一大早就接到开发商的电话,说数据库的CPU使用率为100%,应用相应迟缓。急匆匆的赶到现场发现进行了基本的检查后发现是latch: cache buffers chains 作祟,处理过程还算顺利,当时忘了记录log,这里总结下处理思路,以便下次查看。
故障分析思路
查看等待事件,判断故障起因
SQL>select * from (select sid,event,p1,p2,p3,p1text,WAIT_TIME,SECONDS_IN_WAIT from v$session_wait where wait_class# <> 6
order by wait_time desc) where rownum <=10;
确认为latch: cache buffers chains引起的故障后,查看latch的命中率
SQL>SELECT name, gets, misses, sleeps,
immediate_gets, immediate_misses
FROM v$latch
WHERE name = 'cache buffers chains';
各列名称意义如下
NAME:latch名称
IMMEDIATE_GETS:以Immediate模式latch请求数
IMMEDIATE_MISSES:请求失败数
GETS:以Willing to wait请求模式latch的请求数
MISSES:初次尝试请求不成功次数
SPIN_GETS:第一次尝试失败,但在以后的轮次中成功
SLEEP[x]:成功获取前sleeping次数
WAIT_TIME:花费在等待latch的时间
这里需要注意MISSES/GETS如果在达10%左右,则说明有比较严重的latch争用,也可以通过查询v$latch_children视图查看其他latch信息 ,语句如下
SQL> SELECT *
FROM (SELECT addr, child#, gets, misses, sleeps, immediate_gets igets,
immediate_misses imiss, spin_gets sgets
FROM v$latch_children
WHERE NAME = 'cache buffers chains'
ORDER BY sleeps DESC)
WHERE ROWNUM < 11;
关于latch的统计信息,主要关注以下几部分
misses/gets的比率是多少
获自spinning的misses的百分比是多少
latch请求了多少次
latch休眠了多少次
查看热点对象和访问信息,TCH列表示对象被访问的次数
SQL> SELECT *
FROM ( SELECT addr,
ts#,
file#,
dbarfil,
dbablk,
tch
FROM x$bh
ORDER BY tch DESC)
WHERE ROWNUM < 11;
通过对象的文件号和块号查看具体对象信息
SQL>select owner, segment_name, partition_name, tablespace_name
from dba_extents
where relative_fno = &v_dba_rfile
and &v_dba_block between block_id and block_id + blocks - 1;
也可以通过如下sql查找热点块,主要
SELECT *
FROM (SELECT O.OWNER, O.OBJECT_NAME, O.OBJECT_TYPE, SUM(TCH) TOUCHTIME
FROM X$BH B, DBA_OBJECTS O
WHERE B.OBJ = O.DATA_OBJECT_ID
AND B.TS# > 0
GROUP BY O.OWNER, O.OBJECT_NAME, O.OBJECT_TYPE
ORDER BY SUM(TCH) DESC)
WHERE ROWNUM <= 10;
查看引起latch: cache buffers chains的sql
SQL> select * from (select
count(*),
sql_id,
nvl(o.object_name,ash.current_obj#) objn,
substr(o.object_type,0,10) otype,
3 4 5 6 CURRENT_FILE# fn,
CURRENT_BLOCK# blockn
from v$active_session_history ash
, all_objects o
where event like 'latch: cache buffers chains'
and o.object_id (+)= ash.CURRENT_OBJ#
group by sql_id, current_obj#, current_file#,
current_block#, o.object_name,o.object_type
order by count(*) desc )where rownum <=10;
根据上面得到的sql_id信息查看sql全文
SQL>select sql_fulltext from v$sqlarea where sql_id='&sqlid';
查看SQL的执行计划
SQL>SELECT * FROM table(DBMS_XPLAN.DISPLAY_CURSOR(('&sql_id',0));
在认为sql执行计划不准确的情况也可以通过sql_id查看sql的address和hash_value查看sql的实际执行计划
SQL>SELECT address, hash_value FROM v$sql
WHERE sql_id='&sql_id';
SQL>SELECT operation, options, object_name, cost FROM v$sql_plan
WHERE address = '&addr' AND hash_value = 'hash_v';
当某个会话长时间持有latch时,可以通过联合v$latchholder和v$session视图查看sql信息
SQL>SELECT s.sql_hash_value,s.sql_id,s.address, l.name
FROM V$SESSION s, V$LATCHHOLDER l
WHERE s.sid = l.sid;
故障处理思路
1、根据sql执行计划判断该执行计划是否正确,sql执行过长往往意味着过长时间的持有latch。
2、优化nested loop join,如果有可能使用hash join代替nested loop join。也可以利用对热块索引进行hash分区,或者使用hash簇的方式减缓热块现象。
3、调整表的pctfree值,将数据尽可能的分布到多个块中
4、调整应用
关于热块,可以参阅笔者的如下文章
http://czmmiao.iteye.com/blog/1474472
参考至:https://sites.google.com/site/embtdbo/wait-event-documentation/oracle-latch-cache-buffers-chains
http://space.itpub.net/354732/viewspace-697317
http://www.toadworld.com/KNOWLEDGE/KnowledgeXpertforOracle/tabid/648/TopicID/RCBCL/Default.aspx
http://blog.163.com/wghbeyond@126/blog/static/351661812010619073376/
本文原创,转载请注明出处、作者
如有错误,欢迎指正