在 IBM CAMP 服务器上实现开源 Hadoop HDFS 的高可用性-4

look_w

安装配置 Hadoop2.2.0,实现自动方式 NameNode HA 的切换
下面进行自动切换配置。
在 hdfs-site.xml 里添加：
<!-- "ha.zookeeper.quorum" 打开自动切换 NameNode 的功能。-->

1 2 3 4

<property>         <name>dfs.ha.automatic-failover.enabled</name>         <value>true</value> </property>

在 core-site.xml 里添加：
<!-- "ha.zookeeper.quorum" 指定用于 HA 的 ZooKeeper 集群机器列表，并且列表数目一定是奇数。-->

1 2 3 4

<property>         <name>ha.zookeeper.quorum</name>         <value>plinux09:2181,plinux10:2181,plinux11:2181</value> </property>

配置完上面两个文件，分别在 plinux09、plinux10、plinux11 上启动 ZooKeeper Server。

1 2 3 4 5 6 7 8 9 10 11 12 13 14

[hadoop@plinux09 hadoop]$ zkServer.sh start JMX enabled by default Using config: /bigdata/zookeeper/bin/../conf/zoo.cfg Starting zookeeper ... STARTED [hadoop@plinux10 ~]$ zkServer.sh start JMX enabled by default Using config: /bigdata/zookeeper/bin/../conf/zoo.cfg Starting zookeeper ... STARTED [hadoop@plinux11 ~]$ zkServer.sh start JMX enabled by default Using config: /bigdata/zookeeper/bin/../conf/zoo.cfg Starting zookeeper ... STARTED

然后格式化 ZKFC

1 2 3 4 5 6 7 8 9 10 11 12 13 14

[hadoop@plinux09 hadoop]$  hdfs zkfc -formatZK =============================================== The configured parent znode /hadoop-ha/mycluster already exists. Are you sure you want to clear all failover information from ZooKeeper? WARNING: Before proceeding, ensure that all HDFS services and failover controllers are stopped! =============================================== Proceed formatting /hadoop-ha/mycluster? (Y or N) Y 14/01/13 02:19:24 INFO ha.ActiveStandbyElector: Recursively deleting /hadoop-ha/mycluster from ZK... 14/01/13 02:19:24 INFO ha.ActiveStandbyElector: Successfully deleted /hadoop-ha/mycluster from ZK. 14/01/13 02:19:24 INFO ha.ActiveStandbyElector: Successfully created /hadoop-ha/mycluster in ZK. 14/01/13 02:19:24 INFO zookeeper.ZooKeeper: Session: 0x2438a6d6f0a0000 closed 14/01/13 02:19:24 INFO zookeeper.ClientCnxn: EventThread shut down

* 注意：格式化 ZooKeeper 之前必须先关闭 HDFS Services。运行用 dfs-stop.sh 命令来关闭。
格式化完毕后启动 NameNode,DataNode, journalnode 和 zkfc。

1 2 3 4 5 6 7 8 9 10 11 12 13 14

[hadoop@plinux09 hadoop]$ start-dfs.sh Starting namenodes on [plinux09 plinux10] plinux10: starting namenode, logging to /bigdata/hadoop/logs/hadoop-hadoop-namenode-plinux10.out plinux09: starting namenode, logging to /bigdata/hadoop/logs/hadoop-hadoop-namenode-plinux09.out plinux11: starting datanode, logging to /bigdata/hadoop/logs/hadoop-hadoop-datanode-plinux11.out plinux09: starting datanode, logging to /bigdata/hadoop/logs/hadoop-hadoop-datanode-plinux09.out plinux10: starting datanode, logging to /bigdata/hadoop/logs/hadoop-hadoop-datanode-plinux10.out Starting journal nodes [plinux09 plinux10 plinux11] plinux11: starting journalnode, logging to /bigdata/hadoop/logs/hadoop-hadoop-journalnode-plinux11.out plinux10: starting journalnode, logging to /bigdata/hadoop/logs/hadoop-hadoop-journalnode-plinux10.out plinux09: starting journalnode, logging to /bigdata/hadoop/logs/hadoop-hadoop-journalnode-plinux09.out Starting ZK Failover Controllers on NN hosts [plinux09 plinux10] plinux10: starting zkfc, logging to /bigdata/hadoop/logs/hadoop-hadoop-zkfc-plinux10.out plinux09: starting zkfc, logging to /bigdata/hadoop/logs/hadoop-hadoop-zkfc-plinux09.out

启动好后，其中一台 NameNode 会自动被激活成 active，另一台是 standby 状态。

1 2 3 4

[hadoop@plinux09 hadoop]$ hdfs haadmin -getServiceState nn1 Standby [hadoop@plinux09 hadoop]$ hdfs haadmin -getServiceState nn2 Active

通过 Kill NameNode 的进程来进行 NameNode HA 的自动切换测试。

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

[hadoop@plinux10 ~]$ ps -ef|grep -i namenode hadoop    9378     1  1 02:23 ?        00:00:18 /bigdata/ibm-java-ppc64-70/bin/java – *** org.apache.hadoop.hdfs.server.namenode.NameNode [hadoop@plinux10 ~]$ kill -9 9378 [hadoop@plinux09 conf]$ hdfs haadmin -getServiceState nn1 Active [hadoop@plinux09 conf]$ hdfs haadmin -getServiceState nn2 14/01/13 02:56:09 INFO ipc.Client: Retrying connect to server: plinux10/9.110.75.105:8020. Already tried 0 time(s); retry policy is RetryUpToMaximumCountWithFixedSleep(        maxRetries=1, sleepTime=1 SECONDS) Operation failed: Call From plinux09/9.110.75.104 to plinux10:8020 failed on connection exception: java.net.ConnectException: Connection refused; For more details see:    http://wiki.apache.org/hadoop/ConnectionRefused

当 kill plinux10 上 active NameNode 的进程后，在短时间内 NameNode 自动切换到 plinux09 上。
基于 HDFS HA 来配置 Hbase
基于 HDFS HA 来配置 Hbase 的方法如下：
配置 hbase-site.xml 文件，添加如下内容

1	[hadoop@plinux09 bigdata]$ vim /bigdata/hbase/conf/hbase-site.xml

<!-- "hbase.rootdir"用于 HA 节点的 NameService 来配置 HBase。-->

1 2 3 4 5 6 7 8

<property>                 <name>hbase.rootdir</name>                 <value>hdfs://mycluster/hbase</value>         </property> <property>                 <name>hbase.master</name>                 <value>hdfs://mycluster:60000</value>         </property>

配置好后启动 Hbase 服务。

1 2 3 4 5 6

[hadoop@plinux10 ~]$ start-hbase.sh starting master, logging to /bigdata/hbase/logs/hbase-hadoop-master-plinux10.out plinux12: starting regionserver, logging to /bigdata/hbase/logs/hbase-hadoop-regionserver-plinux12.out plinux11: starting regionserver, logging to /bigdata/hbase/logs/hbase-hadoop-regionserver-plinux11.out plinux09: starting regionserver, logging to /bigdata/hbase/logs/hbase-hadoop-regionserver-plinux09.out plinux10: starting regionserver, logging to /bigdata/hbase/logs/hbase-hadoop-regionserver-plinux10.out

图 7.登录 plinux10 的 Web 查看 HBase 的状态通过以上配置，Hbase 可以在 HDFS HA 的环境中正常工作。这样一套基本的 Hadoop HDFS 的 HA 解决方案配置完成。
总结随着 Hadoop 日渐成熟，越来越多的企业使用它来分析处理日常大量的数据。随之而来的安全性，可靠性和性能问题被提出。本文使用开源的软件结合 IBM CAMP 服务器的强大性能实现高可用性来满足客户的需求。