[MAPREDUCE-6745] Job directories should be clean in staging directorg /tmp/hadoop-yarn/staging after MapReduce job finish successfully

• Type: Bug
• Status: Open
• Priority: Blocker
• Resolution: Unresolved
• Affects Version/s: 2.7.2
• Component/s: mr-am

Environment

Suse 11 sp3

• Assignee: Unassigned
• Reporter: liuxiaoping
• Created: 29/Jul/16 02:13
• Updated: 01/Sep/16 11:51

Description

If MapReduce client set mapreduce.task.files.preserve.failedtasks=true, temporary job directory will not be deleted in staging directory /tmp/hadoop-yarn/staging.
As time goes by, the job files are more and more, eventually lead to below exeception:

org.apache.hadoop.ipc.RemoteException(org.apache.hadoop.hdfs.protocol.FSLimitException$MaxDirectoryItemExceededException):
The directory item limit of /tmp/hadoop-yarn/staging/username/.staging is exceeded: limit=1048576 items=1048576
at org.apache.hadoop.hdfs.server.namenode.FSDirectory.verifyMaxDirItems(FSDirectory.java:936)
at org.apache.hadoop.hdfs.server.namenode.FSDirectory.addLastINode(FSDirectory.java:981)
at org.apache.hadoop.hdfs.server.namenode.FSDirMkdirOp.unprotectedMkdir(FSDirMkdirOp.java:237)
at org.apache.hadoop.hdfs.server.namenode.FSDirMkdirOp.createSingleDirectory(FSDirMkdirOp.java:191)
at org.apache.hadoop.hdfs.server.namenode.FSDirMkdirOp.createChildrenDirectories(FSDirMkdirOp.java:166)
at org.apache.hadoop.hdfs.server.namenode.FSDirMkdirOp.mkdirs(FSDirMkdirOp.java:97)
at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.mkdirs(FSNamesystem.java:3788)
at org.apache.hadoop.hdfs.server.namenode.NameNodeRpcServer.mkdirs(NameNodeRpcServer.java:986)
at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolServerSideTranslatorPB.mkdirs(ClientNamenodeProtocolServerSideTranslatorPB.java:624)
at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolProtos.$ClientNamenodeProtocol$2.callBlockingMethod(ClientNamenodeProtocolProtos.java:624)
at org.apache.hadoop.ipc.ProtobufRpcEngine$Server$ProtoBufRpcInvoker.call(ProtobufRpcEngine.java:616)
at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:973)
at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2088)
at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2084)
at java.security.auth.Subject.doAs(Subject.java:422)
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1672)
at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2082)

The official instructions for the configuration mapreduce.task.files.preserve.failedtasks is below:
Should the files for failed tasks be kept. This should only be used on jobs that are failing, because the storage is never reclaimed.
It also prevents the map outputs from being erased from the reduce directory as they are consumed.

According to the instructions, I think the temporary files for successful tasks shouldn't be kept.