oracle rba相当重要,与oracle恢复及redo内容,检查点进程工作机制,
有密不可分的关系与联系;理解它的构成,是进一步深入学习oracle
必备路途
摘贴网上基于rba的文章:
http://blog.csdn.net/tianlesoftware/article/details/6700080
http://space.itpub.net/9240380/viewspace-751510
http://www.doc88.com/p-331776598706.html ---非常好,通俗易懂
RBA 由以下三部分组成:
(1)the log file sequence number (4 bytes)
(2)the log file block number (4 bytes)
(3)the byte offset into the block at which the redo record starts (2bytes)
由上可知:
1,rba与checkpoint queue关系很大,即与dbwr关系很大。dbwr每次在dirty buffer要把哪些buffer写入到datefile中
2,rba与lgwr也有很大的。
3,rba与恢复也有关系.on-disk rb即实例恢复到这个rba,库就可以运行了
4,各种不同rba的获取方式
5,rba与controlfile
在上面提到的几种RBA 值可以通过X$BH,和 X$KCCRT 视图进行查看:
1,The low and high RBAs for dirty buffers can be seen in X$BH.(There is also a recovery RBA which is used to record the
progress ofpartial block recovery by PMON.)
2,The incremental checkpoint RBA, the target RBA and theon-disk RBA can all be seen in X$TARGETRBA.
3,The incremental checkpointRBA and the on-disk RBA can also be seen in X$KCCCP.
4,The full thread checkpoint RBA can be seen in X$KCCRT.
先看第1种方式:
1,The low and high RBAs for dirty buffers can be seen in X$BH.(There is also a recovery RBA which is used to record the
progress ofpartial block recovery by PMON.)
我的测试方法:
1,构建测试表并插入一条记录,但不提交
2,查看x$bh确认其rba相关信息
3,dump redo rba相关信息
/*************测试开始***************************/
-----------会话1----
SQL> insert into t_rba values(1)
2 ;
1 row created.
SQL>
SQL>
SQL> select dbms_rowid.rowid_object(rowid),dbms_rowid.rowid_relative_fno(rowid),
dbms_rowid.rowid_block_number(rowid) from t_rba;
DBMS_ROWID.ROWID_OBJECT(ROWID) DBMS_ROWID.ROWID_RELATIVE_FNO(ROWID)
------------------------------ ------------------------------------
DBMS_ROWID.ROWID_BLOCK_NUMBER(ROWID)
------------------------------------
70184 10
276646
---会话2-----------
--------------查看dirty buffer信息
--你发现cr相关的列全是0因为此时不需要构建cr块
SQL> select TS#,FILE#,DBARFIL,DBABLK,CHANGES,OBJ,BA,CR_SCN_BAS,CR_SCN_WRP, CR_XID_USN ,CR_XID_SLT, CR_XID_SQN, CR_UBA_FIL ,CR_UBA_BLK, CR_UBA_SEQ ,CR_UBA_REC
2 from x$bh where bj=(select object_id from dba_objects where wner='SCOTT' and object_name='T_RBA');
TS# FILE# DBARFIL DBABLK CHANGES OBJ BA CR_SCN_BAS CR_SCN_WRP CR_XID_USN CR_XID_SLT CR_XID_SQN CR_UBA_FIL CR_UBA_BLK CR_UBA_SEQ CR_UBA_REC
---------- ---------- ---------- ---------- ---------- ---------- ---------------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ----------
8 10 10 276645 1 70184 000007FF01372000 0 0 0 0 0 0 0 0 0
8 10 10 276642 1 70184 000007FF0135C000 0 0 0 0 0 0 0 0 0
8 10 10 276647 1 70184 000007FF01E06000 0 0 0 0 0 0 0 0 0
8 10 10 276644 1 70184 000007FF0123A000 0 0 0 0 0 0 0 0 0
8 10 10 276641 1 70184 000007FF00A7E000 0 0 0 0 0 0 0 0 0
8 10 10 276646 1 70184 000007FF00D4E000 0 0 0 0 0 0 0 0 0
8 10 10 276643 1 70184 000007FF01DFE000 0 0 0 0 0 0 0 0 0
8 10 10 276640 1 70184 000007FF01DFA000 0 0 0 0 0 0 0 0 0
8 rows selected
--我们别启一个会话,运行select * from t_rba;
--cr列有数据了,而且与v$transaction联系起来了,CR_SCN_BAS即cr块构建自这个scn开始构建
SQL> select TS#,FILE#,DBARFIL,DBABLK,CHANGES,OBJ,BA,CR_SCN_BAS,CR_SCN_WRP, CR_XID_USN ,CR_XID_SLT, CR_XID_SQN, CR_UBA_FIL ,CR_UBA_BLK, CR_UBA_SEQ ,CR_UBA_REC
2 from x$bh where bj=(select object_id from dba_objects where wner='SCOTT' and object_name='T_RBA')
3 ;
TS# FILE# DBARFIL DBABLK CHANGES OBJ BA CR_SCN_BAS CR_SCN_WRP CR_XID_USN CR_XID_SLT CR_XID_SQN CR_UBA_FIL CR_UBA_BLK CR_UBA_SEQ CR_UBA_REC
---------- ---------- ---------- ---------- ---------- ---------- ---------------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ----------
8 10 10 276645 1 70184 000007FF01372000 0 0 0 0 0 0 0 0 0
8 10 10 276642 1 70184 000007FF0135C000 0 0 0 0 0 0 0 0 0
8 10 10 276647 1 70184 000007FF01E06000 0 0 0 0 0 0 0 0 0
8 10 10 276644 1 70184 000007FF0123A000 0 0 0 0 0 0 0 0 0
8 10 10 276646 1 70184 000007FF02F26000 10307069 0 72 15 134 3 7610 416 14
8 10 10 276646 1 70184 000007FF02108000 10307068 0 72 15 134 3 7610 416 14
8 10 10 276646 1 70184 000007FF00D4E000 0 0 0 0 0 0 0 0 0
8 10 10 276643 1 70184 000007FF01DFE000 0 0 0 0 0 0 0 0 0
8 rows selected
/********转储data block***********/
---与上述的ba 000007FF00D4E000相对应,而这个ba 000007FF02108000是为了上述的select构造的cr block,它的cr_scn_bas即下述的LSCN
---HSCN即最新的scn,所以说cr block是根据current block构造的一个数据块
BH (0x000007FF00FE4538) file#: 10 rdba: 0x028438a6 (10/276646) class: 1 ba: 0x000007FF00D4E000
set: 11 pool 3 bsz: 8192 bsi: 0 sflg: 1 pwc: 0,25
dbwrid: 0 obj: 70184 objn: 70184 tsn: 8 afn: 10 hint: f
hash: [0x000007FF0C325E88,0x000007FF023E1C58] lru: [0x000007FF00BEF710,0x000007FF013DACF0]
lru-flags: hot_buffer
obj-flags: object_ckpt_list
ckptq: [0x000007FF0C3F8AA8,0x000007FF0C3F8AA8] fileq: [0x000007FF0C3F8BE8,0x000007FF0C3F8BE8] objq: [0x000007FF0783B2F8,0x000007FF0783B2F8]
st: XCURRENT md: NULL tch: 1
flags: buffer_dirty block_written_once redo_since_read
LRBA: [0x31d.69b8.0] LSCN: [0x0.9d45fc] HSCN: [0x0.9d4661] HSUB: [1]
cr pin refcnt: 0 sh pin refcnt: 0
----------转储undo block-------------------------
*-----------------------------
* Rec #0xe slt: 0x0f objn: 70184(0x00011228) objd: 70184 tblspc: 8(0x00000008)
* Layer: 11 (Row) opc: 1 rci 0x00 --rci可以判断是否还往前回溯构造cr block(适用对一个记录或多记录update多次未提交),此地rci为0,不用再往上cr了
Undo type: Regular undo Begin trans Last buffer split: No
Temp Object: No
Tablespace Undo: No
rdba: 0x00000000Ext idx: 0
flg2: 0
*-----------------------------
uba: 0x00c01dba.01a0.0d ctl max scn: 0x0000.009cbdf1 prv tx scn: 0x0000.009cbdf2
txn start scn: scn: 0x0000.009d36ed logon user: 61
prev brb: 12595255 prev bcl: 0
KDO undo record:
KTB Redo
op: 0x03 ver: 0x01
compat bit: 4 (post-11) padding: 0
op: Z
KDO Op code: DRP row dependencies Disabled
xtype: XA flags: 0x00000000 bdba: 0x028438a6 hdba: 0x028438a2
itli: 1 ispac: 0 maxfr: 4858
tabn: 0 slot: 0(0x0)
---同上--------------
SQL> select CR_SFL ,CR_CLS_BAS, CR_CLS_WRP, LRBA_SEQ , LRBA_BNO , HSCN_BAS , HSCN_WRP, HSUB_SCN
from x$bh where bj=(select object_id from dba_objects where wner='SCOTT' and object_name='T_RBA');
CR_SFL CR_CLS_BAS CR_CLS_WRP LRBA_SEQ LRBA_BNO HSCN_BAS HSCN_WRP HSUB_SCN
---------- ---------- ---------- ---------- ---------- ---------- ---------- ----------
0 0 0 0 0 4294967295 65535 1
0 0 0 0 0 4294967295 65535 1
0 0 0 0 0 4294967295 65535 1
0 0 0 0 0 4294967295 65535 1
0 0 0 0 0 4294967295 65535 1
0 0 0 0 0 4294967295 65535 2
0 0 0 0 0 4294967295 65535 1
0 0 0 0 0 4294967295 65535 1
8 rows selected
---查询事务----------
SQL> select * from v$transaction;
ADDR XIDUSN XIDSLOT XIDSQN UBAFIL UBABLK UBASQN UBAREC STATUS START_TIME START_SCNB START_SCNW START_UEXT START_UBAFIL START_UBABLK START_UBASQN START_UBAREC SES_ADDR FLAG SPACE RECURSIVE NOUNDO PTX NAME PRV_XIDUSN PRV_XIDSLT PRV_XIDSQN PTX_XIDUSN PTX_XIDSLT PTX_XIDSQN DSCN-B DSCN-W USED_UBLK USED_UREC LOG_IO PHY_IO CR_GET CR_CHANGE START_DATE DSCN_BASE DSCN_WRAP START_SCN DEPENDENT_SCN XID PRV_XID PTX_XID
---------------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------------- -------------------- ---------- ---------- ---------- ------------ ------------ ------------ ------------ ---------------- ---------- ----- --------- ------ --- ----------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ----------- ---------- ---------- ---------- ------------- ---------------- ---------------- ----------------
000007FF0B146238 72 15 134 3 7610 416 14 ACTIVE 03/24/13 13:34:35 10303213 0 0 3 7610 416 14 000007FF0C86B270 7683 NO NO NO NO 0 0 0 0 0 0 0 0 1 1 46 0 2 0 2013/3/24 1 0 0 10303213 0 48000F0086000000 0000000000000000 0000000000000000
----查看当前redo
SQL> select * from v$log where status='CURRENT';
GROUP# THREAD# SEQUENCE# BYTES BLOCKSIZE MEMBERS ARCHIVED STATUS FIRST_CHANGE# FIRST_TIME NEXT_CHANGE# NEXT_TIME
---------- ---------- ---------- ---------- ---------- ---------- -------- ---------------- ------------- ----------- ------------ -----------
7 1 797 209715200 512 1 NO CURRENT 10299796 2013/3/24 1 281474976710
SQL> select * from v$logfile where group#=( select group# from v$log where status='CURRENT');
GROUP# STATUS TYPE MEMBER IS_RECOVERY_DEST_FILE
---------- ------- ------- -------------------------------------------------------------------------------- ---------------------
7 ONLINE D:\ORACLE11G_64BIT\ORADATA\ORCL\REDO07.LOG NO
---看看lrba的状态变化
---如果你未在其它会话cr或者dml,看不到lrba
SQL> select ba,obj,lrba_seq,lrba_bno from x$bh where bj=(select object_id from dba_objects where wner='SCOTT' and object_name='T_RBA');
BA OBJ LRBA_SEQ LRBA_BNO
---------------- ---------- ---------- ----------
000007FF01372000 70184 0 0
000007FF0135C000 70184 0 0
000007FF01E06000 70184 0 0
000007FF0123A000 70184 0 0
000007FF02106000 70184 0 0
000007FF01DFE000 70184 0 0
6 rows selected
---如何看到lrba呢,在另一会话dml,再回到上述会话,
SQL> /
BA OBJ LRBA_SEQ LRBA_BNO
---------------- ---------- ---------- ----------
000007FF01372000 70184 0 0
000007FF0135C000 70184 0 0
000007FF01E06000 70184 0 0
000007FF0123A000 70184 0 0
000007FF02CBE000 70184 798 35439 即对应 v$log.sequence#
000007FF02CC4000 70184 0 0
000007FF01DFE000 70184 0 0
7 rows selected
小结:可看到lrba,但未在x$bh看到hrba
我们看看第2种方式:
2,The incremental checkpoint RBA, the target RBA and the on-disk RBA can all be seen in X$TARGETRBA.
最后小结一下:
1,rba由三部分组成:
seqno 序列号 4byte
blockno 块号 4byte
startbyte 起始(或叫偏移量)字节 2个字节
2,lrba即数据块第一次变更时在redo logfile中的rba
引:ckpt-q进程即检查点进程根据lrba的升序维护log buffer
ckpt-q进程会按lbra升序的方式从最小的lrba写出这些dirty buffer
哪么ckpt-q写出的lrba就是checkpoint 的位置.
从这个位置lrba一直应用到redo file的on disk rba,这个过程就是
前滚(实例恢复)
3,说了lrba,自然少不了hrba,它就是对数据块最近一次变更在redo
logfile中的rba
(注:如果hrba大于on-disk rba,那么hrba对应的redo entries还在
log buffer,这样一旦数据库挂掉,哈哈,hrba就不能被恢复了)
所以只能从lrba恢复到on-disk rba,别急:on-disk rba在后面说说)
4,checkpoint rba,这个上面已经提过了,就是ckpt-q进程要写入的脏数据块的
lrba,而这lrba之前的脏数据说明已全部写入disk;
换句话说,ckpt就是每次写入上次checkpoint rba到这次checkpoint rba的数据到磁盘的
还有个问题:ckpt进程每次要从哪个rba开始写入到磁盘呢,其实啊,它工作时会记录
要写的第一个rba,这个rba就是checkpoint rba,这样每次就会有针对性的去写了。其实就是
个限制写入的范围
5,再来说下on-disk rba,它就是lgwr由log buffer写到log file最后面的一个rba;
也就是实例恢复前滚的终点(自lrba到on-disk rba)
6,target rba,是不是有些晕,哈哈;这个其实是checkpoint事件结束的rba
也就是说:从ckpt开始产生检查点事件,自checkpoint rba开始写入到终止写入的哪个
checkpoint rba;
oracle为了提升恢复的性能和时间(实例恢复),dbwr进程会尽力把checkpoint rba设置长一点;
(原因很简单:这样每次写入到磁盘多一些;恢复起来就快得多了)
它与参数:fast_start_mttr_target有关
7,相关视图:
1,增量检查点checkpoint rba及on-disk rba
select * from x$kcccp;
注:cpdrt列是检查点队列中脏块数量
cpods列是on disk rba的scn
cpodt列是on disk rba的时间戳
cphbt列是心跳
2,lrba及hrba
select lrba_seq,lrba_bno from x$bh;
select hxfil,fhrba_seq,fhrba_bno,fhrba_bof from x$kcvfh;--x$kcvfh即v$datafile_header的基表
3,查看target rba
select target_rba_seq,target_rba_bno,target_rba_bof from x$targetrba;
4,完全检查点checkpoint rba
select rtckp_rba_seq,rtckp_rba,rtckp_rba_bof from x$kccrt;
8,再总结下:
1,redo和checkpoint关系很大
2,配置log_checkpoint_to_alert=true;checkpoint会写入到alert
3,数据文件头包括checkpoint_change#,
更重要还包括了checkpoint_change#所在logfile的sequence#,这样恢复就知道要用哪个archive log或log