Oracle表碎片整理操作步骤详解

时间:2022-03-08 20:32:11

高水位线(HWL)下的许多数据块都是无数据的,但全表扫描的时候要扫描到高水位线的数据块,也就是说oracle要做许多的无用功!因此oracle提供了shrink space碎片整理功能。对于索引,可以采取rebuild online的方式进行碎片整理,一般来说,经常进行DML操作的对象DBA要定期进行维护,同时注意要及时更新统计信息!

一:准备测试数据,使用HR用户,创建T1表,插入约30W的数据,并根据object_id创建普通索引,表占存储空间34M

 

复制代码代码如下:

SQL> conn /as sysdba 
已连接。 
SQL> select default_tablespace from dba_users where username='HR'; 

DEFAULT_TABLESPACE 
------------------------------------------------------------ 
USERS 

SQL> conn hr/hr 
已连接。 

SQL> insert into t1 select * from t1; 
已创建 74812 行。 

SQL> insert into t1 select * from t1; 
已创建 149624 行。 

SQL> commit; 
提交完成。 

SQL> create index idx_t1_id on t1(object_id); 
索引已创建。 

SQL> exec dbms_stats.gather_table_stats('HR','T1',CASCADE=>TRUE); 
PL/SQL 过程已成功完成。 

SQL> select count(1) from t1; 

  COUNT(1) 
---------- 
    299248 

SQL> select sum(bytes)/1024/1024 from dba_segments where segment_name='T1'; 
SUM(BYTES)/1024/1024 
-------------------- 
             34.0625 

SQL> select sum(bytes)/1024/1024 from dba_segments where segment_name='IDX_T1_ID'; 
SUM(BYTES)/1024/1024 
-------------------- 
                   6 

 

二:估算表在高水位线下还有多少空间可用,这个值应当越低越好,表使用率越接近高水位线,全表扫描所做的无用功也就越少!

DBMS_STATS包无法获取EMPTY_BLOCKS统计信息,所以需要用analyze命令再收集一次统计信息

 

复制代码代码如下:

SQL> SELECT blocks, empty_blocks, num_rows FROM user_tables WHERE table_name ='T1'; 

    BLOCKS EMPTY_BLOCKS   NUM_ROWS 
---------- ------------ ---------- 
      4302            0     299248 

SQL> analyze table t1 compute statistics; 
表已分析。 

SQL> SELECT blocks, empty_blocks, num_rows FROM user_tables WHERE table_name ='T1'; 

    BLOCKS EMPTY_BLOCKS   NUM_ROWS 
---------- ------------ ---------- 
      4302           50     299248 

SQL> col table_name for a20 
SQL> SELECT TABLE_NAME, 
  2         (BLOCKS * 8192 / 1024 / 1024) - 
  3         (NUM_ROWS * AVG_ROW_LEN / 1024 / 1024) "Data lower than HWM in MB" 
  4    FROM USER_TABLES 
  5   WHERE table_name = 'T1'; 

TABLE_NAME           Data lower than HWM in MB 
-------------------- ------------------------- 
T1                                  5.07086182 

 

三: 查看执行计划,全表扫描大概需要消耗CPU 1175

 

复制代码代码如下:

SQL> explain plan for select * from t1; 
已解释。 

SQL> select * from table(dbms_xplan.display); 

PLAN_TABLE_OUTPUT 
-------------------------------------------------------------------------------- 
Plan hash value: 3617692013 
-------------------------------------------------------------------------- 
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     | 
-------------------------------------------------------------------------- 
|   0 | SELECT STATEMENT  |      |   299K|    28M|  1175   (1)| 00:00:15 | 
|   1 |  TABLE ACCESS FULL| T1   |   299K|    28M|  1175   (1)| 00:00:15 | 
-------------------------------------------------------------------------- 

 

四:删除大部分数据,收集统计信息,全表扫描依然需要消耗CPU 1168

 

复制代码代码如下:

SQL> delete from t1 where object_id>100; 
已删除298852行。 

SQL> commit; 
提交完成。 

SQL> select count(*) from t1; 

  COUNT(*) 
---------- 
       396 

SQL>  exec dbms_stats.gather_table_stats('HR','T1',CASCADE=>TRUE); 
PL/SQL 过程已成功完成。 

SQL> analyze table t1 compute statistics; 
表已分析。 

SQL> SELECT blocks, empty_blocks, num_rows FROM user_tables WHERE table_name ='T1'; 

    BLOCKS EMPTY_BLOCKS   NUM_ROWS 
---------- ------------ ---------- 
      4302           50        396 

 
SQL> explain plan for select * from t1; 
已解释。 

SQL> select * from table(dbms_xplan.display); 

PLAN_TABLE_OUTPUT 
------------------------------------------------------------------------------ 
Plan hash value: 3617692013 
-------------------------------------------------------------------------- 
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     | 
-------------------------------------------------------------------------- 
|   0 | SELECT STATEMENT  |      |   396 | 29700 |  1168   (1)| 00:00:15 | 
|   1 |  TABLE ACCESS FULL| T1   |   396 | 29700 |  1168   (1)| 00:00:15 | 
-------------------------------------------------------------------------- 

 

五:估算表在高水位线下还有多少空间是无数据的,但在全表扫描时又需要做无用功的数据

 

复制代码代码如下:

SQL> SELECT TABLE_NAME, 
  2         (BLOCKS * 8192 / 1024 / 1024) - 
  3         (NUM_ROWS * AVG_ROW_LEN / 1024 / 1024) "Data lower than HWM in MB" 
  4    FROM USER_TABLES 
  5   WHERE table_name = 'T1'; 

TABLE_NAME           Data lower than HWM in MB 
-------------------- ------------------------- 
T1                                  33.5791626 

 

六:对表进行碎片整理,重新收集统计信息

 

复制代码代码如下:

SQL> alter table t1 enable row movement; 
表已更改。 

SQL> alter table t1 shrink space cascade; 
表已更改。 

SQL> select sum(bytes)/1024/1024 from dba_segments where segment_name='T1'; 

SUM(BYTES)/1024/1024 
-------------------- 
                .125 

SQL> select sum(bytes)/1024/1024 from dba_segments where segment_name='IDX_T1_ID 
'; 

SUM(BYTES)/1024/1024 
-------------------- 
               .0625 

SQL> SELECT TABLE_NAME, 
  2         (BLOCKS * 8192 / 1024 / 1024) - 
  3         (NUM_ROWS * AVG_ROW_LEN / 1024 / 1024) "Data lower than HWM in MB" 
  4    FROM USER_TABLES 
  5   WHERE table_name = 'T1'; 

TABLE_NAME           Data lower than HWM in MB 
-------------------- ------------------------- 
T1                                  33.5791626 

SQL> exec dbms_stats.gather_table_stats('HR','T1',CASCADE=>TRUE); 
PL/SQL 过程已成功完成。 

这个时候,只剩下0.1M的无用功了,执行计划中,全表扫描也只需要消耗CPU 3 
SQL> SELECT TABLE_NAME, 
  2         (BLOCKS * 8192 / 1024 / 1024) - 
  3         (NUM_ROWS * AVG_ROW_LEN / 1024 / 1024) "Data lower than HWM in MB" 
  4    FROM USER_TABLES 
  5   WHERE table_name = 'T1'; 

TABLE_NAME           Data lower than HWM in MB 
-------------------- ------------------------- 
T1                                  .010738373 

 
SQL> select * from table(dbms_xplan.display); 

PLAN_TABLE_OUTPUT 
-------------------------------------------------------------------------------- 
Plan hash value: 3617692013 
-------------------------------------------------------------------------- 
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     | 
-------------------------------------------------------------------------- 
|   0 | SELECT STATEMENT  |      |   396 | 29700 |     3   (0)| 00:00:01 | 
|   1 |  TABLE ACCESS FULL| T1   |   396 | 29700 |     3   (0)| 00:00:01 | 
-------------------------------------------------------------------------- 

总共只有5个块,空块却有50个,明显empty_blocks信息过期 
SQL> select blocks,empty_blocks,num_rows from user_tables where table_name='T1'; 

    BLOCKS EMPTY_BLOCKS   NUM_ROWS 
---------- ------------ ---------- 
         5           50        396 

SQL> analyze table t1 compute statistics; 
表已分析。 

SQL> select blocks,empty_blocks,num_rows from user_tables where table_name='T1'; 

 
    BLOCKS EMPTY_BLOCKS   NUM_ROWS 
---------- ------------ ---------- 
         5            3        396