千家信息网

分析Oracle Nologging

发表于:2024-11-25 作者:千家信息网编辑
千家信息网最后更新 2024年11月25日,这篇文章主要介绍"分析Oracle Nologging",在日常操作中,相信很多人在分析Oracle Nologging问题上存在疑惑,小编查阅了各式资料,整理出简单好用的操作方法,希望对大家解答"分
千家信息网最后更新 2024年11月25日分析Oracle Nologging

这篇文章主要介绍"分析Oracle Nologging",在日常操作中,相信很多人在分析Oracle Nologging问题上存在疑惑,小编查阅了各式资料,整理出简单好用的操作方法,希望对大家解答"分析Oracle Nologging"的疑惑有所帮助!接下来,请跟着小编一起来学习吧!

Nologging:

在对大表插入数据的时候,经常会用到nologging选项。Nologging并不是不产生redo,

Nologging+append 只是不会对数据产生redo(但依然有其他的redo,例如数据字典变化产生的redo)。

同理 logging+append下undo也是大大地减少,减少的是数据的undo,是数据本身的undo,就如同redo的减少也一样,是数据本身的redo。

这和数据库是否产生redo和undo是不同的概念,比如空间分配的redo和undo,这就不是数据本身的变化。

Nologging 主要影响:

sql loader 直接路径加载

直接路径的insert(append hint)

create table as select

alter table move

创建和重建索引

在非归档模式下, 对于nologging 和 logging模式,使用 append,都不会对数据生成redo。

在归档模式下,只有将表置于nologging 模式,并且使用append 才不会对数据生成redo.

通过v$mystat视图来显示出当前会话产生的redo来进行显示效果:

select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

测试:

测试前一定要确定是否开始force_logging功能:

如果开启force_logging功能,那么nologging是不生效的。

SYS@prod>select force_logging from v$database;

FOR

---

NO

如果结果是YES,那么进行关闭

Alter database no force logging;

归档模式下的测试:

SYS@prod>archive log list;

Database log mode Archive Mode

Automatic archival Enabled

Archive destination /u01/app/oracle/arch/pridb

Oldest online log sequence 230

Next log sequence to archive 232

Current log sequence 232

Create table 测试:

查看当前会话产生的redo值:

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

Nologging测试:

SYS@prod>create table test nologging as select * from dba_objects;

Table created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 54928

产生redo大小为54928。

SYS@prod>drop table test;

Table dropped.

Logging测试:

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>create table test logging as select * from dba_objects;

Table created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 10262796

产生redo数量为10262796.

结论:用nologging 创建表,不会对数据生成redo,仅对数据字典生成redo.

DML测试:

Insert update delete

Delete:

表logging:

SYS@prod>delete from test;

86978 rows deleted.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 32996412

表nologging:

SYS@prod>delete from test;

86978 rows deleted.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 32991352

Insert 与 update测试就省略了,与delete效果一样。

结论:

对于 INSERT/UPDATE/DELETE的DML 操作, nologging 和 logging 模式没有什么区别。

使用直接路径append对insert测试:

表logging 使用 append插入:

SYS@prod>create table test logging as select * from dba_objects;

Table created.

为了测试效果明显,重新启动一个会话。

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>insert /*+APPEND*/ into test select * from test;

86980 rows created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 10239296

表nologging使用append插入:

SYS@prod>create table test nologging as select * from dba_objects;

Table created.

为了测试效果明显,重新启动一个会话。

SYS@prod>insert /*+APPEND*/ into test select * from test;

86980 rows created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 13884

结论:

对于logging 模式, 使用append hint 在生成redo 上没有什么变化

对于nologging模式,使用append hint 对数据没有生成redo,仅对数据字典生成了redo.

Alter table move测试:

表Logging模式下进行移动:

SYS@prod>create table test logging as select * from dba_objects;

Table created.

重启一个会话

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>alter table test move tablespace example

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 10330784

表nologging模式下进行移动:

SYS@prod>create table test nologging as select * from dba_objects;

Table created.

重启一个会话

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>alter table test move tablespace example;

Table altered.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 71712

结论:表在nologging模式下进行移动,不会记录数据redo,仅会记录数据字典变化的redo。

非归档模式下的测试:

SYS@prod>archive log list;

Database log mode No Archive Mode

Automatic archival Disabled

Archive destination /u01/app/oracle/arch/pridb

Oldest online log sequence 250

Current log sequence 252

1.create table测试

Logging建表:

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>create table test logging as select * from dba_objects;

Table created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 54476

Nologging建表:

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>create table test nologging as select * from dba_objects;

Table created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 53700

结论:在非归档模式下,nologging和logging方式建表差距不大。

2.DML测试:(仅测试insert)

Insert:

表Nologging

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>insert into test select * from test;

86980 rows created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 10153240

表Logging:

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>insert into test select * from test;

86980 rows created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 10162700

结论:

对于 INSERT/UPDATE/DELETE的DML 操作, nologging 和 logging 模式没有什么区别

3.Insert+Append测试:

Logging:

SYS@prod>create table test logging as select * from dba_objects;

Table created.

重启一个会话

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>insert /*+APPEND*/ into test select * from test;

86980 rows created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 13752

Nologging:

SYS@prod>create table test nologging as select * from dba_objects;

Table created.

重启一个会话:

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 0

SYS@prod>insert /*+APPEND*/ into test select * from test;

86980 rows created.

SYS@prod>select a.name,b.value from v$statname a,v$mystat b where a.statistic# = b.statistic# and a.name='redo size';

NAME VALUE

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

redo size 13884

总结: 对于非归档模式,对于nologging 和 logging模式,使用 append,都不会对数据生成redo。

对于归档模式,只有nologging+Append,才不会对数据生成redo,仅对数据字典生成redo

两种模式下的测试结论:

归档模式下:

用nologging 创建表,不会对数据生成redo,仅对数据字典生成redo.

insert/update/delete 的DML 操作,在logging和nologging上没有区别

对于logging 模式, 使用append hint 在生成redo 上没有什么变化

对于nologging模式,使用append hint 对数据没有生成redo,仅对数据字典生成了redo。

非归档模式:

在非归档模式下,create table 在nologging 和 logging 模式差别不大。

对于 INSERT/UPDATE/DELETE的DML 操作, nologging 和 logging 上没有区别。

对于nologging 和 logging模式,使用 append,都不会对数据生成redo。

到此,关于"分析Oracle Nologging"的学习就结束了,希望能够解决大家的疑惑。理论与实践的搭配能更好的帮助大家学习,快去试试吧!若想继续学习更多相关知识,请继续关注网站,小编会继续努力为大家带来更多实用的文章!

0