fix

azkaban/app_start.sh

@@ -2,6 +2,7 @@
 
 DT=$(date +%Y%m%d -d "-1 day $ScheduleTime")
 
-LAST_DT=$(date +%Y%m%d -d "-30 day $DT")
+START_DT=$(date +%Y-%m-%d -d "-30 day $DT")
+END_DT=$(date +%Y-%m-%d -d "-1 day $DT")
 
-hive -hivevar dt=${DT} -hivevar last_dt=${LAST_DT} -f app_start.sql
\ No newline at end of file
+hive -hivevar dt=${DT} -hivevar start_dt=${START_DT} -hivevar end_dt=${END_DT} -f app_start.sql
\ No newline at end of file

azkaban/app_start.sql

@@ -9,10 +9,8 @@ SET hive.mapred.mode=nonstrict;
 
 DROP TABLE IF EXISTS package_num;
 CREATE TABLE package_num AS SELECT package_name,count(1) install_num FROM dwh.dm_install_list_v2
-    WHERE dt = '${dt}' AND update_date BETWEEN '${LAST_DT}' AND '${dt}'
-      AND package_name NOT IN (SELECT distinct package_name FROM dwh.dim_package_tags)
-    GROUP BY package_name HAVING count(1) > 30000
-    ORDER BY install_num DESC LIMIT 5000;
+    WHERE dt = '${dt}' AND update_date BETWEEN '${start_dt}' AND '${end_dt}'
+    GROUP BY package_name HAVING count(1) > 30000;
 
 DROP TABLE IF EXISTS package_info;
 CREATE TABLE package_info AS 
@@ -25,6 +23,6 @@ CREATE TABLE package_info AS
 DROP TABLE IF EXISTS package_list;
 CREATE TABLE package_list AS 
     SELECT /*+ mapjoin(t)*/ p.package_name,p.app_name,p.platform,p.track_view_url,p.category_list,t.install_num
-    FROM package_num t 
+    FROM (SELECT * FROM package_num WHERE package_name NOT IN (SELECT DISTINCT package_name FROM dwh.dim_package_tags)) t
     JOIN package_info p 
-    ON t.package_name = p.package_name;
\ No newline at end of file
+    ON t.package_name = p.package_name ORDER BY t.install_num LIMIT 5000;
\ No newline at end of file