环境：
1）3 台机器。我这里是 3 台虚拟机，都安装了 clickhouse。
192.168.0.10 db_server_yayun_01192.168.0.20 db_server_yayun_02192.168.0.30 db_server_yayun_03
If element has 'incl' attribute, then for it's value will be used corresponding substitution from another file.By default, path to file with substitutions is /etc/metrika.xml. It could be changed in config in 'include_from' element.Values for substitutions are specified in /yandex/name_of_substitution elements in that file.
<yandex><clickhouse_remote_servers><perftest_3shards_1replicas><shard><internal_replication>true</internal_replication><replica><host>db_server_yayun_01</host><port>9000</port></replica></shard><shard><replica><internal_replication>true</internal_replication><host>db_server_yayun_02</host><port>9000</port></replica></shard><shard><internal_replication>true</internal_replication><replica><host>db_server_yayun_03</host><port>9000</port></replica></shard></perftest_3shards_1replicas></clickhouse_remote_servers><zookeeper-servers><node index="1"><host>192.168.0.30</host><port>2181</port></node></zookeeper-servers><macros><replica>192.168.0.10</replica></macros><networks><ip>::/0</ip></networks><clickhouse_compression><case><min_part_size>10000000000</min_part_size><min_part_size_ratio>0.01</min_part_size_ratio><method>lz4</method></case></clickhouse_compression></yandex>
3 台机器的配置文件都一样，唯一有区别的是：
<macros><replica>192.168.0.10</replica></macros>
服务器 ip 是多少这里就写多少，其实不写 ip 也没关系，3 台机器不重复就行。这里是复制需要用到的配置。还有 zk 的配置如下：
<zookeeper-servers><node index="1"><host>192.168.0.30</host><port>2181</port></node></zookeeper-servers>
我的 zk 是安装在 30 的机器上面的，只安装了一个实例，生产环境肯定要放到单独的机器，并且配置成集群。配置文件修改好以后 3 台服务器重启。
官方文档给的步骤是：
ClickHouse deployment to clusterClickHouse cluster is a homogenous cluster. Steps to set up:1. Install ClickHouse server on all machines of the cluster2. Set up cluster configs in configuration file3. Create local tables on each instance4. Create a Distributed table
前面 2 步都搞定了，下面创建本地表，再创建 Distributed表。（3台 机器都创建，DDL 不同步，蛋疼，也可以使用 ON Cluster 语法。在一个节点执行即可。）
CREATE TABLE ontime_local (FlightDate Date,Year UInt16) ENGINE = MergeTree(FlightDate, (Year, FlightDate), 8192);CREATE TABLE ontime_all AS ontime_local ENGINE = Distributed(perftest_3shards_1replicas, default, ontime_local, rand())
插入数据（随便一台机器就行）：
:) insert into ontime_all (FlightDate,Year)values('2001-10-12',2001);INSERT INTO ontime_all (FlightDate, Year) VALUESOk.1 rows in set. Elapsed: 0.013 sec.:) insert into ontime_all (FlightDate,Year)values('2002-10-12',2002);INSERT INTO ontime_all (FlightDate, Year) VALUESOk.1 rows in set. Elapsed: 0.004 sec.:) insert into ontime_all (FlightDate,Year)values('2003-10-12',2003);INSERT INTO ontime_all (FlightDate, Year) VALUESOk.
我这里插入了 3 条数据。下面查询看看（任何一台机器都可以）：
:) select * from ontime_all;SELECT *FROM ontime_all┌─FlightDate─┬─Year─┐│ 2001-10-12 │ 2001 │└────────────┴──────┘┌─FlightDate─┬─Year─┐│ 2002-10-12 │ 2002 │└────────────┴──────┘┌─FlightDate─┬─Year─┐│ 2003-10-12 │ 2003 │└────────────┴──────┘→ Progress: 3.00 rows, 12.00 B (48.27 rows/s., 193.08 B/s.)3 rows in set. Elapsed: 0.063 sec.:)
当在其中一台机器上面查询的时候，抓包其他机器可以看见是有请求的。
tcpdump -i any -s 0 -l -w – dst port 9000
那么关闭其中一台机器呢？
:) select * from ontime_all;SELECT *FROM ontime_all┌─FlightDate─┬─Year─┐│ 2001-10-12 │ 2001 │└────────────┴──────┘┌─FlightDate─┬─Year─┐│ 2002-10-12 │ 2002 │└────────────┴──────┘┌─FlightDate─┬─Year─┐│ 2003-10-12 │ 2003 │└────────────┴──────┘↓ Progress: 6.00 rows, 24.00 B (292.80 rows/s., 1.17 KB/s.) Received exception from server:Code: 279. DB::Exception: Received from localhost:9000, ::1. DB::NetException. DB::NetException: All connection tries failed. Log:Code: 210, e.displayText() = DB::NetException: Connection refused: (db_server_yayun_02:9000, 192.168.0.20), e.what() = DB::NetExceptionCode: 210, e.displayText() = DB::NetException: Connection refused: (db_server_yayun_02:9000, 192.168.0.20), e.what() = DB::NetExceptionCode: 210, e.displayText() = DB::NetException: Connection refused: (db_server_yayun_02:9000, 192.168.0.20), e.what() = DB::NetException
可以看见已经抛错了，竟然不是高可用？是的，就是不是高可用，上面的配置是设置 3 个分片，但是没有副本，所以在挂掉一个节点以后，就会存在问题。后面又看到了文档的另外一种配置方法，那就是配置 2 个节点，副本 2 个，经过测试高可用没有问题，另外也是分布式并行查询。感兴趣的同学可以自行测试。
https://clickhouse.yandex/reference_en.html#Distributed
CREATE TABLE ontime_replica (FlightDate Date,Year UInt16) ENGINE = ReplicatedMergeTree('/clickhouse_perftest/tables/ontime_replica','{replica}',FlightDate,(Year, FlightDate),8192);插入数据测试：
insert into ontime_replica (FlightDate,Year)values(‘2018-10-12’,2018);
参考资料：
https://clickhouse.yandex/reference_en.html#Distributed
https://clickhouse.yandex/tutorial.html
关于 ClickHouse 的技术内容，你们还有什么想知道的吗？赶紧留言告诉小编吧！