Sqoop 导数据到HDFS, 用Spark SQL进行查询

1. 启动HDFS:

cd /usr/local/hadoop-2.7.7/sbin

./start-dfs.sh

2.启动Yarn:
cd cd /usr/local/hadoop-2.7.7/sbin

./start-yarn.sh

3.启动Spark:
/usr/local/spark-2.3.3-bin-hadoop2.7/sbin

./start-master.sh -h 192.168.96.12

./start-slave.sh spark://192.168.96.128:7077

4.创建Sqoop导入任务：

./sqoop-job \

--meta-connect jdbc:hsqldb:hsql://192.168.96.128:16000/sqoop \

--create t_order_increment_job \

-- import --connect jdbc:mysql://192.168.96.1:3306/demo_ds_0?serverTimezone=Asia/Shanghai \

--username root -P \

--append \

--table t_order_increment \

--columns "id,name,my_time" \

--incremental lastmodified \

--check-column my_time \

--last-value '2019-08-30 21:36:16' \

--target-dir /increment/t_order_increment

 

5.执行导入任务：

./sqoop-job \

--meta-connect jdbc:hsqldb:hsql://192.168.96.128:16000/sqoop \

--exec t_order_increment_job -- --last-value '2020-08-12 02:12:18'

6.Spark SQL进行查询的Java代码：

public class IncrementApplication {
    public static void main(String[] args) {
        SparkSession spark = SparkSession.builder()
                .appName("SparkApplication")
                .config("spark.master", "spark://192.168.96.128:7077")
                .config("spark.jars", "/usr/local/workspace/spark-test-1.0-SNAPSHOT-shaded.jar")
                .getOrCreate();
JavaRDD<Order> orderRdd = spark.read().text("hdfs://192.168.96.128:9000/increment/t_order_increment/").javaRDD().map(
                line -> {
                    Order order = new Order();
String[] items = line.getString(0).split(",");
Integer orderId = Integer.valueOf(items[0]);
order.setOrderId(orderId);
Integer userId = Integer.valueOf(items[1]);
order.setUserId(userId);
order.setStatus(items[2]);
                    return order;
}
            );
Dataset<Row> orderDataSet =  spark.createDataFrame(orderRdd, Order.class);
orderDataSet.createOrReplaceTempView("order");
Dataset<Row> sqlDF = spark.sql("SELECT * FROM order");
sqlDF.show();
}
}

附录：

删除HDFS文件的命令：

cd /usr/local/hadoop-2.7.7/bin

./hadoop dfs -rm -R /increment/*

 

注意：sqoop1.4.7版本的metastore换成MySQL后，出现锁表的情况，初步怀疑是sqoop的bug，出现的异常如下：
com.mysql.cj.jdbc.exceptions.MySQLTransactionRollbackException: Lock wait timeout exceeded; try restarting transaction

 

如果连的自己部署的MySQL作为metastore，命令如下：

./sqoop-job \

--meta-connect 'jdbc:mysql://127.0.0.1:3306/sqoop?user=root&password=123456' \

--create t_order_increment_job \

-- import --connect jdbc:mysql://127.0.0.1:3306/kevin?serverTimezone=Asia/Shanghai \

--username root --password 123456 \

--append \

--table t_order_increment \

--incremental lastmodified \

--check-column my_time \

--last-value '2019-08-30 21:36:16' \

--target-dir /increment/t_order_increment

 

如果将MySQL数据增量导入Hive，如果MySQL数据有更新，增量导入Hive后，会出现id重复的记录，在Hive中，可以通过以下方法，查出最新的记录：

select t.id, t.name, t.my_time

from (

       select id, name, my_time, row_number() over (partition by id order by my_time desc) num 

        from t_order_increment

     ) t

where t.num = 1; 

 

如果是增量导数据到Hive，每隔一段时间，需要对数据进行整理，去掉id重复的记录：

./sqoop-merge \

--new-data /kevin/new/kevin_new.txt \

--onto /user/hive/warehouse/t_order_increment/ \

--target-dir /user/hive/warehouse/t_order_increment/2020-08-18 \

--merge-key id \

--class-name t_order_increment \

 

--jar-file t_order_increment.jar