#! /bin/bash # # # # # # # # # # # # # # # # # # # # # # # @author : 冯亮 # @date : 2017-08-07 # @desc : 解析MP SDK 天数据 # # # # # # # # # # # # # # # # # # # # # # source ../dmp_env.sh date_path=$(date +"%Y/%m/%d" -d "-1 day $ScheduleTime") dt=$(date +"%Y%m%d" -d "-1 day $ScheduleTime") old_path=$(date +"%Y/%m/%d" -d "-2 day $ScheduleTime") INPUT_PATH="${MP_REQUEST_PATH}/$date_path/*" OUTPUT_PATH="${MP_REQUEST_DAILY_PATH}/$date_path" ID_MAPPING_PATH="${DMP_ADN_REQUEST_DEVICE_IDS}/$date_path/adn_request" check_await ${MP_REQUEST_PATH}/$date_path/23 check_await ${ID_MAPPING_PATH} export HADOOP_CLIENT_OPTS="-Xmx4096m $HADOOP_CLIENT_OPTS" spark-submit --class mobvista.dmp.datasource.mpsdk.ParseMPSDKDaily \ --conf spark.sql.shuffle.partitions=200 \ --master yarn --deploy-mode cluster --executor-memory 8g --driver-memory 4g --executor-cores 4 --num-executors 20 \ ../${JAR} -input $INPUT_PATH -output $OUTPUT_PATH -mappingPath $ID_MAPPING_PATH -parallelism 100 -coalesce 20 if [ $? -ne 0 ];then exit 255 fi mount_partition "etl_mpsdk_request_daily" "dt='$dt'" "$OUTPUT_PATH"