物理读之LRU（最近最少被使用）的深入解析

一组 LRU 链表包括 LRU 主链， LRU 辅助链， LRUW 主链， LRUW 辅助链，称为一个 WorkSet( 工作组 ) 如下图： 650) this.width=650;" src="http://www.68idc.cn/help/uploads/allimg/151214/10060LM3-0.jpg" title="111.png" alt="wKioL1PXLA_RrQwAAAE_O1bng

一组LRU链表包括LRU主链，LRU辅助链，LRUW主链，LRUW辅助链，称为一个WorkSet(工作组)如下图：

sys@ZMDB> selectCNUM_SET,CNUM_REPL,ANUM_REPL,CNUM_WRITE,ANUM_WRITE from x$kcbwds whereCNUM_SET>0;

CNUM_SET CNUM_REPL ANUM_REPL CNUM_WRITE ANUM_WRITE

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

15221 15221 3796 0 0

15221 15221 3783 0 0

CNUM_SET:工作组总的buffer总数量

CNUM_REPL:工作组中LRU的buffer总数量（主LRU+辅LRU）

ANUM_REPL:工作组中辅LRU总BUFFER的数量

通过隐含参数查到BUFFER的总的个数是30442，正好与上面的CNUM_SET=15221+15221

sys@ZMDB>@?/rdbms/admin/show_para

Enter value for p: _db_block_buffers

old 12: AND upper(i.ksppinm) LIKEupper('%&p%')

new 12: AND upper(i.ksppinm) LIKEupper('%_db_block_buffers%')

P_NAME P_DESCRIPTION P_VALUE ISDEFAULT ISMODIFIEDISADJ

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

_db_block_buffers Number of database blocks cached inmemory: hidden 30442 TRUE FALSE FALSE

Parameter

我们用以下语句查下数据库中buffer所在LRU的状态

sys@ZMDB> select lru_flag,count(*) from x$bh group by lru_flag;

LRU_FLAG COUNT(*)

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

6 208

2 10

4 7122

8 15199

0 7646

我们对LRU_FLAG=6，2，4，8，0等做出解释，举个例子，对于6是什么含义呢？

首先要在x$bh中找到lru_flag=6的任意的一个BUFFER

sys@ZMDB> select LRU_FLAG,LOWER(BA)from x$bh where lru_flag=6 andrownum=1;

LRU_FLAG LOWER(BA)

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

6 0000000081dae000

DUMP buffer_cache中BH信息，如下命令：

sys@ZMDB>alter session set events'immediate trace name buffers level 1';

Session altered.

ys@ZMDB> col value for a85

sys@ZMDB> select * from v$diag_info where name='Default TraceFile';

INST_ID NAME VALUE

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

1 Default Trace File /u01/app/oracle/diag/rdbms/zmdb/zmdb/trace/zmdb_ora_13235.trc

通过BA=81dae000搜索trace文件，

/u01/app/oracle/diag/rdbms/zmdb/zmdb/trace/zmdb_ora_13235.trc

得到如下内容：

BH (0x81fe7e38) file#: 1 rdba: 0x0040ace1 (1/44257) class: 1 ba:0x81dae000

set: 6 pool: 3 bsz: 8192 bsi: 0 sflg: 1 pwc:0,25

dbwrid: 0 obj: 421 objn: 423 tsn: 0 afn: 1hint: f

hash: [0x9ef9d710,0x853f8da8] lru:[0x81fe7df0,0x81fe8050]

lru-flags: moved_to_tail on_auxiliary_list

ckptq: [NULL] fileq: [NULL] objq: [NULL]objaq: [NULL]

st: CR md: NULL fpin: 'kdswh06: kdscgr' tch:1

cr: [scn: 0x0.80350f4d],[xid: 0x0.0.0],[uba:0x0.0.0],[cls: 0x0.80350f4d],[sfl: 0x0],[lc: 0x0.8034c532]

flags: block_written_once redo_since_read

LRU_FLAG=6的意思是lru-flags: moved_to_tail on_auxiliary_list，就是向LRU的辅助链表的尾部移动，这有可能是SMON从LRU的主链表上的非脏块、TCH<=1并且状态是非PIN的BUFFER被挂接到LRU辅助链表的尾部。

根据以上的方法同理可以解释出LRU_FLAG的含义：

LRU_FLAG

0==>LRU-主链冷端的头部，这个比较特殊他在DUMP没有显示LRU_FLAG

2==>LRU-主链冷端的尾部，lru-flags:moved_to_tail

4==>LRU-辅助链,lru-flags:on_auxiliary_list

6==>LRU-辅助链的尾部，lru-flags:moved_to_tail on_auxiliary_list

8==>LUR-主链热端，lru-flags:hot_buffer

当发生物理读时，Oracle会从LRU辅助链表找空闲的BUFFER，然后把LRU辅助的链上的BUFFER挂接到LRU主链的冷端头，实验如下：

首先要保证有LRU辅助链上的BUFFER,即有LRU_FLAG=6或LRU_FLAG=4,如果数据库刚刚启来，可能没有LRU_FLAG=6、LRU_FLAG=4，那需要做大量的物理读操作，才会有LRU_FLAG=6或LRU_FLAG=4

sys@ZMDB> alter system flush buffer_cache;

System altered.

sys@ZMDB> selectlru_flag,count(*) from x$bh group by lru_flag;

LRU_FLAG COUNT(*)

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

6 208

4 30009

0 2

第一次DUMP整个BUFFER CACHE:

sys@ZMDB> alter session set events'immediate trace name bufferslevel 1';

/u01/app/oracle/diag/rdbms/zmdb/zmdb/trace/zmdb_ora_13480.trc

发生物理读

gyj@ZMDB> conn gyj/gyj

Connected.

gyj@ZMDB> set autot on;

gyj@ZMDB> select id,name,dbms_rowid.rowid_relative_fno(rowid) file#,dbms_rowid.rowid_block_number(rowid)block# from gyj_t1 where id=1;

ID NAME FILE# BLOCK#

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

1 gyj1 7 139

Execution Plan

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

Plan hash value: 59758809

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

| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |

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

| 0 | SELECT STATEMENT | | 1 | 14 | 68 (0)| 00:00:01 |

|* 1 | TABLE ACCESS FULL| GYJ_T1 | 1| 14 | 68 (0)| 00:00:01 |

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

Predicate Information(identified by operation id):

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

1 - filter("ID"=1)

Statistics

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

1 recursive calls

1 db block gets

254 consistent gets

248 physical reads

0 redo size

733 bytes sent via SQL*Net to client

523 bytes received via SQL*Net from client

2 SQL*Net roundtrips to/from client

0 sorts (memory)

0 sorts (disk)

1 rows processed

sys@ZMDB> selectLRU_FLAG,lower(BA),TCH from x$bh where file#=7 and dbablk=139;

LRU_FLAG LOWER(BA) TCH

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

0 000000007d1b2000 1

4 0000000078558000 0

4 0000000085f68000 0

物理读完成后，再次dump整个buffer cache,

sys@ZMDB>alter session set events'immediate trace name buffers level 1';

/u01/app/oracle/diag/rdbms/zmdb/zmdb/trace/zmdb_ora_13511.trc

拿BA=7d1b2000,搜索第一次DUMP的trace文件

/u01/app/oracle/diag/rdbms/zmdb/zmdb/trace/zmdb_ora_13480.trc

BH (0x7d3e8098) file#: 3 rdba:0x00c0586b (3/22635) class: 34 ba: 0x7d1b2000

set: 5 pool: 3 bsz: 8192 bsi: 0 sflg: 1 pwc:0,25

dbwrid:0 obj: -1 objn: 0 tsn: 2 afn: 3 hint: f

hash: [0x9efa7570,0x9efa7570] lru:[0x7f7f5d30,0x7d3e8050]

lru-flags: on_auxiliary_list

ckptq: [NULL] fileq: [NULL] objq: [NULL]objaq: [NULL]

st: FREE md: NULL fpin: 'ktuwh03: ktugnb'tch: 0 lfb: 33

flags:

拿BA=7d1b2000,搜索第二次DUMP的trace文件

/u01/app/oracle/diag/rdbms/zmdb/zmdb/trace/zmdb_ora_13511.trc

BH (0x7d3e8098) file#: 7 rdba:0x01c0008b (7/139) class: 1 ba: 0x7d1b2000

set: 5 pool: 3 bsz: 8192 bsi: 0 sflg: 1 pwc:0,25

dbwrid: 0 obj: 22919 objn: 19567 tsn: 7 afn:7 hint: f

hash: [0x787e4bd8,0x9e4cda50] lru:[0x7f7f5d30,0x7d3e8050]

ckptq: [NULL] fileq: [NULL] objq:[0x9a88e518,0x7d3e8078] objaq: [0x9a88e508,0x7d3e8088]

st: XCURRENT md: NULL fpin: 'kdswh11:kdst_fetch' tch: 1

flags: only_sequential_access

LRBA: [0x0.0.0] LSCN: [0x0.0] HSCN:[0xffff.ffffffff] HSUB: [65535]

从上面的两个trace可以得出结论ba: 0x7d1b2000

从lru-flags:on_auxiliary_list(LRU_FLAG=4)到LRU-主链冷端的头部，这个比较特殊在DUMP没有显示LRU_FLAG(LRU_FLAG=0)

观察LRUTCH>=2时冷端移到热端

1、BUFFER手动设为100M

ALTER SYSTEM SETmemory_max_target=0 scope=spfile;

ALTER SYSTEM SET memory_target=0;

alter system set sga_target=0;

create table gyj1_t80 (idint,name char(2000));

create table gyj2_t80 (idint,name char(2000));

begin

for i in 1 .. 30000

loop

insert into gyj1_t80 values(i,'gyj'||i);

commit;

end loop;

end;

/

SQL> SQL> selectbytes/1024/1024||'M' from dba_segments where segment_name='GYJ1_T80' andowner='GYJ';

BYTES/1024/1024||'M'

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

80M

begin

for i in 1 .. 30000

loop

insert into gyj2_t80 values(i,'gyj'||i);

commit;

end loop;

end;

/

create index idx_gyj1_t80m ongyj1_t80(id);

create index idx_gyj2_t80m ongyj2_t80(id);

SQL> show user;

USER is "GYJ"

SQL> conn / as sysdba

Connected.

SQL> shutdown immediate;

Database closed.

Database dismounted.

ORACLE instance shut down.

SQL> startup

ORACLE instance started.

第一次dump

SQL> alter session set events'immediate trace name buffers level1';

Session altered.

SQL> select * fromv$diag_info where name='Default Trace File';

INST_ID NAME

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

VALUE

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

1 Default Trace File

/u01/app/oracle/diag/rdbms/jfdb/jfdb/trace/jfdb_ora_7210.trc

发生一个物理读走索引

set autot on

selectid,name,dbms_rowid.rowid_relative_fno(rowid)file#,dbms_rowid.rowid_block_number(rowid) block# from gyj1_t80 where id=1;

SQL> selectid,name,dbms_rowid.rowid_relative_fno(rowid)file#,dbms_rowid.rowid_block_number(rowid) block# from gyj1_t80 where id=1;

ID NAME FILE# BLOCK#

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

1 gyj1 5 581

select LRU_FLAG,lower(BA),TCHfrom x$bh where file#=5 and dbablk=581;

SQL> select LRU_FLAG,lower(BA),TCH,decode(state,0,'free',1,'xcur',2,'scur'

2 ,3,'cr', 4,'read',5,'mrec',6,'irec',7,'write',8,'pi',9,'memory',10,'mwrite',11,

3 'donated', 12,'protected', 13,'securefile', 14,'siop',15,'recckpt', 16, 'flashf

4 ree', 17, 'flashcur', 18,'flashna') from x$bh where file#=5 anddbablk=581;

LRU_FLAG LOWER(BA) TCH DECODE(STA

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

0 000000009fca8000 1 xcur

SQL> selectLRU_FLAG,lower(BA),TCH from x$bh where file#=5 and dbablk=581;

LRU_FLAG LOWER(BA) TCH

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

0 000000009fca8000 5

SQL> set autot traceonly;

SQL> select /*+ index(G) */ count(name) fromgyj1_t80 G where id<=8000;

SQL> selectLRU_FLAG,lower(BA),TCH from x$bh where file#=5 and dbablk=581;

LRU_FLAG LOWER(BA) TCH

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

0 000000009fca8000 6

再次发生物理读，此时LRU_FLAG=0变为8，同时TCH=8重置为0

SQL>select LRU_FLAG,lower(BA),TCH from x$bh where file#=5 and dbablk=581;

LRU_FLAG LOWER(BA) TCH

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

0000000009fca8000 8

SQL> select LRU_FLAG,lower(BA),TCH from x$bh where file#=5 anddbablk=581;

LRU_FLAG LOWER(BA) TCH

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

8000000009fca8000 0

BH (0x9ffe02a8) file#: 5 rdba: 0x01400245 (5/581) class: 1 ba:0x9fca8000

set: 5 pool: 3 bsz: 8192bsi: 0 sflg: 2 pwc: 15,19

dbwrid: 0 obj: 13537 objn:13537 tsn: 5 afn: 5 hint: f

hash:[0xb6a86de0,0xb6a86de0] lru: [0x9ffe0260,0x9ffe9a60]

lru-flags: hot_buffer

ckptq: [NULL] fileq: [NULL]objq: [0x9ffe0618,0x9ffe0028] objaq: [0x9ffe0628,0x9ffe0038]

st: XCURRENT md: NULL fpin:'kdswh05: kdsgrp' tch: 0

flags:

LRBA: [0x0.0.0] LSCN:[0x0.0] HSCN: [0xffff.ffffffff] HSUB: [65535]

当TCH=0时，再发生大量物理读，地址为9fca8000的BUFFER就被重用了，彻底从BUFFER消失

SQL> selectLRU_FLAG,lower(BA),TCH from x$bh where file#=5 and dbablk=581;

LRU_FLAG LOWER(BA) TCH

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

8 000000009fca8000 0

SQL> select LRU_FLAG,lower(BA),TCH from x$bh wherefile#=5 and dbablk=581;

no rows selected

通过实验，我们更清楚地了解到物理读LRU的基本流程，可以进一步理解物理读内部的LRU算法。