1.导入carbondata依赖的jar包

将apache-carbondata-1.5.3-bin-spark2.3.2-hadoop2.7.2.jar导入$SPARKHOME/jars;或将apache-carbondata-1.5.3-bin-spark2.3.2-hadoop2.7.2.jar导入在$SPARKHOME创建的carbondlib目录

2.导入kafka依赖的jar包

接入kafka数据需要依赖kafka的jars，将以下jars导入$SPARKHOME/jars

kafka-clients-0.10.0.1.jar
spark-sql-kafka-0-10_2.11-2.3.2.jar

3.spark-shell启动服务

./bin/spark-shell --master spark://hostname:7077 --jars apache-carbondata-1.5.3-bin-spark2.3.2-hadoop2.7.2.jar

a).导入依赖

import org.apache.spark.sql.SparkSession
import org.apache.spark.sql.CarbonSession._

b).创建session

启动第一个目录是数据存储目录，第二个目录是元数据目录；都可以是hdfs目录

val carbon = SparkSession.builder().config(sc.getConf).getOrCreateCarbonSession("/home/bigdata/carbondata/data","/home/bigdata/carbondata/carbon.metastore")

c).创建source表

carbon.sql(
            s"""
                | CREATE TABLE IF NOT EXISTS kafka_json_source(
                |     id STRING,
                |    name STRING,
                |    age INT,
                |    brithday TIMESTAMP)
                | STORED AS carbondata
                | TBLPROPERTIES(
                |    'streaming'='source',
                |    'format'='kafka',
                |    'kafka.bootstrap.servers'='hostname:9092',
                |    'subscribe'='kafka_json',
                |    'record_format'='json',
                |    'comment'='get kafka data')
            """.stripMargin).show()

d).创建sink表

carbon.sql(
            s"""
                | CREATE TABLE IF NOT EXISTS kafka_json_sink(
                |     id STRING,
                |    name STRING,
                |    age INT,
                |    brithday TIMESTAMP)
                | STORED AS carbondata
                | TBLPROPERTIES(
                |    'streaming'='sink')
            """.stripMargin).show()

e).创建job任务

carbon.sql(
            s"""
                | CREATE STREAM kafka_json_job ON TABLE kafka_json_sink(
                | STMPROPERTIES(
                |    'trigger'='ProcessingTime',
                |    'interval'='10 seconds')
                | AS SELECT * FROM kafka_json_source
            """.stripMargin).show()

f).创建DATAMAP

carbon.sql(
            s"""
                | CREATE DATAMAP agg_kafka_json_sink 
                | ON TABLE kafka_json_sink(
                | USING "preaggregate"
                | AS
                |    SELECT id,name,sum(age),max(age),min(age),avg(age)
                |    FROM kafka_json_sink
                |    GRPUP BY id,name
            """.stripMargin).show()

4.常用SQL命令

a).导入本地数据

carbon.sql("LOAD DATA INPATH '/home/bigdata/carbondata/sample.csv' INTO TABLE kafka_json_source").show()

b).查看表结构

carbon.sql("DESC kafka_json_source").show()

c).查看表数据

carbon.sql("SELECT * FROM kafka_json_source WHERE id=1").show()

d).清理表数据

carbon.sql("TRUNCATE TABLE kafka_json_sink").show()

e).删除表

carbon.sql("DROP TABLE IF EXISTS kafka_json_source").show()

f).查看job任务状态

carbon.sql("SHOW STREAMS ON TABLE kafka_json_sink").show()

g).删除job任务

carbon.sql("DROP STREAM kafka_json_job").show()

h).查询DATAMAP表信息

carbon.sql("DESC agg_kafka_json_sink_kafka_json_sink").show()

i).查询表Segments信息

carbon.sql("SHOW SEGMENTS FOR TABLE kafka_json_sink").show()

j).条件查询

carbon.sql("SELECT * FROM kafka_json_sink WHERE agent_id=499 AND signature=''").show()

k).聚合查询

carbon.sql("SELECT agent_id,signature,method_type,sum(elapse_time),max(elapse_time),min(elapse_time) FROM kafka_json_sink GROUP BY agent_id,signature,method_type").show()

5.注意事项

a).kafka使用配置

由于Carbondata的kafka-consumer反序列化配置如下，所以在kafka-producer应该使用对于配置，否则无法解析数据

key.deserializer = org.apache.kafka.common.serialization.ByteArrayDeserializer
value.deserializer = org.apache.kafka.common.serialization.ByteArrayDeserializer