You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Marta Kuczora (Jira)" <ji...@apache.org> on 2020/10/28 17:12:00 UTC
[jira] [Updated] (HIVE-24322) In case of direct insert, the attempt
ID has to be checked when reading the manifest files
[ https://issues.apache.org/jira/browse/HIVE-24322?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Marta Kuczora updated HIVE-24322:
---------------------------------
Description:
In 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 task 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 paths 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 attempt 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 attempt 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.
was:
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.
> 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
> Priority: Major
> Fix For: 4.0.0
>
>
> In 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 task 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 paths 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 attempt 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 attempt 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)