[20181119]sql语句执行缓慢分析.txt
--//上个星期在优化时遇到sql语句执行缓慢的问题,百思不知其解,做一个记录.
--//我们生产系统有一个表索引有点多,肯定一些索引不需要,我想删除一些不必要的索引(可以先invisiable).再观察情况.
1.环境:
zzzzzz> @ ver1
PORT_STRING VERSION BANNER
------------------------------ -------------- --------------------------------------------------------------------------------
x86_64/Linux 2.4.xx 11.2.0.3.0 Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production
2.执行脚本如下:
$ cat a.txt
SELECT /*+ &&2 */ owner,
index_name,
-- index_type,
table_NAME,
COMPRESSION,
UNIQUENESS,
distinct_keys,
AVG_LEAF_blocks_per_key,
AVG_DATA_BLOCKS_PER_KEY,
blevel,
leaf_blocks,
clustering_factor,
status,
num_rows,
sample_size,
last_analyzed
FROM dba_indexes
WHERE owner LIKE 'XXXXXX_YYY'
AND table_name = UPPER ('&&1')
AND (owner, index_name) IN (SELECT owner, index_name
FROM dba_indexes
WHERE owner LIKE 'XXXXXX_YYY'
AND index_type <> 'LOB'
MINUS
(SELECT object_owner, object_name
FROM v$sql_plan
WHERE object_type = 'INDEX'
AND object_owner LIKE 'XXXXXX_YYY'
UNION
SELECT object_owner, object_name
FROM dba_hist_sql_plan
WHERE object_type = 'INDEX'
AND object_owner LIKE 'XXXXXX_YYY')
MINUS
(SELECT owner, constraint_name
FROM dba_constraints
WHERE owner LIKE 'XXXXXX_YYY'
AND constraint_type IN ('P',
'U',
'R')));
--//主要目的先排除一些现在使用的索引.
3.执行:
--//session 1:
zzzzzz> alter session set statistics_level=all;
Session altered.
zzzzzz> @ a.txt MS_CF01 aaa
...
--//慢.等。。。。
--//session 1,按ctrl+c中断执行,实际上没有执行完成:
zzzzzz> @ dpc '' ''
PLAN_TABLE_OUTPUT
-------------------------------------
SQL_ID frzb5c8tjx9wz, child number 0
...
Plan hash value: 2331593202
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Starts | E-Rows |E-Bytes| Cost (%CPU)| E-Time | A-Rows | A-Time | Buffers | Reads | OMem | 1Mem | Used-Mem |
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | | | 485 (100)| | 1 |00:01:42.24 | 2568K| 93 | | | |
| 1 | NESTED LOOPS OUTER | | 1 | 1 | 179 | 145 (0)| 00:00:02 | 1 |00:01:42.24 | 2568K| 93 | | | |
| 2 | NESTED LOOPS OUTER | | 1 | 1 | 168 | 144 (0)| 00:00:02 | 1 |00:01:42.24 | 2568K| 93 | | | |
| 3 | NESTED LOOPS OUTER | | 1 | 1 | 165 | 143 (0)| 00:00:02 | 1 |00:01:42.24 | 2568K| 93 | | | |
... 太长...
| 38 | SORT UNIQUE | | 3490 | 1 | 375 | 340 (97)| 00:00:05 | 126 |00:03:18.78 | 4909K| 15 | 2048 | 2048 | |
| 39 | UNION-ALL | | 3490 | | | | | 1461 |00:03:18.73 | 4909K| 15 | | | |
|* 40 | FIXED TABLE FULL | X$KQLFXPL | 3490 | 1 | 22 | 3 (100)| 00:00:01 | 826 |00:02:58.63 | 0 | 0 | | | |
|* 41 | TABLE ACCESS FULL | WRH$_SQL_PLAN | 3489 | 1 | 17 | 308 (1)| 00:00:04 | 635 |00:00:20.08 | 4909K| 15 | | | |
--//很明显主要慢在ID=40,41,而实际上主要是循环次数3490导致执行缓慢.
--//很奇怪的是我在toad下很快完成.第1次执行需要实际长一点外,以后都很快.再次查看执行计划
--//注不能在toad下看,toad看仅仅看到child_number=0的执行计划.
zzzzzz> @ dpc frzb5c8tjx9wz ''
....
SQL_ID frzb5c8tjx9wz, child number 1
-------------------------------------
Plan hash value: 1956464785
---------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | E-Rows |E-Bytes| Cost (%CPU)| E-Time | OMem | 1Mem | Used-Mem |
---------------------------------------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | | | 570 (100)| | | | |
|* 1 | FILTER | | | | | | | | |
| 2 | NESTED LOOPS OUTER | | 1 | 179 | 230 (0)| 00:00:03 | | | |
| 3 | NESTED LOOPS OUTER | | 1 | 168 | 229 (0)| 00:00:03 | | | |
| 4 | NESTED LOOPS OUTER | | 1 | 164 | 228 (0)| 00:00:03 | | | |
| 5 | NESTED LOOPS | | 1 | 156 | 227 (0)| 00:00:03 | | | |
....
| 50 | UNION-ALL | | | | | | | | |
|* 51 | FIXED TABLE FULL | X$KQLFXPL | 1 | 22 | 3 (100)| 00:00:01 | | | |
|* 52 | TABLE ACCESS FULL | WRH$_SQL_PLAN | 1 | 17 | 308 (1)| 00:00:04 | | | |
| 53 | SORT UNIQUE NOSORT | | 1 | 211 | 15 (7)| 00:00:01 | | | |
--//开始想是正常的,因为我在sqlplus下设置了alter session set statistics_level=all;.但是仔细看仔细计划对于的操作id不一样,而且Plan hash value也不同.
--//我也尝试在toad下加入alter session set statistics_level=all;按F9执行也会执行缓慢.
zzzzzz> @ share frzb5c8tjx9wz
SQL_TEXT = SELECT /*+ aaa */ owner, index_name, -- index_type, table_NAME, COMPRESSION, UNIQUENESS, distinct_keys, AVG_LEAF_blocks_per_key, AVG_DATA_BLOCKS_PER_KEY,
blevel, leaf_blocks, clustering_factor, status, num_rows, sample_size, last_analyzed FROM dba_indexes WHERE owner LIKE :"SYS_B_00" AND table_name = UPPER (:"SYS_B_01") AND (owner,
index_name) IN (SELECT owner, index_name FROM dba_indexes WHERE owner LIKE :"SYS_B_02" AND index_type <> :"SYS_B_03"
MINUS (SELECT object_owner, object_name FROM v$sql_plan WHERE object_type = :"SYS_B_04" AND
object_owner LIK
SQL_ID = frzb5c8tjx9wz
ADDRESS = 00000000AA459A38
CHILD_ADDRESS = 00000000AB1E7238
CHILD_NUMBER = 0
REASON = <ChildNode><ChildNumber>0</ChildNumber><ID>45</ID><reason>NLS
Settings(0)</reason><size>2x4</size><SessionLengthSemantics>0</SessionLengthSemantics><CursorLengthSemantics>0</CursorLengthSemantics></ChildNode><ChildNode><ChildNumber>0</ChildNumber><ID>3</ID><reason>Optimizer
mismatch(12)</reason><size>2x216</size><statistics_level> all typical </statistics_level></ChildNode>
--------------------------------------------------
SQL_TEXT = SELECT /*+ aaa */ owner, index_name, -- index_type, table_NAME, COMPRESSION, UNIQUENESS, distinct_keys, AVG_LEAF_blocks_per_key, AVG_DATA_BLOCKS_PER_KEY,
blevel, leaf_blocks, clustering_factor, status, num_rows, sample_size, last_analyzed FROM dba_indexes WHERE owner LIKE :"SYS_B_00" AND table_name = UPPER (:"SYS_B_01") AND (owner,
index_name) IN (SELECT owner, index_name FROM dba_indexes WHERE owner LIKE :"SYS_B_02" AND index_type <> :"SYS_B_03"
MINUS (SELECT object_owner, object_name FROM v$sql_plan WHERE object_type = :"SYS_B_04" AND
object_owner LIK
SQL_ID = frzb5c8tjx9wz
ADDRESS = 00000000AA459A38
CHILD_ADDRESS = 00000000AB3775D8
CHILD_NUMBER = 1
OPTIMIZER_MISMATCH = Y
REASON = <ChildNode><ChildNumber>1</ChildNumber><ID>3</ID><reason>Optimizer mismatch(12)</reason><size>2x216</size><statistics_level> typical all </statistics_level></ChildNode>
--//我发现一个奇特的情况,sqlplus执行的语句和toad下执行的语句不能共享,提示LANGUAGE_MISMATCH= Y.
zzzzzz> select sql_id,child_number,PLAN_HASH_VALUE,buffer_gets,executions,ELAPSED_TIME from v$sql where sql_id='frzb5c8tjx9wz';
SQL_ID CHILD_NUMBER PLAN_HASH_VALUE BUFFER_GETS EXECUTIONS ELAPSED_TIME
------------- ------------ --------------- ----------- ---------- ------------
frzb5c8tjx9wz 0 2331593202 4978087 1 201931031
frzb5c8tjx9wz 1 1956464785 30458 2 1486831
--//对比2者NLS*环境:
zzzzzz> select * from nls_session_parameters ;
PARAMETER VALUE
-------------------------- ----------------------------------
NLS_LANGUAGE AMERICAN
NLS_TERRITORY AMERICA
NLS_CURRENCY $
NLS_ISO_CURRENCY AMERICA
NLS_NUMERIC_CHARACTERS .,
NLS_CALENDAR GREGORIAN
NLS_DATE_FORMAT YYYY-MM-DD HH24:MI:SS
NLS_DATE_LANGUAGE AMERICAN
NLS_SORT BINARY
NLS_TIME_FORMAT HH.MI.SSXFF AM
NLS_TIMESTAMP_FORMAT YYYY-MM-DD HH24:MI:SS.FF
NLS_TIME_TZ_FORMAT HH24.MI.SSXFF TZH:TZM
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
NLS_TIMESTAMP_TZ_FORMAT YYYY-MM-DD HH24:MI:SS.FF TZH:TZM
NLS_DUAL_CURRENCY $
NLS_COMP BINARY
NLS_LENGTH_SEMANTICS BYTE
NLS_NCHAR_CONV_EXCP FALSE
17 rows selected.
--//toad下执行:
select * from nls_session_parameters ;
PARAMETER,VALUE
NLS_LANGUAGE,AMERICAN
NLS_TERRITORY,AMERICA
NLS_CURRENCY,$
NLS_ISO_CURRENCY,AMERICA
NLS_NUMERIC_CHARACTERS,.,
NLS_CALENDAR,GREGORIAN
NLS_DATE_FORMAT,YYYY-MM-DD HH24:MI:SS
NLS_DATE_LANGUAGE,AMERICAN
NLS_SORT,BINARY
NLS_TIME_FORMAT,HH.MI.SSXFF AM
NLS_TIMESTAMP_FORMAT,YYYY-MM-DD HH24:MI:SS.FF
NLS_TIME_TZ_FORMAT,HH.MI.SSXFF AM TZR
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
NLS_TIMESTAMP_TZ_FORMAT,YYYY-MM-DD HH24:MI:SS.FF TZH:TZM
NLS_DUAL_CURRENCY,$
NLS_COMP,BINARY
NLS_LENGTH_SEMANTICS,BYTE
NLS_NCHAR_CONV_EXCP,FALSE
--//注意下划线,仅仅参数NLS_TIME_TZ_FORMAT不同.
$ export NLS_TIME_TZ_FORMAT='HH.MI.SSXFF AM TZR'
--//再次登录,问题消失.
alter session set statistics_level=all;
@ a.txt MS_CF01 aaa
--//很慢.登录后执行:
@ a.txt MS_CF01 aaa
--//可以发现很快,因为我在toad下已经执行过一次.
--//真心想不到这个参数,会导致这么奇怪的问题.
--//补充说明一下:
--//如果写成'HH24.MI.SSXFF AM TZR',调用sqlplus报错.
$ export NLS_TIME_TZ_FORMAT='HH24.MI.SSXFF AM TZR'
$ rlsql system/test12@192.168.31.8:1521/hrp430
ERROR:
ORA-01818: 'HH24' precludes use of meridian indicator
SP2-0152: ORACLE may not be functioning properly
ORA-24550: signal received: [si_signo=11] [si_errno=0] [si_code=1] [si_int=0] [si_ptr=(nil)] [si_addr=0x10]
kpedbg_dmp_stack()+362<-kpeDbgCrash()+192<-kpeDbgSignalHandler()+119<-skgesig_sigactionHandler()+218<-__sighandler()<-pthread_mutex_destroy()<-kpufhndl0()+5461<-kpufhndl()+11<-OCIHandleFree()+19<-afioci()+2372<-aficnthdf()+62<-aficexf()+185<-aficex()+370<-afidrv()+321<-frame_dummy()+123<-__libc_start_main()+244
rlwrap: warning: sqlplus crashed, killed by SIGSEGV.
rlwrap itself has not crashed, but for transparency,
it will now kill itself (without dumping core) with the same signal
warnings can be silenced by the --no-warnings (-n) option
Segmentation fault
--//多次尝试,确认实际上这个问题并不是参数NLS_TIME_TZ_FORMAT导致问题,而是执行计划不稳定.每次分析都有可能导致新的执行计划.
--//参数NLS_TIME_TZ_FORMAT不同,仅仅导致产生新的子光标,重新生成的执行计划.两者混合在一起导致问题非常怪异.
--//包括设置alter session set statistics_level=all;
--//如果我改写成如下,无论什么情况都是很快完成.
WITH a
AS (SELECT /*+ MATERIALIZE */ object_owner, object_name
FROM v$sql_plan
WHERE object_type = 'INDEX' AND object_owner LIKE 'XXXXXX_YYY')
SELECT /*+ &&2 */ owner
,index_name
,--index_type,
table_NAME
,COMPRESSION
,UNIQUENESS
,distinct_keys
,AVG_LEAF_blocks_per_key
,AVG_DATA_BLOCKS_PER_KEY
,blevel
,leaf_blocks
,clustering_factor
,status
,num_rows
,sample_size
,last_analyzed
FROM dba_indexes
WHERE owner LIKE 'XXXXXX_YYY'
AND table_name = UPPER ('&&1')
AND (owner, index_name) IN (SELECT owner, index_name
FROM dba_indexes
WHERE owner LIKE 'XXXXXX_YYY'
AND index_type <> 'LOB'
MINUS
(SELECT object_owner, object_name FROM a
UNION
SELECT object_owner, object_name
FROM dba_hist_sql_plan
WHERE object_type = 'INDEX'
AND object_owner LIKE 'XXXXXX_YYY')
MINUS
(SELECT owner, constraint_name
FROM dba_constraints
WHERE owner LIKE 'XXXXXX_YYY'
AND constraint_type IN ('P'
,'U'
,'R')));
zzzzzz> @ aaa.txt MS_CF01 aaa
zzzzzz> select sql_id,child_number,PLAN_HASH_VALUE,buffer_gets,executions,ELAPSED_TIME from v$sql where sql_id='4zapbf2u32kx2';
SQL_ID CHILD_NUMBER PLAN_HASH_VALUE BUFFER_GETS EXECUTIONS ELAPSED_TIME
------------- ------------ --------------- ----------- ---------- ------------
4zapbf2u32kx2 0 2384590172 9680 1 903877
4zapbf2u32kx2 1 115948183 11455 1 847278
zzzzzz> @ share 4zapbf2u32kx2
SQL_TEXT = with a as ( SELECT /*+ MATERIALIZE */ object_owner, object_name FROM v$sql_plan WHERE object_type = :"SYS_B_00" AND object_owner LIKE :"SYS_B_01" ) SELECT /*+ aaa */ owner, index_name, --index_type,
table_NAME, COMPRESSION, UNIQUENESS, distinct_keys, AVG_LEAF_blocks_per_key, AVG_DATA_BLOCKS_PER_KEY, blevel, leaf_blocks, clustering_factor, status, num_rows, sample_size,
last_analyzed FROM dba_indexes WHERE owner LIKE :"SYS_B_02" AND table_name = UPPER (:"SYS_B_03") AND (owner, index_name) IN (SELECT owner, index_name FROM dba_indexes
WHERE owner LIKE :"SYS_B_04" AND index_type <> :"SYS_B_05" MINUS (SELECT object_owner, object_name FROM a
UN
SQL_ID = 4zapbf2u32kx2
ADDRESS = 00000000A6303098
CHILD_ADDRESS = 00000000A68317F8
CHILD_NUMBER = 0
USE_FEEDBACK_STATS = Y
REASON = <ChildNode><ChildNumber>0</ChildNumber><ID>3</ID><reason>Optimizer mismatch(13)</reason><size>3x4</size><kxscflg>32</kxscflg><kxscfl4>4194560</kxscfl4><dnum_kksfcxe>262144</dnum_kksfcxe></ChildNode>
--------------------------------------------------
SQL_TEXT = with a as ( SELECT /*+ MATERIALIZE */ object_owner, object_name FROM v$sql_plan WHERE object_type = :"SYS_B_00" AND object_owner LIKE :"SYS_B_01" ) SELECT /*+ aaa */ owner, index_name, --index_type,
table_NAME, COMPRESSION, UNIQUENESS, distinct_keys, AVG_LEAF_blocks_per_key, AVG_DATA_BLOCKS_PER_KEY, blevel, leaf_blocks, clustering_factor, status, num_rows, sample_size,
last_analyzed FROM dba_indexes WHERE owner LIKE :"SYS_B_02" AND table_name = UPPER (:"SYS_B_03") AND (owner, index_name) IN (SELECT owner, index_name FROM dba_indexes
WHERE owner LIKE :"SYS_B_04" AND index_type <> :"SYS_B_05" MINUS (SELECT object_owner, object_name FROM a
UN
SQL_ID = 4zapbf2u32kx2
ADDRESS = 00000000A6303098
CHILD_ADDRESS = 00000000A841E528
CHILD_NUMBER = 1
REASON =
--------------------------------------------------
PL/SQL procedure successfully completed.
--//不过不能共享的原因是是USE_FEEDBACK_STATS=Y.
总结:
1.环境变量,执行时参数不同导致产生新的子光标,出现一些怪异的情况,感觉toad下执行快,而sqlplus下执行慢.实际上反过来的情况也出
现过.我估计cardinality FEEDBACK的影响.
2.对于复杂视图的执行计划先 MATERIALIZE也许能获得好的效果.
3.我发现一个奇怪的情况就是实际上打开alter session set statistics_level=all;就会选择慢的执行计划.
--//sqlplus开始第1次执行(硬分析)总是选择慢的执行计划.真心搞不懂....暂时放一下.
来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/267265/viewspace-2220517/，如需转载，请注明出处，否则将追究法律责任。
转载于:http://blog.itpub.net/267265/viewspace-2220517/
234
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
