oracle segment undo_71_UNDO扩展学习

UNDO扩展学习

UNDO是Oracle在UNDO SEGMENT(回滚段)中记录的信息。下面使用UNDO这名字可能会包含两种意思，一是前面说的回滚段中的记录、二是UNDO整个机制。

在这部分内容里，没有涉及闪回技术，要知道不少闪回技术是基于UNDO机制来实现的，比如flashback query、flashback table、flashback version query等。

首先看下我的操作环境：

sys@L10GR204> select * from v$version;

BANNER

----------------------------------------------------------------------------------------------------

Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bi

PL/SQL Release 10.2.0.4.0 - Production

CORE 10.2.0.4.0 Production

TNS for Linux: Version 10.2.0.4.0 - Production

NLSRTL Version 10.2.0.4.0 - Production

下面列出本节要讨论的内容：

1) 什么时候联机回滚段？

2) 如何证明SYSTEM回滚段只对系统对象提供服务

3) alter system set undo_tablespace = uuu;

4) 如何创建一个新的UNDO表空间并使用

5) 回滚段增减

6) 了解几个参数

7) 构造一个CR块

8) 快速构造ora-01555 快照过旧

9) enq: US – contention

10) Oracle INSERT、UPDATE、DELETE与Undo

什么时候联机回滚段？

这是一个非常简单的问题，如果大家已经掌握了数据库的启动过程，那就很容易回答上了。

看下面简单实验：

当前实例为MOOUNT状态

sys@L10GR204> select status from v$instance;

STATUS

------------------------

MOUNTED

这时候试图访问和UNDO相关视图，将会报错

sys@L10GR204> select * from v$rollname;

select * from v$rollname

ERROR at line 1:

ORA-01219: database not open: queries allowed on fixed tables/views only

当我们OPEN之后，即可读取回滚段信息

sys@L10GR204> alter database open;

Database altered.

sys@L10GR204> col name for a55

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- ------------------------------------------------------- ------------------------------

0 SYSTEM ONLINE

1 _SYSSMU1$ ONLINE

2 _SYSSMU2$ ONLINE

3 _SYSSMU3$ ONLINE

4 _SYSSMU4$ ONLINE

5 _SYSSMU5$ ONLINE

6 _SYSSMU6$ ONLINE

7 _SYSSMU7$ ONLINE

8 _SYSSMU8$ ONLINE

9 _SYSSMU9$ ONLINE

10 _SYSSMU10$ ONLINE

11 rows selected.

-- 仔细看，USN为0的回滚段为SYSTEM，这是我们创建数据库的时候创建于系统表空间的回滚段，只提供于更改系统对象做服务。

sys@L10GR204> set line 150

sys@L10GR204> col SEGMENT_NAME for a15

sys@L10GR204> col TABLESPACE_NAME for a10

sys@L10GR204> select segment_name, segment_type, tablespace_name from dba_segments where segment_name = 'SYSTEM';

SEGMENT_NAME SEGMENT_TYPE TABLESPACE

--------------- ------------------------------------ ----------

SYSTEM ROLLBACK SYSTEM

如何证明SYSTEM回滚段只对系统对象提供服务

为了证明这句话的正确性，我先将非SYSTEM的回滚段去除，只要将当前UNDO表空间脱机处理即可。

sys@L10GR204> select file_id, tablespace_name from dba_data_files where tablespace_name = 'UNDOTBS1';

FILE_ID TABLESPACE

---------- ----------

2 UNDOTBS1

UNDO表空间脱机需要注意操作：

在非归档模式如下操作

OPEN模式直接操作：

sys@L10GR204> alter database datafile 2 offline drop;

ERROR:

ORA-03114: not connected to ORACLE

alter database datafile 2 offline drop

ERROR at line 1:

ORA-00603: ORACLE server session terminated by fatal error

告警日志频繁输出：

Mon Mar 25 13:35:28 2013

Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:

ORA-00376: file 2 cannot be read at this time

ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'

Mon Mar 25 13:35:38 2013

Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:

ORA-00376: file 2 cannot be read at this time

ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'

Mon Mar 25 13:35:38 2013

Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:

ORA-00376: file 2 cannot be read at this time

ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'

Mon Mar 25 13:35:38 2013

Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:

ORA-00376: file 2 cannot be read at this time

ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'

Mon Mar 25 13:35:38 2013

Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_pmon_3756.trc:

ORA-00376: file 2 cannot be read at this time

ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'

试图关闭数据库

sys@L10GR204> shutdown immediate

-- 系统HANG住，无法正常关闭

解决方法：

sys@L10GR204> shutdown abort

sys@L10GR204> startup mount

sys@L10GR204> recover database;

sys@L10GR204> alter database open;

经过上面一系列操作，系统能open了，但是，当我们再次关闭数据库的时候，将得到错误提示

sys@L10GR204> shutdown immediate

ORA-00376: file 2 cannot be read at this time

ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'

这是因为UNDO表空间当前状态是OFFLINE缘故

sys@L10GR204> select FILE#, status from v$datafile;

FILE# STATUS

---------- --------------

1 SYSTEM

2 OFFLINE

3 ONLINE

4 ONLINE

5 ONLINE

在线处理之后恢复正常

sys@L10GR204> alter database datafile 2 online;

Database altered.

sys@L10GR204> select FILE#, status from v$datafile;

FILE# STATUS

---------- --------------

1 SYSTEM

2 ONLINE

3 ONLINE

4 ONLINE

5 ONLINE

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- ------------------------------------------------------- ------------------------------

0 SYSTEM ONLINE

11 _SYSSMU11$ ONLINE

在数据库open又开启归档模式下，进行如下操作：

sys@L10GR204> alter database datafile 2 offline drop;

ERROR:

ORA-03114: not connected to ORACLE

alter database datafile 2 offline drop

ERROR at line 1:

ORA-00603: ORACLE server session terminated by fatal error

可见依然会挂起，但这时候，我们是无法通过非归档模式下解决方法来恢复，当打开数据库时，直接异常终止，如下：

idle> alter database open;

alter database open

ERROR at line 1:

ORA-01092: ORACLE instance terminated. Disconnection forced

告警日志输出：

Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_smon_22399.trc:

ORA-00604: error occurred at recursive SQL level 1

ORA-00376: file 2 cannot be read at this time

ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'

Mon Mar 25 13:50:42 2013

Errors in file /u01/app/oracle/admin/L10GR204/bdump/l10gr204_smon_22399.trc:

ORA-00604: error occurred at recursive SQL level 1

ORA-00376: file 2 cannot be read at this time

ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'

Mon Mar 25 13:50:43 2013

Errors in file /u01/app/oracle/admin/L10GR204/udump/l10gr204_ora_23025.trc:

ORA-00604: error occurred at recursive SQL level 1

ORA-00376: file 2 cannot be read at this time

ORA-01110: data file 2: '/u01/app/oracle/oradata/L10GR204/undotbs01.dbf'

Error 604 happened during db open, shutting down database

USER: terminating instance due to error 604

Instance terminated by USER, pid = 23025

ORA-1092 signalled during: alter database open...

这时候，我们需要手工介恢复：

idle> conn / as sysdba

idle> startup mount

[oracle@khm21 ~]$ rman target /

RMAN> recover database;

idle> alter database open;

上面是在误操作(OPEN状态下)之后恢复的过程，OPEN状态下将默认回滚表空间脱机处理是较危险的操作，请务必重启到MOUNT模式下操作

sys@L10GR204> alter database datafile 2 offline drop;

-- 不带drop选项，会报ORA-01145

sys@L10GR204> alter database datafile 2 offline;

alter database datafile 2 offline

ERROR at line 1:

ORA-01145: offline immediate disallowed unless media recovery enabled

在系统运行于归档模式下的时候，则如下操作：

sys@L10GR204> shutdown immediate

sys@L10GR204> startup mount

sys@L10GR204> alter database datafile 2 offline;

sys@L10GR204> alter database open;

下面验证system回滚段只为系统对象做服务，使用sys用户操作：

sys@L10GR204> create table test(id int);

Table created.

sys@L10GR204> insert into test values(1);

1 row created.

sys@L10GR204> delete from test;

1 row deleted.

sys@L10GR204> rollback;

Rollback complete.

当前只有一个SYSTEM回滚段，我们操作对象时SYSTEM表空间里的时候，SYSTEM回滚段就为此服务，因此正常。

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- ------------------------------ ------------------------------

0 SYSTEM ONLINE

通过普通用户操作：

ning@L10GR204> show user

USER is "ning"

ning@L10GR204> create table test(id int);

create table test(id int)

ERROR at line 1:

ORA-01552: cannot use system rollback segment for non-system tablespace 'ning'

-提示system回滚段无法为其他表空间提供服务。

将默认回滚表空间在线处理

sys@L10GR204> alter database datafile 2 online;

Database altered.

有了默认回滚段，操作恢复正常

ning@L10GR204> create table test(id int);

Table created.

ning@L10GR204> insert into test values(5);

1 row created.

ning@L10GR204> rollback;

Rollback complete.

这时候发现产生了一个回滚段

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- ------------------------------ ------------------------------

0 SYSTEM ONLINE

1 _SYSSMU1$ ONLINE

通过如下方法可以查看_SYSSMU1$回滚段的使用情况：

sys@L10GR204> select OWNER, SEGMENT_NAME, TABLESPACE_NAME, EXTENT_ID, FILE_ID, BLOCK_ID, STATUS from dba_undo_extents where segment_name='_SYSSMU1$' and rownum <=3;

OWNER SEGMENT_NA TABLESPACE EXTENT_ID FILE_ID BLOCK_ID STATUS

------ ---------- ---------- ---------- ---------- ---------- ------------------

SYS _SYSSMU1$ UNDOTBS1 0 2 9 EXPIRED

SYS _SYSSMU1$ UNDOTBS1 1 2 2569 EXPIRED

SYS _SYSSMU1$ UNDOTBS1 2 2 24217 EXPIRED

alter system set undo_tablespace = uuu;

为什么要研究修改undo表空间，只是想告诉大家，当我们执行alter system set undo_tablespace = uuu;的时候都会发生什么。

首先我们观察当前都有哪些回滚段：

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- ------------------------------ ------------------------------

0 SYSTEM ONLINE

1 _SYSSMU1$ ONLINE

2 _SYSSMU2$ ONLINE

3 _SYSSMU3$ ONLINE

4 _SYSSMU4$ ONLINE

5 _SYSSMU5$ ONLINE

6 _SYSSMU6$ ONLINE

7 _SYSSMU7$ ONLINE

8 _SYSSMU8$ ONLINE

9 _SYSSMU9$ ONLINE

10 _SYSSMU10$ ONLINE

11 rows selected.

创建一个新的undo表空间，注意，这里为下面的实验，我故意创建了非常小的回滚表空间。

sys@L10GR204> create undo tablespace undo_ning datafile '/u01/app/oracle/oradata/L10GR204/undo_ning.dbf' size 150K autoextend off;

Tablespace created.

将当前默认undo tablespace设置成新建的

sys@L10GR204> alter system set undo_tablespace=undo_ning;

System altered.

发现原先的回滚段全部消失，表空间过小，一个回滚段都没创建。

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- ------------------------------ ------------------------------

0 SYSTEM ONLINE

添加数据文件

sys@L10GR204> alter tablespace undo_ning add datafile '/u01/app/oracle/oradata/L10GR204/undo_ning02.dbf' size 200k autoextend on;

Tablespace altered.

生成一个回滚段。

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- ------------------------------ ------------------------------

0 SYSTEM ONLINE

12 _SYSSMU12$ ONLINE

结论：在我们在线修改undo tablespace的时候，并不是原先的回滚段移动到新的undo tablespace，而是脱机重新创建。但如果原undo tablespace里还有active事务，则需要等到这些活动事务全部结束之后才能切换。

如何创建一个新的UNDO表空间并使用

上面已经演示过创建新UNDO表空间并使用的过程，这里分享一下如果创建过程遇到ORA-30015如何去解决。

sys@L10GR204> create undo tablespace l_u_s datafile '/u01/app/oracle/oradata/L10GR204/l_u_s01.dbf' size 1M autoextend off;

Tablespace created.

当我想使用新创建的undo表空间的时候，得到错误提示

sys@L10GR204> alter system set undo_tablespace=l_u_s;

alter system set undo_tablespace=l_u_s

ERROR at line 1:

ORA-02097: parameter cannot be modified because specified value is invalid

ORA-30015: previously offlined undo tablespace 'UNDO_ning' is still pending

查一下当前回滚段状态，发现_SYSSMU19$状态为PENDING OFFLINE，估计还有个活跃事务。

sys@L10GR204> col NAME for a20

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- -------------------- ------------------------------

0 SYSTEM ONLINE

19 _SYSSMU19$ PENDING OFFLINE

34 _SYSSMU34$ ONLINE

35 _SYSSMU35$ ONLINE

36 _SYSSMU36$ ONLINE

37 _SYSSMU37$ ONLINE

38 _SYSSMU38$ ONLINE

39 _SYSSMU39$ ONLINE

40 _SYSSMU40$ ONLINE

9 rows selected.

我们通过下面脚本区找当前活跃事务SESSION信息

sys@L10GR204> col USERNAME for a10

sys@L10GR204> col ROLLBACK for a15

sys@L10GR204> SELECT S.SID, S.SERIAL#, S.USERNAME, R.NAME "ROLLBACK"

2 FROM V$SESSION S, V$TRANSACTION T, V$ROLLNAME R

3 WHERE S.TADDR = T.ADDR AND T.XIDUSN = R.USN

4 AND R.name IN (SELECT segment_name FROM dba_rollback_segs WHERE TABLESPACE_NAME ='UNDO_ning')

5 /

SID SERIAL# USERNAME ROLLBACK

---------- ---------- ---------- ---------------

148 34 ning _SYSSMU19$

直接将该SESSION KILL掉

sys@L10GR204> alter system kill session '148,34';

System altered.

这下就可以正常设置新UNDO表空间了

sys@L10GR204> alter system set undo_tablespace='L_U_S';

System altered.

sys@L10GR204> alter system set undo_retention=1;

System altered.

回滚段增减

在上面，我们已经看出，当Oracle数据库启动的时候，默认就会创建多个回滚段，但是，某些系统这些回滚段是不够用的，表现在事务量非常大的系统中。

不过不用担心，Oracle也知道有这种现象，因此如果当前回滚段不足，则会自动创建新回滚段并提供使用。一段时间(通过指定算法)后如果事务量减少，Oracle的SMON将部分回滚段脱机。

我们在上面环境的基础下，通过实验观察这个效果。

这里用到了如下脚本

[oracle@khm21 ~]$ cat showsysevent.sql

column EVENT for a65

select *

from (

select event, total_waits, time_waited

from v$system_event

where wait_class <> 'Idle'

order by 3 desc

)

where EVENT = 'enq: US - contention'

查看当前回滚段情况：

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- ------------------------------ ------------------------------

0 SYSTEM ONLINE

12 _SYSSMU12$ ONLINE

13 _SYSSMU13$ ONLINE

创建一张测试对象表并插入数据

ning@L10GR204> create table khm(id number, domain varchar2(20));

Table created.

ning@L10GR204> begin

2 for i in 1 .. 100 loop

3 insert into khm values(i,'Integrated Cloud Applications and Platform Services');

5 end loop;

6 commit;

7 end;

8 /

PL/SQL procedure successfully completed.

创建一个过程，对khm表进行delete，然后马上回滚：

create or replace procedure khm_delete

begin

for i in 1 .. 100 loop

delete from khm;

rollback;

end loop;

end;

打开100个会话并执行该存储过程

var jobno number;

begin

for l in 1 .. 100 loop

dbms_job.submit(:jobno, 'khm_delete;');

commit;

end loop;

end;

发现回滚段自动增加了几个

sys@L10GR204> select rn.usn, rn.name, rs.status from v$rollname rn, v$rollstat rs where rn.usn = rs.usn;

USN NAME STATUS

---------- ------------------------------ ------------------------------

0 SYSTEM ONLINE

12 _SYSSMU12$ ONLINE

13 _SYSSMU13$ ONLINE

14 _SYSSMU14$ ONLINE

15 _SYSSMU15$ ONLINE

16 _SYSSMU16$ ONLINE

6 rows selected.

下面是US锁等待统计，具体下面讨论

ning@L10GR204> @showsysevent

EVENT TOTAL_WAITS TIME_WAITED

----------------------------------------------------------------- ----------- -----------

enq: US - contention 26 339

了解几个参数

我不会去解释所有和UNDO相关的参数，UNDO_MANAGEMENT、UNDO_RETENTION、UNDO_TABLESPACE这三个参数我也不解释。

使用下面脚本查看和回滚相关的一些参数，了解一下即可，除非特殊情况不建议修改（尤其是隐含参数）

sys@L10GR204> set pagesize 9999

sys@L10GR204> set line 150

sys@L10GR204> col NAME for a30

sys@L10GR204> col VALUE for a20

sys@L10GR204> col DESCRIB for a80

sys@L10GR204> SELECT x.ksppinm NAME, y.ksppstvl VALUE, x.ksppdesc DESCRIB

2 FROM SYS.x$ksppi x, SYS.x$ksppcv y

3 WHERE x.inst_id = USERENV ('Instance')

4 AND y.inst_id = USERENV ('Instance')

5 AND x.indx = y.indx

6 AND x.ksppinm LIKE '%&par%'

sys@L10GR204> /

Enter value for par: undo

old 6: AND x.ksppinm LIKE '%&par%'

new 6: AND x.ksppinm LIKE '%undo%'

NAME VALUE DESCRIB

------------------------------ -------------------- --------------------------------------------------------------------------------

_gc_undo_affinity TRUE if TRUE, enable dynamic undo affinity

_kcl_undo_locks 128 number of locks per undo segment

_kcl_undo_grouping 32 grouping for undo block locks

_gc_undo_affinity_locks TRUE if TRUE, get affinity locks for undo

_gc_dissolve_undo_affinity FALSE if TRUE, dissolve undo affinity after an offline

_gc_initiate_undo_affinity TRUE if TRUE, initiate undo affinity after an online

undo_management AUTO instance runs in SMU mode if TRUE, else in RBU mode

undo_tablespace UNDO_ning use/switch undo tablespace

_collect_undo_stats TRUE Collect Statistics v$undostat

_undo_debug_mode 0 debug flag for undo related operations

_verify_undo_quota FALSE TRUE - verify consistency of undo quota statistics

_in_memory_undo TRUE Make in memory undo for top level transactions

_undo_autotune TRUE enable auto tuning of undo_retention

undo_retention 900 undo retention in seconds

_smon_undo_seg_rescan_limit 10 limit of SMON continous undo segments re-scan

_undo_debug_usage 0 invoke undo usage functions for testing

_optimizer_undo_cost_change 10.2.0.4 optimizer undo cost change

_optimizer_undo_changes FALSE undo changes to query optimizer

18 rows selected.

sys@L10GR204> SELECT x.ksppinm NAME, y.ksppstvl VALUE, x.ksppdesc DESCRIB

2 FROM SYS.x$ksppi x, SYS.x$ksppcv y

3 WHERE x.inst_id = USERENV ('Instance')

4 AND y.inst_id = USERENV ('Instance')

5 AND x.indx = y.indx

6 AND x.ksppinm LIKE '%&par%'

sys@L10GR204> /

Enter value for par: rollback

old 6: AND x.ksppinm LIKE '%&par%'

new 6: AND x.ksppinm LIKE '%rollback%'

NAME VALUE DESCRIB

------------------------------ -------------------- --------------------------------------------------------------------------------

transactions_per_rollback_segm 5 number of active transactions per rollback segment

ent

rollback_segments undo segment list

_rollback_segment_initial 1 starting undo segment number

_rollback_segment_count 0 number of undo segments

_offline_rollback_segments offline undo segment list

_corrupted_rollback_segments corrupted undo segment list

_cleanup_rollback_entries 100 no. of undo entries to apply per transaction cleanup

_rollback_stopat 0 stop at -position to step rollback

fast_start_parallel_rollback LOW max number of parallel recovery slaves that may be used

9 rows selected.

从中介绍几个隐含参数：

_undo_autotune ：从Oracle 10g R2开始默认为TRUE，意思是根据undo表空间大小或者系统负载情况，自动调整undo_retention。这个功能可以在线关闭，方法：alter system set "_undo_autotune"=false;

_smon_undo_seg_rescan_limit：limit of SMON continous undo segments re-scan，上面介绍了Oracle脱机回滚段是根据算法的，算法里面的其中一个条件（如果前面条件都不满足则采用此条件）就是参考_smon_undo_seg_rescan_limit的值。

_rollback_segment_count：指定在线状态的回滚段个数。

构造一个CR块

下面是简单构造一个CR块的示例。

SESSION A:

ning@L10GR204> select distinct sid from v$mystat;

SID

----------

149

当前KHM表id=95的记录如下：

ning@L10GR204> select * from khm where id = 95;

ID DOMAIN

---------- ----------------------------------------

95 Integrated Cloud Applications and Platform Services

打开一个游标，但不执行

ning@L10GR204> variable l refcursor

ning@L10GR204> exec open :l for select * from khm where id = 95;

PL/SQL procedure successfully completed.

SESSION B:

ning@L10GR204> select distinct sid from v$mystat;

SID

----------

143

在另一个会话将id=95的记录值修改并提交

ning@L10GR204> update khm set domain = 'Create your own social network with the best community website builder - NING' where id = 95;

1 row updated.

ning@L10GR204> commit;

Commit complete.

回到打开游标的会话，执行并读取，这时我们发现读取到的是被SESSION B修改之前的记录，这就从CR块里读出来的记录

SESSION A:

ning@L10GR204> print l

ID DOMAIN

---------- ----------------------------------------

95 Integrated Cloud Applications and Platform Services

再次查询，因为事务已经提交，所以读取到的是UPDATE之后的内容

ning@L10GR204> select * from khm where id = 95;

ID DOMAIN

---------- ----------------------------------------

95 Create your own social network with the best community website builder - NING

这其中SCN的作用是非常重要的，其实SESSION A打开游标的时候就有对应的SCN，这个SCN肯定小于更新并提交时SCN，当我们print的时候还是拿当前SCN去查询，经过跟块上的SCN比较之后它去回滚段读取镜像数据。

下面我们深入了解一下构造CR块并读取过程。

拿下面数据作为参考进行实验：

ning@L10GR204> select rowid, t.* from khm t where t.id >95;

ROWID ID DOMAIN

------------------ ---------- ----------------------------------------

AAANIuAAFAAAAAWABf 96 Create your own social network with the best community website builder - NING

AAANIuAAFAAAAAWABg 97 Create your own social network with the best community website builder - NING

AAANIuAAFAAAAAWABh 98 Create your own social network with the best community website builder - NING

AAANIuAAFAAAAAWABi 99 Integrated Cloud Applications and Platform Services

AAANIuAAFAAAAAWABj 100 Integrated Cloud Applications and Platform Services

SESSION A更新两条记录

ning@L10GR204> update khm set domain='bbs.ning.com' where id=96;

1 row updated.

ning@L10GR204> update khm set domain='ffdg' where id=97;

1 row updated.

查看该事务存在信息

ning@L10GR204> @showtra

Enter value for sid: 149

old 13: WHERE SID=&SID

new 13: WHERE SID=149

XIDUSN XIDSLOT XIDSQN UBAFIL UBABLK UBASQN UBAREC STATUS

---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------------

50 12 3 13 44 3 45 ACTIVE

通过USN号去查看当前使用的回滚段

sys@L10GR204> select USN, NAME from v$rollname where USN=50;

USN NAME

---------- -------------------------

50 _SYSSMU50$

如果想dump该回滚段头信息，则如下方式操作：

sys@L10GR204> alter system dump undo header '_SYSSMU50$';

sys@L10GR204> oradebug setmypid

sys@L10GR204> oradebug tracefile_name

/u01/app/oracle/admin/L10GR204/udump/l10gr204_ora_4085.trc

如果想dump该回滚段内容看，则如下方式操作：

sys@L10GR204> alter system dump datafile 13 block 44;

sys@L10GR204> oradebug setmypid

sys@L10GR204> oradebug tracefile_name

/u01/app/oracle/admin/L10GR204/udump/l10gr204_ora_4277.trc

至于DUMP出来的内容如何去看，我将在下面分析，这里不做解释。

下面是简单演示如何通过v$bh查看cr状态xcur状态：

SESSION A:

当前是free状态

sys@L10GR204> select file#, block#, status from v$bh where file#=5 and block#=23;

FILE# BLOCK# STATUS

---------- ---------- --------------

5 23 free

插入一条记录并查看

ning@L10GR204> insert into khm values(1,'恩墨学院 - 名师传道，授业以专');

1 row created.

ning@L10GR204> @showrowid

Enter value for table_name: khm

old 7: from &table_name

new 7: from khm

Enter value for id: 1

old 8: where id=&id

new 8: where id=1

ROWID OBJECT_ID FILE_ID BLOCK_ID NUM ROWIDTOCHAR(ROWID)

------------------ ---------- ---------- ---------- ---------- ------------------------------------

AAANIuAAFAAAAAXAAA 53806 5 23 0 AAANIuAAFAAAAAXAAA

sys@L10GR204> select file#, block#, status from v$bh where file#=5 and block#=23;

FILE# BLOCK# STATUS

---------- ---------- --------------

5 23 free

5 23 xcur

5 23 cr

-- INSERT一条记录产生了一个XCUR块儿以及CR块儿。

SESSION B:

查询该表：

ning@L10GR204> select * from khm;

ID DOMAIN

---------- ----------------------------------------

91 Create your own social network with the best community website builder - NING

92 Create your own social network with the best community website builder - NING

93 Create your own social network with the best community website builder - NING

94 Create your own social network with the best community website builder - NING

95 Create your own social network with the best community website builder - NING

96 Create your own social network with the best community website builder - NING

97 Create your own social network with the best community website builder - NING

98 Create your own social network with the best community website builder - NING

99 Integrated Cloud Applications and Platform Services

100 Integrated Cloud Applications and Platform Services

10 rows selected.

结果又产生了一个CR块儿

sys@L10GR204> select file#, block#, status from v$bh where file#=5 and block#=23;

FILE# BLOCK# STATUS

---------- ---------- --------------

5 23 free

5 23 xcur

5 23 cr

SESSION C: