大数据集群遇到的问题（Hadoop、Spark、Hive、kafka、Hbase、Phoenix）

大数据平台中遇到的实际问题，整理了一下，使用CDH5.8版本，包括Hadoop、Spark、Hive、kafka、Hbase、Phoenix、Impala、Sqoop、CDH等问题，初步整理下最近遇到的问题，不定期更新。

启动nodemanager失败

2016-09-07 14:28:46,434 FATAL org.apache.hadoop.yarn.server.nodemanager.NodeManager: Error starting NodeManager
org.apache.hadoop.yarn.exceptions.YarnRuntimeException: java.net.BindException: Problem binding to [0.0.0.0:8040] 
java.net.BindException: Address already in use; For more details see:  http://wiki.apache.org/hadoop/BindException
        at org.apache.hadoop.yarn.factories.impl.pb.RpcServerFactoryPBImpl.getServer(RpcServerFactoryPBImpl.java:139)
        at org.apache.hadoop.yarn.ipc.HadoopYarnProtoRPC.getServer(HadoopYarnProtoRPC.java:65)
        at org.apache.hadoop.yarn.ipc.YarnRPC.getServer(YarnRPC.java:54)
        at org.apache.hadoop.yarn.server.nodemanager.containermanager.localizer.ResourceLocalizationService.createServer(ResourceLocalizationService.java:278)
        at org.apache.hadoop.yarn.server.nodemanager.containermanager.localizer.ResourceLocalizationService.serviceStart(ResourceLocalizationService.java:258)
        at org.apache.hadoop.service.AbstractService.start(AbstractService.java:193)
        at org.apache.hadoop.service.CompositeService.serviceStart(CompositeService.java:121)
        at org.apache.hadoop.yarn.server.nodemanager.containermanager.ContainerManagerImpl.serviceStart
(ContainerManagerImpl.java:293)
        at org.apache.hadoop.service.AbstractService.start(AbstractService.java:193)
        at org.apache.hadoop.service.CompositeService.serviceStart(CompositeService.java:121)
        at org.apache.hadoop.yarn.server.nodemanager.NodeManager.serviceStart(NodeManager.java:199)
        at org.apache.hadoop.service.AbstractService.start(AbstractService.java:193)
        at org.apache.hadoop.yarn.server.nodemanager.NodeManager.initAndStartNodeManager(NodeManager.java:339)

问题原因：8040端口被占用，netstat -apn | grep 8040后发现有nodemanager进程再运行。
解决办法：kill正在运行的nodemanager进程，重新启动nodemanager。成功。

Spark on yarn部署遇到的问题
命令行启动spark-shell --master yarn

org.apache.spark.SparkException: Yarn application has already ended! It might have been killed or unable to launch application master.
	at org.apache.spark.scheduler.cluster.YarnClientSchedulerBackend.waitForApplication(YarnClientSchedulerBackend.scala:124)
	at org.apache.spark.scheduler.cluster.YarnClientSchedulerBackend.start(YarnClientSchedulerBackend.scala:64)
	at org.apache.spark.scheduler.TaskSchedulerImpl.start(TaskSchedulerImpl.scala:144)
	at org.apache.spark.SparkContext.(SparkContext.scala:530)
	at org.apache.spark.repl.SparkILoop.createSparkContext(SparkILoop.scala:1017)
	at $iwC$$iwC.(:15)
	at $iwC.(:24)
	at (:26)
	at .(:30)
	at .()
	at .(:7)
	at .()
	at $print()
	at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
	at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)

applicationMaster日志只有：
/bin/bash: {{JAVA_HOME}}/bin/java: No such file or directory
问题原因：Spark使用的Hadoop版本与与线上Hadoop版本不一致
解决办法：使用线上Hadoop版本对应的Spark，解决问题

启动spark sql server

16/09/08 18:45:26 INFO service.AbstractService: Service:HiveServer2 is started.
16/09/08 18:45:26 INFO thriftserver.HiveThriftServer2: HiveThriftServer2 started
16/09/08 18:45:26 ERROR thrift.ThriftCLIService: Error starting HiveServer2: could not start ThriftBinaryCLIService
org.apache.thrift.transport.TTransportException: Could not create ServerSocket on address /172.16.0.43:10002.
        at org.apache.thrift.transport.TServerSocket.(TServerSocket.java:109)
        at org.apache.thrift.transport.TServerSocket.(TServerSocket.java:91)
        at org.apache.thrift.transport.TServerSocket.(TServerSocket.java:87)
        at org.apache.hive.service.auth.HiveAuthFactory.getServerSocket(HiveAuthFactory.java:241)
        at org.apache.hive.service.cli.thrift.ThriftBinaryCLIService.run(ThriftBinaryCLIService.java:66)
        at java.lang.Thread.run(Thread.java:722)
16/09/08 18:45:26 INFO server.HiveServer2: Shutting down HiveServer2
16/09/08 18:45:26 INFO service.AbstractService: Service:ThriftBinaryCLIService is stopped.
16/09/08 18:45:26 INFO service.AbstractService: Service:OperationManager is stopped.
16/09/08 18:45:26 INFO service.AbstractService: Service:SessionManager is stopped.
16/09/08 18:45:26 INFO service.AbstractService: Service:CLIService is stopped.

问题原因：与Hive中的HiveServer2 进程端口冲突。
解决办法：修改spark配置中的hive-site.xml中的
  hive.server2.thrift.bind.host
  172.16.0.46
将地址改到没有hiveserver服务的节点。

安装impala
启动impala-server服务：[hadoop@master impala]$ sudo service impala-server status

Log line format: [IWEF]mmdd hh:mm:ss.uuuuuu threadid file:line] msg
E0913 15:12:06.521616  5070 logging.cc:103] stderr will be logged to this file.
log4j:WARN No appenders could be found for logger (org.apache.hadoop.metrics2.lib.MutableMetricsFactory).
log4j:WARN Please initialize the log4j system properly.
log4j:WARN See http://logging.apache.org/log4j/1.2/faq.html#noconfig for more info.
E0913 15:12:10.286190  5070 impala-server.cc:210] Could not read the HDFS root directory at hdfs://master:50070. Error was:
Failed on local exception: com.google.protobuf.InvalidProtocolBufferException: Protocol message end-group tag did not match expected tag.; Host Details : local host is: "master/172.16.0.39"; destination host is: "master":50070;
E0913 15:12:10.286229  5070 impala-server.cc:212] Aborting Impala Server startup due to improper configuration

问题原因：是版本问题
解决办法：安装Hadoop2.2.0对应版本CDH版本

使用yum install安装软件时，总会找原来的jar包
问题原因：是yum缓存的原因
解决办法：使用yum clean all清除缓存即可。

部署线上CDH5.8集群，Cdh安装中遇到“正在获取安装锁”，安装时出错了，后来重新装，就遇到了这个问题
问题原因：上次安装时有scm的一些文件遗留。
解决办法：进入/tmp 目录，ls -a查看，删除scm_prepare_node.*的文件，以及.scm_prepare_node.lock文件。

DNS反向解析错误，不能正确解析Cloudera Manager Server主机名

正在检测 Cloudera Manager Server...
 BEGIN host -t PTR 172.16.0.240 
 240.0.16.172.in-addr.arpa domain name pointer localhost. 
END (0) 
 using localhost as scm server hostname 
BEGIN which python 
 /usr/bin/python 
END (0) 
BEGIN python -c 'import socket; import sys; s = socket.socket(socket.AF_INET); s.settimeout(5.0); s.connect((sys.argv[1], int(sys.argv[2]))); s.close();' localhost 7182 
 Traceback (most recent call last): 
 File "", line 1, in  
 File "", line 1, in connect 
socket.error: [Errno 111] Connection refused 
END (1) 
 could not contact scm server at localhost:7182, giving up 
 waiting for rollback request

解决办法：将连不上的机器 /usr/bin/host 文件删掉。mv /usr/bin/host /usr/bin/host.bak

安装Hadoop时，启动失败

java.io.IOException: the pathcomponent: '/data' is world-writable. Its permissions are 0777. Pleasefix this or select a different socket path

问题原因： DataNode的root根目录权限设置为0777太高导致不安全

解决办法：修改为755或者默认权限

使用sqoop导msyql数据到hbase中，mysql数据库字段 tinyint 类型导入到hbase中成了true/false
原因：这属于sqoop的一种错误转换
解决办法：url后面拼上tinyInt1isBit=false

hive执行join操作，会失败
语句：use hivedb;insert into table report.sitef2_tmp_payment_hbase select concat_ws('_', member0.site_id,substr(trade0.pay_time, 0, 10),substr(trade0.pay_time, 0, 10)), member0.site_id,substr(trade0.pay_time, 0, 10) pay_time,count(*) trade_num, sum(trade0.curry_amount) trade_volume from trade_trade_payments trade0 left join member_member member0 on (trade0.pay_status = '20' and trade0.sup_id = member0.member_id) group by substr(trade0.pay_time, 0, 10),member0.site_id

return code 3 from MapredLocalTask
Execution failed with exit status: 3
Obtaining error information
Task failed!
Task ID:
  Stage-4
Logs:
/tmp/root/hive.log
FAILED: Execution Error, return code 3 from org.apache.hadoop.hive.ql.exec.mr.MapredLocalTask

问题原因：join基于hbase的映射表时，join进行了转换
解决办法：set hive.auto.convert.join = false;

hive建表时插入注释时，中文乱码。

解决办法：修改hive元数据表的编码设置
//修改字段注释字符集
alter table COLUMNS_V2 modify column COMMENT varchar(256) character set utf8;
alter table TABLE_PARAMS modify column PARAM_VALUE varchar(4000) character set utf8;
//修改分区注释字符集
alter table PARTITION_KEYS modify column PKEY_COMMENT varchar(4000) character set utf8;
//修改表注释字符集

CDH集群总是使用交换区内存，有时候impala会使用交换区内存比较慢：
解决办法：设置交换区使用
http://blog.csdn.net/huaishu/article/details/8762957
[root@rhce ~]# sysctl vm.swappiness=10
vm.swappiness = 10
[root@rhce ~]# cat /proc/sys/vm/swappiness
10
这里我们的修改已经生效，但是如果我们重启了系统，又会变成60.
--永久修改：
在/etc/sysctl.conf 文件里添加如下参数：
vm.swappiness=10
或者：
[root@rhce ~]# echo 'vm.swappiness=10' >>/etc/sysctl.conf

安装kafka，直接安装官方地址的kafka（https://archive.cloudera.com/kafka/parcels/latest/），下载，激活。
安装好之后启动失败。
问题原因：broker_max_heap_size配置太小。这个安装好之后，cdh配置的大小为256MB，其实默认是1G的。
解决办法：将broker_max_heap_size调大，我改为了512MB，启动的成功。根据具体情况可继续调大该参数。

CDH安装Phoenix，启动Phoenix失败
Apache Phoenix unable to connect to HBase
 ./sqlline.py master1:2181/kafka启动报错

6/10/24 18:42:47 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Error: org.apache.hadoop.hbase.DoNotRetryIOException: Class org.apache.phoenix.coprocessor.MetaDataRegionObserver cannot be loaded Set hbase.table.sanity.checks to false at conf or table descriptor if you want to bypass sanity checks
	at org.apache.hadoop.hbase.master.HMaster.warnOrThrowExceptionForFailure(HMaster.java:1707)
	at org.apache.hadoop.hbase.master.HMaster.sanityCheckTableDescriptor(HMaster.java:1568)
	at org.apache.hadoop.hbase.master.HMaster.createTable(HMaster.java:1497)
	at org.apache.hadoop.hbase.master.MasterRpcServices.createTable(MasterRpcServices.java:468)
	at org.apache.hadoop.hbase.protobuf.generated.MasterProtos$MasterService$2.callBlockingMethod(MasterProtos.java:55682)
	at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2170)
	at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:109)
	at org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:133)
	at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:108)
	at java.lang.Thread.run(Thread.java:745) (state=08000,code=101)
org.apache.phoenix.exception.PhoenixIOException: org.apache.hadoop.hbase.DoNotRetryIOException: Class org.apache.phoenix.coprocessor.MetaDataRegionObserver cannot be loaded Set hbase.table.sanity.checks to false at conf or table descriptor if you want to bypass sanity checks

原因：安装Phoenix时，只把Phoenix的包放到hbase的lib中了，但是没有重启hbase，hbase没有加载Phoenix的相关class
解决办法：重启hbase集群

连接远程kafka时，使用ip地址连接报错：
命令：kafka-console-consumer --zookeeper 172.16.0.242:2181/kafka --topic wk_wordlibrary_title --from-beginning

2016-10-26 14:39:57,949] WARN Fetching topic metadata with correlation id 12 for topics [Set(wk_wordlibrary_title)] from broker [BrokerEndPoint(92,node1,9092)] failed (kafka.client.ClientUtils$)
java.nio.channels.ClosedChannelException
	at kafka.network.BlockingChannel.send(BlockingChannel.scala:110)
	at kafka.producer.SyncProducer.liftedTree1$1(SyncProducer.scala:76)
	at kafka.producer.SyncProducer.kafka$producer$SyncProducer$$doSend(SyncProducer.scala:75)
	at kafka.producer.SyncProducer.send(SyncProducer.scala:120)
	at kafka.client.ClientUtils$.fetchTopicMetadata(ClientUtils.scala:58)
	at kafka.client.ClientUtils$.fetchTopicMetadata(ClientUtils.scala:93)
	at kafka.consumer.ConsumerFetcherManager$LeaderFinderThread.doWork(ConsumerFetcherManager.scala:66)
	at kafka.utils.ShutdownableThread.run(ShutdownableThread.scala:63)
[2016-10-26 14:39:57,949] WARN [console-consumer-89764_t2-1477463994654-bcfb2442-leader-finder-thread], Failed to find leader for Set([wk_wordlibrary_title,0]) (kafka.consumer.ConsumerFetcherManager$LeaderFinderThread)

原因：在client使用的元信息是机器名而不是IP。客户端无法解析这个机器名所以出现了前面的异常。 
解决办法：1.在客户端主机配置远程kafka主机-ip映射
2.也可以在server.properties中配置advertised.host.name（broker ip值）

其他部门使用hdfs存储，占用存储空间较大，将磁盘沾满。（副本数为3）

解决办法：先将hdfs副本数配置设为1，然后使用hdfs dfs -setrep -R 1  /tmp1/logs命令将对应目录的文件副本数更新为1，腾出空间，然后当再写hdfs文件时，设置dfs.replication=1(Hadoop dfs -D dfs.replication=1 -put 70M logs/2)---参考链接：http://blog.csdn.net/lskyne/article/details/8898666
hbase regionserver（master1）启动失败,报错日志

 ABORTING region server master1,60020,1478833728358: Unhandled: org.apache.hadoop.hbase.ClockOutOfSyncException: Server master1,60020,1478833728358 has been rejected; Reported time is too far out of sync with master.  Time difference of 36240ms > max allowed of 30000ms
	at org.apache.hadoop.hbase.master.ServerManager.checkClockSkew(ServerManager.java:401)
	at org.apache.hadoop.hbase.master.ServerManager.regionServerStartup(ServerManager.java:267)
	at org.apache.hadoop.hbase.master.MasterRpcServices.regionServerStartup(MasterRpcServices.java:366)
	at org.apache.hadoop.hbase.protobuf.generated.RegionServerStatusProtos$RegionServerStatusService$2.callBlockingMethod(RegionServerStatusProtos.java:8615)
	at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2170)
	at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:109)
	at org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:133)
	at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:108)
	at java.lang.Thread.run(Thread.java:745) 

原因：master1时钟不同步
解决办法：同步master1时钟(ntpdate master1),启动成功

spark sql server（thriftserver.HiveThriftServer2）执行任务时会停掉

报错：ERROR executor.CoarseGrainedExecutorBackend: Driver 172.16.0.241:43286 disassociated! Shutting down.
	INFO util.ShutdownHookManager: Shutdown hook called
	INFO util.ShutdownHookManager: Deleting directory /data/cdh/yarn/nm/usercache/root/appcache/application_1478760707461_0459/spark-ab6b5ac2-0d5c-4e10-90a1-75b948596b7c
	INFO remote.RemoteActorRefProvider$RemotingTerminator: Remoting shut down.
	ERROR executor.CoarseGrainedExecutorBackend: RECEIVED SIGNAL 15: SIGTERM

原因：从信息上看来不知道是什么问题，但是归根结底还是内存的问题。
解决办法：一是，如上面所说，加大excutor-memory的值，减少executor-cores的数量，问题可以解决。二是，加大executor.overhead的值，但是这样其实并没有解决掉根本的问题。所以如果集群的资源是支持的话，就用1的办法吧。
　　另外，这个错误也出现在partitionBy(new HashPartition(partiton-num))时，如果partiton-num太大或者太小的时候会报这种错误，说白了也是内存的原因，不过这个时候增加内存和overhead没有什么用，得去调整这个partiton-num的值。