Hbase节点经常挂掉 异常Premature EOF: no length prefix available

hbase | 2019-09-13 10:02:39

1.异常信息

最近Hbase regionServer节点经常挂掉几个，查了一下挂掉节点hbase-root-regionserver-slave3.log日志，出现如下错误信息

2019-07-11 02:58:34,320 WARN  [regionserver/slave3/192.168.202.19:16020-longCompactions-1562643385477-SendThread(slave3:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 88158ms for sessionid 0x369e6926a042130
2019-07-11 02:58:34,320 WARN  [main-SendThread(slave4:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 91643ms for sessionid 0x16bc2ec8e9e0177
2019-07-11 02:58:34,320 INFO  [regionserver/slave3/192.168.22.159:16020-shortCompactions-1562643620766-SendThread(slave2:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 84959ms for sessionid 0x16bc2ec8e9e017b, closing socket connection and attempting reconnect
2019-07-11 02:58:34,320 INFO  [main-SendThread(slave4:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 91643ms for sessionid 0x16bc2ec8e9e0177, closing socket connection and attempting reconnect
2019-07-11 02:58:34,321 WARN  [regionserver/slave3/192.168.22.159:16020-SendThread(slave3:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 101972ms for sessionid 0x16bc2ec8e9e0178
2019-07-11 02:58:34,321 INFO  [regionserver/slave3/192.168.22.159:16020-SendThread(slave3:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 101972ms for sessionid 0x16bc2ec8e9e0178, closing socket connection and attempting reconnect
2019-07-11 02:58:34,320 INFO  [regionserver/slave3/192.168.202.19:16020-longCompactions-1562643385477-SendThread(slave3:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 88158ms for sessionid 0x369e6926a042130, closing socket connection and attempting reconnect
2019-07-11 02:58:34,321 WARN  [RS_OPEN_REGION-slave3:16020-0-SendThread(slave3:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 88092ms for sessionid 0x369e6926a04212d
2019-07-11 02:58:34,321 INFO  [RS_OPEN_REGION-slave3:16020-0-SendThread(slave3:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 88092ms for sessionid 0x369e6926a04212d, closing socket connection and attempting reconnect
2019-07-11 02:58:34,322 INFO  [slave3,16020,1562643381590_ChoreService_1] hbase.ScheduledChore: Chore: slave3,16020,1562643381590-MemstoreFlusherChore missed its start time
2019-07-11 02:58:34,323 INFO  [slave3,16020,1562643381590_ChoreService_1] hbase.ScheduledChore: Chore: CompactionChecker missed its start time
2019-07-11 02:58:34,324 WARN  [ResponseProcessor for block BP-1486468690-127.0.0.1-1550458385182:blk_1076360281_2619665] hdfs.DFSClient: DFSOutputStream ResponseProcessor exception  for block BP-1486468690-127.0.0.1-1550458385182:blk_1076360281_2619665
java.io.EOFException: Premature EOF: no length prefix available
at org.apache.hadoop.hdfs.protocolPB.PBHelper.vintPrefixed(PBHelper.java:2294)
at org.apache.hadoop.hdfs.protocol.datatransfer.PipelineAck.readFields(PipelineAck.java:244)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer$ResponseProcessor.run(DFSOutputStream.java:847)
2019-07-11 02:58:34,423 WARN  [JvmPauseMonitor] util.JvmPauseMonitor: Detected pause in JVM or host machine (eg GC): pause of approximately 78980ms
GC pool 'ParNew' had collection(s): count=1 time=3603ms
GC pool 'ConcurrentMarkSweep' had collection(s): count=1 time=75581ms
2019-07-11 02:58:34,337 WARN  [regionserver/slave3/192.168.202.19:16020] util.Sleeper: We slept 81967ms instead of 3000ms, this is likely due to a long garbage collecting pause and it's usually bad, see http://hbase.apache.org/book.html#trouble.rs.runtime.zkexpired
2019-07-11 02:58:34,329 WARN  [DataStreamer for file /hbase-1.4.9-data/data/default/DATACENTER.QUEST_SCORE_MODEL/ee69c3e031c6acc061e50f231f828fdb/.tmp/757f314d0b02456081bf101595e81213 block BP-1486468690-127.0.0.1-1550458385182:blk_1076360262_2619646] hdfs.DFSClient: DataStreamer Exception
java.io.IOException: Broken pipe
at sun.nio.ch.FileDispatcherImpl.write0(Native Method)
at sun.nio.ch.SocketDispatcher.write(SocketDispatcher.java:47)
at sun.nio.ch.IOUtil.writeFromNativeBuffer(IOUtil.java:93)
at sun.nio.ch.IOUtil.write(IOUtil.java:65)
at sun.nio.ch.SocketChannelImpl.write(SocketChannelImpl.java:471)
at org.apache.hadoop.net.SocketOutputStream$Writer.performIO(SocketOutputStream.java:63)
at org.apache.hadoop.net.SocketIOWithTimeout.doIO(SocketIOWithTimeout.java:142)
at org.apache.hadoop.net.SocketOutputStream.write(SocketOutputStream.java:159)
at org.apache.hadoop.net.SocketOutputStream.write(SocketOutputStream.java:117)
at java.io.BufferedOutputStream.write(BufferedOutputStream.java:122)
at java.io.DataOutputStream.write(DataOutputStream.java:107)
at org.apache.hadoop.hdfs.DFSPacket.writeTo(DFSPacket.java:176)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.run(DFSOutputStream.java:613)
2019-07-11 02:58:34,329 WARN  [ResponseProcessor for block BP-1486468690-127.0.0.1-1550458385182:blk_1076359688_2619061] hdfs.DFSClient: DFSOutputStream ResponseProcessor exception  for block BP-1486468690-127.0.0.1-1550458385182:blk_1076359688_2619061
java.io.EOFException: Premature EOF: no length prefix available
at org.apache.hadoop.hdfs.protocolPB.PBHelper.vintPrefixed(PBHelper.java:2294)
at org.apache.hadoop.hdfs.protocol.datatransfer.PipelineAck.readFields(PipelineAck.java:244)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer$ResponseProcessor.run(DFSOutputStream.java:847)
2019-07-11 02:58:34,328 INFO  [slave3,16020,1562643381590_ChoreService_3] hbase.ScheduledChore: Chore: slave3,16020,1562643381590-HeapMemoryTunerChore missed its start time
2019-07-11 02:58:34,423 WARN  [ResponseProcessor for block BP-1486468690-127.0.0.1-1550458385182:blk_1076360276_2619660] hdfs.DFSClient: DFSOutputStream ResponseProcessor exception  for block BP-1486468690-127.0.0.1-1550458385182:blk_1076360276_2619660
java.io.EOFException: Premature EOF: no length prefix available
at org.apache.hadoop.hdfs.protocolPB.PBHelper.vintPrefixed(PBHelper.java:2294)
at org.apache.hadoop.hdfs.protocol.datatransfer.PipelineAck.readFields(PipelineAck.java:244)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer$ResponseProcessor.run(DFSOutputStream.java:847)
2019-07-11 02:58:34,431 WARN  [sync.4] hdfs.DFSClient: Error while syncing
java.io.IOException: All datanodes DatanodeInfoWithStorage[192.168.202.19:50010,DS-796e0cf8-31c6-4fb3-8017-70fcf2f09135,DISK] are bad. Aborting...
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.setupPipelineForAppendOrRecovery(DFSOutputStream.java:1224)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.processDatanodeError(DFSOutputStream.java:990)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.run(DFSOutputStream.java:507)
。。。
2019-07-11 02:58:34,440 INFO  [sync.0] wal.FSHLog: Slow sync cost: 105 ms, current pipeline: [DatanodeInfoWithStorage[192.168.202.19:50010,DS-796e0cf8-31c6-4fb3-8017-70fcf2f09135,DISK]]
2019-07-11 02:58:34,440 INFO  [sync.2] wal.FSHLog: Slow sync cost: 103 ms, current pipeline: [DatanodeInfoWithStorage[192.168.202.19:50010,DS-796e0cf8-31c6-4fb3-8017-70fcf2f09135,DISK]]
2019-07-11 02:58:34,598 INFO  [main-SendThread(slave2:2181)] zookeeper.ClientCnxn: Opening socket connection to server slave2/192.168.202.19.155:2181. Will not attempt to authenticate using SASL (unknown error)
2019-07-11 02:58:34,599 INFO  [main-SendThread(slave2:2181)] zookeeper.ClientCnxn: Socket connection established to slave2/192.168.202.19.155:2181, initiating session
2019-07-11 02:58:34,610 WARN  [main-SendThread(slave2:2181)] zookeeper.ClientCnxn: Unable to reconnect to ZooKeeper service, session 0x16bc2ec8e9e0177 has expired
2019-07-11 02:58:34,611 INFO  [main-SendThread(slave2:2181)] zookeeper.ClientCnxn: Unable to reconnect to ZooKeeper service, session 0x16bc2ec8e9e0177 has expired, closing socket connection
2019-07-11 02:58:34,611 FATAL [main-EventThread] regionserver.HRegionServer: ABORTING region server slave3,16020,1562643381590: regionserver:16020-0x16bc2ec8e9e0177, quorum=master:2181,slave1:2181,slave2:2181,slave3:2181,slave4:2181, baseZNode=/hbase regionserver:16020-0x16bc2ec8e9e0177 received expired from ZooKeeper, aborting
org.apache.zookeeper.KeeperException$SessionExpiredException: KeeperErrorCode = Session expired
at org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher.connectionEvent(ZooKeeperWatcher.java:692)
at org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher.process(ZooKeeperWatcher.java:624)
at org.apache.hadoop.hbase.zookeeper.PendingWatcher.process(PendingWatcher.java:40)
at org.apache.zookeeper.ClientCnxn$EventThread.processEvent(ClientCnxn.java:530)
at org.apache.zookeeper.ClientCnxn$EventThread.run(ClientCnxn.java:505)
2019-07-11 02:58:34,611 FATAL [main-EventThread] regionserver.HRegionServer: RegionServer abort: loaded coprocessors are: [org.apache.phoenix.coprocessor.ScanRegionObserver, org.apache.phoenix.coprocessor.UngroupedAggregateRegionObserver, org.apache.phoenix.hbase.index.Indexer, org.apache.phoenix.coprocessor.GroupedAggregateRegionObserver, org.apache.phoenix.coprocessor.ServerCachingEndpointImpl, org.apache.hadoop.hbase.coprocessor.MultiRowMutationEndpoint]
2019-07-11 02:58:34,705 INFO  [regionserver/slave3/192.168.202.19:16020-longCompactions-1562643385477-SendThread(master:2181)] zookeeper.ClientCnxn: Opening socket connection to server master/10.10.22.127:2181. Will not attempt to authenticate using SASL (unknown error)
2019-07-11 02:58:34,706 INFO  [regionserver/slave3/192.168.202.19:16020-longCompactions-1562643385477-SendThread(master:2181)] zookeeper.ClientCnxn: Socket connection established to master/10.10.22.127:2181, initiating session
2019-07-11 02:58:34,780 INFO  [regionserver/slave3/192.168.202.19:16020-longCompactions-1562643385477-SendThread(master:2181)] zookeeper.ClientCnxn: Session establishment complete on server master/10.10.22.127:2181, sessionid = 0x369e6926a042130, negotiated timeout = 90000
2019-07-11 02:58:34,847 WARN  [regionserver/slave3/192.168.202.19:16020-shortCompactions-1562643620766] compactions.DefaultCompactor: Failed to close the writer after an unfinished compaction.
java.io.IOException: All datanodes DatanodeInfoWithStorage[192.168.202.19:50010,DS-796e0cf8-31c6-4fb3-8017-70fcf2f09135,DISK] are bad. Aborting...
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.setupPipelineForAppendOrRecovery(DFSOutputStream.java:1224)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.processDatanodeError(DFSOutputStream.java:990)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer.run(DFSOutputStream.java:507)
2019-07-11 02:58:34,985 FATAL [regionserver/slave3/192.168.202.19:16020.logRoller] regionserver.LogRoller: Aborting
java.io.IOException: cannot get log writer
at org.apache.hadoop.hbase.wal.DefaultWALProvider.createWriter(DefaultWALProvider.java:395)
at org.apache.hadoop.hbase.regionserver.wal.FSHLog.createWriterInstance(FSHLog.java:742)
at org.apache.hadoop.hbase.regionserver.wal.FSHLog.rollWriter(FSHLog.java:707)
at org.apache.hadoop.hbase.regionserver.LogRoller.run(LogRoller.java:173)
at java.lang.Thread.run(Thread.java:745)
Caused by: java.io.FileNotFoundException: Parent directory doesn't exist: /hbase-1.4.9-data/WALs/slave3,16020,1562643381590
at org.apache.hadoop.hdfs.server.namenode.FSDirectory.verifyParentDir(FSDirectory.java:1776)
at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.startFileInternal(FSNamesystem.java:2547)
at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.startFileInt(FSNamesystem.java:2479)
at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.startFile(FSNamesystem.java:2363)
at org.apache.hadoop.hdfs.server.namenode.NameNodeRpcServer.create(NameNodeRpcServer.java:624)
at org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolServerSideTranslatorPB.create(ClientNamenodeProtocolServerSideTranslatorPB.java:398)
at org.apache.hadoop.hdfs.protocol.proto.ClientNamenodeProtocolProtos$ClientNamenodeProtocol$2.callBlockingMethod(ClientNamenodeProtocolProtos.java)
at org.apache.hadoop.ipc.ProtobufRpcEngine$Server$ProtoBufRpcInvoker.call(ProtobufRpcEngine.java:616)
at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:982)
at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2217)
at org.apache.hadoop.ipc.Server$Handler$1.run(Server.java:2213)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAs(Subject.java:422)
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1762)
at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2211)
at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62)
at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
at java.lang.reflect.Constructor.newInstance(Constructor.java:423)
at org.apache.hadoop.ipc.RemoteException.instantiateException(RemoteException.java:106)
at org.apache.hadoop.ipc.RemoteException.unwrapRemoteException(RemoteException.java:73)
at org.apache.hadoop.hdfs.DFSOutputStream.newStreamForCreate(DFSOutputStream.java:1839)
at org.apache.hadoop.hdfs.DFSClient.create(DFSClient.java:1689)
at org.apache.hadoop.hdfs.DFSClient.create(DFSClient.java:1658)
at org.apache.hadoop.hdfs.DistributedFileSystem$8.doCall(DistributedFileSystem.java:493)
at org.apache.hadoop.hdfs.DistributedFileSystem$8.doCall(DistributedFileSystem.java:489)
at org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81)
at org.apache.hadoop.hdfs.DistributedFileSystem.createNonRecursive(DistributedFileSystem.java:504)
at org.apache.hadoop.fs.FileSystem.createNonRecursive(FileSystem.java:1115)
at org.apache.hadoop.fs.FileSystem.createNonRecursive(FileSystem.java:1091)
at org.apache.hadoop.hbase.regionserver.wal.ProtobufLogWriter.init(ProtobufLogWriter.java:90)
at org.apache.hadoop.hbase.wal.DefaultWALProvider.createWriter(DefaultWALProvider.java:384)
... 4 more
。。。


2.异常分析


从异常信息来看，先是有大量Compactions日志信息，然后hdfs.DFSClient警告

紧接着抛出

java.io.EOFException: Premature EOF: no length prefix available
at org.apache.hadoop.hdfs.protocolPB.PBHelper.vintPrefixed(PBHelper.java:2294)
at org.apache.hadoop.hdfs.protocol.datatransfer.PipelineAck.readFields(PipelineAck.java:244)
at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer$ResponseProcessor.run(DFSOutputStream.java:847)


可见是由于hdfs异常抛出来的

所以我去hadoop查看了datanode的日志,搜索了一下ERROR信息

2019-07-11 02:51:18,969 WARN org.apache.hadoop.hdfs.server.datanode.DataNode: The downstream error might be due to congestion in upstream including this node. Propagating the error: 
2019-07-11 02:51:18,989 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1:42028 dst: /192.16.7.1.155:50010
2019-07-11 02:51:19,538 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1.103:51408 dst: /192.16.7.1.155:50010
2019-07-11 02:54:39,243 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1.159:60682 dst: /192.16.7.1.155:50010
2019-07-11 02:54:45,280 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1.159:60788 dst: /192.16.7.1.155:50010
2019-07-11 02:54:45,289 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1:42302 dst: /192.16.7.1.155:50010
2019-07-11 02:54:45,295 WARN org.apache.hadoop.hdfs.server.datanode.DataNode: The downstream error might be due to congestion in upstream including this node. Propagating the error: 
2019-07-11 02:54:45,297 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1:42320 dst: /192.16.7.1.155:50010
2019-07-11 03:06:59,338 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1.127:52564 dst: /192.16.7.1.155:50010
2019-07-11 03:06:59,390 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1.159:33306 dst: /192.16.7.1.155:50010
2019-07-11 03:06:59,392 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1.103:53360 dst: /192.16.7.1.155:50010
2019-07-11 04:39:09,382 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1.155:44490 dst: /192.16.7.1.155:50010
2019-07-11 04:39:22,059 ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: slave2:50010:DataXceiver error processing WRITE_BLOCK operation  src: /192.16.7.1.155:46914 dst: /192.16.7.1.155:50010

字面意思就是dataXceiver在写的时候操作异常

大量hbase compaction操作的时候就导致了hdfs 的data 写错误，也许是hdfs压力过大


3.解决方法


修改数据传输线程个数和文件个数

网上很多帖子说修改dfs.datanode.max.xcievers，这是老版本的参数，目前使用的版本都是dfs.datanode.max.transfer.threads


hdfs-site.xml

<property>
	<name>dfs.datanode.max.transfer.threads</name>
	<value>8192</value>
	<description>
		默认值4091
		Specifies the maximum number of threads to use for transferring data
		in and out of the DN.
	</description>
</property>
<property>
	<name>dfs.namenode.handler.count</name>
	<value>20</value>
	<description>
		默认值10
		The number of server threads for the namenode.
	</description>
</property>
<property>
	<name>dfs.datanode.handler.count</name>
	<value>20</value>
	<description>
		默认值10
		The number of server threads for the datanode.
	</description>
</property>

同时我也调大了用于处理RPC调用的线程数量。


每个节点修改后 刷新配置

hdfs dfsadmin -refreshSuperUserGroupsConfiguration

登录后即可回复 登录 | 注册
    
相关文章
java hadoop异常no filesystem for scheme hdfsspark操作mongodb异常 cannot cast bsonvaluehadoop异常there are 0 datanode s running and no node s are excludedhbase节点经常挂掉 异常premature eof no length prefix available解决spark shell 写入hbase 异常job in state define instead of runningspark rdd写入数据到hbase nullpointerexception异常hbase regionserver挂掉all datanodes datanodeinfowithstorage x disk are bad. abortingjdbc连接phoenix hbase 异常the node /hbase is not in zookeeper如何解决kafka 异常 no resolvable bootstrap urls given in bootstrap.serversjava jdbc连接phoenix hbase异常clientpath null serverpath null finished falsecss length php ncurses 函数 add string with specified length at current positionphp sqlite 函数 returns whether more rows are availablephp mongocursor sets arbitrary flags in case there is no method available the specific flagphp solrquery restrict matches to terms that start with the prefixphp splfileobject truncates the file to a given lengthphp swoole lock try to lock a read write lock for reading and return straight away even the lock is not available.php tidynode checks if a node represents text no markup python异常No module named 'win32com'解决 nacos docker mysql8 no No Datasource set问题
关注编程学问公众号