理解HBase zookeeper和-Root-/.MET表

1 探究Zookeeper

安装过HBase的开发者一定知道，HBase本身就嵌入了Zookeeper，你可以选择安装嵌入的Zookeeper，也可以安装一个分布式的Zookeeper集群。具体的安装步骤，可以到我的博客里查找。

我们可以手工与Zookeeper交互的方式就是使用HBaseshell命令。使用zk_dump来查看HBase的安装等信息。

[plain] view
plaincopy





<span style="font-size:18px;"> hbase(main):002:0>zk_dump

HBaseis rooted at /hbase

Active master address:hadoop-master,60000,1390921674646

Backup master addresses:

Region server holding ROOT:machine-0,60020,1390921686697

Region servers:

machine-1,60020,1390921690387

machine-2,60020,1390921767658

machine-0,60020,1390921686697

Quorum Server Statistics:

machine-1:2222

Zookeeper version: 3.4.5-1392090, built on 09/30/2012 17:52 GMT

Clients:

/192.168.177.158:55294[1](queued=0,recved=121,sent=135)

/192.168.177.172:54763[0](queued=0,recved=1,sent=0)

/192.168.177.168:59896[1](queued=0,recved=48,sent=48)

/192.168.177.158:55299[1](queued=0,recved=47,sent=47)

Latency min/avg/max: 0/2/42

Received: 217

Sent: 230

Connections: 4

Outstanding:0

Zxid: 0x3300000047

Mode: follower

Node count: 29

machine-0:2222

Zookeeper version: 3.4.5-1392090, built on 09/30/2012 17:52 GMT

Clients:

/192.168.177.172:38860[1](queued=0,recved=75,sent=77)

/192.168.177.172:38855[1](queued=0,recved=322,sent=369)

/192.168.177.167:52643[1](queued=0,recved=78,sent=90)

/192.168.177.172:38900[0](queued=0,recved=1,sent=0)

Latency min/avg/max: 0/2/291

Received: 476

Sent: 536

Connections: 4

Outstanding: 0

Zxid: 0x3300000047

Mode: follower

Node count: 29

machine-2:2222

Zookeeper version: 3.4.5-1392090, built on 09/30/2012 17:52 GMT

Clients:

/192.168.177.172:45693[0](queued=0,recved=1,sent=0)

/192.168.177.172:45687[1](queued=0,recved=13,sent=13)

/192.168.177.167:36053[1](queued=0,recved=43,sent=43)

Latency min/avg/max: 0/0/21

Received: 57

Sent: 56

Connections: 3

Outstanding: 0

Zxid: 0x3300000047

Mode: leader

Node count: 29</span>

根据显示的信息，我们可以查阅到HDFS中的HBase根路径，激活的主机地址域名和端口，root表的主机地址和端口。

[plain] view
plaincopy





<span style="font-size:18px;">HBase is rooted at /hbase

Active master address:hadoop-master,60000,1390921674646

Backup master addresses:

Region server holding ROOT:machine-0,60020,1390921686697</span>

区域服务器地址信息

[plain] view
plaincopy





<span style="font-size:18px;">Region servers:

machine-1,60020,1390921690387

machine-2,60020,1390921767658

machine-0,60020,1390921686697</span>

Zookeeper法定服务器的统计信息，当前的法定服务名称以及Zookeeper版本信息，客户端信息等

[plain] view
plaincopy





<span style="font-size:18px;">Quorum Server Statistics:

machine-1:2222

Zookeeper version: 3.4.5-1392090, built on 09/30/2012 17:52 GMT

Clients:

/192.168.177.158:55294[1](queued=0,recved=121,sent=135)

/192.168.177.172:54763[0](queued=0,recved=1,sent=0)

/192.168.177.168:59896[1](queued=0,recved=48,sent=48)

/192.168.177.158:55299[1](queued=0,recved=47,sent=47)

Latency min/avg/max: 0/2/291

Received: 476

Sent: 536

Connections: 4

Outstanding: 0

Zxid: 0x3300000047

Mode: follower

Node count: 29

machine-2:2222

Zookeeper version: 3.4.5-1392090, built on 09/30/2012 17:52 GMT

Clients:

/192.168.177.172:45693[0](queued=0,recved=1,sent=0)

/192.168.177.172:45687[1](queued=0,recved=13,sent=13)

/192.168.177.167:36053[1](queued=0,recved=43,sent=43)

Latency min/avg/max: 0/0/21

Received: 57

Sent: 56

Connections: 3

Outstanding: 0

Zxid: 0x3300000047

Mode: leader

Node count: 29</span>

当我们尝试去理解系统当前的状态时，这些信息显得尤为重要。那些主机加入到集群中了，那个主机扮演什么角色。更重要的是，个主机充当HBase的-root-表服务。HBase客户端需要这些信息来实现读写操作，而Zookeeper正好可以提供这些信息。

客户端能够制动化地与Zookeeper进行操作交流以及找到区域服务器

（RegisionServer）。为了更好的理解这些信息，我们将探究一下HBase中两个特殊的表-root-和.meta.。

2 探究-ROOT-表

扫描-root-表，查看具体的信息。



从输出的信息可以知道，-root-表包含.meta.表信息。-root-表中仅有一行数据，并且所有的信息是关于region, 这就意味着，整个系统由-root-表管理。这我目前的实例中，一个单一的.meta.表中的region能够包含所有信息。从-root-表中，我们可以知道有4列(regioninfo, server,serverstartcode, v.)数据，.meta.的数据在RegionServer 服务机的machine-0上。由于只有一个region，regioninfo字段中的开始字段startkey和结束字段endkey都为空。

3 探究.META.表

从上一节中，我们知道，regioninfo中包含region的名称，开始关键字startkey，结束关键字endkey，编码过的名称（编码过的名称是系统内部使用的）。如果.meta.表中没有表单定义，会有下列的显示：

hbase(main):030:0> scan '.META.'

ROW COLUMN+CELL

0 row(s) in 5.4180 seconds

一旦有表定义在HBase中，呈现效果会是这样：

[plain] view
plaincopy





<span style="font-size:18px;">hbase(main):012:0>scan '.META.'

ROW COLUMN+CELL

car,,1389254794373.41b48a453column=info:regioninfo, timestamp=1389254795518, value={NAME =>'car,,1389254794373.

79036ed6a28620f95bc3e73. 41b48a45379036ed6a28620f95bc3e73.',STARTKEY => '', ENDKEY => '', ENCODED => 41b48a4

5379036ed6a28620f95bc3e73,}

car,,1389254794373.41b48a453column=info:server, timestamp=1390982716032, value=machine-1:60020

79036ed6a28620f95bc3e73.

car,,1389254794373.41b48a453column=info:serverstartcode, timestamp=1390982716032, value=1390982688688

79036ed6a28620f95bc3e73.

pentaho_mappings,,1389456264column=info:regioninfo, timestamp=1389456267362, value={NAME =>'pentaho_mappings,,1

784.1fa2cc86c50ced5fa39630da389456264784.1fa2cc86c50ced5fa39630dae76fb1a2.', STARTKEY => '', ENDKEY=> '', ENCOD

e76fb1a2. ED =>1fa2cc86c50ced5fa39630dae76fb1a2,}

pentaho_mappings,,1389456264column=info:server, timestamp=1390982716247, value=machine-0:60020

784.1fa2cc86c50ced5fa39630da

e76fb1a2.

pentaho_mappings,,1389456264column=info:serverstartcode, timestamp=1390982716247, value=1390982687824

784.1fa2cc86c50ced5fa39630da

e76fb1a2.

users,,1389925800801.6a5a2b0column=info:regioninfo, timestamp=1389925809127, value={NAME =>'users,,138992580080

170dfd0e19df13849706ec035. 1.6a5a2b0170dfd0e19df13849706ec035.',STARTKEY => '', ENDKEY => '', ENCODED => 6a5a2

b0170dfd0e19df13849706ec035,}

users,,1389925800801.6a5a2b0column=info:server, timestamp=1390982718429, value=machine-1:60020

170dfd0e19df13849706ec035.

users,,1389925800801.6a5a2b0column=info:serverstartcode, timestamp=1390982718429, value=1390982688688

170dfd0e19df13849706ec035.

weblogs,,1389255303284.d2973column=info:regioninfo, timestamp=1389255304879, value={NAME =>'weblogs,,1389255303

dcc1b1ff64cc56a197fdd83ec5e.284.d2973dcc1b1ff64cc56a197fdd83ec5e.', STARTKEY => '', ENDKEY => '',ENCODED => d29

73dcc1b1ff64cc56a197fdd83ec5e,}

weblogs,,1389255303284.d2973column=info:server, timestamp=1390982716256, value=machine-0:60020

dcc1b1ff64cc56a197fdd83ec5e.

weblogs,,1389255303284.d2973column=info:serverstartcode, timestamp=1390982716256, value=1390982687824

dcc1b1ff64cc56a197fdd83ec5e.

4 row(s) in0.3280 seconds</span>

正如你看到的，.META.包含的是表的信息，这里就不详述了。看下列这个结构图便可以理解HBase中这两个表的关系。