import org.apache.spark.SparkContext
import org.apache.spark.sql.{DataFrame, SparkSession}
object RemoteDebug {
//设置用户名是防止用户没有权限访问集群,以及hdfs
System.setProperty("user.name", "hadoop")
System.setProperty("HADOOP_USER_NAME", "hive")
val spark: SparkSession = SparkSession.builder()
.appName("remote_test")
.master("spark://master_ip:master_port")
.config("spark.jars","F:\\IdeaProject\\project\\target\\package.jar")
.config("spark.driver.host", "192.168.*.*")
.config("spark.driver.port", "9089")
.enableHiveSupport().getOrCreate()
val sc: SparkContext = spark.sparkContext
sc.setLogLevel("ERROR")
val mysql_url = "jdbc:mysql://server_ip:server_port/instance?useSSL=false"
val mysql_user = "username"
val mysql_password = "password"
def read4mysql(table:String):DataFrame={
spark.read.format("jdbc")
.option("driver", "com.mysql.jdbc.Driver")
.option("url",mysql_url)
.option("user",mysql_user)
.option("password",mysql_password)
.option("dbtable",table)
.load()
}
def main(args: Array[String]) {
val order_df: DataFrame = read4mysql(
s"(select * from a ")
order_df.show(10)
spark.close()
}
}
网友评论