1. clickhouse-client

底层是基于tcp协议

1.1 交互式协议

[root@clickhouse1 ~]#
[root@clickhouse1 ~]# clickhouse-client -h clickhouse1 --port 9000 -u default --password default123 -m -n
ClickHouse client version 21.6.5.37 (official build).
Connecting to clickhouse1:9000 as user default.
Connected to ClickHouse server version 21.6.5 revision 54448.

clickhouse1 :)
clickhouse1 :) select 1;

SELECT 1

Query id: a58a5e62-75cb-4b3c-a8e8-5ed8d2917b85

┌─1─┐
│ 1 │
└───┘

1 rows in set. Elapsed: 0.016 sec. 

clickhouse1 :)
clickhouse1 :) select 2;

SELECT 2

Query id: 35771efe-e306-403d-b39a-7a58abba4412

┌─2─┐
│ 2 │
└───┘

1 rows in set. Elapsed: 0.006 sec. 

clickhouse1 :)
clickhouse1 :) select hostName();

SELECT hostName()

Query id: 7d08f34a-489e-4e05-91be-a2e72f4c08fa

┌─hostName()──┐
│ clickhouse1 │
└─────────────┘

1 rows in set. Elapsed: 0.010 sec. 

clickhouse1 :) 

执行的记录被保存clickhouse-client-history

[root@clickhouse1 ~]# 
[root@clickhouse1 ~]# pwd
/root
[root@clickhouse1 ~]# 
[root@clickhouse1 ~]# cat .clickhouse-client-history
......省略部分......
### 2021-07-09 04:37:28.414
select 1;
### 2021-07-09 04:39:48.674
select 2;
### 2021-07-09 04:39:51.683
select hostName();

1.2 非交互模式

导入数据

[root@clickhouse1 ~]# 
[root@clickhouse1 ~]# pwd
/root
[root@clickhouse1 ~]# 
[root@clickhouse1 ~]# cat insert.csv 
3,san
4,si
[root@clickhouse1 ~]#
[root@clickhouse1 ~]# clickhouse-client -h clickhouse1 --port 9000 -u default --password default123 -m -n --format_csv_delimiter="," --query "insert into distribute_test_all format CSV" < /root/insert.csv 
[root@clickhouse1 ~]#

导出数据

[root@clickhouse1 ~]#
[root@clickhouse1 ~]# clickhouse-client -h clickhouse1 --port 9000 -u default --password default123 -m -n --query "select * from distribute_test_all" > /root/select.csv 
[root@clickhouse1 ~]# 
[root@clickhouse1 ~]# pwd
/root
[root@clickhouse1 ~]# 
[root@clickhouse1 ~]# cat select.csv 
1	一
4	si
1	yi
2	er
3	san
[root@clickhouse1 ~]#

2. Java/Scala使用JDBC连接Clickhouse

底层基于HTTP接口，Java版本如下：

pom.xml

        <dependency>
            <groupId>com.clickhouse</groupId>
            <artifactId>clickhouse-jdbc</artifactId>
            <version>0.3.2-patch8</version>
        </dependency>

Clickhouse_JDBC_test.java

import ru.yandex.clickhouse.BalancedClickhouseDataSource;
import ru.yandex.clickhouse.settings.ClickHouseProperties;

import java.sql.Connection;
import java.sql.PreparedStatement;

public class Clickhouse_JDBC_test {
    

    public static void main(String[] args) throws Exception {
    

        String url = "jdbc:clickhouse://192.168.23.41:8123,192.168.23.42:8123,192.168.23.43:8123,192.168.23.44:8123/default";
        ClickHouseProperties clickHouseProperties = new ClickHouseProperties();
        clickHouseProperties.setUser("default");
        clickHouseProperties.setPassword("default123");

        BalancedClickhouseDataSource balanced = new BalancedClickhouseDataSource(url, clickHouseProperties);
        //对每个host进行ping操作, 排除不可用的连接
        balanced.actualize();
        Connection conn = balanced.getConnection();

        String insertSql = "insert into test.java_table_test_all(id, name) values(?, ?)";
        PreparedStatement pst = conn.prepareStatement(insertSql);


        for (int i = 1; i < 9999; i++) {
    
            pst.setInt(1, i);
            pst.setString(2, "name" + i);
            pst.addBatch();

            if (i % 1000 == 0) {
    
                // 执行conn.setAutoCommit(false);会报Transactions are not supported异常
                // 所以不能执行conn.commit();
                // 只能执行pst.executeBatch();由clickhouse进行后台提交。测试时插入数据条数是正确的
                pst.executeBatch();
                pst.clearBatch();
            }
        }

        // 最后一批数据插入
        pst.executeBatch();
        pst.clearBatch();

        pst.close();
        conn.close();

    }
}

3. Pyspark连接Clickhouse

添加驱动jar包

1. 从maven仓库的clickhouse-jdbc下载clickhouse-jdbc-0.3.2.jar驱动包
2. 从maven仓库的clickhouse-client下载clickhouse-client-0.3.2.jar驱动包
3. 从maven仓库的clickhouse-http-client下载clickhouse-http-client-0.3.2.jar驱动包
4. 将下载的jar包放到所有Spark服务器的spark-3.1.2-bin-hadoop3.2/jars目录下
5. 然后重启Spark服务

pySpark示例程序如下：

from pyspark.sql import SparkSession
from pyspark import SparkConf



if __name__ == '__main__':

    conf = SparkConf() \
        .setAppName("clickhouseTest")
    spark = SparkSession.builder.config(conf=conf).getOrCreate()
    spark.sparkContext.setLogLevel("WARN")


    df = spark.read.format("jdbc") \
        .option("driver", "com.clickhouse.jdbc.ClickHouseDriver") \
        .option("url","jdbc:clickhouse://bigdata001:8123,bigdata002:8123,bigdata003:8123/clickhouse_test_db") \
        .option("user","default") \
        .option("password","default123") \
        .option("dbtable","(select user_id, name, age from user_tb where age between 10 and 30) tempTable") \
        .option("partitionColumn", "age") \
        .option("fetchsize", 25000) \
        .option("lowerBound", 10) \
        .option("upperBound", 30) \
        .option("numPartitions", 15) \
        .option("queryTimeout", 600) \
        .load()

    df.show(n=3, truncate=False)

    spark.stop()