[HDFS-13120] Snapshot diff could be corrupted after concat

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.7.0
• Fix Version/s: 3.1.0, 2.10.0, 2.9.1, 3.0.1, 2.8.4, 2.7.6
• Component/s: namenode, snapshots
• Hadoop Flags: Reviewed
• Assignee: Xiaoyu Yao
• Reporter: Xiaoyu Yao
• Created: 08/Feb/18 00:14
• Updated: 09/Feb/18 06:49
• Resolved: 09/Feb/18 06:49

Description

The snapshot diff can be corrupted after concat files. This could lead to Assertion upon DeleteSnapshot and getSnapshotDiff operations later.

For example, we have seen customers hit stack trace similar to the one below but during loading edit entry of DeleteSnapshotOp. After the investigation, we found this is a regression caused by HDFS-3689 where the snapshot diff is not fully cleaned up after concat.

I will post the unit test to repro this and fix for it shortly.

org.apache.hadoop.ipc.RemoteException(java.lang.AssertionError): Element already exists: element=0.txt, CREATED=[0.txt, 1.txt, 2.txt]
	at org.apache.hadoop.hdfs.util.Diff.insert(Diff.java:196)
	at org.apache.hadoop.hdfs.util.Diff.create(Diff.java:216)
	at org.apache.hadoop.hdfs.util.Diff.combinePosterior(Diff.java:463)
	at org.apache.hadoop.hdfs.server.namenode.snapshot.DirectoryWithSnapshotFeature$DirectoryDiff.combinePosteriorAndCollectBlocks(DirectoryWithSnapshotFeature.java:205)
	at org.apache.hadoop.hdfs.server.namenode.snapshot.DirectoryWithSnapshotFeature$DirectoryDiff.combinePosteriorAndCollectBlocks(DirectoryWithSnapshotFeature.java:162)
	at org.apache.hadoop.hdfs.server.namenode.snapshot.AbstractINodeDiffList.deleteSnapshotDiff(AbstractINodeDiffList.java:100)
	at org.apache.hadoop.hdfs.server.namenode.snapshot.DirectoryWithSnapshotFeature.cleanDirectory(DirectoryWithSnapshotFeature.java:728)
	at org.apache.hadoop.hdfs.server.namenode.INodeDirectory.cleanSubtree(INodeDirectory.java:830)
	at org.apache.hadoop.hdfs.server.namenode.snapshot.DirectorySnapshottableFeature.removeSnapshot(DirectorySnapshottableFeature.java:237)
	at org.apache.hadoop.hdfs.server.namenode.INodeDirectory.removeSnapshot(INodeDirectory.java:292)
	at org.apache.hadoop.hdfs.server.namenode.snapshot.SnapshotManager.deleteSnapshot(SnapshotManager.java:321)
	at org.apache.hadoop.hdfs.server.namenode.FSDirSnapshotOp.deleteSnapshot(FSDirSnapshotOp.java:249)
	at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.deleteSnapshot(FSNamesystem.java:6566)
	at org.apache.hadoop.hdfs.server.namenode.NameNodeRpcServer.deleteSnapshot(NameNodeRpcServer.java:1823)
	at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolServerSideTranslatorPB.deleteSnapshot(ClientNamenodeProtocolServerSideTranslatorPB.java:1200)
	at org.apache.hadoop.hdfs.protocol.proto.ClientNamenodeProtocolProtos$ClientNamenodeProtocol$2.callBlockingMethod(ClientNamenodeProtocolProtos.java)
	at org.apache.hadoop.ipc.ProtobufRpcEngine$Server$ProtoBufRpcInvoker.call(ProtobufRpcEngine.java:524)
	at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:1007)
	at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:873)
	at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:819)
	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:1965)
	at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2679)

Issue Links

• is broken by: HDFS-3689 Add support for variable length block