oracle 11.2.0.2数据库碰到了ora-00600错误，ORA-00600: 内部错误代码, 参数: [kkoipt:invalid join method],是一个bug，查询metalink发现，居然目前唯一确认影响的版本就是11.2.0.2，这个Bug的描述为Bug 12591252 Query with ORDER BY fails with ORA-600 [
oracle 11.2.0.2数据库碰到了ora-00600错误。
详细错误信息如下：
Fri Sep 16 15:23:52 2011
Errors in file /u01/diag/rdbms/ora1/ora1/trace/ora1_ora_20382140.trc (incident=169704):
ORA-00600: 内部错误代码, 参数: [kkoipt:invalid join method], [1], [0], [], [], [], [], [], [], [], [], []
Incident details in: /u01/diag/rdbms/ora1/ora1/incident/incdir_169704/ora1_ora_20382140_i169704.trc
Fri Sep 16 15:24:00 2011
Dumping diagnostic data in directory=[cdmp_20110916152400], requested by (instance=1, sid=20382140), summary=[incident=169704].
Use ADRCI or Support Workbench to package the incident.
See Note 411.1 at My Oracle Support for error and packaging details.
而对应的详细TRACE如下：
bash-3.2$ more /u01/diag/rdbms/ora1/ora1/incident/incdir_169704/ora1_ora_20382140_i169704.trc
Dump file /u01/diag/rdbms/ora1/ora11/incident/incdir_169704/ora1_ora_20382140_i169704.trc
Oracle Database 11g Enterprise Edition Release 11.2.0.2.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options
ORACLE_HOME = /u01/product/11.2.0/dbhome_1
System name: AIX
Node name: node1
Release: 1
Version: 6
Machine: 00F6CD264C00
Instance name: ora1
Redo thread mounted by this instance: 1
Oracle process number: 193
Unix process pid: 20382140, image: oracle@s180
*** 2011-09-16 15:23:52.275
*** SESSION ID:(29.7169) 2011-09-16 15:23:52.275
*** CLIENT ID:() 2011-09-16 15:23:52.275
*** SERVICE NAME:(ora1) 2011-09-16 15:23:52.275
*** MODULE NAME:(TOAD 10.5.0.41) 2011-09-16 15:23:52.275
*** ACTION NAME:() 2011-09-16 15:23:52.275
Dump continued from file: /u01/diag/rdbms/ora1/ora1/trace/ora1_ora_20382140.trc
ORA-00600: 内部错误代码, 参数: [kkoipt:invalid join method], [1], [0], [], [], [], [], [], [], [], [], []
========= Dump for incident 169704 (ORA 600 [kkoipt:invalid join method]) ========
*** 2011-09-16 15:23:52.336
dbkedDefDump(): Starting incident default dumps (flags=0x2, level=3, mask=0x0)
----- Current SQL Statement for this session (sql_id=7ukzmn3p6zby6) -----
SELECT *
FROM (
SELECT a.childpolicyno, a.policyno, e.exportno corpno, a.effectdate,
TO_CHAR (a.lapsedate, 'yyyy-mm-dd') lapsedate, e.chnname,
e.address, e.keyflag, e.specialflag, g.empname,
g.secdeptname, f.TYPE, ROWNUM AS ID
FROM t_schildpolicy a,
t_spolicy f,
t_exportcorp e,
t_nodeinfo d,
v_employeecustomer g
WHERE f.policyno = a.policyno
AND f.insurantno = e.exportno
AND f.policyno = g.productid(+)
AND d.nodeid = f.nodeid
AND e.chnname LIKE '%%'
AND ((d.corpid = '3502'))
AND ROWNUM <= 10
ORDER BY a.policyno)
WHERE ID BETWEEN 1 AND 10
----- Call Stack Trace -----
calling call entry argument values in hex
location type point (? means dubious value)
-------------------- -------- -------------------- ----------------------------
skdstdst()+40 bl 107b6e01c FFFFFFFFFFECCA8 ? 000002004 ?
000000001 ? 000000003 ?
000000000 ? 000000002 ?
000000001 ? 000000000 ?
ksedst1()+104 call skdstdst() FFFFFFFFFFEBCB0 ? 000002004 ?
110A597A0 ? 10A027B2C ?
110A597A0 ? 000000000 ?
FFFFFFFFFFEBDE0 ? 700000007 ?
ksedst()+40 call ksedst1() 3030000000000 ? 002050033 ?
10A027B20 ? 700000000025C ?
000000000 ? 000000000 ?
10A027180 ? 000000000 ?
dbkedDefDump()+2828 call ksedst() FFFFFFFFFFEBE90 ? 000000000 ?
000000000 ? 000000000 ?
000000000 ? 000000000 ?
000000000 ? 300000003 ?
ksedmp()+76 call dbkedDefDump() 310A597A0 ? 1100010C8 ?
FFFFFFFFFFEC490 ?
28444040FFFEC66C ?
100148568 ? 1096635A8 ?
FFFFFFFFFFEC4E0 ? 11064B598 ?
ksfdmp()+88 call ksedmp() 000000000 ? 000000000 ?
0096635C3 ? 109CB2C50 ?
200000000000000 ? 000000000 ?
110C221E8 ? 110A597A0 ?
dbgexPhaseII()+1212 call ksfdmp() 000002004 ? 110A597A0 ?
000000000 ? FFFFFFFFFFEC658 ?
FFFFFFFFFFEC580 ?
FFFFFFFFFFECCA8 ? 1001D04B8 ?
110C221E8 ?
dbgexProcessError() call dbgexPhaseII() 110A597A0 ? 110C203F8 ?
+3604 0000296E8 ? 200000000 ?
FFFFFFFFFFED258 ? 00000006A ?
000000000 ? 000000000 ?
dbgeExecuteForError call dbgexProcessError() 110A597A0 ? 110C221E8 ?
()+72 100000000 ? 000000000 ?
110D01C88 ? 000000000 ?
110CADC78 ? 110C23F30 ?
dbgePostErrorKGE()+ call dbgeExecuteForError FFFFFFFFFFF0830 ?
1152 () B7417335409B9B1B ?
FFFFFFFFFFF06F0 ? 0409B9800 ?
10524EA10 ?
2147AE154168E65F ?
10524EA10 ? 000000000 ?
dbkePostKGE_kgsf()+ call dbgePostErrorKGE() 002050000 ? 001160000 ?
64 25810001330 ? 000000000 ?
110001330 ? FFFFFFFFFFF1510 ?
109613CD0 ? 110CB3F18 ?
这个SQL错误的奇特之处在于，如果将ORDER BY语句中的a.policyno变成其他列，则不会导致错误，甚至是换成与之相等关联的f.policyno，同样不会报错。
这显然是一个bug，查询metalink发现，居然目前唯一确认影响的版本就是11.2.0.2，这个Bug的描述为Bug 12591252 Query with ORDER BY fails with ORA-600 [kkoipt:invalid join method]。而Oracle的解释是，优化器试图通过索引来避免排序操作时，可能引发这个bug。这正好说明了为什么SQL中换成其他的列就不会导致错误。
目前这个bug在Windows平台的11.2.0.2的patch 10中被fixed，其他平台只能通过单独的12591252补丁来解决这个问题。当然如果能通过改写SQL来实现这个目的，无疑代价是最小的。
声明： 此文观点不代表本站立场；转载须要保留原文链接；版权疑问请联系我们。
