一. 副本概述

副本的目的主要是保障數(shù)據(jù)的高可用性，ClickHouse中的副本沒有主從之分。所有的副本都是平等的。

副本寫入流程：

二. 副本搭建

1. 實(shí)驗(yàn)環(huán)境

hadoop1(192.168.47.128)
hadoop2(192.168.47.129)

2. 修改配置文件

修改兩臺主機(jī)/etc/clickhouse-server/config.xml配置文件，指定副本使用的zookeeper

<zookeeper><node><host>hadoop1</host><port>2181</port></node>
</zookeeper>

3. 啟動(dòng)服務(wù)

1. 啟動(dòng)hadoop1上的zookeeper服務(wù)

[root@hadoop1 zookeeper-3.9.2]# ./bin/zkServer.sh  start
ZooKeeper JMX enabled by default
Using config: /opt/zookeeper-3.9.2/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED

2. 啟動(dòng)hadoop1和hadoop2的ClickHouse服務(wù)

[root@hadoop1 ~]# systemctl  start clickhouse-server

[root@hadoop2 ~]# systemctl  start clickhouse-server

三. 數(shù)據(jù)同步驗(yàn)證

1. 在hadoop1和hadoop2建表

   # hadoop1
   hadoop1 :) create table t_test (
   id UInt32,
   name String,
   age UInt32
   ) engine =ReplicatedMergeTree('/clickhouse/table/01/t_test','rep_1')
   primary key (id)
   order by (id,name);# hadoop2
   hadoop2 :) create table t_test (
   id UInt32,
   name String,
   age UInt32
   ) engine =ReplicatedMergeTree('/clickhouse/table/01/t_test','rep_2')
   primary key (id)
   order by (id,name);
   

   副本表的引擎為ReplicatedMergeTree，兩張表在zookeeper的節(jié)點(diǎn)路徑是一樣的

2. 在hadoop1執(zhí)行insert語句

   hadoop1 :) insert into t_test values(1, 'name1', 20), (2, 'name2', 30), (3, 'name3', 15);INSERT INTO t_test FORMAT Values
   Query id: 6c7fb969-fbe4-4199-af27-1dbbe3c8be14
   Ok.3 rows in set. Elapsed: 0.031 sec.
   

3. 查看hadoop2數(shù)據(jù)表中的數(shù)據(jù)

   hadoop2 :) select * from t_test;SELECT *
   FROM t_testQuery id: fef23f16-f122-45e6-bfa2-94e784a9e042┌─id─┬─name──┬─age─┐
   1. │  1 │ name1 │  20 │
   2. │  2 │ name2 │  30 │
   3. │  3 │ name3 │  15 │└────┴───────┴─────┘
   3 rows in set. Elapsed: 0.005 sec.
   

   由此可以看出，hadoop1插入的數(shù)據(jù)已經(jīng)完全同步到hadoop2上了。