ClickHouse 高可用集群安装测试

Posted 2021-03-19 wuning

1. 环境

CentOS 7.6

2. 安装前准备

2.1 检查是否支持 SSE 4.2 指令集

grep -q sse4_2 /proc/cpuinfo && echo "SSE 4.2 supported" || echo "SSE 4.2 not supported"

2.2 CentOS 取消文件打开数限制

文件打开数限制

在 /etc/security/limits.conf 文件尾追加
* soft nofile 65536 
* hard nofile 65536 
* soft nproc 131072 
* hard nproc 131072

# "*" 表示所有用户都生效；
# "soft"表示应用软件级别；
# "hard"表示操作系统级别；
# "nofile"表示最大文件打开数
# "nproc"表示最大进程数

用户进程数限制

vim /etc/security/limits.d/20-nproc.conf 

*          soft    nproc     131072
root       soft    nproc     unlimited

重启服务器后生效

临时生效：ulimit -n 65535

2.3 CentOS 取消 SELINUC

vim /etc/selinux/config
SELINUX=disabled

2.4 安装依赖

yum install -y libtool
yum install -y *unixODBC*
yum install libicu.x86_64

3. 安装

官网：https://clickhouse.yandex/
下载地址：http://repo.red-soft.biz/repos/clickhouse/stable/el7/

3.1 单机安装

3.1.1 上传5个文件到服务器，分别安装着5个 rpm 文件

rpm -ivh clickhouse-server-common-1.1.54236-4.el7.x86_64.rpm 
rpm -ivh clickhouse-server-1.1.54236-4.el7.x86_64.rpm 
rpm -ivh clickhouse-compressor-1.1.54236-4.el7.x86_64.rpm 
rpm -ivh clickhouse-debuginfo-1.1.54236-4.el7.x86_64.rpm 
rpm -ivh clickhouse-client-1.1.54236-4.el7.x86_64.rpm 

3.1.2 启动 ClickHouse

前台启动

clickhouse-server --config-file=/etc/clickhouse-server/config.xml

后台启动

nohup clickhouse-server --config-file=/etc/clickhouse-server/config.xml  >/dev/null 2>&1 &
# /dev/null 表示空设备文件；可以将 /dev/null 看作 "黑洞"，所有写入它的东西都会丢失
# 0 表示 stdin 标准输入
# 1 表示 stdout 标准输出
# 2 表示 stderr 标准错误
# 2>&1 表示将标准错误重定向到标准输出。这里表示标准错误也会输出到 /dev/null

3.1.3 client 连接 server

clickhouse-client

3.2 高可用集群

3.2.1 高可用集群架构

ClickHouse集群：
3个节点
1个节点,2个ClickHouse实例
3分片,2副本

• cdh01: 实例1, 端口: tcp 9002, http 8123, 同步端口9009, 类型: 分片1, 副本1
• cdh01: 实例2, 端口: tcp 9001, http 8124, 同步端口9011, 类型: 分片3, 副本2 (cdh03的副本)
• cdh02: 实例1, 端口: tcp 9002, http 8123, 同步端口9009, 类型: 分片2, 副本1
• cdh02: 实例2, 端口: tcp 9001, http 8124, 同步端口9011, 类型: 分片1, 副本2 (cdh01的副本)
• cdh03: 实例1, 端口: tcp 9002, http 8123, 同步端口9009, 类型: 分片3, 副本1
• cdh03: 实例2, 端口: tcp 9001, http 8124, 同步端口9011, 类型: 分片2, 副本2 (cdh02的副本)

3.2.2 高可用集群部署

1. 启动脚本复制一份

cp /etc/rc.d/init.d/clickhouse-server  /etc/rc.d/init.d/clickhouse-server-1

vim /etc/rc.d/init.d/clickhouse-server-1

配置文件使用 config1.xml,pid使用clickhouse-server-1.pid

2.config.xml cdh01,cdh02,cdh03 的实例1

注意端口号 tcp-port 改为 9092,每个节点的 interserver_http_host 不一样

vim /etc/clickhouse-server/config.xml

<?xml version="1.0"?>
<yandex>
    <logger>
        <level>trace</level>
        <log>/var/log/clickhouse-server/clickhouse-server.log</log>
        <errorlog>/var/log/clickhouse-server/clickhouse-server.err.log</errorlog>
        <size>1000M</size>
        <count>10</count>
    </logger>

    <!--http 端口-->
    <http_port>8123</http_port>

   
    <openSSL>
        <server>
            <certificateFile>/etc/clickhouse-server/server.crt</certificateFile>
            <privateKeyFile>/etc/clickhouse-server/server.key</privateKeyFile>
            <dhParamsFile>/etc/clickhouse-server/dhparam.pem</dhParamsFile>
            <verificationMode>none</verificationMode>
            <loadDefaultCAFile>true</loadDefaultCAFile>
            <cacheSessions>true</cacheSessions>
            <disableProtocols>sslv2,sslv3</disableProtocols>
            <preferServerCiphers>true</preferServerCiphers>
        </server>
        <client>
            <loadDefaultCAFile>true</loadDefaultCAFile>
            <cacheSessions>true</cacheSessions>
            <disableProtocols>sslv2,sslv3</disableProtocols>
            <preferServerCiphers>true</preferServerCiphers>
            <invalidCertificateHandler>
                <name>RejectCertificateHandler</name>
            </invalidCertificateHandler>
        </client>
    </openSSL>

    <!--tcp 端口-->
    <tcp_port>9002</tcp_port>
    <!-- Port for communication between replicas. Used for data exchange. -->
    <interserver_http_port>9009</interserver_http_port>
    
    <!--其他副本用于请求此服务器的主机名--> 
    <interserver_http_host>cdh01</interserver_http_host>
  
    <!--注释掉监听主机-->
    <!--<listen_host>::</listen_host>-->
    <!--<listen_host>::1</listen_host>-->
    <!--<listen_host>127.0.0.1</listen_host>-->

    <!-- 监听IP -->
    <listen_host>0.0.0.0</listen_host>

    <max_connections>4096</max_connections>
    <keep_alive_timeout>3</keep_alive_timeout>

    <!-- 最大并发查询数 -->
    <max_concurrent_queries>100</max_concurrent_queries>

    <!-- 单位是B -->
    <uncompressed_cache_size>8589934592</uncompressed_cache_size>

    <mark_cache_size>5368709120</mark_cache_size>


    <!-- 存储路径 -->
    <path>/data1/clickhouse/</path>

    <!-- 临时存储路径 -->
    <tmp_path>/data1/clickhouse/tmp/</tmp_path>

    <users_config>users.xml</users_config>

    <default_profile>default</default_profile>

    <default_database>default</default_database>

   
    <!--设置时区为东八区-->
    <timezone>Asia/Shanghai</timezone>

    <remote_servers incl="clickhouse_remote_servers" />

    <!--设置扩展配置文件的路径，默认为/etc/metrika.xml-->
    <include_from>/etc/clickhouse-server/metrika.xml</include_from>

    <zookeeper incl="zookeeper-servers" optional="true" />

    <macros incl="macros" optional="true" />

    <builtin_dictionaries_reload_interval>3600</builtin_dictionaries_reload_interval>

    <query_log>
        <database>system</database>
        <table>query_log</table>

        <flush_interval_milliseconds>7500</flush_interval_milliseconds>
    </query_log>

    <dictionaries_config>*_dictionary.xml</dictionaries_config>

    <compression incl="clickhouse_compression">
    </compression>

    <resharding>
        <task_queue_path>/clickhouse/task_queue</task_queue_path>
    </resharding>
 
    <!-- 控制大表(50G)的删除,0为允许删除任何表 -->
    <max_table_size_to_drop>0</max_table_size_to_drop>

    <graphite_rollup_example>
        <pattern>
            <regexp>click_cost</regexp>
            <function>any</function>
            <retention>
                <age>0</age>
                <precision>3600</precision>
            </retention>
            <retention>
                <age>86400</age>
                <precision>60</precision>
            </retention>
        </pattern>
        <default>
            <function>max</function>
            <retention>
                <age>0</age>
                <precision>60</precision>
            </retention>
            <retention>
                <age>3600</age>
                <precision>300</precision>
            </retention>
            <retention>
                <age>86400</age>
                <precision>3600</precision>
            </retention>
        </default>
    </graphite_rollup_example>
</yandex>

3.config1.xml cdh01,cdh02,cdh03的实例2

注意端口号 tcp_port 改为 9001,http_port 改为 8124,同步端口 改为 9011,每个节点的 interserver_http_host 不一样

主要修改内容

    <logger>
        <level>trace</level>
        <log>/var/log/clickhouse-server/clickhouse-server_1.log</log>
        <errorlog>/var/log/clickhouse-server/clickhouse-server_1.err.log</errorlog>
        <size>1000M</size>
        <count>10</count>
    </logger>

    <http_port>8124</http_port>

        <!-- 存储路径 -->
    <path>/data1/clickhouse_1/</path>

    <!-- 临时存储路径 -->
    <tmp_path>/data1/clickhouse_1/tmp/</tmp_path>


    <tcp_port>9001</tcp_port>

    <!-- Port for communication between replicas. Used for data exchange. -->
    <interserver_http_port>9011</interserver_http_port>

    
    <!--设置扩展配置文件的路径，默认为/etc/metrika_1.xml-->
    <include_from>/etc/clickhouse-server/metrika_1.xml</include_from>

4.users.xml 6个实例配置相同

<?xml version="1.0"?>
<yandex>
    <profiles>
        <default>
            <max_memory_usage>10000000000</max_memory_usage>
            <use_uncompressed_cache>0</use_uncompressed_cache>
            <load_balancing>random</load_balancing>
        </default>

        <readonly>
            <max_memory_usage>10000000000</max_memory_usage>
            <use_uncompressed_cache>0</use_uncompressed_cache>
            <load_balancing>random</load_balancing>
            <readonly>1</readonly>
        </readonly>

    </profiles>

    <quotas>
        <!-- Name of quota. -->
        <default>
            <interval>
                <duration>3600</duration>
                <queries>0</queries>
                <errors>0</errors>
                <result_rows>0</result_rows>
                <read_rows>0</read_rows>
                <execution_time>0</execution_time>
            </interval>
        </default>
    </quotas>

    <users>
        <!--读写用户-->
        <default>
            <password_sha256_hex>37a8eec1ce19687d132fe29051dca629d164e2c4958ba141d5f4133a33f0688f</password_sha256_hex>
            <networks incl="networks" replace="replace">
                <ip>::/0</ip>
            </networks>
            <profile>default</profile>
            <quota>default</quota>
        </default>

        <!--只读用户-->
        <ck>
            <password_sha256_hex>d93beca6efd0421b314c081066064ac0e371b306f715cc0935b2879e249ba9df</password_sha256_hex>
            <networks incl="networks" replace="replace">
                <ip>::/0</ip>
            </networks>
            <profile>readonly</profile>
            <quota>default</quota>
        </ck>
    </users>
</yandex>

如何生成密码？
暗文，sha256sum的Hash值

PASSWORD=default
echo "$PASSWORD"; echo -n "$PASSWORD" | sha256sum | tr -d '-'

5.metrika.xml

cdh01，cdh02,cdh03 的实例1

vim /etc/clickhouse-server/metrika.xml

cdh01，cdh02,cdh03 的实例1

vim /etc/clickhouse-server/metrika_1.xml

<yandex>
<clickhouse_remote_servers>
    <!--3分片2备份-->
    <perftest_3shards_2replicas>
        <!--数据分片1-->
        <shard>
            <!--分片权重，即有多大概率落到此分片上-->
            <weight>1</weight>
            <!--表示是否只将数据写入其中一个副本,默认为flase,表示写入所有副本-->
            <internal_replication>true</internal_replication>
            <replica>
                <host>cdh01</host>
                <port>9002</port>
        <user>default</user>
                <password>default</password>
            </replica>
            <replica>
                <host>cdh02</host>
                <port>9001</port>
                <user>default</user>
                <password>default</password>
            </replica>
        </shard>
        <!--数据分片2-->
        <shard>
            <!--分片权重，即有多大概率落到此分片上-->
            <weight>1</weight>
            <!--表示是否只将数据写入其中一个副本,默认为flase,表示写入所有副本-->
            <internal_replication>true</internal_replication>
            <replica>
                <host>cdh02</host>
                <port>9002</port>
                <user>default</user>
                <password>default</password>
            </replica>
            <replica>
                <host>cdh03</host>
                <port>9001</port>
                <user>default</user>
                <password>default</password>
            </replica>
        </shard>
        <!--数据分片3-->
        <shard>
            <!--分片权重，即有多大概率落到此分片上-->
            <weight>1</weight>
            <!--表示是否只将数据写入其中一个副本,默认为flase,表示写入所有副本-->
            <internal_replication>true</internal_replication>
            <replica>
                <host>cdh03</host>
                <port>9002</port>
                <user>default</user>
                <password>default</password>
            </replica>
            <replica>
                <host>cdh01</host>
                <port>9001</port>
                <user>default</user>
                <password>default</password>
            </replica>
        </shard>
    </perftest_3shards_2replicas>
</clickhouse_remote_servers>

<zookeeper-servers>
  <node index="1">
    <host>cdh01</host>
    <port>2181</port>
  </node>

  <node index="2">
    <host>cdh02</host>
    <port>2181</port>
  </node>
  <node index="3">
    <host>cdh03</host>
    <port>2181</port>
  </node>
</zookeeper-servers>

<!--
layer 是双极分片设置,我们是单集群，这里是01
shard 表示分片编号；
replica 使用了cluster{layer}-{shard}-{replica}
-->
<macros>
    <layer>01</layer>
    <shard>03</shard>
    <!--表示cluster01集群的03分片下的1号副本-->
    <replica>cluster01-03-1</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>

注意：6个节点（除 macros 外）metrika.xml 配置相同

4.集群测试

4.1 集群启动

启动6个实例，两种启动方式

nohup clickhouse-server --config-file=/etc/clickhouse-server/config.xml  >/dev/null 2>&1 &

service click-server-1 start

4.2 启动6个客户端

clickhouse-client -h cdh01 -m -u default --password default --port 9002

clickhouse-client -h cdh01 -m -u default --password default --port 9001

clickhouse-client -h cdh02 -m -u default --password default --port 9002

clickhouse-client -h cdh02 -m -u default --password default --port 9001

clickhouse-client -h cdh03 -m -u default --password default --port 9002

clickhouse-client -h cdh03 -m -u default --password default --port 9001

4.3 测试案例参考：https://hzkeung.com/2018/06/30/clickhouse-cluster-test

采用 ReplicatedMergeTree + Distributed引擎作为集群结构的引擎
ReplicatedMergeTree(zoo_path, replica_name,partition,primykey,8192)

• zoo_path,zk路径(自动在zookeeper中创建)，如果要相互复制，必须一样,
• replica_name‘副本名称, 必须不一样,
• partition,分区
• primykey,含有主键相关字段的元组，可以为单独列
• 8192,索引粒度)

Distributed(cluster, datebase, local_table[, sharding_key])

• cluster,需要写成在config里自定义的cluster名称
• database,是分片数据库的名称
• local_table,是分片本地表的名称 -最后一项sharding_key是选填的，可以是一个表达式，例如rand()，也可以是某列 如user_id，不过该列必须是integer类型，通过对该具体的值进行取余进行分片，如果担心这样没法均匀的进行分片，也可以加上hash函数，如intHash64(user_id)

4.4 测试表

create database ontime;
use ontime;

CREATE TABLE ontime (FlightDate Date,Year UInt16) ENGINE = ReplicatedMergeTree('/clickhouse/tables/01-01/ontime','cluster01-01-1',FlightDate,(Year, FlightDate),8192); #cdh01的实例1，分片01，副本1

CREATE TABLE ontime (FlightDate Date,Year UInt16) ENGINE = ReplicatedMergeTree('/clickhouse/tables/01-03/ontime','cluster01-03-2',FlightDate,(Year, FlightDate),8192); #cdh01的实例2，分片03，副本2

CREATE TABLE ontime (FlightDate Date,Year UInt16) ENGINE = ReplicatedMergeTree('/clickhouse/tables/01-02/ontime','cluster01-02-1',FlightDate,(Year, FlightDate),8192); #cdh02的实例1，分片02，副本1

CREATE TABLE ontime (FlightDate Date,Year UInt16) ENGINE = ReplicatedMergeTree('/clickhouse/tables/01-01/ontime','cluster01-01-2',FlightDate,(Year, FlightDate),8192); #cdh02的实例2，分片01，副本2

CREATE TABLE ontime (FlightDate Date,Year UInt16) ENGINE = ReplicatedMergeTree('/clickhouse/tables/01-03/ontime','cluster01-03-1',FlightDate,(Year, FlightDate),8192); #cdh03的实例1，分片03，副本1

CREATE TABLE ontime (FlightDate Date,Year UInt16) ENGINE = ReplicatedMergeTree('/clickhouse/tables/01-02/ontime','cluster01-02-2',FlightDate,(Year, FlightDate),8192); #cdh03的实例2，分片02，副本2

CREATE TABLE ontime_all (FlightDate Date,Year UInt16) ENGINE= Distributed(perftest_3shards_2replicas, ontime, ontime, rand());  #每个实例都执行

4.5 测试过程

# 在 cdh01 的实例1 (分片01，副本1) 上执行该语句
insert into ontime (FlightDate,Year)values('2001-10-12',2001);

# 在 6个实例上执行该语句
select * from ontime;

可以发现cdh01 实例1(分片01,副本1) 和 cdh02 实例2(分片01,副本2) 两个节点的本地表查询到数据；数据成功复制副本，当其中1个副本挂掉后，集群还可继续使用

# 在 6个实例上执行该语句
select * from ontime_all;

在所有实例上均可查询到数据；任一节点都可读取分布表