Marta Kuczora created HIVE-24322:
------------------------------------
Summary: In case of direct insert, the attempt ID has to be
checked when reading the manifest files
Key: HIVE-24322
URL: https://issues.apache.org/jira/browse/HIVE-24322
Project: Hive
Issue Type: Bug
Affects Versions: 4.0.0
Reporter: Marta Kuczora
Assignee: Marta Kuczora
Fix For: 4.0.0
In [IMPALA-10247|https://issues.apache.org/jira/browse/IMPALA-10247] there was
an exception from Hive when tyring to load the data:
{noformat}
2020-10-13T16:50:53,424 ERROR [HiveServer2-Background-Pool: Thread-23832]
exec.Task: Job Commit failed with exception
'org.apache.hadoop.hive.ql.metadata.HiveException(java.io.EOFException)'
org.apache.hadoop.hive.ql.metadata.HiveException: java.io.EOFException
at
org.apache.hadoop.hive.ql.exec.FileSinkOperator.jobCloseOp(FileSinkOperator.java:1468)
at org.apache.hadoop.hive.ql.exec.Operator.jobClose(Operator.java:798)
at org.apache.hadoop.hive.ql.exec.Operator.jobClose(Operator.java:803)
at org.apache.hadoop.hive.ql.exec.Operator.jobClose(Operator.java:803)
at org.apache.hadoop.hive.ql.exec.tez.TezTask.close(TezTask.java:627)
at org.apache.hadoop.hive.ql.exec.tez.TezTask.execute(TezTask.java:342)
at org.apache.hadoop.hive.ql.exec.Task.executeTask(Task.java:213)
at org.apache.hadoop.hive.ql.exec.TaskRunner.runSequential(TaskRunner.java:105)
at org.apache.hadoop.hive.ql.Executor.launchTask(Executor.java:357)
at org.apache.hadoop.hive.ql.Executor.launchTasks(Executor.java:330)
at org.apache.hadoop.hive.ql.Executor.runTasks(Executor.java:246)
at org.apache.hadoop.hive.ql.Executor.execute(Executor.java:109)
at org.apache.hadoop.hive.ql.Driver.runInternal(Driver.java:721)
at org.apache.hadoop.hive.ql.Driver.run(Driver.java:488)
at org.apache.hadoop.hive.ql.Driver.run(Driver.java:482)
at org.apache.hadoop.hive.ql.reexec.ReExecDriver.run(ReExecDriver.java:166)
at
org.apache.hive.service.cli.operation.SQLOperation.runQuery(SQLOperation.java:225)
at
org.apache.hive.service.cli.operation.SQLOperation.access$700(SQLOperation.java:87)
at
org.apache.hive.service.cli.operation.SQLOperation$BackgroundWork$1.run(SQLOperation.java:322)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAs(Subject.java:422)
at
org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1876)
at
org.apache.hive.service.cli.operation.SQLOperation$BackgroundWork.run(SQLOperation.java:340)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at
java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
Caused by: java.io.EOFException
at java.io.DataInputStream.readInt(DataInputStream.java:392)
at
org.apache.hadoop.hive.ql.exec.Utilities.handleDirectInsertTableFinalPath(Utilities.java:4587)
at
org.apache.hadoop.hive.ql.exec.FileSinkOperator.jobCloseOp(FileSinkOperator.java:1462)
... 29 more
{noformat}
The reason of the exception was that Hive was trying to read an empty manifest
file. Manifest files are used in case of direct insert to determine which files
needs to be kept and which one needs to be cleaned up. They are created by the
tasks and they use the tast attempt Id as postfix. In this particular test what
happened is that one of the container ran out of memory so Tez decided to kill
it right after the manifest file got created but before the pathes got written
into the manifest file. This was the manifest file for the task attempt 0. Then
Tez assigned a new container to the task, so a new attemp was made with
attemptId=1. This one was successful, and wrote the manifest file correctly.
But Hive didn't know about this, since this out of memory issue got handled by
Tez under the hood, so there was no exception in Hive, therefore no clean-up in
the manifest folder. And when Hive is reading the manifest files, it just reads
every file from the defined folder, so it tried to read the manifest files for
attemp 0 and 1 as well.
If there are multiple manifest files with the same name but different
attemptId, Hive should only read the one with the biggest attempt Id.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
