Oracle checkpoint 详解

版权声明:站在巨人的肩膀上,才可以看的更高更远。 https://blog.csdn.net/u011078141/article/details/88729333

Oracle checkpoint 详解

https://blog.csdn.net/tianlesoftware/article/details/6700085版权声明: https://blog.csdn.net/tianlesoftware/article/detail/6700085


   

一. Oracle Checkpoint 说明
 1.1  Checkpoint
(1)A synchronization event at aspecific point in time
(2)Causes some or all dirty blockimages to be written to the database thereby guaranteeing that blocks dirtiedprior to that point in time get written
(3)Brings administration up to date
(4)Several types of checkpoint exist
  RedoLog Checkpoint 和 SCN关系
http://blog.csdn.net/tianlesoftware/article/details/5251916
  Oracle 实例恢复时 前滚(roll forward) 后滚(rollback) 问题
http://www.cndba.cn/Dave/article/1410
  1.2  the point of Oracle Checkpoints
       The point of Oracle checkpoints is to synchronize all datafiles, some datafiles orsome objects to a point in time for consistency, performance and recoverabilitypurposes.
   1.3 Buffer Cache
       checkpoint与Buffer Cache 的关系很大,有关Buffer cache,之前整理的相关文章如下:
Oracle Buffer Cache 原理
http://www.cndba.cn/Dave/article/1398
  Oracle Buffer Cache 中的Recycle Pool 说明
http://blog.csdn.net/tianlesoftware/article/details/6584110
  Oracle Buffer Cache 中 KeepPool 说明
http://www.cndba.cn/Dave/article/1395
  1.3.1 buffer cache 组成
The buffer cache component structures are:
(1)Buffers
       Eachbuffer may hold an image of one data block at any one time
(2)Buffer headers
       =>Storemetadata about contents of the buffers
       =>Actas cache management structures
(3)Buffer pools
       Collectionof buffers used for the same purpose and managed accordingly
(4)Working set
       =>Allor part of a buffer pool
       =>Assigned to a DBWn process
  1.3.2 Buffer Management
(1) Cached buffers managed by doublylinked lists:
(2) REPL
              =>Bufferscontaining block images being used
(3) REPL-AUX
             =>Buffersready to be used for I/O or CR build
              关于CR 参考:
              CR (consistent read) blockscreate 说明
              http://blog.csdn.net/tianlesoftware/article/details/6529401
(4) WRITE and CKPT-Q
             =>DirtyBuffers requiring I/O
(5) WRITE-AUX
             =>DirtyBuffers with I/O in progress
(6) Touch count is used to decide theinitial insertion location in the REPL chain
(7) AUX lists avoid wasteful scanning
  1.3.3 Redo and the Buffer Cache
(1) Block modification dirties thebuffer containing the block image and generates redo
(2) A buffer becomes dirty at aparticular RBA which is a point in the redo stream
       关于RBA,参考我的blog:
       Oracle RBA(Redo Byte Address) 说明
       http://blog.csdn.net/tianlesoftware/article/details/6700080
(3) Redo written by LGWR makes thecorresponding part of the redo log file “active”
(4) Dirty block images written by DBWnmakes the corresponding part of the redo log file “inactive”
(5) Redo is always written priorto the corresponding block images
(6) Size of active redo in the logfile influences instance and crash recovery time
(7) Trade-off between performance andrecovery time
  1.3.4 Buffer Cache I/O
(1)Servers look for an availablebuffer on REPL-AUX then read a data block into selected buffer
       –Buffer gets moved from REPL-AUX to REPL
       –If block is modified, buffer is added to CKPT-Q
       –Servers move dirty buffers to WRITE during free buffer search
(2) DBWn writes dirty buffer contentsto database
       –Buffer gets moved from WRITE to WRITE-AUX
       –Once block written:
       –Buffer is moved back to REPL-AUX
       –Buffer taken off CKPT-Q
(3) DBWn writes upon request
       –Make free buffers
       –Checkpoint
  1.4 触发checkpoint 的条件
       与checkpoint 触发相关的视图:v$instance_recovery。 关于这个视图的具体说明可以参考官方文档:
       http://download.oracle.com/docs/cd/E11882_01/server.112/e17110/dynviews_2004.htm#REFRN30106
  几个相关的字段如下:
(1)WRITES_MTTR:fast_start_mttr_target
(2)WRITES_LOGFILE_SIZE:90% of the smallest online redo logfile
(3)WRITES_LOG_CHECKPOINT_SETTINGS:log_checkpoint_timeout
(4)WRITES_OTHER_SETTINGS:fast_start_io_target
(5)WRITES_AUTOTUNE:10g self tuning checkpoints
(6) WRITES_FULL_THREAD_CKPT:Manual checkpoints
  CKPT发送CHECKPOINT信号的触发条件有如下几条:
       1.log_checkpoint_timeout时间达到
       2. 当前redo日志已经写够log_checkpoint_internavl*操作系统块大小
       3. redo log switch
       4. alter system checkpoint
       5. alter tablespace XXX begin backup,end backup的時候
       6. alter tablespace , datafileoffline, shutdown immediate, direct read的時候;
         Oracle DBWR,LGWR,CKPT,ARCH 触发条件 总结
       http://www.cndba.cn/Dave/article/1396
         在下面具体的checkpoint 分类时有更详细的说明。


  二. Checkpoint 分类
Checkpoint 可以分为以下几类:
(1) Full Checkpoint
(2) Thread Checkpoint
(3) File Checkpoint
(4) Object “Checkpoint”
(5) Parallel Query Checkpoint
(6) Incremental Checkpoint
(7) Log Switch Checkpoint
         如果想看到具体的checkpoint 类型,可以设置log_checkpoints_to_alert 参数为true。 设置为true之后,checkpoint 发生时会写入alert log里。
如:
SYS@dave2(db2)>alter system set log_checkpoints_to_alert=true scope=both;
System altered.
SYS@dave2(db2)> alter system switchlogfile;
System altered.
  log里的信息如下:
Thu Aug 18 18:46:18 2011
ALTER SYSTEM SETlog_checkpoints_to_alert=TRUE SCOPE=BOTH;
Thu Aug 18 18:48:07 2011
Beginning logswitch checkpoint up to RBA [0xa.2.10], SCN: 2148380730
Thread 1 advanced to log sequence 10
 Current log# 3 seq# 10 mem# 0: /u01/app/oracle/oradata/dave2/redo03.log
    2.1 Full Checkpoint
• Writes block images to the database forall dirty buffers from all instances
• Statistics updated:
– DBWR checkpoints
– DBWR checkpoint buffers written
– DBWR thread checkpoint buffers written
• Caused by:
– Alter system checkpoint [global]
– Alter database close
– Shutdown
• Controlfile and datafile headers areupdated
– CHECKPOINT_CHANGE#
  2.2 Thread Checkpoint
• Writes block images to the database forall dirty buffers from one instance
• Statistics updated:
– DBWR checkpoints
– DBWR checkpoint buffers written
– DBWR thread checkpoint buffers written
• Caused by:
– Alter system checkpoint local
• Controlfile and datafile headers areupdated
– CHECKPOINT_CHANGE#
  2.3 File Checkpoint
• Writes block images to the database forall dirty buffers for all files of a tablespace from all instances
• Statistics updated:
– DBWR tablespace checkpoint bufferswritten
– DBWR checkpoint buffers written
– DBWR checkpoints
• Caused by:
– Alter tablespace XXX offline
– Alter tablespace XXX begin backup
– Alter tablespace XXX read only
• Controlfile and datafile headers areupdated
– CHECKPOINT_CHANGE#
  2.4 Parallel Query Checkpoint
• Writes block images to the database forall dirty buffers belonging to objects accessed by the query from all instances
• Statistics updated:
– DBWR checkpoint buffers written
– DBWR checkpoints
• Caused by:
– Parallel Query
– Parallel Query component of PDML or PDDL
– Mandatory for consistency
  2.5 Object “Checkpoint”
• Writes block images to the database forall dirty buffers belonging to an object from all instances
• Statistics updated:
– DBWR object drop buffers written
– DBWR checkpoints
• Caused by:
– Drop table XXX
– Drop table XXX purge
– Truncate table XXX
• Mandatory for media recovery purposes
  2.6 Incremental Checkpoint
• Writes the contents of “some” dirty buffers to the database from CKPT-Q
• Block images written in SCN order
• Checkpoint RBA updated in SGA
• Statistics updated:
– DBWR checkpoint buffers written
• Controlfile is updated every 3 seconds byCKPT
– Checkpoint progress record
  Definition of “Some”
• Every 3 seconds CKPT calculates thecheckpoint
target RBA based on:
– The most current RBA
– log_checkpoint_timeout
– log_checkpoint_interval
– fast_start_mttr_target
– fast_start_io_target
– 90% of the size of the smallest onlineredo log file
• All buffers dirtied prior to the timecorresponding to the target RBA are written to the database
  2.7 Log Switch Checkpoint
• Writes the contents of “some” dirtybuffers to the database
• Statistics updated:
– DBWR checkpoints
– DBWR checkpoint buffers written
– background checkpoints started
– background checkpoints completed
• Controlfile and datafile headers areupdated
– CHECKPOINT_CHANGE#
  2.8 Checkpoint Administration
• Useful checkpoint administration views:
– V$INSTANCE_RECOVERY
– V$SYSSTAT
– V$DATABASE
– V$INSTANCE_LOG_GROUP
– V$THREAD
– V$DATAFILE
– V$DATAFILE_HEADER

ps://img-blog.csdnimg.cn/20190217105710569.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTEwNzgxNDE=,size_16,color_FFFFFF,t_70《算法导论 第三版英文版》_高清中文版.pdf
https://pan.baidu.com/s/17D1kXU6dLdU0YwHM2cvNMw
《深度学习入门:基于Python的理论与实现》_高清中文版.pdf
https://pan.baidu.com/s/1IeVs35f3gX5r6eAdiRQw4A
《深入浅出数据分析》_高清中文版.pdf
https://pan.baidu.com/s/1GV-QNbtmjZqumDkk8s7z5w
《Python编程:从入门到实践》_高清中文版.pdf
https://pan.baidu.com/s/1GUNSg4mdpeOf1LC_MjXunQ
《Python科学计算》_高清中文版.pdf
https://pan.baidu.com/s/1-hDKhK-7rDDFll_UFpKmpw

猜你喜欢

转载自blog.csdn.net/u011078141/article/details/88729333