Mysql取分组中前N条记录

表结构如下：
CREATE TABLE `dwb_rmirror_req_d` (
`thedate` varchar(10) NOT NULL DEFAULT '',
`node` varchar(15) NOT NULL DEFAULT '',
`req_num` bigint(20) DEFAULT NULL,
PRIMARY KEY (`thedate`,`node`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8

表中的记录如下：
mysql> select * from dwb_rmirror_req_d;
+----------+------+------------+
| thedate | node | req_num |
+----------+------+------------+
| 20160215 | f | 2 |
| 20160215 | i | 1 |
| 20160215 | l | 3 |
| 20160217 | f | 2 |
| 20160217 | i | 1 |
| 20160217 | l | 3 |
| 20160218 | f | 2 |
| 20160218 | i | 1 |
| 20160218 | l | 3 |
| 20160219 | f | 2 |
| 20160219 | i | 1 |
| 20160219 | l | 3 |
| 20160220 | f | 2 |
| 20160220 | i | 1 |
| 20160220 | l | 3 |
| 20160221 | f | 2 |
| 20160221 | i | 1 |
| 20160221 | l | 3 |
+----------+------+------------+
18 rows in set (0.00 sec)

1.获取每天查询量最大的记录：
select a.thedate,a.node,a.req_num from dwb_rmirror_req_d a left join dwb_rmirror_req_d b
on a.thedate = b.thedate and a.req_num <= b.req_num
group by a.thedate,a.node,a.req_num
having count(b.node)<=1;
结果如下：
+----------+------+------------+
| thedate | node | req_num |
+----------+------+------------+
| 20160215 | l | 3 |
| 20160217 | l | 3 |
| 20160218 | l | 3 |
| 20160219 | l | 3 |
| 20160220 | l | 3 |
| 20160221 | l | 3 |
+----------+------+------------+
6 rows in set (0.01 sec)

2. 获取每天查询量最高的两条记录：
select a.thedate,a.node,a.req_num from dwb_rmirror_req_d a left join dwb_rmirror_req_d b
on a.thedate = b.thedate and a.req_num <= b.req_num
group by a.thedate,a.node,a.req_num
having count(b.node)<=2
order by a.thedate,a.req_num;
结果如下：
+----------+------+------------+
| thedate | node | req_num |
+----------+------+------------+
| 20160215 | f | 2 |
| 20160215 | l | 3 |
| 20160217 | f | 2 |
| 20160217 | l | 3 |
| 20160218 | f | 2 |
| 20160218 | l | 3 |
| 20160219 | f | 2 |
| 20160219 | l | 3 |
| 20160220 | f | 2 |
| 20160220 | l | 3 |
| 20160221 | f | 2 |
| 20160221 | l | 3 |
+----------+------+------------+
12 rows in set (0.01 sec)