From: http://www.zhujiangroad.com/html/soft/s8076.html
为1.7亿记录表创建快速索引
本文讲述了在大表上创建索引需要注意的事项，以及整个过程。
需求：在STAT_SUBMIT_CENTER表的RECORDTIME字段上面创建一索引。
环境：SunOS 5.9
oracle 9204
8 cpu
3G mem
1. 查看表的具体情况
是不是分区表，有多少个分区，分区字段：
SQL> col table_name for a20
SQL> col column_name for a20
SQL> select a.table_name,a.partitioned,b.partition_count,c.column_name
2 from user_tables a, user_part_tables b, user_part_key_columns c
3 where a.table_name='STAT_SUBMIT_CENTER'
4 and b.table_name='STAT_SUBMIT_CENTER'
5 and c.name='STAT_SUBMIT_CENTER';
TABLE_NAME PAR PARTITION_COUNT COLUMN_NAME
-------------------- --- --------------- --------------------
STAT_SUBMIT_CENTER YES 50 MSGDATE
已使用的每个分区的大小：
SQL> select segment_name,partition_name,round(bytes/1024/1024) from user_segments where segment_name ='STAT_SUBMIT_CENTER' and bytes/1024/1024>0.25 order by 3 desc;
SEGMENT_NAME PARTITION_NAME ROUND(BYTES/1024/1024)
-------------------------- ------------------------------ ----------------------
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20051101 1722
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20051021 1488
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20051111 1440
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20051121 1355
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20051221 1335
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20050911 1309
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20051211 1253
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20051201 1247
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20050921 1198
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20060101 1151
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20060111 1068
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20051001 1018
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20051011 865
STAT_SUBMIT_CENTER STAT_SUBMIT_CENTER_20060121 796
14 rows selected.
整个表的大小：
SQL> select segment_name,sum(bytes/1024/1024) from user_segments where segment_name ='STAT_SUBMIT_CENTER' group by segment_name;
SEGMENT_NAME SUM(BYTES/1024/1024)
-------------------------------- --------------------
STAT_SUBMIT_CENTER 17234
表的记录数：
SQL> set timing on
SQL> select count(*) from STAT_SUBMIT_CENTER;
COUNT(*)
----------
170341007
Elapsed: 00:14:18.60
还有这个表上的索引情况如下：
table STAT_SUBMIT_CENTER 17234 Mindex IDX_SUBCEN_ADDRUSER 5155 M ADDRUSER PK_STAT_SUBMIT_CENTER 10653 M MSGDATE,ADDRUSER,MSGID
然后查看一些数据库参数情况：
SQL> show parameter work
NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
workarea_size_policy string AUTO
SQL> show parameter pga
NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
pga_aggregate_target big integer 209715200
SQL> select * from dba_temp_files;
FILE_NAME
------------------------------------------------------------------------------------------------------------------------
FILE_ID TABLESPACE_NAME BYTES BLOCKS STATUS RELATIVE_FNO AUT MAXBYTES MAXBLOCKS
---------- ------------------------------ ---------- ---------- --------- ------------ --- ---------- ----------
INCREMENT_BY USER_BYTES USER_BLOCKS
------------ ---------- -----------
/bgdata/oracle/temp01.dbf 1 TEMP 3563061248 434944 AVAILABLE 1 YES 4294967296 524288 6400 3562012672 434816
2. 需要考虑的几个方面
1）创建的索引需要几个G的磁盘空间
2）创建索引需要排序，使用pga_aggregate_target，要把这个值从200M加大到2G
3）如果内存不够，需要temp表空间，把temp表空间加大到8G。itpub上有一个帖子说过，15亿条记录用了34G空间。
4）在线创建，时间会比较长。讨论后，停这个表的操作，非online创建。
3. 实际操作过程
1）数据文件够，不扩展；temp数据文件扩展：
alter database tempfile '/bgdata/oracle/temp01.dbf' resize 8192m;
2）在workarea_size_policy=AUTO的情况下，改pga_aggregate_target=2048m，对于串行操作，一个session能使用的pga=MIN(5%PGA_AGGREGATE_TARGET,100MB)，这样可以使得pga用到最大的值：
alter system set pga_aggregate_target=2048m;
3）因为这是一个比较长的过程，所以写脚本让后台运行：
createind.sh
vi createind.sh
#!/bin/sh
sqlplus user/password <<EOF
create index IDX_SUBMIT_RECORDTIME on STAT_SUBMIT_CENTER(RECORDTIME) local;
exit
EOF
4）创建过程中可以观察v$sort_segment，v$sort_usage看排序情况：
select tablespace_name,current_users,total_blocks,used_blocks,free_blocks from v$sort_segment;
select * from v$sort_usage;
5）创建完成后，把tempfile和pga_aggregate_target改回原值：
alter database tempfile '/bgdata/oracle/temp01.dbf' resize 4096m;
alter system set pga_aggregate_target=500m;
4. 实际创建过程中观察到的情况
1）开始之前：
SQL> select tablespace_name,current_users,total_blocks,used_blocks,free_blocks from v$sort_segment;
TABLESPACE_NAME CURRENT_USERS TOTAL_BLOCKS USED_BLOCKS FREE_BLOCKS
------------------------------- ------------- ------------ ----------- -----------
TEMP 0 431360 0 431360
SQL> select * from v$sort_usage;
no rows selected
2）创建之初，抓到这么一条sql：
insert into obj$(owner#,name,namespace,obj#,type#,ctime,mtime,stime,status,remoteowner,linkname,subname,dataobj#,flags,oid$,spare1,spare2)values(:1,:2,:3,:4,:5,:6,:7,:8,:9,:10,:11,:12,:13,:14,:15,:16, :17)
3）然后v$sort_segment.USED_BLOCKS变大，v$sort_usage.BLOCKS变大，一直增长到：
SQL> select tablespace_name,current_users,total_blocks,used_blocks,free_blocks from v$sort_segment;
TABLESPACE_NAME CURRENT_USERS TOTAL_BLOCKS USED_BLOCKS FREE_BLOCKS
------------------------------- ------------- ------------ ----------- -----------
TEMP 1 431360 46720 384640
SQL> select * from v$sort_usage;
USERNAME USER SESSION_ADDR SESSION_NUM SQLADDR SQLHASH
------------------------------ ------------------------------ ---------------- ----------- ---------------- ----------
TABLESPACE CONTENTS SEGTYPE SEGFILE# SEGBLK# EXTENTS BLOCKS SEGRFNO#
------------------------------- --------- --------- ---------- ---------- ---------- ---------- ----------
DPC DPC 00000003974CFFB0 6134 0000000399CAB288 1254950678TEMP TEMPORARY SORT 201
431113 365 46720 1
这个过程中抓到的sql：select file# from file$ where ts#=:1
4）v$sort_segment.USED_BLOCKS变为0，v$sort_usage.BLOCKS变为0
5）重复3，4两步，估计这个是创建一个分区的索引
需要解释一下的是，上面的sql只是我随机抓到的运行时间比较长的，整个create index过程会复杂很多，具体怎么样可以用sqltrace跟踪。这里主要看的是temp表空间的使用情况。
同时，在创建的过程中：
SQL> select segment_name,partition_name from user_segments where segment_name='IDX_SUBMIT_RECORDTIME';
no rows selected
SQL> select index_name,partition_name from user_ind_partitions where INDEX_NAME='IDX_SUBMIT_RECORDTIME';
no rows selected
当时忘了查user_segments中其实是有一个segment_name为一串数字的记录，那个才是正在创建的索引；如果这个事务失败了，将回滚。
最后耗时99分钟完成。
5. 创建完成后分析索引
但是接下来还有一件事。创建完成后要分析索引，否则就是走了索引，查询也巨慢无比。
SQL> explain plan for select count(*) from stat_submit_center where recordtime>trunc(sysdate);
Explained.
SQL> @?/rdbms/admin/utlxplp.sql
PLAN_TABLE_OUTPUT
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost | Pstart| Pstop |
-------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 9 | 4 | | || 1 | SORT AGGREGATE | | 1 | 9 | | | || 2 | PARTITION RANGE ALL | | | | | 1 | 50 ||* 3 | INDEX FAST FULL SCAN| IDX_SUBMIT_RECORDTIME | 8878K| 76M| 4 | 1 | 50 |
-------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
3 - filter("STAT_SUBMIT_CENTER"."RECORDTIME">TRUNC(SYSDATE@!))Note: cpu costing is off16
rows selected.
SQL> set autotrace on explain
SQL> set timing on
SQL> select count(*) from stat_submit_center where recordtime>trunc(sysdate);aa^Cselect count(*) from stat_submit_center where recordtime>trunc(sysdate)*ERROR at line 1:ORA-01013: user requested cancel of current operationElapsed: 00:11:49.85
SQL>
SQL> set autotrace off
上面可以看到，因为没有分析索引，虽然它走的是新建的IDX_SUBMIT_RECORDTIME这个索引，但是查询巨慢，10分钟后也没有结果。下面我们分析一下。
SQL> Analyze index IDX_SUBMIT_RECORDTIME estimate statistics;
Index analyzed.
Elapsed: 00:00:06.84
SQL> set autotrace on explain
SQL> select count(*) from stat_submit_center where recordtime>trunc(sysdate);COUNT(*)
----------
926736
Elapsed: 00:00:05.37
Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=4360 Card=1 Bytes=9)1 0 SORT (AGGREGATE)2 1 PARTITION RANGE (ALL)3 2 INDEX (RANGE SCAN) OF 'IDX_SUBMIT_RECORDTIME' (NON-UNIQUE) (Cost=4360 Card=8878740 Bytes=79908660)
SQL> set autotrace off
索引分析之后，查询时间在5分钟，效率大大提高。
10万+
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
