTableMetaClient.scanHoodieInstantsFromFileSystem(HoodieTableMetaClient.java:625) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.timeline.HoodieActiveTimeline.(HoodieActiveTimeline.java:163) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.timeline.HoodieActiveTimeline.(HoodieActiveTimeline.java:155) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.timeline.HoodieActiveTimeline.(HoodieActiveTimeline.java:175) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.HoodieTableMetaClient.getActiveTimeline(HoodieTableMetaClient.java:352) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.HoodieTableMetaClient.(HoodieTableMetaClient.java:153) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.HoodieTableMetaClient.newMetaClient(HoodieTableMetaClient.java:689) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.HoodieTableMetaClient.access$000(HoodieTableMetaClient.java:81) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.common.table.HoodieTableMetaClient$Builder.build(HoodieTableMetaClient.java:770) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.table.HoodieFlinkTable.create(HoodieFlinkTable.java:62) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.client.HoodieFlinkTableServiceClient.getHoodieTable(HoodieFlinkTableServiceClient.java:173) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.client.HoodieFlinkTableServiceClient.writeTableMetadata(HoodieFlinkTableServiceClient.java:179) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.client.HoodieFlinkWriteClient.writeTableMetadata(HoodieFlinkWriteClient.java:279) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.client.BaseHoodieWriteClient.commit(BaseHoodieWriteClient.java:282) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.client.BaseHoodieWriteClient.commitStats(BaseHoodieWriteClient.java:233) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.client.HoodieFlinkWriteClient.commit(HoodieFlinkWriteClient.java:111) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.client.HoodieFlinkWriteClient.commit(HoodieFlinkWriteClient.java:74) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.client.BaseHoodieWriteClient.commit(BaseHoodieWriteClient.java:199) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.sink.StreamWriteOperatorCoordinator.doCommit(StreamWriteOperatorCoordinator.java:537) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0] at org.apache.hudi.sink.StreamWriteOperatorCoordinator.commitInstant(StreamWriteOperatorCoordinator.java:513) ~[hudi-flink1.16-bundle-0.13.0.jar:0.13.0]

[torvalds-dev-testbot[bot]]

torvalds.dev is analyzing the ticket

[torvalds-dev-testbot[bot]]

This seems like an issue with possible memory leaks in the org.apache.hadoop.hive.conf.HiveConf class when using Flink together with Hudi.

To address this, you'll need to:

1. Reproduce the issue on your local machine using the information provided in the "To Reproduce" section.
2. If the issue is reproducible, analyze the code which is specifically using the org.apache.hadoop.hive.conf.HiveConf class and try to identify where it may be creating a memory leak.
3. Document what you find during your analysis and propose a solution. The seemingly problematic code provided might be part of the issue, so look into its functionality and how it's using org.apache.hadoop.hive.conf.HiveConf.
4. Test your solution by running the memory intensive operations and monitor memory use.
5. If the issue is solved, document the steps to resolve it. If it's not solved, try to refine your analysis and attempt a different solution.

Please remember to share your findings with the user and ask for permission if you need to run any resource-intensive tests.