使用 10046 查看执行计划并读懂 trace 文件

查看 sql 执行计划的方法有许多种, 10046 事件就是其中的一种. 与其他查看 sql 执行计划不同, 当我们遇到比较复杂的 sql 语句, 我们可以通过 10046 跟踪 sql 得到执行计划中每一个步骤的逻辑读, 物理读以及花费的时间等. 这种细粒度的跟踪对于我们分析 sql 性能尤其有用.

一般来说, 使用 10046 事件得到 sql 执行计划的步骤如下:
1. 激活当前 session 10046 事件
2. 在当前 session 中执行 sql 语句
3. 关闭当前 session 10046 事件

执行完上述步骤后, 通常会自动生成一个 trace 文件. 在 oracle 11g 中, trace 文件一般放在$ORACLE_BASE/diag/rdbms/{database_name}/$ORACLE_SID/trace 目录下. 如果使用 oradebug 激活跟踪 10046后, 可以使用 oradebug tracefile_name 得到刚刚生成的 trace 文件的完整路径.

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
background_dump_dest                 string      g:appdavidddiag
dbmsdavid
                                                 david	race

刚刚提到的 oradebug 激活跟踪 10046 事件, 我想大部分 dba 都会使用. oradebug 是个功能强大非常好用的工具, 使用 oradebug help 将会看到它的功能很多

SQL> oradebug help
HELP           [command]                        Describe one or all commands
SETMYPID                                        Debug current process
SETOSPID       <ospid>                        Set OS pid of process to debug
SETORAPID      <orapid> ['force']             Set Oracle pid of process to debug
SETORAPNAME    <orapname>                     Set Oracle process name to debug
SHORT_STACK                                     Get abridged OS stack
CURRENT_SQL                                     Get current SQL
DUMP           <dump_name> <lvl> [addr]         Invoke named dump
DUMPSGA        [bytes]                          Dump fixed SGA
DUMPLIST                                        Print a list of available dumps
EVENT          <text>                         Set trace event in process
SESSION_EVENT  <text>                         Set trace event in session
DUMPVAR        <p|s|uga> <name> [level]         Print/dump a fixed PGA/SGA/UGA variable
DUMPTYPE       <address> <type> <count>       Print/dump an address with type info
SETVAR         <p|s|uga> <name> <value>       Modify a fixed PGA/SGA/UGA variable
PEEK           <addr> <len> [level]             Print/Dump memory
POKE           <addr> <len> <value>           Modify memory
WAKEUP         <orapid>                       Wake up Oracle process
SUSPEND                                         Suspend execution
RESUME                                          Resume execution
FLUSH                                           Flush pending writes to trace file
CLOSE_TRACE                                     Close trace file
TRACEFILE_NAME                                  Get name of trace file
LKDEBUG                                         Invoke global enqueue service debugger
NSDBX                                           Invoke CGS name-service debugger
-G             <Inst-List | def | all>        Parallel oradebug command prefix
-R             <Inst-List | def | all>        Parallel oradebug prefix (return output
SETINST        <instance# .. | all>           Set instance list in double quotes
SGATOFILE      <SGA dump dir>                 Dump SGA to file; dirname in double quotes
DMPCOWSGA      <SGA dump dir>                 Dump & map SGA as COW; dirname in double quotes
MAPCOWSGA      <SGA dump dir>                 Map SGA as COW; dirname in double quotes
HANGANALYZE    [level] [syslevel]               Analyze system hang
FFBEGIN                                         Flash Freeze the Instance
FFDEREGISTER                                    FF deregister instance from cluster
FFTERMINST                                      Call exit and terminate instance
FFRESUMEINST                                    Resume the flash frozen instance
FFSTATUS                                        Flash freeze status of instance
SKDSTTPCS      <ifname>  <ofname>               Helps translate PCs to names
WATCH          <address> <len> <self|exist|all|target>  Watch a region of memory
DELETE         <local|global|target> watchpoint <id>    Delete a watchpoint
SHOW           <local|global|target> watchpoints        Show  watchpoints
DIRECT_ACCESS  <set/enable/disable command | select query> Fixed table access
CORE                                            Dump core without crashing process
IPC                                             Dump ipc information
UNLIMIT                                         Unlimit the size of the trace file
PROCSTAT                                        Dump process statistics
CALL           <func> [arg1] ... [argn]       Invoke function with arguments

使用 oradebug 跟踪 10046 命令如下:

SQL> oradebug setmypid
Statement processed.

// 激活 10046 事件
SQL> oradebug event 10046 trace name context forever,level 12;
Statement processed.

SQL> select /*+ leading(t3) use_merge(t4) */ *
  2  from t3, t4
  3  where t3.id = t4.t3_id and t3.n = 1100;

10 rows selected.

// 在当前 session 关闭 10046 事件
SQL> oradebug event 10046 trace name context off;
Statement processed.

// 使用 oradebug tracefile_name 可以直接看到生成的 trace 文件的位置
SQL> oradebug tracefile_name;
g:appdavidddiag
dbmsdaviddavid	racedavid_ora_2176.trc

其中, 10046 按照收集信息的内容分为以下等级:

Level 0	停用SQL跟踪，相当于SQL_TRACE=FALSE
Level 1	标准SQL跟踪，相当于SQL_TRACE=TRUE
Level 4	在level 1的基础上增加绑定变量的信息
Level 8	在level 1的基础上增加等待事件的信息
Level 12	在level 1的基础上增加绑定变量和等待事件的信息

分析读懂 trace 文件

现在我们打开 g:appdavidddiag dbmsdaviddavid racedavid_ora_2176.trc 看看生成的 trace 文件的内容

<pre name="code" class="sql"><pre name="code" class="sql"><pre name="code" class="sql"><pre name="code" class="sql" style="font-size:14px;">PARSING IN CURSOR #22 len=92 dep=0 uid=0 oct=3 lid=0 tim=900460923321 hv=1624778336 ad='34671d90' sqlid='g0rdyg9hdh9m0'
select /*+ leading(t3) use_merge(t4) */ *
from t3, t4
where t3.id = t4.t3_id and t3.n = 1100
END OF STMT
PARSE #22:c=0,e=10777,p=0,cr=0,cu=0,mis=1,r=0,dep=0,og=1,plh=3831111046,tim=900460923319
EXEC #22:c=0,e=29,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=1,plh=3831111046,tim=900460923482
WAIT #22: nam='SQL*Net message to client' ela= 2 driver id=1111838976 #bytes=1 p3=0 obj#=-1 tim=900460923512
FETCH #22:c=15625,e=23922,p=0,cr=119,cu=0,mis=0,r=1,dep=0,og=1,plh=3831111046,tim=900460947462
WAIT #22: nam='SQL*Net message from client' ela= 221 driver id=1111838976 #bytes=1 p3=0 obj#=-1 tim=900460947755
WAIT #22: nam='SQL*Net message to client' ela= 2 driver id=1111838976 #bytes=1 p3=0 obj#=-1 tim=900460947803
FETCH #22:c=0,e=71,p=0,cr=0,cu=0,mis=0,r=9,dep=0,og=1,plh=3831111046,tim=900460947864
STAT #22 id=1 cnt=10 pid=0 pos=1 obj=0 op='MERGE JOIN  (cr=119 pr=0 pw=0 time=28 us cost=193 size=1280 card=10)'
STAT #22 id=2 cnt=1 pid=1 pos=1 obj=0 op='SORT JOIN (cr=15 pr=0 pw=0 time=0 us cost=6 size=63 card=1)'
STAT #22 id=3 cnt=1 pid=2 pos=1 obj=83550 op='TABLE ACCESS FULL T3 (cr=15 pr=0 pw=0 time=0 us cost=5 size=63 card=1)'
STAT #22 id=4 cnt=10 pid=1 pos=2 obj=0 op='SORT JOIN (cr=104 pr=0 pw=0 time=11 us cost=187 size=650000 card=10000)'
STAT #22 id=5 cnt=10000 pid=4 pos=1 obj=83552 op='TABLE ACCESS FULL T4 (cr=104 pr=0 pw=0 time=8603 us cost=29 size=650000 card=10000)'

从上面的 trace 文件我们可以看出 sql 语句经过了 parse(解析) -> exec(执行) -> fetch(从游标中获取数据) 几个过程, 其中第一句说明了当前跟踪执行的 sql 语句的概况,比如使用游标号, sql 语句的长度, 递归深度等等基本信息:

PARSING IN CURSOR #22 len=92 dep=0 uid=0 oct=3 lid=0 tim=900460923321 hv=1624778336 ad='34671d90' sqlid='g0rdyg9hdh9m0'

cursor	cursor number
len	sql 语句长度
dep	sql 语句递归深度
uid	user id
oct	oracle command type
lid	privilege user id
tim	timestamp，时间戳
hv	hash id
ad	sql address 地址, 用在 v$sqltext
sqlid	sql id

接着, 下面的语句说明了 sql 语句具体的执行过程以及每一个步骤消耗 CPU 的时间等性能指标

PARSE #22:c=0,e=10777,p=0,cr=0,cu=0,mis=1,r=0,dep=0,og=1,plh=3831111046,tim=900460923319
EXEC #22:c=0,e=29,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=1,plh=3831111046,tim=900460923482
FETCH #22:c=15625,e=23922,p=0,cr=119,cu=0,mis=0,r=1,dep=0,og=1,plh=3831111046,tim=900460947462

c	CPU 消耗的时间
e	Elapsed time
p	number of physical reads 物理读的次数
cr	number of buffers retrieved for CR reads   逻辑读的数据块
cu	number of buffers retrieved in current mode (current 模式读取的数据块)
mis	cursor missed in the cache 库缓存中丢失的游标, 硬解析次数
r	number of rows processed 处理的行数
dep	递归深度
og	optimizer mode 【1：all_rows, 2：first_rows, 3：rule, 4：choose】
plh	plan hash value
tim	timestamp 时间戳

以及执行过程中的发生的等待事件

WAIT #22: nam='SQL*Net message to client' ela= 2 driver id=1111838976 #bytes=1 p3=0 obj#=-1 tim=900460923512

nam	an event that we waited for 等待事件
ela	此操作消耗的时间
p3	block 块号
trm	timestamp 时间戳

最后显示的是该游标的执行计划

STAT #22 id=1 cnt=10 pid=0 pos=1 obj=0 op='MERGE JOIN  (cr=119 pr=0 pw=0 time=28 us cost=193 size=1280 card=10)'
STAT #22 id=2 cnt=1 pid=1 pos=1 obj=0 op='SORT JOIN (cr=15 pr=0 pw=0 time=0 us cost=6 size=63 card=1)'
STAT #22 id=3 cnt=1 pid=2 pos=1 obj=83550 op='TABLE ACCESS FULL T3 (cr=15 pr=0 pw=0 time=0 us cost=5 size=63 card=1)'
STAT #22 id=4 cnt=10 pid=1 pos=2 obj=0 op='SORT JOIN (cr=104 pr=0 pw=0 time=11 us cost=187 size=650000 card=10000)'
STAT #22 id=5 cnt=10000 pid=4 pos=1 obj=83552 op='TABLE ACCESS FULL T4 (cr=104 pr=0 pw=0 time=8603 us cost=29 size=650000 card=10000

cnt	当前行源返回的行数
pid	parent id of this row source 当前行源的父结点 id
pos	position in explain plan 执行计划的位置
obj	object id of row source (if this is a base object)
op	the row source access operation

例如, 执行步骤 merge join 消耗的逻辑读为 119, 物理读为 0, 耗费的时间为 28 us, 成本 cost 193，返回 10 条记录

使用 tkprof 命令翻译 trace 文件

我们也可以使用 tkprof 命令对 trace 文件进行翻译，得到一个容易理解的 trace 汇总报表文件

C:Documents and Settingsdavidd> tkprof g:appdavidddiag
dbmsdaviddavid	racedavid_ora_2176.trc d:	race.trc

TKPROF: Release 11.2.0.1.0 - Development on Thu Dec 18 18:51:44 2014

Copyright (c) 1982, 2009, Oracle and/or its affiliates.  All rights reserved.

tkprof 翻译的 trace 文件的汇总报表如下：

Trace file: g:appdavidddiag
dbmsdaviddavid	racedavid_ora_2176.trc
Sort options: default


********************************************************************************
count    = number of times OCI procedure was executed
cpu      = cpu time in seconds executing
elapsed  = elapsed time in seconds executing
disk     = number of physical reads of buffers from disk
query    = number of buffers gotten for consistent read
current  = number of buffers gotten in current mode (usually for update)
rows     = number of rows processed by the fetch or execute call
********************************************************************************


select /*+ leading(t3) use_merge(t4) */ *
from t3, t4
where t3.id = t4.t3_id and t3.n = 1100


call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.00       0.00          0          0          0           0
Execute      1      0.00       0.00          0          0          0           0
Fetch        2      0.00       0.00          0        119          0          10
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        4      0.00       0.00          0        119          0          10


Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: SYS


Rows     Row Source Operation
-------  ---------------------------------------------------
     10  MERGE JOIN  (cr=119 pr=0 pw=0 time=0 us cost=193 size=1280 card=10)
      1   SORT JOIN (cr=15 pr=0 pw=0 time=0 us cost=6 size=63 card=1)
      1    TABLE ACCESS FULL T3 (cr=15 pr=0 pw=0 time=0 us cost=5 size=63 card=1)
     10   SORT JOIN (cr=104 pr=0 pw=0 time=0 us cost=187 size=650000 card=10000)
  10000    TABLE ACCESS FULL T4 (cr=104 pr=0 pw=0 time=8733 us cost=29 size=650000 card=10000)




Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  SQL*Net message to client                       2        0.00          0.00
  SQL*Net message from client                     2       20.23         20.23






********************************************************************************


OVERALL TOTALS FOR ALL NON-RECURSIVE STATEMENTS


call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        1      0.00       0.00          0          0          0           0
Execute      1      0.00       0.00          0          0          0           0
Fetch        2      0.00       0.00          0        119          0          10
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        4      0.00       0.00          0        119          0          10


Misses in library cache during parse: 1


Elapsed times include waiting on following events:
  Event waited on                             Times   Max. Wait  Total Waited
  ----------------------------------------   Waited  ----------  ------------
  SQL*Net message to client                       3        0.00          0.00
  SQL*Net message from client                     3       20.23         30.20




OVERALL TOTALS FOR ALL RECURSIVE STATEMENTS


call     count       cpu    elapsed       disk      query    current        rows
------- ------  -------- ---------- ---------- ---------- ----------  ----------
Parse        0      0.00       0.00          0          0          0           0
Execute      0      0.00       0.00          0          0          0           0
Fetch        0      0.00       0.00          0          0          0           0
------- ------  -------- ---------- ---------- ---------- ----------  ----------
total        0      0.00       0.00          0          0          0           0


Misses in library cache during parse: 0


    1  user  SQL statements in session.
    0  internal SQL statements in session.
    1  SQL statements in session.
********************************************************************************
Trace file: g:appdavidddiag
dbmsdaviddavid	racedavid_ora_2176.trc
Trace file compatibility: 11.1.0.7
Sort options: default


       1  session in tracefile.
       1  user  SQL statements in trace file.
       0  internal SQL statements in trace file.
       1  SQL statements in trace file.
       1  unique SQL statements in trace file.
     122  lines in trace file.
       0  elapsed seconds in trace file.

其中，Misses in library cache during parse ：1   意思是解析的时候库缓存丢失游标,  也就是说发生了一次硬解析