哈尔滨做网站公司哪家好seo是什么味

背景

量级庞大的日志通过mysql不足以支撑业务需求，以前通过任务调度定时跑批从mysql同步到hive存储，这种方式时效性为T+1，也就是说今天的日志，明天才能同步到hive，总而言之时效性不高。为了提高时效性，改为流式计算flink实时同步

• 那么作为测试人员，我们如何保证切换同步方式后的数据正确性呢？通过对比新旧表数据是否一致显然是最简单的方法
• 这次改动涉及600多张表，每一张表的字段数基本在千以上，甚至部分表字段数达万以上，面对如此庞大的数据量，通过人眼一个个去对比显然不太现实

探索与实践

方案一：sql脚本

SELECT column_names, COUNT(*) AS count_diff 
FROM (SELECT CONCAT_WS(',',A,B) FROM udc_test.s000 WHERE dt='20230814'UNION ALL SELECT CONCAT_WS(',',A,B) FROM test.s000 WHERE dt = '20230814' and rule_log_id in (select rule_log_id from udc_test.s000) 
) AS combined
GROUP BY column_names
HAVING COUNT(column_names) = 1select * from (select 'table1',A,B from udc_test.s000 WHERE dt='20230814' and rule_log_id in ('123456')union all select 'table2',A,B from test.s000 WHERE dt='20230814' and rule_log_id in ('123456')
)a order by a.table1 asc

方案二：python脚本

from pyhive import hive
from datetime import datetimeif __name__ == '__main__':#换成生产的连接conn = hive.Connection(host="xxx", port='xxx', auth="xxx", database='xxx', username='xxx',password='xxx')#这里换成需要比较的表名tableName1 = 'test.ssc_python_compare_fields1'tableName2 = 'test.ssc_python_compare_fields2'current_time = datetime.now()hash_code = str(hash(current_time))# 获取表结构query1 = 'desc ' + tableName1query2 = 'desc ' + tableName2cursor = conn.cursor()cursor.execute(query1)columns1 = [row[0] for row in cursor.fetchall()]cursor.execute(query2)columns2 = [row[0] for row in cursor.fetchall()]# 去除掉不需要比较的字段columns1.remove('# Partition Information')columns1.remove('# col_name')columns1.remove('dt')columns2.remove('# Partition Information')columns2.remove('# col_name')columns2.remove('dt')set1 = set(columns1)set2 = set(columns2)# 取出来表1特有的字段，可以保存到文件diffrence1 = set1 - set2print(diffrence1)# 取出来表2特有的字段，可以保存到文件diffrence2 = set2 - set1print(diffrence2)# 取表1和表2共有的字段，用于比较差异intersection = set1 & set2# 生成比较的sqlsql = 'select  'for element in intersection:sql = sql + 'if( nvl(t1.' + element + ',' + hash_code + ' )!= nvl( t2.' + element + ',' + hash_code + ') , \'no\',\'yes\') as ' + element + ' , '#print(sql)sql = sql[:-2]#print(sql)#sql中的dt可以改成具体需要比较的日期sql = sql + ' from ' + tableName1 + ' as t1 left join  ' + tableName2 \+ ' as t2  on  t1.rule_log_id=t2.rule_log_id ' \' and t1.dt= \'20230815\' and t2.dt = \'20230815\'  and t1.apply_type=t2.apply_type   where  'for element in intersection:sql = sql + ' t1.' + element + '!=t2.' + element + ' or 'sql = sql[:-3]print(sql)sql = sql + ' limit 1 '# 执行sql,获取到结果，如果两列不相等的话，值为no，相等的话值为yescursor.execute(sql)result = cursor.fetchone()# print(result)# 获取上述sql的元数据信息metadatas = cursor.descriptionprint('============================================================')# 遍历结果集，查找出比较结果不相同的数据，拿到列名index = 0while index < len(metadatas):if (result[index] != 'yes'):print(metadatas[index][0])index += 1print('============================================================')