使用Oracle Partition Table对日志表进行分区
某个系统中有个日志型的表,现在的大小大概超过500MB,与该表相关的语句的执行计划都是对这个表进行全表扫描。
下面是该表的定义:
01 create table p_siteaccesslog
02 (
03;;logIDvarchar(40) not null,
04;;infoType;;;;;varchar(20) default 'site',
05;;siteId; varchar(30),
06;;columnId;;;;;varchar(30),
07;;infoId; varchar(30),
08;;url;;varchar(300),
09;;location;;;;;varchar(80),
10;;action; varchar(20),
11;;accessDate;;;varchar(10),
12;;accessTime;;;varchar(8),
13;;sessionID;;;;varchar(90),
14;;userName;;;;;varchar(30),
15;;userBrowser;;varchar(50),
16;;userOS; varchar(30),
17;;userIp; varchar(30),
18;;localization;varchar(100),
19;;constraint PK_p_siteaccesslog primary key (logID)
20 )
该表有一个主键,为每条日志分配一个主键,感觉上一百年都不会用上这个主键来查一次数据。所以在新的分区表上将其去掉。
1 test$ora8i@4.20 SQL> l
21* select segment_name,sum(bytes/1048576) MB from user_extents group by segment_name order by 2 desc
3 test$ora8i@4.20 SQL> /
4
5 SEGMENT_NAME;;MB
6 ---------------------------------------- ----------
7 P_SITEACCESSLOG 536
由于是日志型的表,第一感觉就是按照时间分区,并淘汰(切换)旧的数据。
于是沿着这个思路,按照每月一个分区的策略构建分区表:
01 create table my_p_siteaccesslog
02 (
03;;logIDvarchar(40) not null,
04;;infoType;;;;;varchar(20) default 'site',
05;;siteId; varchar(30),
06;;columnId;;;;;varchar(30),
07;;infoId; varchar(30),
08;;url;;varchar(300),
09;;location;;;;;varchar(80),
10;;action; varchar(20),
11;;accessDate;;;varchar(10),
12;;accessTime;;;varchar(8),
13;;sessionID;;;;varchar(90),
14;;userName;;;;;varchar(30),
15;;userBrowser;;varchar(50),
16;;userOS; varchar(30),
17;;userIp; varchar(30),
18;;localization;varchar(100)
19 )
20 partition by range (accessDate)
21 (
22;;partition pbf2009 values less than ('2009-01-01'),
23;;partition p200901 values less than ('2009-02-01'),
24;;partition p200902 values less than ('2009-03-01'),
25;;partition p200903 values less than ('2009-04-01'),
26;;partition p200904 values less than ('2009-05-01'),
27;;partition p200905 values less than ('2009-06-01'),
28;;partition p200906 values less than ('2009-07-01'),
29;;partition p200907 values less than ('2009-08-01'),
30;;partition p200908 values less than ('2009-09-01'),
31;;partition p200909 values less than ('2009-10-01'),
32;;partition p200910 values less than ('2009-11-01'),
33;;partition p200911 values less than ('2009-12-01'),
34;;partition p200912 values less than ('2010-01-01'),
35;;partition p201001 values less than ('2010-02-01'),
36;;partition p201002 values less than ('2010-03-01'),
37;;partition p201003 values less than ('2010-04-01'),
38;;partition p201004 values less than ('2010-05-01'),
39;;partition p201005 values less than ('2010-06-01'),
40;;partition p201006 values less than ('2010-07-01'),
41;;partition p201007 values less than ('2010-08-01'),
42;;partition p201008 values less than ('2010-09-01'),
43;;partition p201009 values less than ('2010-10-01'),
44;;partition p201010 values less than ('2010-11-01'),
45;;partition p201011 values less than ('2010-12-01'),
46;;partition p201012 values less than ('2011-01-01')
47 )
由于数据库是Oracle 10g,不是11g,无法支持interval分区,只能多创建几个未来的分区了。
表创建好之后,就导数据流,由于该表不算太大,500多兆的插入还是可以接受的。
1 insert into my_p_siteaccesslog select * from P_SITEACCESSLOG ;
2 commit;
下一步使用分区切换技术,将旧的数据搬走(不是删掉)
创建用于存放2008年数据大表
01 create table my_p_siteaccesslog_2008
02 (
03;;logIDvarchar(40) not null,
04;;infoType;;;;;varchar(20) default 'site',
05;;siteId; varchar(30),
06;;columnId;;;;;varchar(30),
07;;infoId; varchar(30),
08;;url;;varchar(300),
09;;location;;;;;varchar(80),
10;;action; varchar(20),
11;;accessDate;;;varchar(10),
12;;accessTime;;;varchar(8),
13;;sessionID;;;;varchar(90),
14;;userName;;;;;varchar(30),
15;;userBrowser;;varchar(50),
16;;userOS; varchar(30),
17;;userIp; varchar(30),
18;;localization;varchar(100)
19 );
使用分区切换,将2008年的数据切换出来:
1 ALTER TABLE my_p_siteaccesslog EXCHANGE PARTITION pbf2009 WITH TABLE my_p_siteaccesslog_2008;
创建专门用于存放2009年数据的分区表:
01 create table my_p_siteaccesslog_2009
02 (
03;;logIDvarchar(40) not null,
04;;infoType;;;;;varchar(20) default 'site',
05;;siteId; varchar(30),
06;;columnId;;;;;varchar(30),
07;;infoId; varchar(30),
08;;url;;varchar(300),
09;;location;;;;;varchar(80),
10;;action; varchar(20),
11;;accessDate;;;varchar(10),
12;;accessTime;;;varchar(8),
13;;sessionID;;;;varchar(90),
14;;userName;;;;;varchar(30),
15;;userBrowser;;varchar(50),
16;;userOS; varchar(30),
17;;userIp; varchar(30),
18;;localization;varchar(100)
19 )
20 partition by range (accessdate)
21 (
22;;partition p200901 values less than ('2009-02-01'),
23;;partition p200902 values less than ('2009-03-01'),
24;;partition p200903 values less than ('2009-04-01'),
25;;partition p200904 values less than ('2009-05-01'),
26;;partition p200905 values less than ('2009-06-01'),
27;;partition p200906 values less than ('2009-07-01'),
28;;partition p200907 values less than ('2009-08-01'),
29;;partition p200908 values less than ('2009-09-01'),
30;;partition p200909 values less than ('2009-10-01'),
31;;partition p200910 values less than ('2009-11-01'),
32;;partition p200911 values less than ('2009-12-01'),
33;;partition p200912 values less than ('2010-01-01')
34 )
令人非常郁闷的是Oracle竟然不支持两个分区表之间的分区切换,一定要借助一个中间表,这个功能连SQL Server都有了啊。
于是创建一个专门用于切换的中间表:
01 create table my_p_siteaccesslog_exchange
02 (
03;;logIDvarchar(40) not null,
04;;infoType;;;;;varchar(20) default 'site',
05;;siteId; varchar(30),
06;;columnId;;;;;varchar(30),
07;;infoId; varchar(30),
08;;url;;varchar(300),
09;;location;;;;;varchar(80),
10;;action; varchar(20),
11;;accessDate;;;varchar(10),
12;;accessTime;;;varchar(8),
13;;sessionID;;;;varchar(90),
14;;userName;;;;;varchar(30),
15;;userBrowser;;varchar(50),
16;;userOS; varchar(30),
17;;userIp; varchar(30),
18;;localization;varchar(100)
19 );
下面就是毫无技术含量并且不断重复的就借助中间表将my_p_siteaccesslog中的2009年数据切换到中my_p_siteaccesslog_2009中的代码:
01 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200901 WITH TABLE my_p_siteaccesslog_exchange;
02 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200901 WITH TABLE my_p_siteaccesslog_exchange;
03
04 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200902 WITH TABLE my_p_siteaccesslog_exchange;
05 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200902 WITH TABLE my_p_siteaccesslog_exchange;
06
07 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200903 WITH TABLE my_p_siteaccesslog_exchange;
08 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200903 WITH TABLE my_p_siteaccesslog_exchange;
09
10 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200904 WITH TABLE my_p_siteaccesslog_exchange;
11 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200904 WITH TABLE my_p_siteaccesslog_exchange;
12
13 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200905 WITH TABLE my_p_siteaccesslog_exchange;
14 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200905 WITH TABLE my_p_siteaccesslog_exchange;
15
16 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200906 WITH TABLE my_p_siteaccesslog_exchange;
17 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200906 WITH TABLE my_p_siteaccesslog_exchange;
18
19 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200907 WITH TABLE my_p_siteaccesslog_exchange;
20 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200907 WITH TABLE my_p_siteaccesslog_exchange;
21
22 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200908 WITH TABLE my_p_siteaccesslog_exchange;
23 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200908 WITH TABLE my_p_siteaccesslog_exchange;
24
25 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200909 WITH TABLE my_p_siteaccesslog_exchange;
26 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200909 WITH TABLE my_p_siteaccesslog_exchange;
27
28 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200910 WITH TABLE my_p_siteaccesslog_exchange;
29 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200910 WITH TABLE my_p_siteaccesslog_exchange;
30
31 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200911 WITH TABLE my_p_siteaccesslog_exchange;
32 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200911 WITH TABLE my_p_siteaccesslog_exchange;
33
34 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p200912 WITH TABLE my_p_siteaccesslog_exchange;
35 ALTER TABLE my_p_siteaccesslog_2009 EXCHANGE PARTITION p200912 WITH TABLE my_p_siteaccesslog_exchange;
顺便创建存放2010年数据的分区表
01 create table my_p_siteaccesslog_2010
02 (
03;;logIDvarchar(40) not null,
04;;infoType;;;;;varchar(20) default 'site',
05;;siteId; varchar(30),
06;;columnId;;;;;varchar(30),
07;;infoId; varchar(30),
08;;url;;varchar(300),
09;;location;;;;;varchar(80),
10;;action; varchar(20),
11;;accessDate;;;varchar(10),
12;;accessTime;;;varchar(8),
13;;sessionID;;;;varchar(90),
14;;userName;;;;;varchar(30),
15;;userBrowser;;varchar(50),
16;;userOS; varchar(30),
17;;userIp; varchar(30),
18;;localization;varchar(100)
19 )
20 partition by range (accessdate)
21 (
22;;partition p201001 values less than ('2010-02-01'),
23;;partition p201002 values less than ('2010-03-01'),
24;;partition p201003 values less than ('2010-04-01'),
25;;partition p201004 values less than ('2010-05-01'),
26;;partition p201005 values less than ('2010-06-01'),
27;;partition p201006 values less than ('2010-07-01'),
28;;partition p201007 values less than ('2010-08-01'),
29;;partition p201008 values less than ('2010-09-01'),
30;;partition p201009 values less than ('2010-10-01'),
31;;partition p201010 values less than ('2010-11-01'),
32;;partition p201011 values less than ('2010-12-01'),
33;;partition p201012 values less than ('2011-01-01')
34 )
并将1月份的数据切换到表my_p_siteaccesslog_2010中:
1 ALTER TABLE my_p_siteaccesslog;;;EXCHANGE PARTITION p201001 WITH TABLE my_p_siteaccesslog_exchange;
2 ALTER TABLE my_p_siteaccesslog_2010 EXCHANGE PARTITION p201001 WITH TABLE my_p_siteaccesslog_exchange;
经过一轮折腾之后,我们看看现各个表的情况:
01 test$ora8i@4.20 SQL> l
021* select segment_name,sum(bytes/1048576) MB from user_extents group by segment_name order by 2 desc
03 test$ora8i@4.20 SQL> /
04
05 SEGMENT_NAME;;MB
06 ---------------------------------------- ----------
07 P_SITEACCESSLOG 536
08 MY_P_SITEACCESSLOG_2009 435
09 MY_P_SITEACCESSLOG_2008; 51
10 MY_P_SITEACCESSLOG_2010;;45.6875
11 MY_P_SITEACCESSLOG; 35.1875
12 MY_P_SITEACCESSLOG_EXCHANGE.0625
表MY_P_SITEACCESSLOG以后将会改名为P_SITEACCESSLOG正式上线使用,其中仅保留1~2个月的数据用于查询,当需要查询旧数据的时候,应用可以不动,然后将旧的数据切换到MY_P_SITEACCESSLOG(也就是以后的P_SITEACCESSLOG)中就可以使用了。
分区表的切换产生的redo非常少,主要都是用于数据字典,因此系统的灵活性变得更高了。
相关文章: