Decode360's Blog

业精于勤而荒于嬉 QQ:150355677 MSN:decode360@hotmail.com

  BlogJava :: 首页 :: 新随笔 :: 联系 ::  :: 管理 ::
  302 随笔 :: 26 文章 :: 82 评论 :: 0 Trackbacks
    Statspack的Report文档其实是将Oracle系统的各项系统个指标分项统计成各个报表,之后放入到Report文档中,所以可以根据自己的需要截取其中的一部分进行分析。里面的数据非常之多,要一次分析完全是不太现实的。
 
    下面截取一下常见的学习一下:
 
 
1、报表头信息
 
    报表头显示了数据库实例的相关信息,包括数据库名称、ID、版本号、主机等信息;以及起止snap的信息等:
 
STATSPACK report for
 
Database    DB Id    Instance     Inst Num Startup Time    Release     RAC
~~~~~~~~ ----------- ------------ -------- --------------- ----------- ---
           486494833 dodo                1 21-2月 -09 16:2 10.2.0.1.0  NO
                                           7
 
Host  Name:   HHZ-0099         Num CPUs:    2        Phys Memory (MB):    2,046
~~~~
 
Snapshot       Snap Id     Snap Time      Sessions Curs/Sess Comment
~~~~~~~~    ---------- ------------------ -------- --------- -------------------
Begin Snap:          1 21-2月 -09 20:27:36      18       3.4
  End Snap:          4 21-2月 -09 21:05:50      19       3.1
   Elapsed:               38.23 (mins)
 
Cache Sizes                       Begin        End
~~~~~~~~~~~                  ---------- ----------
               Buffer Cache:       404M             Std Block Size:         8K
           Shared Pool Size:       164M                 Log Buffer:     6,860K
 
 
2、负载间档
 
    该部分提供每秒和每个事物的统计信息,是监控系统吞吐量和负载变化的重要部分。
 
Load Profile                            Per Second       Per Transaction
~~~~~~~~~~~~                       ---------------       ---------------
                  Redo size:              1,888.85             22,805.41
              Logical reads:                 51.72                624.40
              Block changes:                  5.49                 66.27
             Physical reads:                  0.01                  0.13
            Physical writes:                  1.52                 18.33
                 User calls:                  0.51                  6.13
                     Parses:                  2.21                 26.72
                Hard parses:                  0.25                  3.00
                      Sorts:                  0.79                  9.59
                     Logons:                  0.03                  0.39
                   Executes:                  4.12                 49.78
               Transactions:                  0.08
 
  % Blocks changed per Read:   10.61    Recursive Call %:    97.58
 Rollback per transaction %:    0.00       Rows per Sort:    51.29
 
    说明:
    ⑴ Redo size:每秒产生的日志大小(单位字节),可标志数据变更频率, 数据库任务的繁重与否
    ⑵ Logical reads:平决每秒产生的逻辑读,单位是block
    ⑶ Block changes:每秒block变化数量,数据库事物带来改变的块数量
    ⑷ Physical reads:平均每秒数据库从磁盘读取的block数
    ⑸ Physical writes:平均每秒数据库写磁盘的block数
    ⑹ User calls:每秒用户call次数
    ⑺ Parses: 每秒解析次数,近似反应每秒语句的执行次数, 软解析每秒超过300次意味着你的“应用程序”效率不高,调整session_cursor_cache
    ⑻ Hard parses:每秒产生的硬解析次数, 每秒超过100次,就可能说明你绑定使用的不好
    ⑼ Sorts:每秒产生的排序次数
    ⑽ Logons:每秒的登陆次数
    ⑾ Executes:每秒的执行次数
    ⑿ Transactions:每秒产生的事务数,反映数据库任务繁重与否
    ⒀ Recursive Call %: 如果有很多PLSQL,那么他就会比较高
    ⒁ Rollback per transaction %:看回滚率是不是很高,因为回滚很耗资源
 
    注:回滚率=Round(User rollbacks / (user commits + user rollbacks) ,4)* 100% , 如果回滚率过高,可能说明你的数据库经历了太多的无效操作, 过多的回滚可能还会带来Undo Block的竞争。
 

3、实例命中率
 
    该部分可以提前找出ORACLE潜在将要发生的性能问题,很重要
 
Instance Efficiency Percentages
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
            Buffer Nowait %:   99.99       Redo NoWait %:  100.00
            Buffer  Hit   %:   99.98    In-memory Sort %:  100.00
            Library Hit   %:   94.65        Soft Parse %:   88.77
         Execute to Parse %:   46.34         Latch Hit %:  100.00
Parse CPU to Parse Elapsd %:   97.83     % Non-Parse CPU:   95.37
 
 Shared Pool Statistics        Begin   End
                               ------  ------
             Memory Usage %:   60.34   71.99
    % SQL with executions>1:   63.66   66.00
  % Memory for SQL w/exec>1:   82.19   73.37
 
    说明:
    ⑴ Buffer Nowait %:在缓冲区中获取Buffer的未等待比率, Buffer Nowait<99%说明,有可能是有热块(查找x$bh的tch和v$latch_children的cache buffers chains)
    ⑵ Redo NoWait %:在Redo缓冲区获取Buffer的未等待比率
    ⑶ Buffer Hit %:数据块在数据缓冲区中得命中率,通常应在90%以上,否则需要调整,小于95%,重要的参数,小于90%可能是要加db_cache_size,但是大量的非选择的索引也会造成该值很高(大量的db file sequential read)
    ⑷ In-memory Sort %:在内存中的排序率
    ⑸ Library Hit %:主要代表sql在共享区的命中率,通常在95%以上,否则需要要考虑加大共享池,绑定变量,修改cursor_sharing等参数。
    ⑹ Soft Parse %:近似看作sql在共享区的命中率,小于95%需要考虑到绑定,如果低于80%,那么就可能sql基本没有被重用
    ⑺ Execute to Parse %:sql语句解析后被重复执行的次数,如果过低,可以考虑设置session_cached_cursors参数, 公式为100 * (1 - Parses/Executions) = Execute to Parse 所以如果系统Parses > Executions,就可能出现该比率小于0的情况, 该值<0通常说明shared pool设置或效率存在问题造成反复解析,reparse可能较严重,或者可是同snapshot有关如果该值为负值或者极低,通常说明数据库性能存在问题。
    ⑻ Latch Hit %: 要确保>99%,否则存在严重的性能问题,比如绑定等会影响该参数
    ⑼ Parse CPU to Parse Elapsd %:解析实际运行事件/(解析实际运行时间+解析中等待资源时间)越高越好
    ⑽ % Non-Parse CPU:查询实际运行时间/(查询实际运行时间+sql解析时间),太低表示解析消耗时间过多。100*(parse time cpu / parse time elapsed)= Parse CPU to Parse Elapsd %
 
    注:
    如果一个经常访问的列上的索引被删除,可能会造成Buffer Hit显著的下降
    如果增加了索引,但是他影响了Oracle正确的选择表连接时的驱动顺序,那么可能会导致Buffer Hit显著增高
    如果你的命中率变化幅度很大,说明你要改变SQL模式
    Shared Pool相关统计数据
    ⑴ Memory Usage %:共享池内存使用率,应该稳定在70%-90%间,太小浪费内存,太大则内存不足。
    ⑵ % SQL with executions>1:执行次数大于1的sql比率,若太小可能是没有使用bind variables。
    ⑶ % Memory for SQL w/exec>1:也即是memory for sql with execution>1:执行次数大于1的sql消耗内存/所有sql消耗的内存
 
 
4、首要等待事件(Top 5)
 
    Oracle等待事件是衡量Oracle运行状况的重要依据及指示,主要有空闲等待事件和非空闲等待事件。
    TIMED_STATISTICS = TRUE 等待事件按等待的时间排序
    TIMED_STATISTICS = FALSE 等待事件按等待的数量排序
    Statspack中必须在Session上设置TIMED_STATISTICS = TRUE

    空闲等待事件是Oracle正等待某种工作,在诊断和优化数据库时候,不用过多注意这部分事件,非空闲等待事件专门针对Oracle的活动,指数据库任务或应用程序运行过程中发生的等待,这些等待事件是我们在调整数据库应该关注的。
Top 5 Timed Events                                                    Avg %Total
~~~~~~~~~~~~~~~~~~                                                   wait   Call
Event                                            Waits    Time (s)   (ms)   Time
----------------------------------------- ------------ ----------- ------ ------
db file parallel write                           1,575          20     13   53.8
CPU time                                                        10          26.0
control file sequential read                     2,001           3      2    8.7
control file parallel write                        805           3      4    7.7
log file parallel write                            676           1      1    1.8
          -------------------------------------------------------------
    比较影响性能常见等待事件有:

    ① db file scattered read
    该事件通常与全表扫描有关。因为全表扫描是被放入内存中进行的,通常情况下它不可能被放入连续的缓冲区中,所以就散布在缓冲区的缓存中。该指数的数量过大说明缺少索引或者限制了索引的使用(也可以调整optimizer_index_cost_adj) 。这种情况也可能是正常的,因为执行全表扫描可能比索引扫描效率更高。当系统存在这些等待时,需要通过检查来确定全表扫描是否必需的来调整。如果经常必须进行全表扫描,而且表比较小, 把该表存人keep池。如果是大表经常进行全表扫描,那么应该是OLAP系统,而不是OLTP的.
    ② db file sequential read
    该事件说明在单个数据块上大量等待,该值过高通常是由于表间连接顺序很糟糕,或者使用了非选择性索引。通过将这种等待与statspack报表中已知其它问题联系起来(如效率不高的sql),通过检查确保索引扫描是必须的,并确保多表连接的连接顺序来调整, DB_CACHE_SIZE可以决定该事件出现的频率
    ③ buffer busy wait
    当缓冲区以一种非共享方式或者如正在被读入到缓冲时,就会出现该等待。该值不应该大于1%,确认是不是由于热点块造成(如果是可以用反转索引,或者用更小块大小)
    ④ latch free
    常跟应用没有很好的应用绑定有关。闩锁是底层的队列机制(更加准确的名称应该是互斥机制),用于保护系统全局区(SGA)共享内存结构闩锁用于防止对内存结构的并行访问。如果闩锁不可用,就会记录一次闩锁丢失。绝大多数得闩锁问题都与使用绑定变量失败(库缓存闩锁)、生成重作问题(重执行分配闩锁)、缓存的争用问题(缓存LRU链) 以及缓存的热数据宽块(缓存链)有关。当闩锁丢失率高于0.5%时,需要调整这个问题。
    ⑤ log buffer space
    日志缓冲区写的速度快于LGWR写REDOFILE的速度,可以增大日志文件大小,增加日志缓冲区的大小,或者使用更快的磁盘来写数据。
    ⑥ logfile switch
    通常是因为归档速度不够快,需要增大重做日志
    ⑦ log file sync
    当一个用户提交或回滚数据时,LGWR将会话得重做操作从日志缓冲区填充到日志文件中,用户的进程必须等待这个填充工作完成。在每次提交时都出现,如果这个等待事件影响到数据库性能,那么就需要修改应用程序的提交频率, 为减少这个等待事件,须一次提交更多记录,或者将重做日志REDO LOG 文件访在不同的物理磁盘上。
 
    更加具体的说明可以参见eygle大师的《Statspack-v3.0》
 
 
5、实例的CPU及内存信息
 
Host CPU  (CPUs: 2)
~~~~~~~~              Load Average
                      Begin     End      User  System    Idle     WIO     WCPU
                    ------- -------   ------- ------- ------- ------- --------
                                         1.62    1.99   96.39
 
Instance CPU
~~~~~~~~~~~~
              % of total CPU for Instance:    0.27
              % of busy  CPU for Instance:    7.54
  %DB time waiting for CPU - Resource Mgr:
 
Memory Statistics                       Begin          End
~~~~~~~~~~~~~~~~~                ------------ ------------
                  Host Mem (MB):      2,046.4      2,046.4
                   SGA use (MB):        584.0        584.0
                   PGA use (MB):         40.3         40.5
    % Host Mem used for SGA+PGA:         30.5         30.5
          -------------------------------------------------------------
 
Time Model System Stats  DB/Inst: DODO/dodo  Snaps: 1-4
-> Ordered by % of DB time desc, Statistic name
 
Statistic                                       Time (s) % of DB time
----------------------------------- -------------------- ------------
sql execute elapsed time                             9.4        100.9
DB CPU                                               8.2         87.5
PL/SQL execution elapsed time                        2.3         24.2
parse time elapsed                                   1.7         17.8
hard parse elapsed time                              1.5         15.5
PL/SQL compilation elapsed time                      0.1          1.6
failed parse elapsed time                            0.0           .3
hard parse (sharing criteria) elaps                  0.0           .2
connection management call elapsed                   0.0           .1
repeated bind elapsed time                           0.0           .1
hard parse (bind mismatch) elapsed                   0.0           .0
sequence load elapsed time                           0.0           .0
DB time                                              9.4
background elapsed time                             32.4
background cpu time                                  4.3
          -------------------------------------------------------------
 
 
6、数据库用户程序发生的所有等待事件
 
Wait Events  DB/Inst: DODO/dodo  Snaps: 1-4
-> s - second, cs - centisecond,  ms - millisecond, us - microsecond
-> %Timeouts:  value of 0 indicates value was < .5%.  Value of null is truly 0
-> Only events with Total Wait Time (s) >= .001 are shown
-> ordered by Total Wait Time desc, Waits desc (idle events last)
 
                                                                    Avg
                                                %Time Total Wait   wait    Waits
Event                                    Waits  -outs   Time (s)   (ms)     /txn
--------------------------------- ------------ ------ ---------- ------ --------
db file parallel write                   1,575      0         20     13      8.3
control file sequential read             2,001      0          3      2     10.5
control file parallel write                805      0          3      4      4.2
log file parallel write                    676      0          1      1      3.6
os thread startup                           35      0          0     10      0.2
db file sequential read                     21      0          0     13      0.1
log file sync                               43      0          0      2      0.2
direct path write                           35      0          0      2      0.2
rdbms ipc reply                             28      0          0      0      0.1
SQL*Net more data to client                 45      0          0      0      0.2
Streams AQ: qmn slave idle wait             82      0      2,297  28012      0.4
Streams AQ: qmn coordinator idle           163     52      2,297  14092      0.9
virtual circuit status                      76    100      2,274  29921      0.4
jobq slave wait                            712     99      2,127   2988      3.7
Streams AQ: waiting for time mana            1    100      1,109 ######      0.0
SQL*Net message from client                682      0        745   1092      3.6
class slave wait                             2    100         10   4999      0.0
SQL*Net message to client                  683      0          0      0      3.6
          -------------------------------------------------------------
 
 
7、数据库后台进程发生的等待事件
 
Background Wait Events  DB/Inst: DODO/dodo  Snaps: 1-4
-> %Timeouts:  value of 0 indicates value was < .5%.  Value of null is truly 0
-> Only events with Total Wait Time (s) >= .001 are shown
-> ordered by Total Wait Time desc, Waits desc (idle events last)
 
                                                                    Avg
                                                %Time Total Wait   wait    Waits
Event                                    Waits  -outs   Time (s)   (ms)     /txn
--------------------------------- ------------ ------ ---------- ------ --------
db file parallel write                   1,575      0         20     13      8.3
control file parallel write                805      0          3      4      4.2
control file sequential read               729      0          2      3      3.8
log file parallel write                    677      0          1      1      3.6
os thread startup                           35      0          0     10      0.2
direct path write                            5      0          0     13      0.0
db file sequential read                      5      0          0     10      0.0
events in waitclass Other                   54      0          0      0      0.3
rdbms ipc message                        8,090     92     23,167   2864     42.6
Streams AQ: qmn slave idle wait             82      0      2,297  28012      0.4
Streams AQ: qmn coordinator idle           163     52      2,297  14092      0.9
pmon timer                                 789    100      2,292   2906      4.2
smon timer                                  14     43      2,267 ######      0.1
Streams AQ: waiting for time mana            1    100      1,109 ######      0.0
          -------------------------------------------------------------
 
 
8、等待事件柱状图
 
Wait Event Histogram  DB/Inst: DODO/dodo  Snaps: 1-4
-> Total Waits - units: K is 1000, M is 1000000, G is 1000000000
-> % of Waits - column heading: <=1s is truly <1024ms, >1s is truly >=1024ms
-> % of Waits - value: .0 indicates value was <.05%, null is truly 0
-> Ordered by Event (idle events last)
 
                           Total ----------------- % of Waits ------------------
Event                      Waits  <1ms  <2ms  <4ms  <8ms <16ms <32ms  <=1s   >1s
-------------------------- ----- ----- ----- ----- ----- ----- ----- ----- -----
LGWR wait for redo copy      17  100.0
SQL*Net break/reset to cli    2  100.0
SQL*Net more data to clien   45  100.0
buffer busy waits             5  100.0
control file parallel writ  805   22.1  69.6    .5   1.0   1.5   2.6   2.7
control file sequential re 2001   91.9    .4    .6   3.1   1.8    .8   1.3
db file parallel write     1575   47.3   3.5   8.1   6.4  12.8   7.3  14.6
db file scattered read        1  100.0
db file sequential read      21   23.8              19.0  33.3   9.5  14.3
direct path read              5  100.0
direct path write            35   97.1                                 2.9
latch free                    5  100.0
latch: library cache          2  100.0
latch: shared pool            1  100.0
log file parallel write     676   96.2    .7    .6    .3    .4   1.0    .7
log file sync                43   86.0         2.3   2.3   2.3   4.7   2.3
os thread startup            35                      5.7  91.4   2.9
rdbms ipc reply              28  100.0
undo segment extension        3  100.0
SQL*Net message from clien  682   67.9   7.6  14.1   5.1   2.1    .9   1.0   1.3
SQL*Net message to client   683  100.0
SQL*Net more data from cli   61  100.0
Streams AQ: qmn coordinato  163   48.5                                      51.5
Streams AQ: qmn slave idle   82                                            100.0
Streams AQ: waiting for ti    1                                            100.0
class slave wait              2                                            100.0
dispatcher timer             38                                            100.0
jobq slave wait             712                                         .3  99.7
pmon timer                  789    1.1          .1                      .3  98.5
rdbms ipc message          8083    5.2    .6    .4    .6    .5    .1  29.3  63.2
smon timer                   13    7.7               7.7   7.7        15.4  61.5
virtual circuit status       76                                            100.0
          -------------------------------------------------------------
 
 
9、TOP SQL
 
    调整首要的25个缓冲区读操作和首要的25个磁盘读操作做的查询,将可对系统性能产生5%到5000%的增益。
 
    内容太多不再截取...
 
 
10、实例活动
 
Instance Activity Stats  DB/Inst: DODO/dodo  Snaps: 1-4
 
Statistic                                      Total     per Second    per Trans
--------------------------------- ------------------ -------------- ------------
redo synch time                                   11            0.0          0.1
redo synch writes                                614            0.3          3.2
redo wastage                                 174,084           75.9        916.2
redo write time                                   61            0.0          0.3
redo writer latching time                          2            0.0          0.0
redo writes                                      677            0.3          3.6
rollback changes - undo records a                 10            0.0          0.1
rollbacks only - consistent read                   8            0.0          0.0
rows fetched via callback                     14,999            6.5         78.9
session connect time                           2,312            1.0         12.2
session cursor cache hits                      3,352            1.5         17.6
session logical reads                        118,636           51.7        624.4
session pga memory                        34,734,200       15,141.3    182,811.6
session pga memory max                    32,964,728       14,370.0    173,498.6
session uga memory                   158,915,954,392   69,274,609.6 ############
session uga memory max                    44,277,484       19,301.4    233,039.4
shared hash latch upgrades - no w             33,011           14.4        173.7
sorts (memory)                                 1,823            0.8          9.6
sorts (rows)                                  93,505           40.8        492.1
sql area purged                                    2            0.0          0.0
switch current to new buffer                       2            0.0          0.0
table fetch by rowid                          23,047           10.1        121.3
table fetch continued row                         61            0.0          0.3
table scan blocks gotten                       9,066            4.0         47.7
table scan rows gotten                       400,817          174.7      2,109.6
table scans (short tables)                       760            0.3          4.0
total number of times SMON posted                  8            0.0          0.0
transaction rollbacks                              5            0.0          0.0
undo change vector size                    1,329,952          579.8      6,999.8
user I/O wait time                                34            0.0          0.2
user calls                                     1,165            0.5          6.1
user commits                                     190            0.1          1.0
workarea executions - optimal                    884            0.4          4.7
write clones created in backgroun                  0            0.0          0.0
          -------------------------------------------------------------
 
 
11、IO情况统计
 
Tablespace IO Stats  DB/Inst: DODO/dodo  Snaps: 1-4
->ordered by IOs (Reads + Writes) desc
 
File IO Stats  DB/Inst: DODO/dodo  Snaps: 1-4
->Mx Rd Bkt: Max bucket time for single block read
->ordered by Tablespace, File
 
File Read Histogram Stats  DB/Inst: DODO/dodo  Snaps: 1-4
->Number of single block reads in each time range
->ordered by Tablespace, File
 
 
12、缓冲区
 
Buffer Pool Statistics  DB/Inst: DODO/dodo  Snaps: 1-4
-> Standard block size Pools  D: default,  K: keep,  R: recycle
-> Default Pools for other block sizes: 2k, 4k, 8k, 16k, 32k
-> Buffers: the number of buffers.  Units of K, M, G are divided by 1000
 
Instance Recovery Stats  DB/Inst: DODO/dodo  Snaps: 1-4
-> B: Begin snapshot,  E: End snapshot
 
Buffer Pool Advisory  DB/Inst: DODO/dodo  End Snap: 4
-> Only rows with estimated physical reads >0 are displayed
-> ordered by Pool, Block Size, Buffers For Estimate
 
Buffer wait Statistics  DB/Inst: DODO/dodo  Snaps: 1-4
-> ordered by wait time desc, waits desc
 
 
13、PGA
 
PGA Aggr Target Stats  DB/Inst: DODO/dodo  Snaps: 1-4
-> B: Begin snap   E: End snap (rows identified with B or E contain data
   which is absolute i.e. not diffed over the interval)
-> PGA cache hit % - percentage of W/A (WorkArea) data processed only in-memory
-> Auto PGA Target - actual workarea memory target
-> W/A PGA Used    - amount of memory used for all Workareas (manual + auto)
-> %PGA W/A Mem    - percentage of PGA memory allocated to workareas
-> %Auto W/A Mem   - percentage of workarea memory controlled by Auto Mem Mgmt
-> %Man W/A Mem    - percentage of workarea memory under manual control
 
 
14、进程该要
 
Process Memory Summary Stats  DB/Inst: DODO/dodo  Snaps: 1-4
-> B: Begin snap   E: End snap
-> All rows below contain absolute values (i.e. not diffed over the interval)
-> Max Alloc is Maximum PGA Allocation size at snapshot time
   Hist Max Alloc is the Historical Max Allocation for still-connected processes
-> Num Procs or Allocs:  For Begin/End snapshot lines, it is the number of
   processes. For Category lines, it is the number of allocations
-> ordered by Begin/End snapshot, Alloc (MB) desc
 
 
15、回滚段
 
Undo Segment Summary  DB/Inst: DODO/dodo  Snaps: 1-4
-> Min/Max TR (mins) - Min and Max Tuned Retention (minutes)
-> STO - Snapshot Too Old count,  OOS - Out Of Space count
-> Undo segment block stats:
   uS - unexpired Stolen,   uR - unexpired Released,   uU - unexpired reUsed
   eS - expired   Stolen,   eR - expired   Released,   eU - expired   reUsed
 
 
16、锁
 
...
 




-The End-

posted on 2009-01-20 20:13 decode360-3 阅读(568) 评论(0)  编辑  收藏 所属分类: Oracle

只有注册用户登录后才能发表评论。


网站导航:
博客园   IT新闻   Chat2DB   C++博客   博问