安装ambari的时候遇到的ambari和hadoop问题集

5.在安装的时候遇到的问题

5.1使用ambari-server start的时候出现ERROR: Exiting with exit code -1.

5.1.1REASON: Ambari Server java process died with exitcode 255. Check /var/log/ambari-server/ambari-server.out for more information

 

解决：

由于是重新安装，所以在使用/etc/init.d/postgresql  initdb初始化数据库的时候会出现这个错误，所以需要

先用yum –y remove postgresql*命令把postgresql卸载

然后把/var/lib/pgsql/data目录下的文件全部删除

然后再配置postgresql数据库（执行1.6章节内容）

然后再次安装（3章节内容）

 

5.1.2在日志中有如下错误：ERROR [main] AmbariServer:820 - Failed to run the Ambari Server

 

com.google.inject.ProvisionException: Guice provision errors:

 

1) Error injecting method, java.lang.NullPointerException

  at org.apache.ambari.server.api.services.AmbariMetaInfo.init(AmbariMetaInfo.java:243)

  at org.apache.ambari.server.api.services.AmbariMetaInfo.class(AmbariMetaInfo.java:125)

  while locating org.apache.ambari.server.api.services.AmbariMetaInfo

    for field at org.apache.ambari.server.controller.AmbariServer.ambariMetaInfo(AmbariServer.java:145)

  at org.apache.ambari.server.controller.AmbariServer.class(AmbariServer.java:145)

  while locating org.apache.ambari.server.controller.AmbariServer

 

1 error

        at com.google.inject.internal.InjectorImpl$4.get(InjectorImpl.java:987)

        at com.google.inject.internal.InjectorImpl.getInstance(InjectorImpl.java:1013)

        at org.apache.ambari.server.controller.AmbariServer.main(AmbariServer.java:813)

Caused by: java.lang.NullPointerException

        at org.apache.ambari.server.stack.StackModule.processRepositories(StackModule.java:665)

        at org.apache.ambari.server.stack.StackModule.resolve(StackModule.java:158)

        at org.apache.ambari.server.stack.StackManager.fullyResolveStacks(StackManager.java:201)

        at org.apache.ambari.server.stack.StackManager.(StackManager.java:119)

        at org.apache.ambari.server.stack.StackManager$$FastClassByGuice$$33e4ffe0.newInstance()

        at com.google.inject.internal.cglib.reflect.$FastConstructor.newInstance(FastConstructor.java:40)

        at com.google.inject.internal.DefaultConstructionProxyFactory$1.newInstance(DefaultConstructionProxyFactory.java:60)

        at com.google.inject.internal.ConstructorInjector.construct(ConstructorInjector.java:85)

        at com.google.inject.internal.ConstructorBindingImpl$Factory.get(ConstructorBindingImpl.java:254)

        at com.google.inject.internal.InjectorImpl$4$1.call(InjectorImpl.java:978)

        at com.google.inject.internal.InjectorImpl.callInContext(InjectorImpl.java:1031)

        at com.google.inject.internal.InjectorImpl$4.get(InjectorImpl.java:974)

        at com.google.inject.assistedinject.FactoryProvider2.invoke(FactoryProvider2.java:632)

        at com.sun.proxy.$Proxy26.create(Unknown Source)

        at org.apache.ambari.server.api.services.AmbariMetaInfo.init(AmbariMetaInfo.java:247)

5.2安装HDFS和HBASE的时候出现/usr/hdp/current/hadoop-client/conf  doesn't exist

5.2.1/etc/Hadoop/conf文件链接存在

是由于/etc/hadoop/conf和/usr/hdp/current/hadoop-client/conf目录互相链接，造成死循环，所以要改变一个的链接

cd /etc/hadoop

rm -rf conf

ln -s /etc/hadoop/conf.backup /etc/hadoop/conf

 

HBASE也会遇到同样的问题，解决方式同上

cd /etc/hbase

rm -rf conf

ln -s /etc/hbase/conf.backup /etc/hbase/conf

 

ZooKeeper也会遇到同样的问题，解决方式同上

cd /etc/zookeeper

rm -rf conf

ln -s /etc/zookeeper/conf.backup /etc/zookeeper/conf

 

5.2.2/etc/Hadoop/conf文件链接不存在

查看正确的配置，发现缺少两个目录文件config.backup和2.4.0.0-169，把文件夹拷贝到/etc/hadoop目录下

 

重新创建/etc/hadoop目录下的conf链接：

cd /etc/hadoop

rm -rf conf

ln -s /usr/hdp/current/hadoop-client/conf conf

 

问题解决

 

5.3在认证机器(Confirm Hosts)的时候出现错误Ambari agent machine hostname (localhost) does not match expected ambari server hostname

Ambari配置时在Confirm Hosts的步骤时，中间遇到一个很奇怪的问题：总是报错误：

Ambari agent machine hostname (localhost.localdomain) does not match expected ambari server hostname (xxx).

后来修改的/etc/hosts文件中

 

修改前：

127.0.0.1   localhost dsj-kj1
::1         localhost dsj-kj1

10.13.39.32     dsj-kj1

10.13.39.33     dsj-kj2

10.13.39.34     dsj-kj3

10.13.39.35     dsj-kj4

10.13.39.36     dsj-kj5

修改后：

127.0.0.1  localhost localhost.localdomain localhost4 localhost4.localdomain4
::1          localhost localhost.localdomain localhost6 localhost6.localdomain6

10.13.39.32     dsj-kj1

10.13.39.33     dsj-kj2

10.13.39.34     dsj-kj3

10.13.39.35     dsj-kj4

10.13.39.36     dsj-kj5

感觉应该是走的ipv6协议，很奇怪，不过修改后就可以了。

5.4ambary-server重装

删除使用脚本删除

注意删除后要安装两个系统组件

yum -y install ruby*

yum -y install redhat-lsb*

yum -y install snappy*

 

安装参考3

 

5.5Ambari连接mysql设置

在主节点把mysql数据库连接包拷贝在/var/lib/ambary-server/resources目录下并改名为mysql-jdbc-driver.jar

cp /usr/share/java/mysql-connector-java-5.1.17.jar /var/lib/ambari-server/resources/mysql-jdbc-driver.jar

 

再在图形界面下启动hive

5.6在注册机器(Confirm Hosts)的时候出现错误Failed to start ping port listener of: [Errno 98] Address already in use

 

某个端口或者进程一直陪占用

解决方法：
发现df命令一直执行没有完成，

[root@testserver1 ~]# netstat -lanp|grep 8670
tcp        0      0 0.0.0.0:8670                0.0.0.0:*                   LISTEN      2587/df

[root@testserver1 ~]# kill -9 2587
kill后，再重启ambari-agent问题解决

[root@testserver1 ~]# service ambari-agent restart
Verifying Python version compatibility...
Using python  /usr/bin/python2.6
ambari-agent is not running. No PID found at /var/run/ambari-agent/ambari-agent.pid
Verifying Python version compatibility...
Using python  /usr/bin/python2.6
Checking for previously running Ambari Agent...
Starting ambari-agent
Verifying ambari-agent process status...
Ambari Agent successfully started
Agent PID at: /var/run/ambari-agent/ambari-agent.pid
Agent out at: /var/log/ambari-agent/ambari-agent.out
Agent log at: /var/log/ambari-agent/ambari-agent.log

5.7在注册机器(Confirm Hosts)的时候出现错误The following hosts have Transparent HugePages (THP) enabled。THP should be disabled to avoid potential Hadoop performance issues

解决方法：
在Linux下执行：

echo never >/sys/kernel/mm/redhat_transparent_hugepage/defrag

echo never >/sys/kernel/mm/redhat_transparent_hugepage/enabled

echo never >/sys/kernel/mm/transparent_hugepage/enabled

echo never >/sys/kernel/mm/transparent_hugepage/defrag

 

5.8启动hive的时候出现错误unicodedecodeerror ambari in position 117

 

查看/etc/sysconfig/i18n文件，发现内容如下：

LANG=”zh_CN.UTF8”

原来系统字符集设置成了中文，改成如下内容，问题解决：

LANG="en_US.UTF-8"

 

 

5.9安装Metrics的时候报如下错误，安装包找不到

1.failure: Updates-ambari-2.2.1.0/ambari/ambari-metrics-monitor-2.2.1.0-161.x86_64.rpm from HDP-UTILS-1.1.0.20: [Errno 256] No more mirrors to try.

 

在ftp源服务器上执行命令：

cd /var/www/html/ambari/HDP-UTILS-1.1.0.20/repos/centos6

mkdir Updates-ambari-2.2.1.0

cp -r /var/www/html/ambari/Updates-ambari-2.2.1.0/ambari /var/www/html/ambari/HDP-UTILS-1.1.0.20/repos/centos6/Updates-ambari-2.2.1.0

 

然后重新生成repodata

cd /var/www/html/ambari

rm -rf repodata

createrepo ./

 

2.failure: HDP-UTILS-1.1.0.20/repos/centos6/Updates-ambari-2.2.1.0/ambari/ambari-metrics-monitor-2.2.1.0-161.x86_64.rpm from HDP-UTILS-1.1.0.20: [Errno 256] No more mirrors to try.

 

在/etc/yum.repos.d目录下删除mnt.repo，并使用yum clean all命令来清空yum的缓存

cd /ec/yum.repos.d

rm -rf mnt.repo

yum clean all

5.11jps 报process information unavailable解决办法

4791 -- process information unavailable

 

解决办法：

 

进入tmp目录，

 

cd /tmp

 

删除该目录下

 

名称为hsperfdata_{ username}的文件夹

 

然后jps,清净了。

 

脚本：

cd /tmp

ls -l | grep hsperf | xargs rm -rf

ls -l | grep hsperf

 

5.12namenode启动报错在日志文件中ERROR namenode.NameNode (NameNode.java:main(1712)) - Failed to start namenode

日志中还有java.net.BindException: Port in use: gmaster:50070

Caused by: java.net.BindException: Address already in use

判断原因是50070上一次没有释放，端口占用

 

netstat下time_wait状态的tcp连接： 
1.这是一种处于连接完全关闭状态前的状态； 
2.通常要等上4分钟（windows server）的时间才能完全关闭； 
3.这种状态下的tcp连接占用句柄与端口等资源，服务器也要为维护这些连接状态消耗资源； 
4.解决这种time_wait的tcp连接只有让服务器能够快速回收和重用那些TIME_WAIT的资源：修改注册表[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\services\Tcpip\Parameters]添加dword值TcpTimedWaitDelay=30（30也为微软建议值；默认为2分钟）和MaxUserPort：65534(可选值5000 - 65534)； 
5.具体tcpip连接参数配置还可参照这里：http://technet.microsoft.com/zh-tw/library/cc776295%28v=ws.10%29.aspx 
6.linux下： 
vi /etc/sysctl.conf 
新增如下内容： 
net.ipv4.tcp_tw_reuse = 1 
net.ipv4.tcp_tw_recycle = 1 
net.ipv4.tcp_syncookies=1 

net.ipv4.tcp_fin_timeout=30

net.ipv4.tcp_keepalive_time=1800

net.ipv4.tcp_max_syn_backlog=8192

使内核参数生效： 
[root@web02 ~]# sysctl -p 
readme: 
net.ipv4.tcp_syncookies=1 打开TIME-WAIT套接字重用功能，对于存在大量连接的Web服务器非常有效。 
net.ipv4.tcp_tw_recyle=1 
net.ipv4.tcp_tw_reuse=1 减少处于FIN-WAIT-2连接状态的时间，使系统可以处理更多的连接。 
net.ipv4.tcp_fin_timeout=30 减少TCP KeepAlive连接侦测的时间，使系统可以处理更多的连接。 
net.ipv4.tcp_keepalive_time=1800 增加TCP SYN队列长度，使系统可以处理更多的并发连接。 
net.ipv4.tcp_max_syn_backlog=8192

5.13在启动的时候报错误resource_management.core.exceptions.Fail: Execution of 'ambari-sudo.sh  -H -E /usr/hdp/current/hadoop-client/sbin/hadoop-daemon.sh

在日志中有如下内容：

2016-03-31 13:55:28,090 INFO  security.ShellBasedIdMapping (ShellBasedIdMapping.java:updateStaticMapping(322)) - Not doing static UID/GID mapping because '/etc/nfs.map' does not exist.

2016-03-31 13:55:28,096 INFO  nfs3.WriteManager (WriteManager.java:(92)) - Stream timeout is 600000ms.

2016-03-31 13:55:28,096 INFO  nfs3.WriteManager (WriteManager.java:(100)) - Maximum open streams is 256

2016-03-31 13:55:28,096 INFO  nfs3.OpenFileCtxCache (OpenFileCtxCache.java:(54)) - Maximum open streams is 256

2016-03-31 13:55:28,259 INFO  nfs3.RpcProgramNfs3 (RpcProgramNfs3.java:(205)) - Configured HDFS superuser is

2016-03-31 13:55:28,261 INFO  nfs3.RpcProgramNfs3 (RpcProgramNfs3.java:clearDirectory(231)) - Delete current dump directory /tmp/.hdfs-nfs

2016-03-31 13:55:28,269 WARN  fs.FileUtil (FileUtil.java:deleteImpl(187)) - Failed to delete file or dir [/tmp/.hdfs-nfs]: it still exists.

说明hdfs这个用户对/tmp没有权限

赋予权限给hdfs用户：

chown  hdfs:hadoop /tmp

 

再启动问题解决

5.14在安装ranger组件的时候出现错误连接不上mysql数据库rangeradmin用户和不能赋权的问题

在数据库中先删除所有rangeradmin用户，注意使用drop user命令：

drop user 'rangeradmin'@'%';

drop user 'rangeradmin'@'localhost';

drop user 'rangeradmin'@'gmaster';

drop user 'rangeradmin'@'gslave1';

drop user 'rangeradmin'@'gslave2';

FLUSH PRIVILEGES;

 

再创建用户（注意gmaster是ranger安装的服务器机器名）