SGA中Latch 的分类和查询--结合v$latch视图
SGA中Latch 的分类和查询
SGA中Latch分两类：（1）愿意等的LATCH(willing-to_wait)
（2）不愿意等 immediate
CBC 多个BH组成 Cache Buffer Chain (CBC)
例如：CBC Latch的产生次数的查询:
SYS@ bys3>col name for a20SYS@ bys3>select NAME,GETS ,MISSES ,SLEEPS ,IMMEDIATE_GETS, IMMEDIATE_MISSES from v$latch where name like '%cache buffers chains%';
NAME GETS MISSES SLEEPS IMMEDIATE_GETS IMMEDIATE_MISSES
-------------------- ---------- ---------- ---------- -------------- ----------------
cache buffers chains 2006932 0 0 78095 0
v$latch各字段描述：--来自官方文档
GETS Number of times the latch was requested in willing-to-wait mode在愿意等待模式获取锁的次数
MISSES Number of times the latch was requested in willing-to-wait mode and the requestor had to wait愿意等待模式，请求者请求LATCH不成功后等待的次数
SLEEPS Number of times a willing-to-wait latch request resulted in a session sleeping while waiting for the latch在愿意等待模式导致一个会话SLEEP的次数
IMMEDIATE_GETS Number of times a latch was requested in no-wait mode 立即获取latch成功的次数
IMMEDIATE_MISSES Number of times a no-wait latch request did not succeed (that is, missed)立即获取latch不成功的次数
###################################
v$latch视图的查询
SYS@ bys3>
select count(*) from v$latch; --ORACLE版本为11.2.0.4，锁有582种
COUNT(*)
----------
582
SYS@ bys3> select count(*) from v$latch where IMMEDIATE_GETS+IMMEDIATE_MISSES>0; --愿意等待willing-to-wait的latch
COUNT(*)
----------
35
SYS@ bys3> select count(*) from v$latch where IMMEDIATE_GETS+IMMEDIATE_MISSES=0; ---不愿意等待immediate的latch
COUNT(*)
----------
547
################
关于willing-to-wait模式时请求者请求LATCH不成功后等待的次数涉及_spin_count参数：默认值2000，即SPIN2000次仍不能获取latch才sleep.
SYS@ bys3> @?/rdbms/admin/show_para --脚本具体内容见：http://blog.csdn.net/haibusuanyun/article/details/17140197 第二小节。
Enter value for p: spin
old 3: WHERE i.inst_id = USERENV ('Instance') AND CV.inst_id = USERENV ('Instance') AND i.indx = CV.indx AND upper(i.ksppinm) LIKE upper('%&p%') ORDER BY REPLACE (i.ksppinm, '_', '')
new 3: WHERE i.inst_id = USERENV ('Instance') AND CV.inst_id = USERENV ('Instance') AND i.indx = CV.indx AND upper(i.ksppinm) LIKE upper('%spin%') ORDER BY REPLACE (i.ksppinm, '_', '')
P_NAME P_DESCRIPTION P_VALUE ISDEFAULT ISMODIFIED ISADJ
------------------ ------------------------ ---------- --------- ---------- -----
_lm_lms_spin make lms not sleep FALSE TRUE FALSE FALSE
_mutex_spin_count Mutex spin count 255 TRUE FALSE FALSE
_spin_count Amount to spin waiting for a latch 1 TRUE FALSE FALSE
这里因为是虚拟机上，使用的单核心，所以_spin_count为1，双核时即为默认值：2000
show parameter cpu 可查看当前CPU情况。
#############
一个latch对应几个桶buckets呢？默认是32，我这里是16.计算如下：
SYS@ bys3>@?/rdbms/admin/show_paraEnter value for p: db_block_hash
old 3: WHERE i.inst_id = USERENV ('Instance') AND CV.inst_id = USERENV ('Instance') AND i.indx = CV.indx AND upper(i.ksppinm) LIKE upper('%&p%') ORDER BY REPLACE (i.ksppinm, '_', '')
new 3: WHERE i.inst_id = USERENV ('Instance') AND CV.inst_id = USERENV ('Instance') AND i.indx = CV.indx AND upper(i.ksppinm) LIKE upper('%db_block_hash%') ORDER BY REPLACE (i.ksppinm, '_', '')
P_NAME P_DESCRIPTION P_VALUE ISDEFAULT ISMODIFIED ISADJ
---------------------------------------- -------------------------------------------------- ------------------------------ --------- ---------- -----
_db_block_hash_buckets Number of database block hash buckets 16384 TRUE FALSE FALSE
_db_block_hash_latches Number of database block hash latches 1024 TRUE FALSE FALSE
16384/1024=16
###################
一个latch的大小是：
select to_number(b.addr,'xxxxxxxxxxxxxxxxxxxx')-to_number(a.addr,'xxxxxxxxxxxxx')from
(select rownum rn,addr from v$latch_children
where name='cache buffers chains' order by addr) a,
(select rownum rn,addr from v$latch_children
where name='cache buffers chains' order by addr) b
WHERE a.rn=b.rn+1 and rownum<100;
在ORACLE版本为11.2.0.4中为124byte.如果连续多个的latch，在查询中会被合并计算，显示为124的整数倍。