[HDFS-12567] BlockPlacementPolicyRackFaultTolerant fails with racks with very few nodes

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0-alpha1
• Fix Version/s: 3.0.0
• Component/s: erasure-coding
• Labels: hdfs-ec-3.0-must-do
• Target Version/s: 3.0.0
• Assignee: Andrew Wang
• Reporter: Andrew Wang
• Created: 29/Sep/17 20:00
• Updated: 26/Oct/17 20:50
• Resolved: 05/Oct/17 23:59

Description

Found this while doing some testing on an internal cluster with an unusual setup. We have a rack with ~20 nodes, then a few more with just a few nodes. It would fail to get (# data blocks) datanodes even though there were plenty of DNs on the rack with 20 DNs.

I managed to reproduce this same issue in a unit test, stack trace like this:

java.io.IOException: File /testfile0 could only be written to 5 of the 6 required nodes for RS-6-3-1024k. There are 9 datanode(s) running and no node(s) are excluded in this operation.
	at org.apache.hadoop.hdfs.server.blockmanagement.BlockManager.chooseTarget4NewBlock(BlockManager.java:2083)
	at org.apache.hadoop.hdfs.server.namenode.FSDirWriteFileOp.chooseTargetForNewBlock(FSDirWriteFileOp.java:286)
	at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.getAdditionalBlock(FSNamesystem.java:2609)
	at org.apache.hadoop.hdfs.server.namenode.NameNodeRpcServer.addBlock(NameNodeRpcServer.java:863)
	at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolServerSideTranslatorPB.addBlock(ClientNamenodeProtocolServerSideTranslatorPB.java:548)
	at org.apache.hadoop.hdfs.protocol.proto.ClientNamenodeProtocolProtos$ClientNamenodeProtocol$2.callBlockingMethod(ClientNamenodeProtocolProtos.java)
	at org.apache.hadoop.ipc.ProtobufRpcEngine$Server$ProtoBufRpcInvoker.call(ProtobufRpcEngine.java:523)
	at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:991)
	at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:869)
	at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:815)
	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:1962)
	at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2675)

This isn't a very critical bug since it's an unusual rack configuration, but it can easily happen during testing.

Issue Links

• is related to: HDFS-12725 BlockPlacementPolicyRackFaultTolerant fails with very uneven racks