1、创建测试表
SQL> CREATE TABLE t_M
2 AS
3 SELECT * FROM all_objects;
Table created
2、查询测试表中记录
SQL> select count(*) from t_m;
COUNT(*)
———-
48941
3、创建中间表
SQL> create table t_m_n as
2 select * from t_m where 1=0;
Table created
4、查询中间表记录
SQL> select count(*) from t_m_n;
COUNT(*)
———-
5、创建刷新物化视图
SQL> CREATE MATERIALIZED VIEW t_m_n
on prebuilt TABLE WITH REDUCED PRECISION
REFRESH FORCE
ON DEMAND
AS
SELECT * FROM t_m;
Materialized view created
6、执行物化视图刷新
SQL> exec dbms_mview.refresh(‘T_M_N’);
PL/SQL procedure successfully completed
7、查询物化视图中记录数
SQL> select count(*) from t_m_n;
COUNT(*)
———-
48941
8、删除物化视图
SQL> DROP MATERIALIZED VIEW T_M_N;
Materialized view dropped
9、查询中间表中条数
SQL> select count(*) from t_m_n;
COUNT(*)
———-
48941
10、后续可能操作
1）t_m和t_m_n相互重命名，实现在线修移动表的位置、改表结构、降低高水位等操作，同（shrink）
2）和dblink结合，实现数据的跨版本迁移

1、T2表结构

SQL> desc t2
Name           Type          Nullable Default Comments
-------------- ------------- -------- ------- --------
OWNER          VARCHAR2(30)  Y
OBJECT_NAME    VARCHAR2(128) Y
SUBOBJECT_NAME VARCHAR2(30)  Y
OBJECT_ID      NUMBER        Y
DATA_OBJECT_ID NUMBER        Y
OBJECT_TYPE    VARCHAR2(19)  Y
CREATED        DATE          Y
LAST_DDL_TIME  DATE          Y
TIMESTAMP      VARCHAR2(19)  Y
STATUS         VARCHAR2(7)   Y
TEMPORARY      VARCHAR2(1)   Y
GENERATED      VARCHAR2(1)   Y
SECONDARY      VARCHAR2(1)   Y

2、创建中间表

CREATE TABLE T2_1
AS
SELECT * FROM t2 WHERE 1=0;

3、验证T2是否用于重定义（因没有主键，采用rowid实现）

EXEC DBMS_REDEFINITION.CAN_REDEF_TABLE(user, 'T2', DBMS_REDEFINITION.cons_use_rowid);

4、执行表的在线重定义

EXEC DBMS_REDEFINITION.START_REDEF_TABLE(USER, 'T2', 'T2_1','OWNER OWNER, OBJECT_NAME OBJECT_NAME, SUBOBJECT_NAME  SUBOBJECT_NAME, OBJECT_ID OBJECT_ID, DATA_OBJECT_ID DATA_OBJECT_ID, OBJECT_TYPE OBJECT_TYPE, CREATED CREATED, LAST_DDL_TIME LAST_DDL_TIME, TIMESTAMP TIMESTAMP, STATUS STATUS, TEMPORARY TEMPORARY, GENERATED GENERATED, SECONDARY SECONDARY',DBMS_REDEFINITION.cons_use_rowid);

说明：
1）采用单引号列出T2与T2_1表列的对应关系
2）列与列之间采用单引号分割，单引号后面要有空格
5、同步数据（可选）

exec dbms_redefinition.sync_interim_table(user, 'T2', 'T2_1');

6、执行结束在线定义过程

EXEC DBMS_REDEFINITION.FINISH_REDEF_TABLE(user, 'T2', 'T2_1');

7、删除中间表

drop table t2_1 purge;

8、处理T2表（删除隐藏列）

SQL> select col#,name,type# from SYS.COL$ WHERE OBJ#=(select object_id from dba_objects where object_name='T2');
      COL# NAME                                TYPE#
---------- ------------------------------ ----------
         1 OWNER                                   1
         2 OBJECT_NAME                             1
         3 SUBOBJECT_NAME                          1
         4 OBJECT_ID                               2
         5 DATA_OBJECT_ID                          2
         6 OBJECT_TYPE                             1
         7 CREATED                                12
         8 LAST_DDL_TIME                          12
         9 TIMESTAMP                               1
        10 STATUS                                  1
        11 TEMPORARY                               1
        12 GENERATED                               1
        13 SECONDARY                               1
         0 SYS_C00014_11081015:39:40$              1
--发现一个多余隐藏列SYS_C00014_11081015:39:40$，我们需要删除
SQL>  alter table t2 set unused ("SYS_C00014_11081015:39:40$");
Table altered
SQL>  alter table t2 drop unused columns;
Table altered

表在线重定义（有主键）

1、现象
Mon Aug 8 08:16:07 2011
The value (30) of MAXTRANS parameter ignored.
kupprdp: master process DM00 started with pid=91, OS id=21125
to execute – SYS.KUPM$MCP.MAIN(‘SYS_EXPORT_TABLE_01’, ‘YDZJ’, ‘KUPC$C_1_20110808081609’, ‘KUPC$S_1_20110808081609’, 0);
kupprdp: worker process DW01 started with worker id=1, pid=57, OS id=21127
to execute – SYS.KUPW$WORKER.MAIN(‘SYS_EXPORT_TABLE_01’, ‘YDZJ’);
kupprdp: worker process DW02 started with worker id=2, pid=95, OS id=21502
to execute – SYS.KUPW$WORKER.MAIN(‘SYS_EXPORT_TABLE_01’, ‘YDZJ’);
2、原因
这个时间段正在使用expdp进行数据库导出，当然从错误描述也可以看出和导出操作有关，不过当时导出是成功，没有报任何错误。 在metalink查询 ，发现这是Version: 10.1.0.2 to 11.1.0.7 (任何平台) 的一个bug，Bug号为：6347775
这是由于在创建备份主表时使用了如下语句：
CREATE TABLE table_name (columns_list) INITRANS 10 MAXTRANS 30
而Maxtrans参数在10gR1时已经废弃，所以在告警日志中就提示：
The value (30) of MAXTRANS parameter ignored. ， 此bug没有什么影响，所以可以忽略。
3、重现
SQL> create table xff(xifenfei varchar2(100)) initrans 10 maxtrans 20;
Table created.
查看alert日志
Mon Aug 8 16:11:28 2011
The value (20) of MAXTRANS parameter ignored.

如果你比较细心，会注意到两件事情：
1、添加一个32g的数据文件会报错
SQL> alter tablespace users add datafile ‘/opt/oracle/oradata/test/user32g.dbf’ size 32g;
alter tablespace users add datafile ‘/opt/oracle/oradata/test/user32g.dbf’ size 32g
*
ERROR at line 1:
ORA-01144: File size (4194304 blocks) exceeds maximum of 4194303 blocks
SQL> !oerr ora 1144
01144, 00000, “File size (%s blocks) exceeds maximum of %s blocks”
// *Cause: Specified file size is larger than maximum allowable size value.
// *Action: Specify a smaller size.
2、添加一个maxsize为unlimited的数据文件
SQL> alter tablespace users add datafile ‘/opt/oracle/oradata/test/user32g.dbf’ size 10M AUTOEXTEND ON maxsize unlimited;
Tablespace altered.
SQL> select MAXBYTES/1024/1024/1024 from dba_data_files where file_name=’/opt/oracle/oradata/test/user32g.dbf’;
MAXBYTES/1024/1024/1024
———————–
31.9999847
你是不是有疑惑，为什么我添加32g的数据文件报错，为什么我添加一个数据文件maxsize设置为unlimited了，还是没有突破32g这个坎
原因分析：
由于Oracle的Rowid中使用22位来代表Block号，这22位最多只能代表2^22-1(4194303)个数据块，而在我们一般情况下使用的数据块大小为8k，所以数据文件的理论大小最大为： 31.9999924G
至于maxsize为unlimited时候，数据文件的大小为什么只有31.9999847G(blocks：4194301.99)，比最大块数(4194303)少了一块，也许是和数据库和系统之间的衔接原因导致。
由上面的分析我们可以知道，数据文件大小和db_block_size有关，那我们可以得到这样的数据文件最大理论值

注：以上规则适用于smallfile tablespace 下的数据文件，bigfile tablespace 下的数据文件不受此限制

1、alert文件中现象
Thu Jul 21 09:49:38 2011
performing DML/DDL operation over object in bin.
Thu Jul 21 09:51:02 2011
performing DML/DDL operation over object in bin.
2、开始模拟
1）确认回收站功能启用
SQL> show parameter recyclebin;
NAME TYPE VALUE
———————————— ———– ——————————
recyclebin string ON
如果是OFF，使用alter system set recyclebin=on;开启回收站功能
2）创建和删除表
SQL> create table t_drop
2 as
3 select * from tab;
Table created.
SQL> drop table t_drop;
Table dropped.
SQL> show recyclebin;
ORIGINAL NAME RECYCLEBIN NAME OBJECT TYPE DROP TIME
—————- —————————— ———— ——————-
T_DROP BIN$qIuF4JHvDJDgQKjADAsarA==$0 TABLE 2011-07-21:09:46:52
3）查看回收站表中数据
SQL> select * from “BIN$qIuF4JHvDJDgQKjADAsarA==$0”;
TNAME TABTYPE CLUSTERID
—————————— ——- ———-
T TABLE
TAB_CRM_CLIENT TABLE
TEST_COUNT TABLE
V_TEST VIEW
TEST TABLE
T_DROP TABLE
TAB_OLCM TABLE
7 rows selected.
4）对回收站中数据进行dml操作
SQL> delete from “BIN$qIuF4JHvDJDgQKjADAsarA==$0”;
delete from “BIN$qIuF4JHvDJDgQKjADAsarA==$0”
*
ERROR at line 1:
ORA-38301: can not perform DDL/DML over objects in Recycle Bin
5）对回收站中数据进行ddl操作
SQL> drop table “BIN$qIuF4JHvDJDgQKjADAsarA==$0”;
drop table “BIN$qIuF4JHvDJDgQKjADAsarA==$0”
*
ERROR at line 1:
ORA-38301: can not perform DDL/DML over objects in Recycle Bin
6）查看回收站，错误重现
Thu Jul 21 09:56:44 2011
performing DML/DDL operation over object in bin.
Thu Jul 21 09:57:19 2011
performing DML/DDL operation over object in bin.
7）说明
由于对回收站中的对象执行了dml或者ddl操作导致alert报类此做。在回收站中不能执行ddl或者dml操作

如果想以中文显示，则需要修改一些配置文件。
包括三个目录：
$ORACLE_HOME/jdk/jre/lib
$ORACLE_HOME/jre/1.4.2/lib
$ORACLE_HOME/javavm/lib/ojvmfonts
这三个目录下都有 font.properties 和 font.properties.zh_CN.Redhat 文件。
如果DB中没有找够三个目录，那只要修改找到的目录下面的就可以啦。
font.properties文件备份：
cp $ORACLE_HOME/jdk/jre/lib/font.properties $ORACLE_HOME/jdk/jre/lib/font.properties.bak
cp $ORACLE_HOME/jre/1.4.2/lib/font.properties $ORACLE_HOME/jre/1.4.2/lib/font.properties.bak
cp $ORACLE_HOME/javavm/lib/ojvmfonts/font.properties $ORACLE_HOME/javavm/lib/ojvmfonts/font.properties.bak
用font.properties.zh_CN.Redhat替换font.properties
cp $ORACLE_HOME/jdk/jre/lib/font.properties.zh_CN.Redhat $ORACLE_HOME/jdk/jre/lib/font.properties
cp $ORACLE_HOME/jre/1.4.2/lib/font.properties.zh_CN.Redhat $ORACLE_HOME/jre/1.4.2/lib/font.properties
cp $ORACLE_HOME/javavm/lib/ojvmfonts/font.properties.zh_CN.Redhat $ORACLE_HOME/javavm/lib/ojvmfonts/font.properties
修改font.properties最后一行
filename.-misc-zysong18030-medium-r-normal–*-%d-*-*-c-*-iso10646-1=/usr/share/fonts/zh_CN/TrueType/zysong.ttf
我们发现字体文件 /usr/share/fonts/zh_CN/TrueType/uming.ttf 根本是不存在的，有些系统可以直接做一个链接文件链接到系统存在的字体文件就可以解决掉乱码问题，但是我的系统做了链接以后还是没能解决，只好修改三个目录下修改后的 font.properties 文件的最后一行为如下内容：
filename.-misc-zysong18030-medium-r-normal–*-%d-*-*-c-*-iso10646-1=/usr/share/fonts/chinese/TrueType/uming.ttf
删除OEM缓存文件
rm -rf $ORACLE_HOME/oc4j/j2ee/oc4j_applications/applications/em/em/cabo/images/cache/zhs/*
重启OEM
emctl stop dbconsole
emctl start dbconsole
说明：
修改的前提必须保证系统里存在这个字体文件
ls /usr/share/fonts/chinese/TrueType/fonts.dir fonts.scale ukai.ttf uming.ttf
自己可以找本系统对应的中文字体文件。

来自Thomas Kyte 《Oracle9i/10g/11g编程艺术》 12章节中.由于long的操作限制,那么在操作long之前可以将该类型的结果转换为varchar2然后再使用,long_help.substr_of的基本意思是将long结果的的前4000字节转换为varchar2类型.如果long的数据超过了4000字节,那么可以将循环调用此函数

create or replace package long_help
authid current_user
as
function substr_of
( p_query in varchar2,
p_from  in number,
p_for   in number,
p_name1 in varchar2 default NULL,
p_bind1 in varchar2 default NULL,
p_name2 in varchar2 default NULL,
p_bind2 in varchar2 default NULL,
p_name3 in varchar2 default NULL,
p_bind3 in varchar2 default NULL,
p_name4 in varchar2 default NULL,
p_bind4 in varchar2 default NULL )
return varchar2;
end;
create or replace package body long_help
as
    g_cursor number := dbms_sql.open_cursor;
    g_query  varchar2(32765);
procedure bind_variable( p_name in varchar2, p_value in varchar2 )
is
begin
    if ( p_name is not null )
    then
        dbms_sql.bind_variable( g_cursor, p_name, p_value );
    end if;
end;
function substr_of
( p_query in varchar2,
  p_from  in number,
  p_for   in number,
  p_name1 in varchar2 default NULL,
  p_bind1 in varchar2 default NULL,
  p_name2 in varchar2 default NULL,
  p_bind2 in varchar2 default NULL,
  p_name3 in varchar2 default NULL,
  p_bind3 in varchar2 default NULL,
  p_name4 in varchar2 default NULL,
  p_bind4 in varchar2 default NULL )
return varchar2
as
    l_buffer       varchar2(4000);
    l_buffer_len   number;
begin
    if ( nvl(p_from,0) <= 0 )
    then
        raise_application_error
        (-20002, 'From must be >= 1 (positive numbers)' );
    end if;
    if ( nvl(p_for,0) not between 1 and 4000 )
    then
        raise_application_error
        (-20003, 'For must be between 1 and 4000' );
    end if;
    if ( p_query <> g_query or g_query is NULL )
    then
        if ( upper(trim(nvl(p_query,'x'))) not like 'SELECT%')
        then
            raise_application_error
            (-20001, 'This must be a select only' );
        end if;
        dbms_sql.parse( g_cursor, p_query, dbms_sql.native );
        g_query := p_query;
    end if;
    bind_variable( p_name1, p_bind1 );
    bind_variable( p_name2, p_bind2 );
    bind_variable( p_name3, p_bind3 );
    bind_variable( p_name4, p_bind4 );
    dbms_sql.define_column_long(g_cursor, 1);
    if (dbms_sql.execute_and_fetch(g_cursor)>0)
    then
        dbms_sql.column_value_long
        (g_cursor, 1, p_for, p_from-1,
         l_buffer, l_buffer_len );
    end if;
    return l_buffer;
end substr_of;
end;

使用方法:查询DBA_TAB_PARTITIONS中的HIGH_VALUE

SELECT *
  FROM (SELECT TABLE_OWNER,
                TABLE_NAME,
                PARTITION_NAME,
                LONG_HELP.SUBSTR_OF('SELECT HIGH_VALUE
FROM   DBA_TAB_PARTITIONS WHERE TABLE_OWNER=:TABLE_OWNER
AND TABLE_NAME=:TABLE_NAME
AND PARTITION_NAME=:PARTITION_NAME',
                                     1,
                                     4000,
                                     'TABLE_OWNER',
                                     TABLE_OWNER,
                                     'TABLE_NAME',
                                     TABLE_NAME,
                                     'PARTITION_NAME',
                                     PARTITION_NAME) HIGH_VALUE
           FROM DBA_TAB_PARTITIONS);

前几天被人问到聚簇因子被问住了，所以今天做了个试验说明这个问题
1、准备试验条件

--创建表t_1
CREATE TABLE t_1
AS
SELECT ROWNUM rn,a.* FROM all_objects a ORDER BY object_name DESC;
--创建t_1表关于rownum索引
CREATE INDEX ind_t_1 ON t_1(rn);
--创建表表t_2
CREATE TABLE t_2
AS
SELECT * FROM (
SELECT ROWNUM rn,a.* FROM all_objects a ) ORDER BY rn ASC;
--创建t_2表关于rownum索引
CREATE INDEX ind_t_2 ON t_2(rn);
--分析两张表及其索引
EXEC DBMS_STATS.gather_table_stats(USER, 'T_1');
EXEC DBMS_STATS.gather_table_stats(USER, 'T_2');
EXEC DBMS_STATS.gather_index_stats(USER, 'IND_T_1');
EXEC DBMS_STATS.gather_index_stats(USER, 'IND_T_2');
--说明：两个表的区别就是t_2表中的rn是有序的，刚刚建立t_2表的索引一致

2、执行查询操作
SQL> set autot traceonly stat;
SQL> SELECT * FROM t_1 WHERE rn BETWEEN 100 AND 120;
已选择21行。
统计信息
———————————————————-
0 recursive calls
0 db block gets
17 consistent gets
0 physical reads
0 redo size
1807 bytes sent via SQL*Net to client
357 bytes received via SQL*Net from client
3 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
21 rows processed
SQL> SELECT * FROM t_2 WHERE rn BETWEEN 100 AND 120;
已选择21行。
统计信息
———————————————————-
0 recursive calls
0 db block gets
7 consistent gets
0 physical reads
0 redo size
1807 bytes sent via SQL*Net to client
357 bytes received via SQL*Net from client
3 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
21 rows processed
3、观察试验结果
通过执行统计信息观察，t_1表的查询一致读是17，而t_2表的一致读只有7，尽然t_1的一致读尽然是t_2的2倍还多，是不是有点奇怪，同样的表结构，同样的数据（t_2多两条数据）
4、分析原因
通过查询聚簇因子发现，两个表的聚簇因子差别很大，基于rn的索引在rn是顺序排列的表中，clustering_factor的值相差很大。
在表中数据有时候属于无序状态，这个时候的CLUSTERING_FACTOR比较接近NUM_ROWS，说明如 果扫描整个表，每次都要根据Index来读取相应行的RowID，这个时候的IO操作很多，自然检索时间会比较长。如果数据有序的 话，CLUSTERING_FACTOR比较接近BLOCKS，说明相邻的数据在一个块中，减少了IO操作数量，自然检索时间会大大降低。

当一个数据块读入sga区，相应的buffer header会被放置到hash列表上,我们称其这hash chains,chain在中文的意为链条或串的意思，表达就是关连性.如果一个进程想访问或修改hash chain上的block,它首先要获得”cache buffers chains” latch。
原因一：低效率的SQL语句（主要体现在逻辑读过高）
cache buffers chains latch很大程度与逻辑读有关，所以要观注v$sql中BUFFER_GETS/EXECUTIONS大的语句。
同时每一个逻辑读需要一个latch get 操作及一个cpu操作，这样的sql也会很耗cpu资源。
原因二：热块（访问过于频繁）
找出热点块方法一：

--找出p1raw
select p1,p1raw from v$session_wait where event='latch: cache buffers chains';
--找到对象
SELECT /*+ RULE */
 E.OWNER || '.' || E.SEGMENT_NAME SEGMENT_NAME,
 E.PARTITION_NAME,
 E.EXTENT_ID EXTENT#,
 X.DBABLK - E.BLOCK_ID + 1 BLOCK#,
 X.TCH,
 L.CHILD#
  FROM SYS.V$LATCH_CHILDREN L, SYS.X$BH X, SYS.DBA_EXTENTS E
 WHERE X.HLADDR = '00000002576EE018'--p1raw
   AND E.FILE_ID = X.FILE#
   AND X.HLADDR = L.ADDR
   AND X.DBABLK BETWEEN E.BLOCK_ID AND E.BLOCK_ID + E.BLOCKS - 1
 ORDER BY X.TCH DESC;

找出热点块方法二：

--直接找出热点块
SELECT OBJECT_NAME, SUBOBJECT_NAME
  FROM DBA_OBJECTS
 WHERE DATA_OBJECT_ID IN
       (SELECT DATA_OBJECT_ID
          FROM (SELECT OBJ DATA_OBJECT_ID, FILE#, DBABLK, CLASS, STATE, TCH
                  FROM X$BH
                 WHERE HLADDR IN (SELECT ADDR
                                    FROM (SELECT ADDR
                                            FROM V$LATCH_CHILDREN
                                           ORDER BY (GETS + MISSES + SLEEPS) DESC)
                                   WHERE ROWNUM < 10)
                 ORDER BY TCH DESC)
         WHERE ROWNUM < 10);

LSNRCTL> change_password
Old password:
New password:
Reenter new password:
Connecting to (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=ecp-uc-db1)(PORT=1521)))
Password changed for LISTENER
The command completed successfully
LSNRCTL> set password
Password:
The command completed successfully
LSNRCTL> save_config
Connecting to (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=ecp-uc-db1)(PORT=1521)))
Saved LISTENER configuration parameters.
Listener Parameter File /opt/oracle/product/10.2.0/db_1/network/admin/listener.ora
Old Parameter File /opt/oracle/product/10.2.0/db_1/network/admin/listener.bak
The command completed successfully
[oracle@ecp-uc-db1 admin]$ cat listener.ora
#—-ADDED BY TNSLSNR 10-JUN-2011 18:13:24—
PASSWORDS_LISTENER = 6D7AA003392C436A
#——————————————–
note：10g数据库上需要上添加（重启监听）
LOCAL_OS_AUTHENTICATION_LISTENER = OFF
1、添加LOCAL_OS_AUTHENTICATION_LISTENER = OFF之前
Security ON: Password or Local OS Authentication
2、添加LOCAL_OS_AUTHENTICATION_LISTENER = OFF之后
Security ON: Password
LSNRCTL> status
Connecting to (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=ecp-uc-db1)(PORT=1521)))
TNS-01169: The listener has not recognized the password
LSNRCTL> stop
Connecting to (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=ecp-uc-db1)(PORT=1521)))
TNS-01169: The listener has not recognized the password
LSNRCTL> set password 123456
The command completed successfully
LSNRCTL> status
Connecting to (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=ecp-uc-db1)(PORT=1521)))
STATUS of the LISTENER
————————
Alias LISTENER
Version TNSLSNR for Linux: Version 10.2.0.4.0 – Production
Start Date 10-JUN-2011 18:15:49
Uptime 0 days 0 hr. 1 min. 16 sec
Trace Level off
Security ON: Password
SNMP OFF
Listener Parameter File /opt/oracle/product/10.2.0/db_1/network/admin/listener.ora
Listener Log File /opt/oracle/product/10.2.0/db_1/network/log/listener.log
Listening Endpoints Summary…
(DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=ECP-UC-DB1)(PORT=1521)))
(DESCRIPTION=(ADDRESS=(PROTOCOL=ipc)(KEY=EXTPROC0)))
Services Summary…
Service “PLSExtProc” has 1 instance(s).
Instance “PLSExtProc”, status UNKNOWN, has 1 handler(s) for this service…
Service “ecp” has 1 instance(s).
Instance “ecp”, status READY, has 1 handler(s) for this service…
Service “ecpXDB” has 1 instance(s).
Instance “ecp”, status READY, has 1 handler(s) for this service…
Service “ecp_XPT” has 1 instance(s).
Instance “ecp”, status READY, has 1 handler(s) for this service…
The command completed successfully