ORA-07445，帮忙分析一下，谢谢

stephen

请教飞总，一个ORA-07445的错,谢谢，根据错误堆栈，之前搜过MOS Doc

ID 378334.1，按照说明该了

"_optimizer_push_pred_cost_based" = FALSE，现在还是报了7445的错误，查了几次错误的trace，都是由update语句引起的，log如附档，帮我分析一下，多谢多谢



环境：
AIX 6100-07 64bit


oracle：10.2.0.4  4节点RAC，没有打补丁

xifenfei

[size=14.3000001907349px]ORA-07445 [_memcpy] And/Or [_memmove] And/Or ORA-07445 [kgidum] And/Or [kgscDump]during updates in 9.2.0.8 or 10.2.0.4 (Doc ID 729206.1)

To Bottom

---

In this Document Symptoms Changes Cause Solution References APPLIES TO:Oracle Database - Enterprise Edition - Version 10.2.0.4 and later Information in this document applies to any platform. ***Checked for relevance on 17-Nov-2011*** SYMPTOMSYou could be getting either of the following errors in the alert.log or all of them while running an update statement: ORA-07445: exception encountered: core dump [_memmove()+2369] [SIGSEGV] [Address not mapped to object] [0x9FFFFFFFBF200000] [] [] ORA-07445: exception encountered: core dump [kgidum()+1536] [SIGSEGV] [Address not mapped to object] [0x9FFFFFFFBF200000] [] [] ORA-07445: exception encountered: core dump [kgscDump()+1376] [SIGSEGV] [Address not mapped to object] [0xB38F00000000B1] [] ORA-7445: exception encountered: core dump [_memcpy()+2369] [SIGSEGV] [Invalid permissions for mapped object] [0x9FFFFFFFBF448028] [] [] ORA-07445: exception encountered: core dump[_intel_fast_memcpy.A()+10] [SIGSEGV] [Invalid permissions for mapped object] [0x2A9768C000] [] [] The call stack extracted from the trace file for the above error(s) will look similar to the one below: updgrh upduaw kdusru kauupd updrow qerupRowProcedure qerupFetch updaul updThreePhaseExe updexe opiexe kpoal8 opiodr ttcpip opitsk opiino opiodr opidrv sou2o opimai_real main main_opd_entry If you enable the 1551 event tracing, a trace file is generated at the time of the internal error reporting the following: Errors in file /u10/app/oracle/admin/sid/udump/sid_ora_x.trc: ORA-01551: extended rollback segment, pinned blocks released CHANGESUpgraded to 9.2.0.8 or 10.2.0.4 CAUSEThis problem is due to unpublished BUG 5868257 : Dump / memory corruption from UPDATE DML Details: There is a problem with the fix from Bug 4549673 which is included in 9.2.0.8 and 10.2.0.4. With that fix applied it is possible to get a dump under updgrh and/or private memory corruption. This issue can also cause malformed RCI information in the redo stream causing problems for log miner and Streams. eg: Corruption / dumps can occur when updating a migrated row where one of the columns is to be updated to NULL and an internal 1551 error occurs during the DML.(1551 errors are not visible to customers and are caught and handled internally) This problem is raised in the following bugs which are closed as Duplicate of Unpublished BUG 5868257. BUG 7138239 - ORA-07445[_INTEL_FAST_MEMCPY.A()+10] FOR UPDATE BUG 7118779 - VARIOUS ORA-600 / ORA-7445 ERRORS FROM REMOTE APPLY SIDE BUG 7136414 - ORA-07445 [_MEMMOVE()+2464] ERROR RUNNING UPDATE BUG 7423332 - MEMORY CORRUPTIONS WITH VARIOUS ORA-600 AND ORA-7445'S UNTIL PMON FAILURE Bug 7390551 ORA-7445 EXCEPTION ENCOUNTERED: CORE DUMP [FFFFFFFF7F40079C] SOLUTION1. Unpublished Bug 5868257 is fixed in patchsets 10.2.0.5 and 11.1.0.6 and higher. Therefore upgrade to one of the Oracle Versions where the bug has been fixed. OR 2. Apply the one-off Patch 5868257 if available on My Oracle Support for your Oracle Version and platform. If your database version is 10.2.0.4 on HP-UX Itanium, then make sure you apply the patch which is dated 23-OCT-2008 (or later) because previous versions of this patch on this platform & DB version showed some problems as seen in Bug 7423332. OR 3. As a workaround, you can help to avoid getting migrated rows by ensuring that PCTFREE is set sufficiently high such that subsequent UPDATE of rows in a data block has space to fit. It seems likely that you could minimize the risk of this issue by rebuilding the main affected table with a sufficient PCTFREE to avoid row migration. 建议：有点类似该bug，你尝试重建该表把pctfree设置为50%（具体自己把握这个比例）试试看