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)

Reply via email to