[HDFS-13757] After HDFS-12886, close() can throw AssertionError "Negative replicas!"

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.1.0, 2.10.0, 2.9.1, 3.2.0, 3.0.3
• Component/s: namenode
• Assignee: Unassigned
• Reporter: Wei-Chiu Chuang
• Created: 20/Jul/18 20:36
• Updated: 02/Oct/19 17:12

Description

While investigating a data corruption bug caused by concurrent recoverLease() and close(), I found HDFS-12886 may cause close() to throw AssertionError under a corner case, because the block has zero live replica, and client calls recoverLease() immediately followed by close().

org.apache.hadoop.ipc.RemoteException(java.lang.AssertionError): Negative replicas!
at org.apache.hadoop.hdfs.server.blockmanagement.LowRedundancyBlocks.getPriority(LowRedundancyBlocks.java:197)
at org.apache.hadoop.hdfs.server.blockmanagement.LowRedundancyBlocks.update(LowRedundancyBlocks.java:422)
at org.apache.hadoop.hdfs.server.blockmanagement.BlockManager.updateNeededReconstructions(BlockManager.java:4274)
at org.apache.hadoop.hdfs.server.blockmanagement.BlockManager.commitOrCompleteLastBlock(BlockManager.java:1001)
at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.commitOrCompleteLastBlock(FSNamesystem.java:3471)
at org.apache.hadoop.hdfs.server.namenode.FSDirWriteFileOp.completeFileInternal(FSDirWriteFileOp.java:713)
at org.apache.hadoop.hdfs.server.namenode.FSDirWriteFileOp.completeFile(FSDirWriteFileOp.java:671)
at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.completeFile(FSNamesystem.java:2854)
at org.apache.hadoop.hdfs.server.namenode.NameNodeRpcServer.complete(NameNodeRpcServer.java:928)
at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolServerSideTranslatorPB.complete(ClientNamenodeProtocolServerSideTranslatorPB.java:607)
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:1025)
at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:876)
at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:822)
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:1689)
at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2682)

I have a test case to reproduce it.

lukmajercak elgoiri would you please take a look at it? I think we should add a check to reject completeFile() if the block is under recovery, similar to what's proposed in HDFS-10240.

Issue Links

• is caused by: HDFS-12886 Ignore minReplication for block recovery