根据statspack报表优化oracle数据库实例之“DB file sequential read”

根据statspack报表优化oracle数据库实例之“DB file sequential read”

oracle的等待事件是衡量oracle运行状况的重要依据及指标。

等待事件的概念是在oracle7.0.1.2中引入的，大致有100个等待事件。在oracle8.0中这个数目增加到了大约150个，在oracle8i中大约有200个事件，到oracle9i时，等待事件增加到360个。

Oracle的等待事件主要有两种类型，即空闲（idle）等待事件和非空闲（non-idle）等待事件。空闲事件指oracle正在等待某种工作，在诊断和优化数据库的时候，我们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的时候应该关注与研究的。

常见的非空闲等待事件有：db file scattered read; db file sequential read; buffer busy waits; free buffer waits; enqueue; latch free; log file parallel write; log file sync.

Db file sequential read的产生

本文主要解释了db file sequence read文件分散读取等待事件产生的原因与优化的方法。

db file sequential read等待时间是由于执行对索引，回滚（undo）段，和表（当借助rowid来访问），控制文件和数据文件头的单块读操作SQL语句（用户和递归）引起的。

对于这些对象的物理I/O请求是很正常的，因此db file sequential read等待的存在不是一定意味库或应用出错了。如果会话在这事件上花了好长事件，它可能也不是一个糟糕的事情。相反，如果会话花了大量时间在equeue或latch free上，那么一定是有问题。这儿单块读变的复杂了。

如果这个等待事件比较显著，可能表示在多表连接中，表的连接顺序存在问题，可能没有正确的使用驱动表；或者可能说明不加选择地进行索引。

    在大多数情况下我们说，通过索引可以更为快速的获取记录，所以对于一个编码规范、调整良好的数据库，这个等待很大是正常的。

    但在很多情况下，使用索引并不总是最佳选择，比如读取较大表中大量的数据，全表扫描可能会明显快于索引扫描，所以在开发中我们就应该注意，对于这样的查询应该进行避免使用索引扫描。

对于db file sequential read中,p1指数据文件ID,p2指block#号,p3指读取的block数量,这个事件一般不可避免，大多由于ＳＱＬ使用索引不当，造成从磁盘上读取连续的数据，接近于全表扫描.可以通过db_file_multiblock_read_count参数来调整每次读取的block数,减少IO.

Statspack的报表

Top 5 Timed Events

~~~~~~~~~~~~~~~~~~                                                     % Total

Event                                               Waits    Time (s) Ela Time

-------------------------------------------- ------------ ----------- --------

CPU time                                                          246    50.79

db file sequential read                            98,012         208    43.01

db file scattered read                              1,001          11     2.20

direct path write                                   2,171           7     1.52

control file parallel write                         1,404           3      .56

―――――――――

获得db file sequential read等待时间占总等待时间的比例

    动态性能视图v$session_event中存储了系统库缓冲池中存储的sql语句的所有等待事件的时间。关联v$session可以获得当前连接的等待时间和该时间所占总等待时间的比例。

    select a.sid,

       a.event,

       a.time_waited,

       a.time_waited / c.sum_time_waited * 100 pct_wait_time,

       round((sysdate - b.logon_time) * 24) hours_connected

from   v$session_event a, v$session b,

       (select sid, sum(time_waited) sum_time_waited

        from   v$session_event

        where  event not in (

                    'Null event',

                    'client message',

                    'KXFX: Execution Message Dequeue - Slave',

                    'PX Deq: Execution Msg',

                    'KXFQ: kxfqdeq - normal deqeue',

                    'PX Deq: Table Q Normal',

                    'Wait for credit - send blocked',

                    'PX Deq Credit: send blkd',

                    'Wait for credit - need buffer to send',

                    'PX Deq Credit: need buffer',

                    'Wait for credit - free buffer',

                    'PX Deq Credit: free buffer',

                    'parallel query dequeue wait',

                    'PX Deque wait',

                    'Parallel Query Idle Wait - Slaves',

                    'PX Idle Wait',

                    'slave wait',

                    'dispatcher timer',

                    'virtual circuit status',

                    'pipe get',

                    'rdbms ipc message',

                    'rdbms ipc reply',

                    'pmon timer',

                    'smon timer',

                    'PL/SQL lock timer',

                    'SQL*Net message from client',

                    'WMON goes to sleep')

        having sum(time_waited) > 0 group by sid) c

where  a.sid         = b.sid

and    a.sid         = c.sid

and    a.time_waited > 0

and    a.event       = 'db file sequential read'

order by hours_connected desc, pct_wait_time;

―――――――――――――――――――――――――――――

98    db file sequential read     1     100  0

94    db file sequential read     298  100  26

95    db file sequential read     1     100  0

92    db file sequential read     5     100  0

20    db file sequential read     19998     94.6695701571672 482

89    db file sequential read     26641     93.6414762741652 482

86    db file sequential read     3866       92.9997594419052 164

60    db file sequential read     21699     92.6199419498037 385

25    db file sequential read     15    88.2352941176471 381

105  db file sequential read     17630     88.2294064658192 481

39    db file sequential read     13782     84.2111694977392 482

54    db file sequential read     294095    82.978999551381   143

11    db file sequential read     11874      82.4297119055883 481

58    db file sequential read     889  81.4848762603116 24

61    db file sequential read     7436       80.7470952329243 482

48    db file sequential read     231  76.2376237623762 455

52    db file sequential read     12    75    471

47    db file sequential read     3101       70.2219202898551 385

31    db file sequential read     2749       64.5911654135338 385

66    db file sequential read     78    63.4146341463415 138

50    db file sequential read     9001       62.3079053025059 210

69    db file sequential read     12505     61.5767185345677 210

40    db file sequential read     55    60.4395604395604 138

91    db file sequential read     80    56.3380281690141 2

14    db file sequential read     199  54.2234332425068 471

99    db file sequential read     59    54.1284403669725 138

82    db file sequential read     67    53.6 138

35    db file sequential read     85    50.8982035928144 138

83    db file sequential read     192  49.6124031007752 471

59    db file sequential read     25811      47.9758364312268 457

9     db file sequential read     5858       47.5024326954265 210

43    db file sequential read     63    47.3684210526316 138

21    db file sequential read     71    46.7105263157895 138

49    db file sequential read     36    43.9024390243902 138

36    db file sequential read     98    43.1718061674009 40

8     db file sequential read     180  38.7096774193548 471

97    db file sequential read     35    35.3535353535354 138

100  db file sequential read     409  32.9307568438003 483

76    db file sequential read     348  32.6148078725398 483

22    db file sequential read     26    29.8850574712644 116

64    db file sequential read     314  28.8602941176471 483

72    db file sequential read     313  27.4561403508772 483

79    db file sequential read     270  27.1084337349398 483

75    db file sequential read     238  26.5033407572383 482

41    db file sequential read     293  26.3489208633094 483

63    db file sequential read     245  25.7082896117524 483

65    db file sequential read     351  25.0178189593728 482

30    db file sequential read     189  24.1687979539642 143

44    db file sequential read     21    24.1379310344828 116

57    db file sequential read     17    23.943661971831   2

24    db file sequential read     275  22.3395613322502 482

26    db file sequential read     308  20.0782268578879 482

62    db file sequential read     203  19.7663096397274 483

19    db file sequential read     297  19.5910290237467 482

90    db file sequential read     251  16.3517915309446 482

71    db file sequential read     397  15.8736505397841 482

55    db file sequential read     218  14.8907103825137 482

33    db file sequential read     407  14.7785039941903 482

74    db file sequential read     249  14.5359019264448 482

80    db file sequential read     265  14.3243243243243 482

77    db file sequential read     251  14.2532651902328 482

5     db file sequential read     7734       12.6872159976378 1941

56    db file sequential read     238  12.2997416020672 482

84    db file sequential read     22    11.8279569892473 2

96    db file sequential read     285  11.552492906364   482

38    db file sequential read     236  9.33544303797468 482

7     db file sequential read     65    8.38709677419355 1941

29    db file sequential read     193  7.81376518218623 482

12    db file sequential read     1     7.14285714285714 0

42    db file sequential read     229  7.03533026113671 482

37    db file sequential read     303  5.47623350804265 482

6     db file sequential read     2931       4.48796472101427 1941

87    db file sequential read     292  1.64247946900664 483

28    db file sequential read     197  1.44047967241884 482

32    db file sequential read     302  0.889674473412874      482

17    db file sequential read     1     0.606060606060606      0

45    db file sequential read     41    0.108110958759625      138

―――――――――――――――――――――――――――――

获取等待事件单块读等待的对象名和分区名

    动态性能视图v$SESSION_WAIT中的p1，和p2两个参数得到对象名和分区名。使用v$bh的缺点是你必须等待块被读入到buffer cache中，否则x$bh视图在buffer中没有p1，p2参数所指的信息。Dba_objects视图也不包含P1和P2所指的rollback和undo段对象。

   SELECT b.Sid,

          Nvl(Substr(a.Object_Name, 1, 30),

              'P1=' || b.P1 || ' P2=' || b.P2 || ' P3=' || b.P3) Object_Name,

          a.Subobject_Name,

          a.Object_Type

     FROM Dba_Objects    a,

          V$session_Wait b,

          Sys.X$bh       c

    WHERE c.Obj = a.Object_Id(+)

      AND b.P1 = c.File#(+)

      AND b.P2 = c.Dbablk(+)

      AND b.Event = 'db file sequential read'

      AND b.sid   = 12

   UNION

   SELECT b.Sid,

          Nvl(Substr(a.Object_Name, 1, 30),

              'P1=' || b.P1 || ' P2=' || b.P2 || ' P3=' || b.P3) Object_Name,

          a.Subobject_Name,

          a.Object_Type

     FROM Dba_Objects    a,

          V$session_Wait b,

          X$bh           c

    WHERE c.Obj = a.Data_Object_Id(+)

      AND b.P1 = c.File#(+)

      AND b.P2 = c.Dbablk(+)

      AND b.Event = 'db file sequential read'

      AND b.sid   = 12

    ORDER BY 1;

查找具有高disk read的语句

    我们可以通过如下两种方式来最小化db file sequential read事件：降低physical和logical read；降低平均等待时间。

    由于当前正在运行的sql可能也会导致wait，所以，没有历史数据的交互式诊断经常无法找出准确的等待事件和sql语句。DBA可以查询v$sql视图来查找有高平均disk_reads的语句。

select disk_reads,HASH_VALUE

from (select HASH_VALUE,disk_reads,

   dense_rank() over

     (order by disk_reads desc) disk_reads_rank

   from v$sql s)

where disk_reads_rank <=100

ORDER BY disk_reads_rank ;

 

SELECT * FROM v$sqltext sl

WHERE sl.HASH_VALUE = 384909134

ORDER BY piece

针对索引的sequential read解决方案

    使用上面的方式得到disk read较多的语句后，如果该语句的执行计划是table access by index rowed,检查索引的clustering factor是非常必要的。

select id.index_name,tb.table_name,id.clustering_factor,tb.num_rows,tb.blocks
 from dba_indexes id,dba_tables tb
 where id.table_name=tb.table_name
 and tb.table_name='&1' and tb.owner='&2'

在上述sql语句的输出结果中，如果dba_indexes.clustering_factor接近表中块的数量，那么表中大多数行是排序的。这是期望的，然而，如果clustering factor接近表中行的数量，它意味着表中的行是随机排列，这种情况对于同样叶块中的索引块来说，指向同样的数据块中的行是不可能的，因此它会导致更多的I/O来完成操作。你可以采取rebuilding表来改善索引clustering factor，为了行根据索引键来排序，其后重建索引。