外键加索引是常识，必须牢记。本来不想写这样的简单案例。可是连续遇到好几起外键不加索引导致性能问题，所以还是写一下。

一个兄弟问我 delete from Sa_Sales_Comm_Detail s where s.sales_commission_id=24240;  ---删除105条数据很慢。要跑几十秒到上百秒

这个表总数据才35万行，sales_commission_id 列有索引，运行计划也确实是走了索引。 走索引返回105 条数据。不可能跑几十秒跑上百秒的。

之后我问他 select * from Sa_Sales_Comm_Detail s where s.sales_commission_id=24240;   这个跑得慢吗。他回答 0.06秒。

select * from Sa_Sales_Comm_Detail s where s.sales_commission_id=24240;  ---0.06秒

delete from Sa_Sales_Comm_Detail s where s.sales_commission_id=24240;   ---几十秒几百秒

遇到这样的，直接做10046 trace，部分的trace文件例如以下

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

PARSING IN CURSOR #4 len=111 dep=1 uid=0 oct=3 lid=0 tim=1374414810328412 hv=4234506700 ad='99cc8678'

 select /*+ all_rows */ count(1) from "CMM"."SA_SALES_PER_SPLIT_DETAIL" where "SALES_COMMISSION_DETAIL_ID" = :1

END OF STMT

PARSE #4:c=4000,e=4619,p=0,cr=0,cu=2,mis=1,r=0,dep=1,og=1,tim=1374414810328406

EXEC #4:c=999,e=1841,p=0,cr=0,cu=0,mis=1,r=0,dep=1,og=1,tim=1374414810330416

FETCH #4:c=226965,e=221384,p=0,cr=6867,cu=0,mis=0,r=1,dep=1,og=1,tim=1374414810551844

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

PARSING IN CURSOR #2 len=106 dep=1 uid=0 oct=3 lid=0 tim=1374414810557316 hv=1936840180 ad='8ae35660'

 select /*+ all_rows */ count(1) from "CMM"."SA_SALES_COMM_REPROT" where "SALES_COMMISSION_DETAIL_ID" = :1

END OF STMT

PARSE #2:c=5000,e=5152,p=0,cr=0,cu=2,mis=1,r=0,dep=1,og=1,tim=1374414810557310

EXEC #2:c=3000,e=2081,p=0,cr=0,cu=0,mis=1,r=0,dep=1,og=1,tim=1374414810559571

WAIT #2: nam='db file scattered read' ela= 384 file#=5 block#=83604 blocks=5 obj#=104150 tim=1374414810560401

WAIT #2: nam='db file scattered read' ela= 154 file#=5 block#=87785 blocks=8 obj#=104150 tim=1374414810561125

WAIT #2: nam='db file scattered read' ela= 141 file#=5 block#=87802 blocks=7 obj#=104150 tim=1374414810561384

WAIT #2: nam='db file scattered read' ela= 204 file#=5 block#=94633 blocks=8 obj#=104150 tim=1374414810561704

WAIT #2: nam='db file scattered read' ela= 160 file#=5 block#=94642 blocks=7 obj#=104150 tim=1374414810562029

WAIT #2: nam='db file scattered read' ela= 167 file#=5 block#=94649 blocks=8 obj#=104150 tim=1374414810562309

WAIT #2: nam='db file scattered read' ela= 184 file#=5 block#=94674 blocks=7 obj#=104150 tim=1374414810562602

WAIT #2: nam='db file scattered read' ela= 172 file#=5 block#=94705 blocks=8 obj#=104150 tim=1374414810562882

WAIT #2: nam='db file scattered read' ela= 135 file#=5 block#=94714 blocks=7 obj#=104150 tim=1374414810563132

WAIT #2: nam='db file scattered read' ela= 162 file#=5 block#=97529 blocks=8 obj#=104150 tim=1374414810563412

WAIT #2: nam='db file scattered read' ela= 234 file#=5 block#=98314 blocks=7 obj#=104150 tim=1374414810563758

WAIT #2: nam='db file scattered read' ela= 154 file#=5 block#=99513 blocks=8 obj#=104150 tim=1374414810564008

WAIT #2: nam='db file scattered read' ela= 143 file#=5 block#=101666 blocks=7 obj#=104150 tim=1374414810564260

WAIT #2: nam='db file scattered read' ela= 166 file#=5 block#=101681 blocks=8 obj#=104150 tim=1374414810564533

WAIT #2: nam='db file scattered read' ela= 157 file#=5 block#=101690 blocks=7 obj#=104150 tim=1374414810564797

WAIT #2: nam='db file scattered read' ela= 128 file#=5 block#=101697 blocks=8 obj#=104150 tim=1374414810565025

WAIT #2: nam='db file scattered read' ela= 335 file#=5 block#=102027 blocks=16 obj#=104150 tim=1374414810565576

WAIT #2: nam='db file scattered read' ela= 355 file#=5 block#=102043 blocks=16 obj#=104150 tim=1374414810566148

WAIT #2: nam='db file scattered read' ela= 302 file#=5 block#=102059 blocks=16 obj#=104150 tim=1374414810566690

WAIT #2: nam='db file scattered read' ela= 323 file#=5 block#=102075 blocks=16 obj#=104150 tim=1374414810567221

WAIT #2: nam='db file scattered read' ela= 310 file#=5 block#=102091 blocks=16 obj#=104150 tim=1374414810567720

WAIT #2: nam='db file scattered read' ela= 270 file#=5 block#=102107 blocks=16 obj#=104150 tim=1374414810568243

WAIT #2: nam='db file scattered read' ela= 378 file#=5 block#=102123 blocks=16 obj#=104150 tim=1374414810568814

WAIT #2: nam='db file scattered read' ela= 253 file#=5 block#=102139 blocks=14 obj#=104150 tim=1374414810569252

WAIT #2: nam='db file scattered read' ela= 527 file#=5 block#=108043 blocks=16 obj#=104150 tim=1374414810570016

WAIT #2: nam='db file scattered read' ela= 309 file#=5 block#=108059 blocks=16 obj#=104150 tim=1374414810570543

WAIT #2: nam='db file scattered read' ela= 281 file#=5 block#=108075 blocks=16 obj#=104150 tim=1374414810571075

WAIT #2: nam='db file scattered read' ela= 356 file#=5 block#=108091 blocks=16 obj#=104150 tim=1374414810571658

WAIT #2: nam='db file scattered read' ela= 273 file#=5 block#=108107 blocks=16 obj#=104150 tim=1374414810572138

WAIT #2: nam='db file scattered read' ela= 381 file#=5 block#=108123 blocks=16 obj#=104150 tim=1374414810572715

WAIT #2: nam='db file scattered read' ela= 318 file#=5 block#=108139 blocks=16 obj#=104150 tim=1374414810573241

WAIT #2: nam='db file scattered read' ela= 302 file#=5 block#=108155 blocks=14 obj#=104150 tim=1374414810573745

WAIT #2: nam='db file scattered read' ela= 280 file#=5 block#=109195 blocks=16 obj#=104150 tim=1374414810574226

WAIT #2: nam='db file scattered read' ela= 362 file#=5 block#=109211 blocks=16 obj#=104150 tim=1374414810574795

WAIT #2: nam='db file scattered read' ela= 333 file#=5 block#=109227 blocks=16 obj#=104150 tim=1374414810575357

WAIT #2: nam='db file scattered read' ela= 331 file#=5 block#=109243 blocks=16 obj#=104150 tim=1374414810575904

WAIT #2: nam='db file scattered read' ela= 377 file#=5 block#=109259 blocks=16 obj#=104150 tim=1374414810576483

WAIT #2: nam='db file scattered read' ela= 349 file#=5 block#=109275 blocks=16 obj#=104150 tim=1374414810577059

WAIT #2: nam='db file scattered read' ela= 344 file#=5 block#=109291 blocks=16 obj#=104150 tim=1374414810577601

WAIT #2: nam='db file scattered read' ela= 320 file#=5 block#=109307 blocks=14 obj#=104150 tim=1374414810578133

WAIT #2: nam='db file scattered read' ela= 385 file#=7 block#=2699 blocks=16 obj#=104150 tim=1374414810578830

从trace文件中面看到。运行delete的时候。隐含的调用了

select /*+ all_rows */ count(1) from "CMM"."SA_SALES_COMM_REPROT" where "SALES_COMMISSION_DETAIL_ID" = :1

这个sql 导致了 db file scattered read(多块读) ，也就是说 上面的sql 没走索引，走了全表扫描 ，于是让他在 SALES_COMMISSION_DETAIL_ID 列上面建立索引

建立索引之后，delete仅仅须要0.几秒就可以。

反正记住。外键上面要建立索引。