You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Stamatis Zampetakis (Jira)" <ji...@apache.org> on 2021/10/12 22:09:00 UTC
[jira] [Commented] (HIVE-25611) OOM when running MERGE query on
wide transactional table with many buckets
[ https://issues.apache.org/jira/browse/HIVE-25611?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17427934#comment-17427934 ]
Stamatis Zampetakis commented on HIVE-25611:
--------------------------------------------
The OOM manifestates only when we pick a large value for {{hive.exec.reducers.bytes.per.reducer}} before running the {{MERGE}} query. This forces the reducer writting to ORC files to run with parallelism 1 simulating a scenario where we have many buckets in the table and the choice of parallelism does not take them into account.
Essentially if we set {{hive.exec.reducers.bytes.per.reducer}} the value to {{10}} for the entire scenario in {{merge_wide_acid_bucketed_table.q}} the query runs fine since the problematic reducer gets a parallelism close to the number of buckets. However, Hive versions with HIVE-22538 but without HIVE-25071 have no way to tune the parallelism.
> OOM when running MERGE query on wide transactional table with many buckets
> ---------------------------------------------------------------------------
>
> Key: HIVE-25611
> URL: https://issues.apache.org/jira/browse/HIVE-25611
> Project: Hive
> Issue Type: Bug
> Components: Query Processor
> Reporter: Stamatis Zampetakis
> Assignee: Stamatis Zampetakis
> Priority: Major
> Attachments: merge_query_plan.txt, merge_wide_acid_bucketed_table.q, wide_table_100_char_cols.csv
>
>
> Running a {{MERGE}} statement over a wide transactional/ACID table with many buckets leads to {{OutOfMemoryError}} during the execution of the query.
> A step-by-step reproducer is attached to the case ( [^merge_wide_acid_bucketed_table.q] [^wide_table_100_char_cols.csv] ) but the main idea is outlined below.
> {code:sql}
> CREATE TABLE wide_table_txt (
> w_id_col int,
> w_char_col0 char(20),
> ...
> w_char_col99 char(20)) STORED AS ORC TBLPROPERTIES ('transactional'='true')
> -- Load data into the table in a way that it gets bucketed
> CREATE TABLE simple_table_txt (id int, name char(20)) STORED AS TEXTFILE;
> -- Load data into simple_table_txt overlapping with the data in wide_table_txt
> MERGE INTO wide_table_orc target USING simple_table_txt source ON (target.w_id_col = source.id)
> WHEN MATCHED THEN UPDATE SET w_char_col0 = source.name
> WHEN NOT MATCHED THEN INSERT (w_id_col, w_char_col1) VALUES (source.id, 'Actual value does not matter');
> {code}
> A sample stacktrace showing the memory pressure is given below:
> {noformat}
> java.lang.OutOfMemoryError: GC overhead limit exceeded
> at org.apache.orc.OrcProto$RowIndexEntry$Builder.create(OrcProto.java:8962) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.OrcProto$RowIndexEntry$Builder.access$12100(OrcProto.java:8931) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.OrcProto$RowIndexEntry.newBuilder(OrcProto.java:8915) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriterBase.<init>(TreeWriterBase.java:98) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.StringBaseTreeWriter.<init>(StringBaseTreeWriter.java:66) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.CharTreeWriter.<init>(CharTreeWriter.java:40) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.createSubtree(TreeWriter.java:163) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.create(TreeWriter.java:133) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.StructTreeWriter.<init>(StructTreeWriter.java:41) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.createSubtree(TreeWriter.java:181) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.create(TreeWriter.java:133) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.StructTreeWriter.<init>(StructTreeWriter.java:41) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.createSubtree(TreeWriter.java:181) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.create(TreeWriter.java:133) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.WriterImpl.<init>(WriterImpl.java:216) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.hadoop.hive.ql.io.orc.WriterImpl.<init>(WriterImpl.java:95) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcFile.createWriter(OrcFile.java:396) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcRecordUpdater.initWriter(OrcRecordUpdater.java:615) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcRecordUpdater.addSimpleEvent(OrcRecordUpdater.java:442) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcRecordUpdater.addSplitUpdateEvent(OrcRecordUpdater.java:495) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcRecordUpdater.update(OrcRecordUpdater.java:519) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.FileSinkOperator.process(FileSinkOperator.java:1200) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.vector.VectorFileSinkOperator.process(VectorFileSinkOperator.java:111) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.Operator.vectorForward(Operator.java:919) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.vector.VectorSelectOperator.process(VectorSelectOperator.java:158) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.ReduceRecordSource.processVectorGroup(ReduceRecordSource.java:497) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.ReduceRecordSource.pushRecordVector(ReduceRecordSource.java:399) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.ReduceRecordSource.pushRecord(ReduceRecordSource.java:256) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.ReduceRecordProcessor.run(ReduceRecordProcessor.java:291) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.TezProcessor.initializeAndRunProcessor(TezProcessor.java:311) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.TezProcessor.run(TezProcessor.java:277) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.tez.runtime.LogicalIOProcessorRuntimeTask.run(LogicalIOProcessorRuntimeTask.java:381) ~[tez-runtime-internals-0.10.1.jar:0.10.1]
> 2021-10-12T10:31:57,723 INFO [TezTR-775147_1_3_4_0_0] task.TezTaskRunner2: Received notification of a failure which will cause the task to die
> java.lang.OutOfMemoryError: GC overhead limit exceeded
> at org.apache.orc.OrcProto$RowIndexEntry$Builder.create(OrcProto.java:8962) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.OrcProto$RowIndexEntry$Builder.access$12100(OrcProto.java:8931) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.OrcProto$RowIndexEntry.newBuilder(OrcProto.java:8915) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriterBase.<init>(TreeWriterBase.java:98) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.StringBaseTreeWriter.<init>(StringBaseTreeWriter.java:66) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.CharTreeWriter.<init>(CharTreeWriter.java:40) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.createSubtree(TreeWriter.java:163) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.create(TreeWriter.java:133) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.StructTreeWriter.<init>(StructTreeWriter.java:41) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.createSubtree(TreeWriter.java:181) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.create(TreeWriter.java:133) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.StructTreeWriter.<init>(StructTreeWriter.java:41) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.createSubtree(TreeWriter.java:181) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.writer.TreeWriter$Factory.create(TreeWriter.java:133) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.orc.impl.WriterImpl.<init>(WriterImpl.java:216) ~[orc-core-1.6.9.jar:1.6.9]
> at org.apache.hadoop.hive.ql.io.orc.WriterImpl.<init>(WriterImpl.java:95) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcFile.createWriter(OrcFile.java:396) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcRecordUpdater.initWriter(OrcRecordUpdater.java:615) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcRecordUpdater.addSimpleEvent(OrcRecordUpdater.java:442) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcRecordUpdater.addSplitUpdateEvent(OrcRecordUpdater.java:495) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.io.orc.OrcRecordUpdater.update(OrcRecordUpdater.java:519) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.FileSinkOperator.process(FileSinkOperator.java:1200) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.vector.VectorFileSinkOperator.process(VectorFileSinkOperator.java:111) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.Operator.vectorForward(Operator.java:919) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.vector.VectorSelectOperator.process(VectorSelectOperator.java:158) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.ReduceRecordSource.processVectorGroup(ReduceRecordSource.java:497) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.ReduceRecordSource.pushRecordVector(ReduceRecordSource.java:399) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.ReduceRecordSource.pushRecord(ReduceRecordSource.java:256) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.ReduceRecordProcessor.run(ReduceRecordProcessor.java:291) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> at org.apache.hadoop.hive.ql.exec.tez.TezProcessor.initializeAndRunProcessor(TezProcessor.java:311) ~[hive-exec-4.0.0-SNAPSHOT.jar:4.0.0-SNAPSHOT]
> {noformat}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)