package mobvista.dmp.datasource.taobao import mobvista.dmp.common.CommonSparkJob import mobvista.dmp.util.PropertyUtil import mobvista.dmp.utils.clickhouse.ClickHouseConnectionFactory import mobvista.dmp.utils.clickhouse.ClickHouseSparkExt._ import org.apache.commons.cli.Options import org.apache.spark.sql.types.{StringType, StructField, StructType} import org.apache.spark.sql.{Row, SparkSession} import ru.yandex.clickhouse.ClickHouseDataSource class UCOaidLaHuoCK extends CommonSparkJob with Serializable { override protected def buildOptions(): Options = { val options = new Options options.addOption("oaidInput", true, "oaidInput") options.addOption("date", true, "date") options.addOption("host", true, "host") options.addOption("cluster", true, "cluster") options.addOption("database", true, "database") options.addOption("table", true, "table") options } val schema = StructType(Array( StructField("device_ids", StringType))) override protected def run(args: Array[String]): Int = { val commandLine = commParser.parse(options, args) if (!checkMustOption(commandLine)) { printUsage(options) return -1 } else printOptions(commandLine) val oaidInput = commandLine.getOptionValue("oaidInput") val date = commandLine.getOptionValue("date") val cluster = commandLine.getOptionValue("cluster") val host = commandLine.getOptionValue("host") val database = commandLine.getOptionValue("database") val table = commandLine.getOptionValue("table") val spark = SparkSession .builder() .appName(s"UCOaidLaHuoCK.$date") .config("spark.rdd.compress", "true") .config("spark.io.compression.codec", "lz4") .config("spark.sql.warehouse.dir", "s3://mob-emr-test/spark-warehouse") .config("spark.serializer", "org.apache.spark.serializer.KryoSerializer") .config("spark.clickhouse.driver", "ru.yandex.clickhouse.ClickHouseDriver") .config("spark.clickhouse.url", PropertyUtil.getProperty("config.properties", "spark.clickhouse.url")) .config("spark.clickhouse.connection.per.executor.max", "5") .config("spark.clickhouse.metrics.enable", "true") .config("spark.clickhouse.socket.timeout.ms", "300000") .config("spark.clickhouse.cluster.auto-discovery", "true") .enableHiveSupport() .getOrCreate() val sc = spark.sparkContext try { try { val oaidRDD = sc.textFile(oaidInput).map(r => { Row(r) }) implicit val clickhouseDataSource: ClickHouseDataSource = ClickHouseConnectionFactory.get(host) val clusterName = Some(cluster): Option[String] //存储oaid到clickhouse val device_type = "OAID_MD5" val dt = mobvista.dmp.common.MobvistaConstant.sdf1.format(mobvista.dmp.common.MobvistaConstant.sdf2.parse(date)) val oaidDF = spark.emptyDataFrame oaidDF.dropPartition(database, table, s"($date,'$device_type')", clusterName) spark.createDataFrame(oaidRDD, schema).saveToClickHouse(database, table, Seq(dt, device_type), Seq("dt", "device_type"), clusterName, batchSize = 1000000) } finally { sc.stop() spark.stop() } 0 } } } object UCOaidLaHuoCK { def main(args: Array[String]): Unit = { new UCOaidLaHuoCK().run(args) } }