ClickHouse

本文主要介绍如何使用DLA Spark访问云ClickHouse。

1. 准备测试数据，命名为ck.csv上传至OSS。

   name,age
   fox,18
   tiger,20
   alice,36

2. 准备以下读写ClickHouse的代码，下面的代码是读取OSS的CSV文件，写入到新建的ClickHouse表中，然后从ClickHouse表中读取数据，打印到控制台。
   • 参考POM文件

         <dependencies>
             <dependency>
                 <groupId>ru.yandex.clickhouse</groupId>
                 <artifactId>clickhouse-jdbc</artifactId>
                 <version>0.2.4</version>
             </dependency>
             <dependency>
                 <groupId>org.apache.spark</groupId>
                 <artifactId>spark-sql_2.11</artifactId>
                 <version>2.4.5</version>
                 <scope>provided</scope>
             </dependency>
          </dependencies>

   • 参考代码片段。完整代码请参考代码。

     package com.aliyun.spark
     
     import java.sql.{Connection, DriverManager}
     import java.util.Properties
     
     import org.apache.spark.sql.execution.datasources.jdbc.JDBCOptions
     import org.apache.spark.sql.{SaveMode, SparkSession}
     
     object SparkWriteToCK {
       val ckProperties = new Properties()
       val url = "jdbc:clickhouse://<您的clickhouse VPC 地址>:8123/default"
       ckProperties.put("driver", "ru.yandex.clickhouse.ClickHouseDriver")
       ckProperties.put("user", "<用户名>")
       ckProperties.put("password", "<密码>")
       ckProperties.put("batchsize","100000")
       ckProperties.put("socket_timeout","300000")
       ckProperties.put("numPartitions","8")
       ckProperties.put("rewriteBatchedStatements","true")
     
       //创建ClickHouse表
       def createCKTable(table: String): Unit ={
         Class.forName(ckProperties.getProperty("driver"))
         var conn : Connection = null
         try {
           conn = DriverManager.getConnection(url, ckProperties.getProperty("user"), ckProperties.getProperty("password"))
           val stmt = conn.createStatement()
           val sql =
             s"""
                |create table if not exists default.${table}  on cluster default(
                |    `name` String,
                |    `age`  Int32)
                |ENGINE = MergeTree() ORDER BY `name` SETTINGS index_granularity = 8192;
                |""".stripMargin
           stmt.executeQuery(sql)
         } finally {
           if(conn != null)
             conn.close()
         }
       }
       
       def main(args: Array[String]): Unit = {
         val table = "ck_test"
         //使用jdbc建ClickHouse表
         createCKTable(table)
         val spark = SparkSession.builder().getOrCreate()
         //将csv的数据写入ClickHouse
         val csvDF = spark.read.option("header","true").csv("oss://<path/to>/ck.csv").toDF("name", "age")
         csvDF.printSchema()
         csvDF.write.mode(SaveMode.Append).option(JDBCOptions.JDBC_BATCH_INSERT_SIZE, 100000).jdbc(url, table, ckProperties)
         //读ck表数据
         val ckDF = spark.read.jdbc(url, table, ckProperties)
         ckDF.show()
       }
     
     }

3. 将以上代码打包编译上传至OSS。
4. 进入DLA控制台，选择Serverless Spark > 作业管理，单击创建作业模板，选择前提条件中创建的虚拟集群。配置作业运行参数后点击执行。

   {
       "file": "oss://<path/to/your/jar>",
       "name": "SparkWriteToCK",
       "className": "com.aliyun.spark.SparkWriteToCK",
       "conf": {
           "spark.driver.resourceSpec": "medium",
           "spark.executor.instances": 5,
           "spark.executor.resourceSpec": "medium",
           "spark.dla.job.log.oss.uri": "oss://<指定您存放SparkUI日志的目录/>",
           "spark.dla.connectors": "oss",
           "spark.dla.eni.enable": "true",
           "spark.dla.eni.security.group.id": "<您前提条件中选定的安全组id>",
           "spark.dla.eni.vswitch.id": "<您前提条件中选定的交换机id>"
       }
   }

   说明 更多配置说明，请参考Spark配置指南。
5. 作业运行后，可在作业界面查看作业日志和SparkUI。