监控长时间运行的查询(监控数据库性能的SQL )

查询v$session_longops动态性能视图可以查看运行时间大于6秒的查询，如果想让运行时间大于6秒的查询被数据库引擎收集，必须满足以下条件：1 参数timed_statistics或sql_trace为true
2 涉及的对象（表或索引）必须被分析（analyze或dbms_stats）

监控数据库性能的SQL

[Ref: http://xzh2000.itpub.net/post/96/102574 ]

在v$session_longops视图中，sofar字段表示已经扫描的块数，totalwork表示总得需要扫描的块数，所以我们即可以对正在运行的长查询进行监控，比如在索引创建时，查看索引创建的进度，也可以查看系统中以往的长查询。。。

col opname format a32
col target_desc format a32
col perwork format a12
set lines 131
select sid,OPNAME,TARGET_DESC,sofar,TOTALWORK,trunc(sofar/totalwork*100,2)||'%' as perwork 
from v$session_longops where sofar!=totalwork;

set lines 121
set pages 999
col opname format a29
col target format a29
col target_desc format a12
col perwork format a12
col remain format 99
col start_time format a21
col sofar format 99999999
col totalwork format 99999999
col sql_text format a101
col bufgets format 99999999

select opname,target,to_char(start_time,'yy-mm-dd:hh24:mi:ss') start_time,elapsed_seconds elapsed,
executions execs,buffer_gets/decode(executions,0,1,executions) bufgets,module,sql_text
from v$session_longops sl,v$sqlarea sa
where sl.sql_hash_value = sa.hash_value and upper(substr(module,1,4)) <> 'RMAN' and substr(opname,1,4) <> 'RMAN' 
and module <> 'SQL*Plus' and sl.start_time>trunc(sysdate)
order by start_time;

==========================================================

1. 监控事例的等待
    select event,sum(decode(wait_Time,0,0,1)) "Prev", 
   sum(decode(wait_Time,0,1,0)) "Curr",count(*) "Tot" 
    from v$session_Wait 
    group by event order by 4;
2. 回滚段的争用情况
    select name, waits, gets, waits/gets "Ratio" 
    from v$rollstat a, v$rollname b 
    where a.usn = b.usn; 
3. 监控表空间的 I/O 比例
    select df.tablespace_name name,df.file_name "file",f.phyrds pyr,
    f.phyblkrd pbr,f.phywrts pyw, f.phyblkwrt pbw
    from v$filestat f, dba_data_files df
    where f.file# = df.file_id
    order by df.tablespace_name;
4. 监控文件系统的 I/O 比例
    select substr(a.file#,1,2) "#", substr(a.name,1,30) "Name", 
    a.status, a.bytes, b.phyrds, b.phywrts 
    from v$datafile a, v$filestat b 
    where a.file# = b.file#; 
5.在某个用户下找所有的索引
    select user_indexes.table_name, user_indexes.index_name,uniqueness, column_name
    from user_ind_columns, user_indexes
    where user_ind_columns.index_name = user_indexes.index_name
    and user_ind_columns.table_name = user_indexes.table_name 
    order by user_indexes.table_type, user_indexes.table_name,
    user_indexes.index_name, column_position;
6. 监控 SGA 的命中率
    select a.value + b.value "logical_reads", c.value "phys_reads",
    round(100 * ((a.value+b.value)-c.value) / (a.value+b.value)) "BUFFER HIT RATIO" 
    from v$sysstat a, v$sysstat b, v$sysstat c
    where a.statistic# = 38 and b.statistic# = 39 
    and c.statistic# = 40; 
7. 监控 SGA 中字典缓冲区的命中率
    select parameter, gets,Getmisses , getmisses/(gets+getmisses)*100 "miss ratio",
    (1-(sum(getmisses)/ (sum(gets)+sum(getmisses))))*100 "Hit ratio"
    from v$rowcache 
    where gets+getmisses <>0
    group by parameter, gets, getmisses; 
8. 监控 SGA 中共享缓存区的命中率，应该小于1%
    select sum(pins) "Total Pins", sum(reloads) "Total Reloads",
    sum(reloads)/sum(pins) *100 libcache
    from v$librarycache;
    select sum(pinhits-reloads)/sum(pins) "hit radio",sum(reloads)/sum(pins) "reload percent" 
    from v$librarycache;
9. 显示所有数据库对象的类别和大小
    select count(name) num_instances ,type ,sum(source_size) source_size ,
    sum(parsed_size) parsed_size ,sum(code_size) code_size ,sum(error_size) error_size,
    sum(source_size) +sum(parsed_size) +sum(code_size) +sum(error_size) size_required 
    from dba_object_size 
    group by type order by 2;
10. 监控 SGA 中重做日志缓存区的命中率，应该小于1%
    SELECT name, gets, misses, immediate_gets, immediate_misses,
    Decode(gets,0,0,misses/gets*100) ratio1,
    Decode(immediate_gets+immediate_misses,0,0,
    immediate_misses/(immediate_gets+immediate_misses)*100) ratio2
    FROM v$latch WHERE name IN ('redo allocation', 'redo copy'); 
11.  监控内存和硬盘的排序比率，最好使它小于 .10，增加 sort_area_size 
    SELECT name, value FROM v$sysstat WHERE name IN ('sorts (memory)', 'sorts (disk)'); 
12. 监控当前数据库谁在运行什么SQL语句
    SELECT osuser, username, sql_text from v$session a, v$sqltext b
    where a.sql_address =b.address order by address, piece;
13. 监控字典缓冲区
    SELECT (SUM(PINS - RELOADS)) / SUM(PINS) "LIB CACHE" FROM V$LIBRARYCACHE;
    SELECT (SUM(GETS - GETMISSES - USAGE - FIXED)) / SUM(GETS) "ROW CACHE" FROM V$ROWCACHE;
    SELECT SUM(PINS) "EXECUTIONS", SUM(RELOADS) "CACHE MISSES WHILE EXECUTING" FROM V$LIBRARYCACHE; 
    后者除以前者,此比率小于1%,接近0%为好。
    SELECT SUM(GETS) "DICTIONARY GETS",SUM(GETMISSES) "DICTIONARY CACHE GET MISSES"
    FROM V$ROWCACHE
14. 找ORACLE字符集
    select * from sys.props$ where name='NLS_CHARACTERSET'; 
15. 监控 MTS
    select busy/(busy+idle) "shared servers busy" from v$dispatcher;
    此值大于0.5时，参数需加大
    select sum(wait)/sum(totalq) "dispatcher waits" from v$queue where type='dispatcher';
    select count(*) from v$dispatcher;
    select servers_highwater from v$mts;
    servers_highwater接近mts_max_servers时，参数需加大
16. 碎片程度
    select tablespace_name,count(tablespace_name) from dba_free_space group by tablespace_name 
    having count(tablespace_name)>10;
    alter tablespace name coalesce;
    alter table name deallocate unused;
    create or replace view ts_blocks_v as
    select tablespace_name,block_id,bytes,blocks,'free space' segment_name from dba_free_space
    union all
    select tablespace_name,block_id,bytes,blocks,segment_name from dba_extents;
    select * from ts_blocks_v;
    select tablespace_name,sum(bytes),max(bytes),count(block_id) from dba_free_space 
    group by tablespace_name;
    查看碎片程度高的表
    SELECT segment_name table_name , COUNT(*) extents
    FROM dba_segments WHERE owner NOT IN ('SYS', 'SYSTEM') GROUP BY segment_name
    HAVING COUNT(*) = (SELECT MAX( COUNT(*) ) FROM dba_segments GROUP BY segment_name);
17. 表、索引的存储情况检查
    select segment_name,sum(bytes),count(*) ext_quan from dba_extents where 
    tablespace_name='&tablespace_name' and segment_type='TABLE' group by tablespace_name,segment_name;
    select segment_name,count(*) from dba_extents where segment_type='INDEX' and owner='&owner'
    group by segment_name;
18、找使用CPU多的用户session
    12是cpu used by this session
    select a.sid,spid,status,substr(a.program,1,40) prog,a.terminal,osuser,value/60/100 value
    from v$session a,v$process b,v$sesstat c
    where c.statistic#=12 and c.sid=a.sid and a.paddr=b.addr order by value desc;
转自：http://xsb.itpub.net/post/419/105941
-----------

查看等待事件

--v$session_wait视图中的p1、p2、p3表示等待事件的具体含义，如果Wait Event是db file scattered read,那么p1=file_id/p2=block_id/p3=blocks,然后通过dba_extents即可

确定出热点对象.
--如果是latch free的话,那么p2为闩锁号,它指向v$latch.

--求等待事件及其对应的latch
col event format a32;
col name format a32;
select sid,event,p1 as "p1 as file_id", p2 as "p2 as block_id/latch", p3 as "p3 as blocks",l.name
from v$session_wait sw,v$latch l 
where event not like '%SQL%' and event not like '%rdbms%' and event not like '%mon%' and sw.p2 = l.latch#(+);

--求等待事件及其热点对象
col owner format a18;
col segment_name format a32;
col segment_type format a32;
select owner, segment_name, segment_type from dba_extents
where file_id = &file_id and &block_id between block_id and block_id + &blocks - 1;

--综合以上两条SQL,同时显示latch及热点对象(速度较慢)
select sw.sid, event, l.name, de.segment_name from v$session_wait sw, v$latch l, dba_extents de
where event not like '%SQL%' and event not like '%rdbms%' and event not like '%mon%'
and sw.p2 = l.latch#(+) and sw.p1 = de.file_id(+) and p2 between de.block_id and de.block_id + de.blocks - 1;

--如果是非空闲等待事件,通过等待会话的SID可以求出该会话在执行的SQL
select sql_text from v$sqltext_with_newlines st, v$session se
where st.address = se.sql_address and st.hash_value = se.sql_hash_value and se.sid = &wait_sid order by piece;

----

如何查看session级的等待事件

===========================================================

作者: smartcobra(http://smartcobra.itpub.net)
发表于: 2004.11.22 11:06
分类: Performance Tuning 
出处: http://smartcobra.itpub.net/post/487/6179
--------------------------------------------------------------- 

当我们对数据库的性能进行调整时，一个最重要的参考指标就是系统等待事件。

当我们对数据库的性能进行调整时，一个最重要的参考指标就是系统等待事件。$system_event,v$session_event,v$session_wait这三个视图里记录的就是系统级和session级的等待事件，通过查询这些视图你可以发现数据库的一些操作到底在等待什么？是磁盘I/O，缓冲区忙，还是插锁等等。

通过如下sql你可以查询你的每个应用程序到底在等待什么，从而针对这些信息对数据库的性能进行调整。

Select s.username,s.program,s.status,se.event,se.total_waits,se.total_timeouts,se.time_waited,se.average_wait from v$session s,v$session_event se

Where s.sid=se.sid

And se.event not like ‘SQl*Net%’

And s.status = ‘ACTIVE’

And s.username is not null；