[2023-04-18]增量收集统计信息
在11g之前，收集global的统计信息时，oracle需要扫描每个分区的统计信息，才能得出 global的统计信息。而在11g以后，Oracle会简单记录分区的概要信息，（synopsis，记录在 WRI$_OPTSTAT_SYNOPSIS_HEAD$ 和 WRI$_OPTSTAT_SYNOPSIS$中），对于global的统计信息，可以不再扫描每个分区的，只需通过计算概要信息，即可得出global 的统计信息。
1、创建分区表
SQL> create table p(id number, age date)
2 partition by range(id)
3 (partition p1 values less than(100),
4 partition p2 values less than(200),
5 partition p3 values less than(300),
6 partition p4 values less than(400)
7 );
Table created.
SQL> insert into p select level lv,sysdate from dual connect by level < 400;
399 rows created.
SQL> commit;
Commit complete.
SQL> exec dbms_stats.gather_table_stats(ownname=>'scott',tabname=>'p');
PL/SQL procedure successfully completed.
SQL> select table_owner,table_name,partition_name from dba_tab_partitions where table_owner='SCOTT' and table_name='P';
TABLE_OWNER TABLE_NAME PARTITION_NAME
------------------------------ ------------------------------ ------------------------------
SCOTT P P1
SCOTT P P2
SCOTT P P3
SCOTT P P4
SQL> alter session set nls_date_format = 'yyyy-mm-dd hh24:mi:ss';
Session altered.
SQL> select table_name,num_rows,last_analyzed from dba_tables where table_name='P';
TABLE_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ---------- -------------------
P 399 2023-04-18 16:33:47
SQL>
SQL> select table_name,partition_name,num_rows,last_analyzed from dba_tab_partitions where table_name='P';
TABLE_NAME PARTITION_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ------------------------------ ---------- -------------------
P P1 99 2023-04-18 16:33:46
P P2 100 2023-04-18 16:33:47
P P3 100 2023-04-18 16:33:47
P P4 100 2023-04-18 16:33:47
2、重新收集统计信息
SQL> exec dbms_stats.gather_table_stats(ownname=>'scott',tabname=>'p');
PL/SQL procedure successfully completed.
SQL> select table_name,num_rows,last_analyzed from dba_tables where table_name='P';
TABLE_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ---------- -------------------
P 399 2023-04-18 16:35:12
SQL> select table_name,partition_name,num_rows,last_analyzed from dba_tab_partitions where table_name='P';
TABLE_NAME PARTITION_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ------------------------------ ---------- -------------------
P P1 99 2023-04-18 16:35:12
P P2 100 2023-04-18 16:35:12
P P3 100 2023-04-18 16:35:12
P P4 100 2023-04-18 16:35:12
SQL> 对表收集统计信息，虽然分区中的数据没有变化，依然会收集全部统计信息。
3、开启增量统计信息的条件
SQL> set lin 300
SQL> col "publish" for a30
SQL> col "granularity" for a30
SQL> col "incremental" for a30
SQL> select
2 dbms_stats.get_prefs('publish','&owner','&tablename') "publish",
3 dbms_stats.get_prefs('granularity','&&owner','&&tablename') "granularity",
4 dbms_stats.get_prefs('INCREMENTAL','&&owner','&&tablename') "incremental"
5 from dual;
Enter value for owner: scott
Enter value for tablename: p
old 2: dbms_stats.get_prefs('publish','&owner','&tablename') "publish",
new 2: dbms_stats.get_prefs('publish','scott','p') "publish",
Enter value for owner: scott
Enter value for tablename: p
old 3: dbms_stats.get_prefs('granularity','&&owner','&&tablename') "granularity",
new 3: dbms_stats.get_prefs('granularity','scott','p') "granularity",
old 4: dbms_stats.get_prefs('INCREMENTAL','&&owner','&&tablename') "incremental"
new 4: dbms_stats.get_prefs('INCREMENTAL','scott','p') "incremental"
publish granularity incremental
------------------------------ ------------------------------ ------------------------------
TRUE AUTO FALSE
SQL>
SQL> exec dbms_stats.set_table_prefs('SCOTT', 'P', 'INCREMENTAL', 'TRUE');
PL/SQL procedure successfully completed.
SQL> col "publish" for a30
SQL> col "granularity" for a30
SQL> col "incremental" for a30
SQL> select
2 dbms_stats.get_prefs('publish','&owner','&tablename') "publish",
3 dbms_stats.get_prefs('granularity','&&owner','&&tablename') "granularity",
4 dbms_stats.get_prefs('INCREMENTAL','&&owner','&&tablename') "incremental"
5 from dual;
old 2: dbms_stats.get_prefs('publish','&owner','&tablename') "publish",
new 2: dbms_stats.get_prefs('publish','scott','p') "publish",
old 3: dbms_stats.get_prefs('granularity','&&owner','&&tablename') "granularity",
new 3: dbms_stats.get_prefs('granularity','scott','p') "granularity",
old 4: dbms_stats.get_prefs('INCREMENTAL','&&owner','&&tablename') "incremental"
new 4: dbms_stats.get_prefs('INCREMENTAL','scott','p') "incremental"
publish granularity incremental
------------------------------ ------------------------------ ------------------------------
TRUE AUTO TRUE
SQL> 4、开启增量后第一次收集统计信息
SQL> exec dbms_stats.gather_table_stats(ownname=>'scott',tabname=>'p');
PL/SQL procedure successfully completed.
SQL> select table_name,num_rows,last_analyzed from dba_tables where table_name='P';
TABLE_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ---------- -------------------
P 399 2023-04-18 16:39:26
SQL> select table_name,partition_name,num_rows,last_analyzed from dba_tab_partitions where table_name='P';
TABLE_NAME PARTITION_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ------------------------------ ---------- -------------------
P P1 99 2023-04-18 16:39:26
P P2 100 2023-04-18 16:39:26
P P3 100 2023-04-18 16:39:26
P P4 100 2023-04-18 16:39:26
SQL> 设置增量收集后，第一次收集统计信息会将全部历史分区重新收集，以生成历史分区的大纲信息，分区数越多，耗时越长
5、开启增量后第二次收集统计信息
SQL> exec dbms_stats.gather_table_stats(ownname=>'scott',tabname=>'p');
PL/SQL procedure successfully completed.
SQL> select table_name,num_rows,last_analyzed from dba_tables where table_name='P';
TABLE_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ---------- -------------------
P 399 2023-04-18 16:41:03
SQL> select table_name,partition_name,num_rows,last_analyzed from dba_tab_partitions where table_name='P';
TABLE_NAME PARTITION_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ------------------------------ ---------- -------------------
P P1 99 2023-04-18 16:39:26
P P2 100 2023-04-18 16:39:26
P P3 100 2023-04-18 16:39:26
P P4 100 2023-04-18 16:39:26
SQL> 全局统计信息时间更新了，但是分区统计信息还是旧的。
6、对某个分区收集统计信息，全局统计信息也会更新
SQL> exec dbms_stats.gather_table_stats(ownname=>'scott',tabname=>'p',PARTNAME=>'p1');
PL/SQL procedure successfully completed.
SQL> select table_name,num_rows,last_analyzed from dba_tables where table_name='P';
TABLE_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ---------- -------------------
P 399 2023-04-18 16:42:26
SQL> select table_name,partition_name,num_rows,last_analyzed from dba_tab_partitions where table_name='P';
TABLE_NAME PARTITION_NAME NUM_ROWS LAST_ANALYZED
------------------------------ ------------------------------ ---------- -------------------
P P1 99 2023-04-18 16:42:26
P P2 100 2023-04-18 16:39:26
P P3 100 2023-04-18 16:39:26
P P4 100 2023-04-18 16:39:26
SQL>