Oracle Share Pool内部管理机制
SHARE POOL利用堆（HEAP)的内存管理方式管理，在物理上由多个内存区（EXTENT)组成，内存区又由多个不同大小的CHUNK组成。而CHUNK又有可重用和空闲之分，并且它们分别有LRU LIST、FREE LIST、RESERVED LIST串联起来。
堆管理
Shared Pool是利用堆内存管理方式管理的（KGH:Kernel Generic Heap).从Oracle 9i开始，可以有多个最高级堆（TOP-LEVLE HEAP)，最高级堆可以分成多个副堆，副堆下面还拥有子堆。堆和副对结构基本相同。从物理上讲，一个堆由多个内存区已link list的形式连接组成。一个内存区物理上使用一个Granule，一个内存区由多个chunk组成，所以chunk是heap的最小内存单位。
Chunk的使用情况可由X$KSMSP内部视图查看。每个堆头上则包含了可使用的chunk列表和已使用的chunk列表。通过dump heap命令可以在trace文件中观察heap和extent的关系。
alter system set events 'immediate trace name heapdump level 2';
案例：
SQL> select name,bytes/1024/1024 from v$sgainfo; NAME BYTES/1024/1024 -------------------------------- --------------- Fixed SGA Size 1.27566528 Redo Buffers 5.59765625 Buffer Cache Size 180 Shared Pool Size 104 Large Pool Size 4 Java Pool Size 4 Streams Pool Size 0 Shared IO Pool Size 0 Granule Size 4 Maximum SGA Size 498.875 Startup overhead in Shared Pool 52 NAME BYTES/1024/1024 -------------------------------- --------------- Free SGA Memory Available 200 12 rows selected. [oracle@node1 ~]$ more /u01/app/oracle/diag/rdbms/prod/prod/trace/prod_ora_14983.trc Trace file /u01/app/oracle/diag/rdbms/prod/prod/trace/prod_ora_14983.trc Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production With the Partitioning, OLAP, Data Mining and Real Application Testing options ORACLE_HOME = /u01/app/oracle/product/11.2.0/db_1 System name: Linux Node name: node1 Release: 2.6.18-194.el5 Version: #1 SMP Tue Mar 16 21:52:43 EDT 2010 Machine: i686 Instance name: prod Redo thread mounted by this instance: 1 Oracle process number: 32 Unix process pid: 14983, p_w_picpath: oracle@node1 (TNS V1-V3) *** 2014-07-02 14:58:47.532 *** SESSION ID:(32.758) 2014-07-02 14:58:47.532 *** CLIENT ID:() 2014-07-02 14:58:47.532 *** SERVICE NAME:(SYS$USERS) 2014-07-02 14:58:47.532 *** MODULE NAME:(sqlplus@node1 (TNS V1-V3)) 2014-07-02 14:58:47.532 *** ACTION NAME:() 2014-07-02 14:58:47.532 KGH Latch Directory Information ldir state: 2 last allocated slot: 99 Slot [ 1] Latch: 0x200065ec Index: 1 Flags: 3 State: 2 next: (nil) Slot [ 2] Latch: 0x3e75571c Index: 1 Flags: 3 State: 2 next: (nil) ......(省略） Slot [ 99] Latch: 0x2002616c Index: 1 Flags: 3 State: 2 next: 0x1 ****************************************************** HEAP DUMP heap name="sga heap" desc=0x200010b4 extent sz=0x7ad4 alt=124 het=32767 rec=9 flg=-126 opc=0 parent=(nil) owner=(nil) nex=(nil) xsz=0x0 heap=(nil) fl2=0x60, nex=(nil) ds for latch 1: 0x2002a990 0x2002b5c8 0x2002c200 0x2002ce38 reserved granule count 0 (granule size 4194304) ****************************************************** HEAP DUMP heap name="sga heap(1,0)" desc=0x2002a990 extent sz=0xfc4 alt=124 het=32767 rec=9 flg=-126 opc=0 parent=(nil) owner=(nil) nex=(nil) xsz=0x400000 heap=(nil) fl2=0x20, nex=(nil) latch set 1 of 1 durations enabled for this heap reserved granules for root 0 (granule size 4194304) EXTENT 0 addr=0x3b800000 Chunk 3b800038 sz= 24 R-freeable "reserved stoppe" Chunk 3b800050 sz= 212888 R-free " " Chunk 3b833fe8 sz= 24 R-freeable "reserved stoppe" Chunk 3b834000 sz= 3981312 perm "perm " alo=3290272 EXTENT 1 addr=0x3bc00000 Chunk 3bc00038 sz= 24 R-freeable "reserved stoppe" Chunk 3bc00050 sz= 212888 R-free " " Chunk 3bc33fe8 sz= 24 R-freeable "reserved stoppe" Chunk 3bc34000 sz= 3510272 perm "perm " alo=3510272 Chunk 3bf8d000 sz= 465920 perm "perm " alo=465920 Chunk 3bffec00 sz= 5120 free " " EXTENT 2 addr=0x3c000000 Chunk 3c000038 sz= 24 R-freeable "reserved stoppe" Chunk 3c000050 sz= 212888 R-free " " Chunk 3c033fe8 sz= 24 R-freeable "reserved stoppe" Chunk 3c034000 sz= 3850176 perm "perm " alo=3850176 Chunk 3c3dffc0 sz= 131088 perm "perm " alo=131088 Chunk 3c3fffd0 sz= 48 free " " EXTENT 3 addr=0x3c400000 Chunk 3c400038 sz= 24 R-freeable "reserved stoppe" Chunk 3c400050 sz= 212888 R-free " " Chunk 3c433fe8 sz= 24 R-freeable "reserved stoppe" Chunk 3c434000 sz= 3980368 perm "perm " alo=3980368 Chunk 3c7ffc50 sz= 944 free " " EXTENT 4 addr=0x3c800000 Chunk 3c800038 sz= 24 R-freeable "reserved stoppe" Chunk 3c800050 sz= 212888 R-free " " Chunk 3c833fe8 sz= 24 R-freeable "reserved stoppe" Chunk 3c834000 sz= 3672700 perm "perm " alo=3672700 Chunk 3cbb4a7c sz= 308240 perm "perm " alo=308240 Chunk 3cbffe8c sz= 372 free " "
CHUNK
Chunk是以链条（chain）的方式存在于内存区，每个chunk包含header和body两部分，chunk的状态大体上可以分为free（可立即使用）、recreatable（可再生）、freeable（只有在session或call期间内保存必要的对象状态）、permanent（永久、不可再生）等，其中free和recreatable状态下的chunk可以重复使用。
FREE LIST
Free list主要用于管理空闲的chunk，并且是用bucket管理的。从Oracle 9i开始，一个heap总共255个bucket，bucket所包含的free chunk大小随bucket的编号增加而递增，个bucket下的free chunk已linked list的形式链接。
（1）获得shared pool latch在free list中查找合适大小的空闲chunk。如果在获取shared pool latch时发生了争用，则会出现latch：shared pool等待事件。这时Oracle会一直持有shared pool latch，直到获得所需的内存为止。所在内存碎片化比较严重的shared pool中，进程持有shared pool latch的时间也会相应变长。
（2）如果不存在合适大小的空闲chunk，则在查找到更大的空闲chunk后分割（split）使用，分割后剩下的内存区域则重新登记到free list中。分割内存意味着内存中的碎片开始增多。由于每个cursor所需内存的大小不同，所以shared pool的空闲内存不像buffer cache中的空闲内存一样具有固定大小。
（3）如果检索了free list也没有找到所有合适的空闲chunk，则检索lru list。lru list上的chunk是重建（recreatable）的，而且是当前不使用的（没有处于pin状态）。
（4）如果在lru list上也没有找到合适的chunk，且所请求的内存还没有达到隐含参数_shared_pool_reserved_min_alloc的阈值，则追加分配share pool中剩余的内存空间。
（5）如果以上请求的内存均失败，则出现ORA-4031错误
FREE LISTS: Bucket 0 size=16 Bucket 1 size=20 Bucket 2 size=24 Bucket 3 size=28 Bucket 4 size=32 Bucket 5 size=36 Bucket 6 size=40 Bucket 7 size=44 Bucket 8 size=48 Chunk 3c3fffd0 sz= 48 free " " Chunk 3ebfffd0 sz= 48 free " " Bucket 9 size=52 Bucket 10 size=56 Bucket 11 size=60 Bucket 12 size=64 Bucket 13 size=68 Bucket 14 size=72 Bucket 15 size=76 Bucket 16 size=80 Bucket 17 size=84 Bucket 18 size=88 Chunk 3e7fffa8 sz= 88 free " " Bucket 19 size=92 Bucket 20 size=96 Bucket 21 size=100 Bucket 22 size=104 Bucket 23 size=108 Bucket 24 size=112 Bucket 25 size=116 Bucket 26 size=120 Bucket 27 size=124 ......
LRU LIST
Lru list主要保存着当前未使用而且可以重建的chunk。当会话在free list中找不到空闲的chunk时，在会在lru list中寻找。
UNPINNED RECREATABLE CHUNKS (lru first): Chunk 3b1f4000 sz= 4096 recreate "CCUR^3de2d7f5 " latch=(nil) Chunk 3a649000 sz= 4096 recreate "CCUR^f5a8106a " latch=(nil) Chunk 383b3000 sz= 4096 recreate "CCUR^2b855a04 " latch=(nil) Chunk 383b1000 sz= 4096 recreate "CCUR^f65cc3ed " latch=(nil) Chunk 3ee4c524 sz= 332 recreate "KGLHD " latch=(nil) Chunk 3af76da8 sz= 4096 recreate "SQLA^13df1501 " latch=(nil) Chunk 3a72e000 sz= 4096 recreate "CCUR^13df1501 " latch=(nil) Chunk 39ff96e0 sz= 332 recreate "KGLHD " latch=(nil) Chunk 3a37e740 sz= 4096 recreate "SQLA^7a9f4a60 " latch=(nil) Chunk 38330000 sz= 4096 recreate "CCUR^7a9f4a60 " latch=(nil) Chunk 39f8732c sz= 332 recreate "KGLHD " latch=(nil) Chunk 3ab0dec4 sz= 4096 recreate "SQLA^8005cf3c " latch=(nil) Chunk 3a670000 sz= 4096 recreate "CCUR^8005cf3c " latch=(nil) Chunk 3ed28b1c sz= 1524 recreate "KGLHD " latch=(nil) Chunk 39cb0d20 sz= 332 recreate "KGLHD " latch=(nil) Chunk 3aaeb214 sz= 4096 recreate "SQLA^63c15ff " latch=(nil) Chunk 3a646000 sz= 4096 recreate "CCUR^63c15ff " latch=(nil) Chunk 39ff9a38 sz= 540 recreate "KQR PO " latch=0x3ca28438 Chunk 3ee4c88c sz= 540 recreate "KQR PO " latch=0x3ca28438 Chunk 3ee4caa8 sz= 540 recreate "KQR PO " latch=0x3ca28438 Chunk 39fabd2c sz= 540 recreate "KQR PO " latch=0x3ca28438 Chunk 39fabf48 sz= 540 recreate "KQR PO " latch=0x3ca28438 Chunk 3ee481f8 sz= 540 recreate "KQR PO " latch=0x3ca28438 Chunk 3ee48414 sz= 540 recreate "KQR PO " latch=0x3ca28438 Chunk 3ee48630 sz= 540 recreate "KQR PO " latch=0x3ca28438 Chunk 3a1a1228 sz= 4096 recreate "KGLS^9b4819a8 " latch=(nil) Chunk 39b4a504 sz= 4096 recreate "KGLS^6ccccfe " latch=(nil) Chunk 3abf0fb4 sz= 4096 recreate "KGLS^8db54da " latch=(nil) Chunk 3aa65508 sz= 4096 recreate "KGLS^e2d7481e " latch=(nil) Chunk 3a2839d0 sz= 4096 recreate "KGLS^8dd845ff " latch=(nil) Chunk 3abd4f78 sz= 4096 recreate "KGLS^f79d229a " latch=(nil)
RESERVED FREE LIST
Oracle 设置了保留内存区域，该区域用 Reserved free list管理。其大小由参数shared_pool_reserved_size决定（最小为5000字节，最大不能超过shared pool的50%）。
RESERVED FREE LISTS: Reserved bucket 0 size=16 Chunk 3dfffc14 sz= 980 R-free " " Chunk 3dbffbe0 sz= 1032 R-free " " Reserved bucket 1 size=4400 Reserved bucket 2 size=8204 Reserved bucket 3 size=8460 Reserved bucket 4 size=8464 Reserved bucket 5 size=8468 Reserved bucket 6 size=8472 Reserved bucket 7 size=9296 Reserved bucket 8 size=9300 Reserved bucket 9 size=12320 Reserved bucket 10 size=12324 Reserved bucket 11 size=16396 Reserved bucket 12 size=32780 Reserved bucket 13 size=65548 Chunk 3b800050 sz= 212888 R-free " " Chunk 3bc00050 sz= 212888 R-free " " Chunk 3c000050 sz= 212888 R-free " " Chunk 3c400050 sz= 212888 R-free " " Chunk 3c800050 sz= 212888 R-free " " Chunk 3cc00050 sz= 212888 R-free " " Chunk 3d3cb8b8 sz= 214832 R-free " " Chunk 3d400050 sz= 212888 R-free " " Chunk 3e000050 sz= 212888 R-free " " Chunk 3e400050 sz= 212888 R-free " " Chunk 3e800050 sz= 212888 R-free " " Chunk 3f000050 sz= 212888 R-free " " Reserved bucket 14 size=1990630 Total reserved free space = 2558612 UNPINNED RECREATABLE CHUNKS (lru first): Chunk 3b1f4000 sz= 4096 recreate "CCUR^3de2d7f5 " latch=(nil) Chunk 3a649000 sz= 4096 recreate "CCUR^f5a8106a " latch=(nil) ......
Oracle对进入reserved free list的对象大小有限制，即只有大于_shared_pool_reserved_min_alloc隐含参数阈值（默认值4400）的cursor才能进入到reserved free list。
SHARED POOL的SUB POOL技术
从Oracle 9i开始，shared pool可以分为多个sub pool，其数量受一下几个因素影响：
1、系统的CPU数量，默认情况下，在Oracle中每4个CPU分配一个sub pool，最多不能超过7个。
2、共享池的大小。sub pool的最小容量随着Oracle版本不同而不同。
9i ----- 128M（Minimum subpool size）
10g<10.2.0.3 ---- 256M
10.2.0.3 and higher ---512M
3、隐含参数_kghdisdx_count值
每个sub pool拥有独立的free list 、lru list和shared pool latch。从这个角度来讲，当系统拥有足够的内存和CPU时，将shared pool分为多个sub pool时能有效的减少shared pool latch的争用。
查看sub pool的数量：
kghlushrpool： =1: shared pool subpools =0: java pool SQL> select count(kghluidx) num_pools 2 from x$kghlu 3 where kghlushrpool=1 4 / NUM_POOLS ---------- 1 SQL> show parameter cpu NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ cpu_count integer 1
查看sub pool latch信息：
SQL> col name for a40 SQL> set linesize 120 SQL> r 1 select addr,name,gets,misses,spin_gets 2 from v$latch_children 3* where name ='shared pool' ADDR NAME GETS MISSES SPIN_GETS -------- ---------------------------------------- ---------- ---------- ---------- 200A80FC shared pool 21 0 0 200A8098 shared pool 21 0 0 200A8034 shared pool 21 0 0 200A7FD0 shared pool 21 0 0 200A7F6C shared pool 21 0 0 200A7F08 shared pool 21 0 0 200A7EA4 shared pool 493378 7 0 7 rows selected.
查看sub pool 内存分配：
X$KSMSS([K]ernal [S]torage [M]emory Management [S]GA [S]atistics (lengths of SGA objects)
SQL> select
2 'shared pool(' || nvl(decode(to_char(ksmdsidx),'0','0-unused',ksmdsidx),'Total')
3 || '):' subpool,
4 sum(ksmsslen) bytes,round(sum (ksmsslen)/1048576 ,2 ) mb
5 from x$ksmss
6 where ksmsslen >0
7 group by rollup (ksmdsidx) order by subpool asc;
SUBPOOL BYTES MB
------------------------------------------------------ ---------- ----------
shared pool(1): 121639892 116
shared pool(Total): 121639892 116
SQL>
Oracle 硬解析
1、获得shared pool latch ，从free list的bucket 中查找合适大小的free chunk。如果free list中的bucket list过长或者shared pool碎片化严重，那么在多个进程同时请求分配内存时，则会发生shared pool latch的争用。
2、如果不存在大小合适的free chunk，则分割较大的free chunk，分割后的free chunk重新挂载到适当大小的bucket下。如果不存在free chunk，则检索lru list。若在lru list中也不能获得合适大小的bucket，则从shared pool的剩余空闲内存中分配。如果cursor大小大于_shared_pool_reserved_min_alloc隐含参数设定的阀值，那么在reserved list中寻找free chunk。若果以上过程均失败，则出现ORA-4031错误。
3、若找到合适大小的chunk，则对cursor相应的handle（library cache handle）以exclusive 模式获得library cache lock，并创建LCO信息。在创建LCO信息后，library cache lock变换为null模式，然后以exclusive模式获得library cache pin，并创建执行计划等信息。硬解析成功后Oracle增加parse count（hard）统计值。
4、对sql cursor已shared模式获得library cache lock和library cache pin，并执行sql，这个阶段称之为执行阶段。
5、sql cursor执行结束后进入fetch阶段。在fetch阶段，sql cursor将library cache pin变为null模式，并释放library cache pin。
软软解析
由于在软解析过程中需要获得library cache latch，所以在高并发软解析的系统中，依然会出现与latch：library cache相关的等待事件，从而导致性能缓慢。
软软解析（Tom Kety）核心原理是通过设置session_cache_cursors参数将某个会话中常用的sql放入UGA的会话缓存区中，当会话发起相同的sql时，可以快速的从UGA取得cursor的信息，从而减少共享池的争用。当一个cursor被解析3次以上（包括3次）就会被放入到UG会话缓存区中。
案例：shared pool latch 争用
案例1：
业务运行前：
17:07:30 SYS@ prod>select name,GETS,MISSES from v$latch where upper(name) like '%LIBRARY%' OR upper(name) like '%SHARE%';
NAME GETS MISSES
---------------------------------------------------------------- ---------- ----------
test shared non-parent l0 0 0
ksxp shared latch 0 0
kcfis stats shared latch 0 0
shared pool 126676 61
library cache load lock 0 0
shared pool simulator 6576 0
shared pool sim alloc 45 0
Shared B-Tree 302 0
shared server configuration 6 0
shared server info 1 0
运行业务：
17:08:34 SCOTT@ prod>begin
17:08:38 2 for i in 1..100000 loop
17:08:52 3 execute immediate 'insert into t1 values ('||i||')';
17:09:18 4 end loop;
17:09:26 5 end;
17:09:27 6 /
PL/SQL procedure successfully completed.
业务运行后：
17:11:05 SYS@ prod>select name,GETS,MISSES from v$latch where upper(name) like '%LIBRARY%' OR upper(name) like '%SHARE%'
NAME GETS MISSES
---------------------------------------------------------------- ---------- ----------
test shared non-parent l0 0 0
ksxp shared latch 0 0
kcfis stats shared latch 0 0
shared pool 4526672 214
library cache load lock 0 0
shared pool simulator 1086437 0
shared pool sim alloc 2048 0
Shared B-Tree 316 0
shared server configuration 6 0
shared server info 1 0
10 rows selected.
17:15:42 SYS@ prod>select sid,event,WAIT_TIME,state from v$session_wait where sid=42
SID EVENT WAIT_TIME STATE
---------- ---------------------------------------------------------------- ---------- -------------------
42 latch: shared pool -1 WAITED SHORT TIME
Elapsed: 00:00:00.08
案例2：
业务运行前：
17:18:35 SYS@ prod>select sid,EVENT,TOTAL_WAITS,AVERAGE_WAIT from v$session_event where sid in (42,46);
SID EVENT TOTAL_WAITS AVERAGE_WAIT
---------- ---------------------------------------------------------------- ----------- ------------
42 Disk file operations I/O 4 .03
42 log file switch (private strand flush incomplete) 1 10.03
42 log file sync 4 1.76
42 db file sequential read 385 .23
42 latch: row cache objects 5 .44
42 latch: shared pool 194 .25
42 SQL*Net message to client 24 0
42 SQL*Net message from client 23 5318.9
42 SQL*Net break/reset to client 2 .08
42 events in waitclass Other 1 0
46 Disk file operations I/O 1 .03
46 db file sequential read 33 .02
46 SQL*Net message to client 13 0
46 SQL*Net message from client 12 79.9
14 rows selected.
运行业务：
17:16:39 SYS@ prod>select sid ,username from v$session where username is not null;
SID USERNAME
---------- ------------------------------
1 SYS
42 SCOTT
46 HR
17:17:22 SCOTT@ prod>begin
17:20:46 2 for i in 1..100000 loop
17:20:52 3 execute immediate 'insert into t1 values ('||i||')';
17:20:58 4 end loop;
17:21:02 5 end;
17:21:05 6 /
PL/SQL procedure successfully completed.
17:17:42 HR@ prod>begin
17:21:16 2 for i in 1..100000 loop
17:21:24 3 execute immediate 'insert into scott.t1 values ('||i||')';
17:21:49 4 end loop;
17:21:51 5 end;
17:21:52 6 /
PL/SQL procedure successfully completed.
业务运行后：
17:22:32 SYS@ prod>select sid,EVENT,TOTAL_WAITS,AVERAGE_WAIT from v$session_event where sid in (42,46);
SID EVENT TOTAL_WAITS AVERAGE_WAIT
---------- ---------------------------------------------------------------- ----------- ------------
42 Disk file operations I/O 4 .03
42 latch: cache buffers chains 16 .18
42 buffer busy waits 2 .15
42 log file switch (private strand flush incomplete) 1 10.03
42 log file sync 4 1.76
42 db file sequential read 413 .21
42 latch: row cache objects 58 .13
42 latch: shared pool 1008 .19
42 library cache: mutex X 123 .33
42 SQL*Net message to client 24 0
42 SQL*Net message from client 24 6044.43
42 SQL*Net break/reset to client 2 .08
42 events in waitclass Other 87 .09
46 Disk file operations I/O 3 .03
46 latch: cache buffers chains 13 .21
46 buffer busy waits 1 .35
46 latch: redo copy 1 1.26
SID EVENT TOTAL_WAITS AVERAGE_WAIT
---------- ---------------------------------------------------------------- ----------- ------------
46 db file sequential read 38 .02
46 enq: HW - contention 1 .01
46 latch: row cache objects 58 .14
46 row cache lock 1 .08
46 latch: shared pool 666 .17
46 library cache: mutex X 99 .29
46 SQL*Net message to client 13 0
46 SQL*Net message from client 13 2010.63
46 events in waitclass Other 68 .14
26 rows selected.
Elapsed: 00:00:00.37
17:22:42 SYS@ prod>
17:22:02 SYS@ prod>select sid,event,WAIT_TIME,state from v$session_wait where sid=42
17:22:25 2 or sid=46;
SID EVENT WAIT_TIME STATE
---------- ---------------------------------------------------------------- ---------- -------------------
42 latch: shared pool -1 WAITED SHORT TIME
46 latch: shared pool -1 WAITED SHORT TIME
----以上部分内容整理自《Oracle DBA实战攻略》 周亮著，感谢作者的辛勤付出！
转载于:https://blog.51cto.com/tiany/1433452
3833
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
