以前我有个想法，我只想对其中的某个或者某几个分区上创建索引，其他分区不想创建，在12C之前的版本，无论是Local还是Global index，都不能实现该需求，但是从ORACLE 12C开始引进了Partial Global/Local Indexes for Partitioned Tables，解决了该问题，可以在指定的分区上创建本地索引或者全局索引，主要语法是在表或者分区,子分区级别设置[INDEXING { ON | OFF }]
创建测试表

CDB_PDB@CHF> SELECT BANNER FROM V$VERSION;

BANNER
--------------------------------------------------------------------------------
Oracle Database 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production
PL/SQL Release 12.1.0.1.0 - Production
CORE    12.1.0.1.0      Production
TNS for 64-bit Windows: Version 12.1.0.1.0 - Production
NLSRTL Version 12.1.0.1.0 - Production

CDB_PDB@CHF> CREATE TABLE xifenfei_orders (
  2  order_id NUMBER(12),
  3  order_address varchar2(100),
  4  order_mode VARCHAR2(20))
  5  INDEXING OFF
  6  PARTITION BY RANGE (order_id)
  7  (PARTITION ord_p1 VALUES LESS THAN (100) INDEXING ON,
  8  PARTITION ord_p2 VALUES LESS THAN (200) INDEXING OFF,
  9  PARTITION ord_p3 VALUES LESS THAN (300) INDEXING ON,
 10  PARTITION ord_p4 VALUES LESS THAN (400),
 11  PARTITION ord_p5 VALUES LESS THAN (500));

表已创建。

创建LOCAL INDEX

CDB_PDB@CHF> CREATE INDEX ind_lox on xifenfei_orders(order_address) LOCAL INDEXING PARTIAL;

索引已创建。

CDB_PDB@CHF> COL INDEX_NAME FOR A10
CDB_PDB@CHF> COL PARTITION_NAME FOR A15
CDB_PDB@CHF> select index_name, partition_name,STATUS
  2   from user_ind_partitions
  3   where index_name = 'IND_LOX';

INDEX_NAME PARTITION_NAME  STATUS
---------- --------------- --------
IND_LOX    ORD_P5          UNUSABLE
IND_LOX    ORD_P4          UNUSABLE
IND_LOX    ORD_P3          USABLE
IND_LOX    ORD_P2          UNUSABLE
IND_LOX    ORD_P1          USABLE

--设置INDEXING OFF对应的index 状态为UNUSABLE，分区继承表


CDB_PDB@CHF> select partition_name,indexing from user_tab_partitions where table_name='XIFENFEI_ORDERS';

PARTITION_NAME  INDE
--------------- ----
ORD_P5          OFF
ORD_P4          OFF
ORD_P3          ON
ORD_P2          OFF
ORD_P1          ON

--因为segment 延迟，无数据，所以无分区和索引记录
CDB_PDB@CHF> SELECT PARTITION_NAME,SEGMENT_NAME FROM USER_SEGMENTS WHERE SEGMENT_NAME='IND_LOX';

未选定行

CDB_PDB@CHF> SELECT PARTITION_NAME,SEGMENT_NAME FROM USER_SEGMENTS WHERE SEGMENT_NAME='XIFENFEI_ORDERS';

未选定行

CDB_PDB@CHF> begin
  2    for i in 1 .. 449 loop
  3      insert into xifenfei_orders
  4      values
  5        (i,'www.orasos.com'||i,'惜分飞'||i);
  6    end loop;
  7  commit;
  8  end;
  9  /

PL/SQL 过程已成功完成。

--插入记录后，分区表有相关记录
CDB_PDB@CHF> SELECT PARTITION_NAME,SEGMENT_NAME,blocks FROM USER_SEGMENTS WHERE SEGMENT_NAME='XIFENFEI_ORDERS';

PARTITION_NAME  SEGMENT_NAME        BLOCKS
--------------- --------------- ----------
ORD_P5          XIFENFEI_ORDERS       1024
ORD_P4          XIFENFEI_ORDERS       1024
ORD_P3          XIFENFEI_ORDERS       1024
ORD_P2          XIFENFEI_ORDERS       1024
ORD_P1          XIFENFEI_ORDERS       1024

--注意:这里只有user_tab_partitions.indexing为on的有记录,也就是说，至于这些分区的索引被创建，其他的未被创建
CDB_PDB@CHF> SELECT PARTITION_NAME,SEGMENT_NAME FROM USER_SEGMENTS WHERE SEGMENT_NAME='IND_LOX';

PARTITION_NAME  SEGMENT_NAME
--------------- ---------------
ORD_P1          IND_LOX
ORD_P3          IND_LOX

分析执行计划

CDB_PDB@CHF> EXEC DBMS_STATS.GATHER_TABLE_STATS(USER,'XIFENFEI_ORDERS',CASCADE=>TRUE);

PL/SQL 过程已成功完成。

CDB_PDB@CHF> SET AUTOT TRACE
CDB_PDB@CHF> SELECT ORDER_MODE FROM XIFENFEI_ORDERS WHERE ORDER_ADDRESS='www.orasos.com99';

未选定行


执行计划
----------------------------------------------------------
Plan hash value: 2800545636

------------------------------------------------------------------------------------------------------------------------
--------
| Id  | Operation                                    | Name            | Rows  | Bytes | Cost (%CPU)| Time     | Pstart|
 Pstop |
------------------------------------------------------------------------------------------------------------------------
--------
|   0 | SELECT STATEMENT                             |                 |     1 |    30 |    40   (0)| 00:00:01 |       |
       |
|   1 |  VIEW                                        | VW_TE_2         |     2 |    24 |    40   (0)| 00:00:01 |       |
       |
|   2 |   UNION-ALL                                  |                 |       |       |            |          |       |
       |
|   3 |    PARTITION RANGE OR                        |                 |     1 |    34 |     1   (0)| 00:00:01 |KEY(OR)|
KEY(OR)|
|*  4 |     TABLE ACCESS BY LOCAL INDEX ROWID BATCHED| XIFENFEI_ORDERS |     1 |    34 |     1   (0)| 00:00:01 |KEY(OR)|
KEY(OR)|
|*  5 |      INDEX RANGE SCAN                        | IND_LOX         |     1 |       |     1   (0)| 00:00:01 |KEY(OR)|
KEY(OR)|
|   6 |    PARTITION RANGE OR                        |                 |     1 |    34 |    39   (0)| 00:00:01 |KEY(OR)|
KEY(OR)|
|*  7 |     TABLE ACCESS FULL                        | XIFENFEI_ORDERS |     1 |    34 |    39   (0)| 00:00:01 |KEY(OR)|
KEY(OR)|
------------------------------------------------------------------------------------------------------------------------
--------

Predicate Information (identified by operation id):
---------------------------------------------------

   4 - filter("XIFENFEI_ORDERS"."ORDER_ID"<100 OR "XIFENFEI_ORDERS"."ORDER_ID">=200 AND
              "XIFENFEI_ORDERS"."ORDER_ID"<300)
   5 - access("ORDER_ADDRESS"='www.orasos.com99')
   7 - filter("ORDER_ADDRESS"='www.orasos.com99' AND ("XIFENFEI_ORDERS"."ORDER_ID">=300 AND 
"XIFENFEI_ORDERS"."ORDER_ID"<500 OR "XIFENFEI_ORDERS"."ORDER_ID"<200 AND "XIFENFEI_ORDERS"."ORDER_ID">=100))

--这里可以看到，当我们没有指定分区范围的时候，显示的执行计划是有分区index的部分直接走index，没有分区index的部分是扫描分区


统计信息
----------------------------------------------------------
         34  recursive calls
          0  db block gets
        120  consistent gets
          1  physical reads
          0  redo size
        347  bytes sent via SQL*Net to client
        533  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          5  sorts (memory)
          0  sorts (disk)
          0  rows processed

CDB_PDB@CHF> SELECT ORDER_MODE FROM XIFENFEI_ORDERS t WHERE ORDER_ADDRESS='www.orasos.com499' and order_id>200 and ord
er_id<300;

未选定行


执行计划
----------------------------------------------------------
Plan hash value: 3337708912

------------------------------------------------------------------------------------------------------------------------
------
| Id  | Operation                                  | Name            | Rows  | Bytes | Cost (%CPU)| Time     | Pstart| P
stop |
------------------------------------------------------------------------------------------------------------------------
------
|   0 | SELECT STATEMENT                           |                 |     1 |    34 |     2   (0)| 00:00:01 |       |
     |
|   1 |  PARTITION RANGE SINGLE                    |                 |     1 |    34 |     2   (0)| 00:00:01 |     3 |
   3 |
|*  2 |   TABLE ACCESS BY LOCAL INDEX ROWID BATCHED| XIFENFEI_ORDERS |     1 |    34 |     2   (0)| 00:00:01 |     3 |
   3 |
|*  3 |    INDEX RANGE SCAN                        | IND_LOX         |     1 |       |     1   (0)| 00:00:01 |     3 |
   3 |
------------------------------------------------------------------------------------------------------------------------
------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("ORDER_ID">200)
   3 - access("ORDER_ADDRESS"='www.orasos.com499')

--指定分区查询，可以明确的看到，该sql直接使用了分区索引


统计信息
----------------------------------------------------------
          1  recursive calls
          0  db block gets
          1  consistent gets
          0  physical reads
          0  redo size
        347  bytes sent via SQL*Net to client
        533  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed

CDB_PDB@CHF> SELECT ORDER_MODE FROM XIFENFEI_ORDERS t WHERE ORDER_ADDRESS='www.orasos.com499' and order_id>300 and ord
er_id<400;

未选定行


执行计划
----------------------------------------------------------
Plan hash value: 2072227240

----------------------------------------------------------------------------------------------------------
| Id  | Operation              | Name            | Rows  | Bytes | Cost (%CPU)| Time     | Pstart| Pstop |
----------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT       |                 |     1 |    34 |    14   (0)| 00:00:01 |       |    |
|   1 |  PARTITION RANGE SINGLE|                 |     1 |    34 |    14   (0)| 00:00:01 |     4 |     4 |
|*  2 |   TABLE ACCESS FULL    | XIFENFEI_ORDERS |     1 |    34 |    14   (0)| 00:00:01 |     4 |     4 |
----------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("ORDER_ADDRESS"='www.orasos.com499' AND "ORDER_ID">300)

--当指定的分区无index之时，直接判断走全表扫描

统计信息
----------------------------------------------------------
          1  recursive calls
          0  db block gets
         22  consistent gets
          0  physical reads
          0  redo size
        347  bytes sent via SQL*Net to client
        533  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed

创建Global index

CDB_PDB@CHF> ALTER INDEX ind_lox INVISIBLE;

索引已更改。

CDB_PDB@CHF> CREATE INDEX IND_G_LOX ON XIFENFEI_ORDERS(ORDER_ADDRESS) Global  INDEXING PARTIAL;

索引已创建。

CDB_PDB@CHF> select index_name,indexing from dba_indexes where index_name='IND_G_LOX';

INDEX_NAME INDEXIN
---------- -------
IND_G_LOX  PARTIAL

执行计划

CDB_PDB@CHF> SELECT ORDER_MODE FROM XIFENFEI_ORDERS t WHERE ORDER_ADDRESS='www.orasos.com99';

未选定行


执行计划
----------------------------------------------------------
Plan hash value: 1912382893

------------------------------------------------------------------------------------------------------------------------
--------
| Id  | Operation                                    | Name            | Rows  | Bytes | Cost (%CPU)| Time     | Pstart|
 Pstop |
------------------------------------------------------------------------------------------------------------------------
--------
|   0 | SELECT STATEMENT                             |                 |     1 |    30 |    41   (0)| 00:00:01 |       |
       |
|   1 |  VIEW                                        | VW_TE_2         |     2 |    24 |    41   (0)| 00:00:01 |       |
       |
|   2 |   UNION-ALL                                  |                 |       |       |            |          |       |
       |
|*  3 |    TABLE ACCESS BY GLOBAL INDEX ROWID BATCHED| XIFENFEI_ORDERS |     1 |    34 |     2   (0)| 00:00:01 | ROWID |
 ROWID |
|*  4 |     INDEX RANGE SCAN                         | IND_G_LOX       |     1 |       |     1   (0)| 00:00:01 |       |
       |
|   5 |    PARTITION RANGE OR                        |                 |     1 |    34 |    39   (0)| 00:00:01 |KEY(OR)|
KEY(OR)|
|*  6 |     TABLE ACCESS FULL                        | XIFENFEI_ORDERS |     1 |    34 |    39   (0)| 00:00:01 |KEY(OR)|
KEY(OR)|
------------------------------------------------------------------------------------------------------------------------
--------

Predicate Information (identified by operation id):
---------------------------------------------------

   3 - filter("T"."ORDER_ID"<100 OR "T"."ORDER_ID">=200 AND "T"."ORDER_ID"<300)
   4 - access("ORDER_ADDRESS"='www.orasos.com99')
   6 - filter("ORDER_ADDRESS"='www.orasos.com99' AND ("T"."ORDER_ID">=300 AND "T"."ORDER_ID"<500 OR
              "T"."ORDER_ID"<200 AND "T"."ORDER_ID">=100))

--这里可以看到因为没有指定分区范围，该sql在含index的分区使用全局index，在没有index的分区直接使用全表扫描

统计信息
----------------------------------------------------------
          1  recursive calls
          0  db block gets
         67  consistent gets
          0  physical reads
          0  redo size
        347  bytes sent via SQL*Net to client
        533  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed


CDB_PDB@CHF> SELECT ORDER_MODE FROM XIFENFEI_ORDERS t WHERE ORDER_ADDRESS='www.orasos.com99' and order_id<100;


执行计划
----------------------------------------------------------
Plan hash value: 3717359654

------------------------------------------------------------------------------------------------------------------------
------
| Id  | Operation                                  | Name            | Rows  | Bytes | Cost (%CPU)| Time     | Pstart| P
stop |
------------------------------------------------------------------------------------------------------------------------
------
|   0 | SELECT STATEMENT                           |                 |     1 |    31 |     2   (0)| 00:00:01 |       |
     |
|*  1 |  TABLE ACCESS BY GLOBAL INDEX ROWID BATCHED| XIFENFEI_ORDERS |     1 |    31 |     2   (0)| 00:00:01 |     1 |
   1 |
|*  2 |   INDEX RANGE SCAN                         | IND_G_LOX       |     1 |       |     1   (0)| 00:00:01 |       |
     |
------------------------------------------------------------------------------------------------------------------------
------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("ORDER_ID"<100)
   2 - access("ORDER_ADDRESS"='www.orasos.com99')

--指定了分区范围，而且该分区又有index，直接使用全局index

统计信息
----------------------------------------------------------
          1  recursive calls
          0  db block gets
          2  consistent gets
          0  physical reads
          0  redo size
        551  bytes sent via SQL*Net to client
        544  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

CDB_PDB@CHF> SELECT ORDER_MODE FROM XIFENFEI_ORDERS t WHERE ORDER_ADDRESS='www.orasos.com99' and order_id>400

未选定行


执行计划
----------------------------------------------------------
Plan hash value: 2072227240

----------------------------------------------------------------------------------------------------------
| Id  | Operation              | Name            | Rows  | Bytes | Cost (%CPU)| Time     | Pstart| Pstop |
----------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT       |                 |     1 |    34 |    14   (0)| 00:00:01 |       |    |
|   1 |  PARTITION RANGE SINGLE|                 |     1 |    34 |    14   (0)| 00:00:01 |     5 |     5 |
|*  2 |   TABLE ACCESS FULL    | XIFENFEI_ORDERS |     1 |    34 |    14   (0)| 00:00:01 |     5 |     5 |
----------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("ORDER_ADDRESS"='www.orasos.com99' AND "ORDER_ID">400)

--指定了分区范围，但是该分区无index，直接使用全表扫描

统计信息
----------------------------------------------------------
          1  recursive calls
          0  db block gets
         22  consistent gets
          0  physical reads
          0  redo size
        347  bytes sent via SQL*Net to client
        533  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed

结论
通过测试，证明Partial Global/Local Indexes for Partitioned Tables确实能够实现对部分分区创建index。
1)如果查询条件确定的分区无index，那直接走全表扫描；
2)如果查询条件可以确定的分区范围内有index，会直接使用index(无论是Local还是GLobal)；
3)如果查询条件未确定分区范围(含index和无index分区情况)，那该sql会在有index分区使用index，在没有index区域走全表扫描

从ORACLE 12C开始允许在同一列上创建不同类型的index，用来做性能调优，不同类型的index主要包括（B*TREE vs BITMAP，Local vs Global)
ORACLE 11G

SQL> select * from v$version;

BANNER
--------------------------------------------------------------------------------
Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - Production
PL/SQL Release 11.2.0.3.0 - Production
CORE    11.2.0.3.0      Production
TNS for Linux: Version 11.2.0.3.0 - Production
NLSRTL Version 11.2.0.3.0 - Production

SQL> CREATE TABLE T_XIFENFEI
  2    (
  3    EMPNO      NUMBER(9)       PRIMARY KEY ,
  4    EMPNAME    VARCHAR2(15)    NOT NULL,
  5    )
  6    /

Table created.

SQL> CREATE INDEX EMP_DEPT_BTREE_IDX ON EMP(DEPTNO);

Index created.

SQL> CREATE BITMAP INDEX EMP_DEPTNO_BITMAP_IDX ON EMP(DEPTNO) INVISIBLE  ;
CREATE BITMAP INDEX EMP_DEPTNO_BITMAP_IDX ON EMP(DEPTNO) INVISIBLE
                                                 *
ERROR at line 1:
ORA-01408: such column list already indexed

ORACLE 12C

CDB_PDB@CHF> select * from v$version;

BANNER                                                                               CON_ID
-------------------------------------------------------------------------------- ----------
Oracle Database 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production              0
PL/SQL Release 12.1.0.1.0 - Production                                                    0
CORE    12.1.0.1.0      Production                                                        0
TNS for 64-bit Windows: Version 12.1.0.1.0 - Production                                   0
NLSRTL Version 12.1.0.1.0 - Production                                                    0

CDB_PDB@CHF> CREATE TABLE T_XIFENFEI
  2    (
  3    EMPNO      NUMBER(9)       PRIMARY KEY ,
  4    EMPNAME    VARCHAR2(15)    NOT NULL
  5    )
  6    /

表已创建。

CDB_PDB@CHF> CREATE INDEX EMP_DEPT_BTREE_IDX ON EMP(DEPTNO);

索引已创建。

CDB_PDB@CHF> CREATE BITMAP INDEX EMP_DEPTNO_BITMAP_IDX ON EMP(DEPTNO);
CREATE BITMAP INDEX EMP_DEPTNO_BITMAP_IDX ON EMP(DEPTNO)
                                                 *
第 1 行出现错误:
ORA-01408: 此列列表已索引


CDB_PDB@CHF> CREATE BITMAP INDEX EMP_DEPTNO_BITMAP_IDX ON EMP(DEPTNO) INVISIBLE ;

索引已创建。

ORACLE 8i中引进了全局临时表,而且把临时表的数据存储在临时表空间以减少资源开销。全局临时表在临时表空间的数据会随着一个事务或者一个会话而结束。一个临时表上执行dml操作自身不产生redo,因为这些数据是存储在临时表空间，但是这些操作会产生uundo，而undo存储是永久表空间，进而undo反过来会产生redo。而对于这些临时表是不需要恢复的，因此这些redo是不是可以不要产生，在ORACLE 12C中引进了Temporary undo的功能(temp_undo_enabled参数控制)，很大程度上解决了该问题

创建临时表

CDB_CDB$ROOT@SYS> conn chf/xifenfei@pdb
已连接。
CDB_PDB@CHF> select * from v$version;

BANNER                                                                               CON_ID
-------------------------------------------------------------------------------- ----------
Oracle Database 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production              0
PL/SQL Release 12.1.0.1.0 - Production                                                    0
CORE    12.1.0.1.0      Production                                                        0
TNS for 64-bit Windows: Version 12.1.0.1.0 - Production                                   0
NLSRTL Version 12.1.0.1.0 - Production                                                    0

temp_undo_enabled=false测试

CDB_PDB@CHF> show parameter temp_undo_enabled;

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
temp_undo_enabled                    boolean     FALSE
CDB_PDB@CHF> create global temporary table g_t_xifenfei(c1 number(8), c2 char(100));

表已创建。

CDB_PDB@CHF> select name, value from v$mystat natural join v$statname
  2    where name = 'redo size' or name = 'redo entries';

NAME                                                                  VALUE
---------------------------------------------------------------- ----------
redo entries                                                             29
redo size                                                              8596

CDB_PDB@CHF> insert into g_t_xifenfei select level, level from dual connect by level <= 10000;

已创建 10000 行。

CDB_PDB@CHF> select name, value from v$mystat natural join v$statname
  2    where name = 'redo size' or name = 'redo entries';

NAME                                                                  VALUE
---------------------------------------------------------------- ----------
redo entries                                                            314
redo size                                                             69892

这里可以发现当temp_undo_enabled=false之时,redo的量在该执行过程中还是增加的很明显,从8596增加到69892(增加了61296 byte的redo)，redo entries也增加了285

temp_undo_enabled=true测试

CDB_PDB@CHF> alter session set temp_undo_enabled = true;

会话已更改。

CDB_PDB@CHF> select name, value from v$mystat natural join v$statname
  2    where name = 'redo size' or name = 'redo entries';

NAME                                                                  VALUE
---------------------------------------------------------------- ----------
redo entries                                                              2
redo size                                                               720

CDB_PDB@CHF> insert into g_t_xifenfei select level, level from dual connect by level <= 10000;

已创建 10000 行。

CDB_PDB@CHF> select name, value from v$mystat natural join v$statname
  2    where name = 'redo size' or name = 'redo entries';

NAME                                                                  VALUE
---------------------------------------------------------------- ----------
redo entries                                                              3
redo size                                                              1000

这里可以看到redo的量之时增加了280byte,redo entries增加了1,增加量非常小,完全和temp_undo_enabled=false不是一个等量级

补充说明
在临时表使用比较多的系统中,设置temp_undo_enabled=true可以减少不少的redo产生,正是因为这样的机制产生,可以实现ORACLE 12C ADG中可以实现全局临时表的DML操作

上篇文章(某个pdb可以在root pdb open状态下进行恢复)发布后,还是有不少朋友有各种争议,比如我的cdb本来就是open的,比如与oracle 12c 新特性文档描述不符等等,这里根据weibo上的各种争论和各位想的测试情况,我继续测试如下
查询数据库相关状态

C:\Users\XIFENFEI>sqlplus / as sysdba

SQL*Plus: Release 12.1.0.1.0 Production on 星期二 8月 13 12:49:47 2013

Copyright (c) 1982, 2013, Oracle.  All rights reserved.


连接到:
Oracle Database 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production
With the Partitioning, OLAP, Advanced Analytics and Real Application Testing options

SQL> show pdbs;

    CON_ID CON_NAME                       OPEN MODE  RESTRICTED
---------- ------------------------------ ---------- ----------
         2 PDB$SEED                       READ ONLY  NO
         3 PDB                            READ WRITE NO
SQL> select name from v$datafile where con_id=3;

NAME
--------------------------------------------------------------------------------
E:\APP\XIFENFEI\ORADATA\CDB\PDB\SYSTEM01.DBF
E:\APP\XIFENFEI\ORADATA\CDB\PDB\SYSAUX01.DBF
E:\APP\XIFENFEI\ORADATA\CDB\PDB\PDB_USERS01.DBF

删除pdb system01.dbf文件

SQL> host dir E:\APP\XIFENFEI\ORADATA\CDB\PDB\SYSTEM01.DBF
 驱动器 E 中的卷没有标签。
 卷的序列号是 000C-3B41

 E:\APP\XIFENFEI\ORADATA\CDB\PDB 的目录

找不到文件

做checkpoint操作

--删除pdb的数据文件(本测试中是system01.dbf),因为dbwr无法写入数据到文件，导致crash，在实际中也有可能是ckpt,lgwr进程等
SQL> alter system checkpoint ;
alter system checkpoint
*
第 1 行出现错误:
ORA-03113: 通信通道的文件结尾
进程 ID: 9316
会话 ID: 136 序列号: 3165

alert日志中写明是dbwr进程无法写文件导致整个cdb crash

Tue Aug 13 12:51:10 2013
Thread 1 advanced to log sequence 316 (LGWR switch)
  Current log# 4 seq# 316 mem# 0: E:\APP\XIFENFEI\ORADATA\CDB\REDO04.LOG
Tue Aug 13 12:51:12 2013
Archived Log entry 5 added for thread 1 sequence 315 ID 0x7377d8de dest 1:
Tue Aug 13 12:51:16 2013
KCF: read, write or open error, block=0x6ad1 online=1
        file=7 'E:\APP\XIFENFEI\ORADATA\CDB\PDB\SYSTEM01.DBF'
        error=27070 txt: 'OSD-04016: 异步 I/O 请求排队时出错。'
Tue Aug 13 12:51:16 2013
Errors in file E:\APP\XIFENFEI\diag\rdbms\cdb\cdb\trace\cdb_dbw0_8612.trc:
Tue Aug 13 12:51:16 2013
Errors in file E:\APP\XIFENFEI\diag\rdbms\cdb\cdb\trace\cdb_dbw0_8612.trc:
ORA-63999: 数据文件出现介质故障
ORA-01114: 将块写入文件 7 时出现 IO 错误 (块 # 27345)
ORA-01110: 数据文件 7: 'E:\APP\XIFENFEI\ORADATA\CDB\PDB\SYSTEM01.DBF'
ORA-27070: 异步读取/写入失败
OSD-04016: 异步 I/O 请求排队时出错。
USER (ospid: 8612): terminating the instance due to error 63999
Tue Aug 13 12:51:17 2013
System state dump requested by (instance=1, osid=8612 (DBW0)), summary=[abnormal instance termination].
System State dumped to trace file E:\APP\XIFENFEI\diag\rdbms\cdb\cdb\trace\cdb_diag_9720.trc
Dumping diagnostic data in directory=[cdmp_20130813125117], requested by (instance=1, osid=8612 (DBW0)), summary=[abnormal instance termination].
Tue Aug 13 12:51:29 2013
Instance terminated by USER, pid = 8612

open cdb数据库

SQL> conn / as sysdba
已连接到空闲例程。
SQL> startup
ORACLE 例程已经启动。

Total System Global Area  521936896 bytes
Fixed Size                  2404552 bytes
Variable Size             293605176 bytes
Database Buffers          218103808 bytes
Redo Buffers                7823360 bytes
数据库装载完毕。
ORA-01157: 无法标识/锁定数据文件 7 - 请参阅 DBWR 跟踪文件
ORA-01110: 数据文件 7: 'E:\APP\XIFENFEI\ORADATA\CDB\PDB\SYSTEM01.DBF'

--直接在cdb中无法offline pdb数据文件
SQL> alter database datafile 7 offline;
alter database datafile 7 offline
*
第 1 行出现错误:
ORA-01516: 不存在的日志文件, 数据文件或临时文件 "7"


SQL> alter session set container=pdb;

会话已更改。

SQL> alter database datafile 7 offline;

数据库已更改。

SQL> conn / as sysdba
已连接。
SQL> alter database open;

数据库已更改。

SQL> show pdbs;

    CON_ID CON_NAME                       OPEN MODE  RESTRICTED
---------- ------------------------------ ---------- ----------
         2 PDB$SEED                       READ ONLY  NO
         3 PDB                            MOUNTED
SQL> exit
从 Oracle Database 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production
With the Partitioning, OLAP, Advanced Analytics and Real Application Testing options 断开

恢复并open pdb

C:\Users\XIFENFEI>rman target /

恢复管理器: Release 12.1.0.1.0 - Production on 星期二 8月 13 12:56:54 2013

Copyright (c) 1982, 2013, Oracle and/or its affiliates.  All rights reserved.

已连接到目标数据库: CDB (DBID=1937199326)

RMAN> restore datafile 7;

启动 restore 于 13-8月 -13
使用目标数据库控制文件替代恢复目录
分配的通道: ORA_DISK_1
通道 ORA_DISK_1: SID=361 设备类型=DISK

通道 ORA_DISK_1: 正在开始还原数据文件备份集
通道 ORA_DISK_1: 正在指定从备份集还原的数据文件
通道 ORA_DISK_1: 将数据文件 00007 还原到 E:\APP\XIFENFEI\ORADATA\CDB\PDB\SYSTEM01.DBF
通道 ORA_DISK_1: 正在读取备份片段 D:\PDB_SYS_01OH54LF_1_1.RMAN
通道 ORA_DISK_1: 段句柄 = D:\PDB_SYS_01OH54LF_1_1.RMAN 标记 = TAG20130812T223943
通道 ORA_DISK_1: 已还原备份片段 1
通道 ORA_DISK_1: 还原完成, 用时: 00:00:16
完成 restore 于 13-8月 -13

RMAN> recover datafile 7;

启动 recover 于 13-8月 -13
使用通道 ORA_DISK_1

正在开始介质的恢复

线程 1 序列 314 的归档日志已作为文件 E:\APP\XIFENFEI\FAST_RECOVERY_AREA\CDB\ARCHIVELOG\2013_08_12\O1_MF_1_314_90KXOZSF_.
ARC 存在于磁盘上
线程 1 序列 315 的归档日志已作为文件 E:\APP\XIFENFEI\FAST_RECOVERY_AREA\CDB\ARCHIVELOG\2013_08_13\O1_MF_1_315_90MGSZ0X_.
ARC 存在于磁盘上
线程 1 序列 316 的归档日志已作为文件 E:\APP\XIFENFEI\FAST_RECOVERY_AREA\CDB\ARCHIVELOG\2013_08_13\O1_MF_1_316_90MH1FGD_.
ARC 存在于磁盘上
归档日志文件名=E:\APP\XIFENFEI\FAST_RECOVERY_AREA\CDB\ARCHIVELOG\2013_08_12\O1_MF_1_314_90KXOZSF_.ARC 线程=1 序列=314
介质恢复完成, 用时: 00:00:02
完成 recover 于 13-8月 -13

RMAN> exit


恢复管理器完成。


C:\Users\XIFENFEI>sqlplus / as sysdba

SQL*Plus: Release 12.1.0.1.0 Production on 星期二 8月 13 12:58:42 2013

Copyright (c) 1982, 2013, Oracle.  All rights reserved.


连接到:
Oracle Database 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production
With the Partitioning, OLAP, Advanced Analytics and Real Application Testing options

SQL> alter session set container=pdb;

会话已更改。

SQL> alter database open;
alter database open
*
第 1 行出现错误:
ORA-01147: SYSTEM 表空间文件 7 处于脱机状态


SQL> alter database datafile 7 online;

数据库已更改。

SQL> alter database open;

数据库已更改。

SQL> conn / as sysdba
已连接。
SQL> show pdbs;

    CON_ID CON_NAME                       OPEN MODE  RESTRICTED
---------- ------------------------------ ---------- ----------
         2 PDB$SEED                       READ ONLY  NO
         3 PDB                            READ WRITE NO

整个操作过程说明一下问题
1.当pdb丢失数据文件或者影响dbwr,lgwr,ckpt等后台进程工作之时,将导致整个cdb crash,因为这些进程都是公用的
2.如果pdb丢失数据文件,通过在pdb中offline该文件,可以正常open cdb和其他pdb,不会长时间影响其他pdb工作
3.在cdb open的情况下,可以恢复pdb的任何数据文件(特殊pdb除外,root,seed之类)