0209天津问题总结2

二、查看metalink信息

 
主题:Ora-00600:Arguments:[Ktte_append_file_info-1]atdatabaseopen 
文档ID:382871.1类型:PROBLEM 
ModifiedDate:20-SEP-2006 状态:MODERATED 
InthisDocument 
Symptoms 
Cause 
Solution 
References 
-------------------------------------------------------------------------------- 
ThisdocumentisbeingdeliveredtoyouviaOracleSupport'sRapidVisibility(RaV)process,andthereforehasnotbeensubjecttoanindependenttechnicalreview. 
Appliesto:OracleServer-EnterpriseEdition-Version:10.2.0.1.0 
Thisproblemcanoccuronanyplatform. 
Oracle10.2.0.164-bit 
LinuxItaniumserverSymptomsTwonodeRACdatabase 
UsingASM,MTS 
Gotinternalerrorsfollowinganopenofthedatabase 
Theerrorinthealertlogis: 
ORA-00600:internalerrorcode,arguments:[ktte_append_file_info-1],[12],[79],[12],[28],[],[],[] 
Thiserrorwasprecededbyanothererror: 
ORA-01157:cannotidentify/lockdatafile80-seeDBWRtracefile 
ORA-01110:datafile80:'+NDMID_DATA/oracle/data/ndmid/ndmid_data_27.dbf' 
Thetracefileshowsnofailingstatement,nocurrentcursor. 
thecustomerisunawareofapossiblecause,orapossibleimpact. 
Callstack: 
ksedstksedmpksfdmpkgerinvkgeasnmierr 
ktte_append_file_infoktte_insert_tsktte_prime_finfo 
ktte_monitor_tsthktte_monitor_tsksbctiksbabs 
kebm_mmon_mainksbrdpopiripopidrvsou2oopimai_real 
mainlibc_start_mainstart 
CauseThisisinternalBug4696967 
ABSTRACT:ORA-600[KTTE_APPEND_FILE_INFO-1]INMMON 
Occurswhenadatafileisinaccessibleatinstancestartup. 
Thealertlogshowsthatadatafilewasinaccessibleasfollows: 
SatMay2011:22:272006 
Errorsinfile/local/oracle/admin/ndmid/bdump/ndmid2_smon_32104.trc: 
ORA-01157:cannotidentify/lockdatafile80-seeDBWRtracefile 
ORA-01110:datafile80:'+NDMID_DATA/oracle/data/ndmid/ndmid_data_27.dbf' 
Thenafewsecondslater,theinstancewasopened,andtheerroroccurred 
SatMay2011:22:412006 
Completed:ALTERDATABASEOPEN 
SatMay2011:29:412006 
Errorsinfile/local/oracle/admin/ndmid/bdump/ndmid2_mmon_32110.trc: 
ORA-00600:internalerrorcode,arguments:[ktte_append_file_info-1],[12],[79],[12],[28], 
[],[],[] 
Itappearswhentheinstancestartsup,itcouldnotfindadatafileas 
+NDMID_DATAvolumewasnotaccessible,MMONreportstheOERI 
ktte_append_file_info-1andcrashing. 
. 
Theproblemhasnotoccurredsincebecausethisvolumeisaccessible.This 
lookslikethesamesituationasbug4696967andnolongeroccursinthe 
10.2.0.2patchset. 
Solution 
Noactionisrequiredexcepttoensurethatalldatafilesareaccessiblebeforestartingthedatabase. 
In10.2.0.2,insteadofaninternalerror,thedatabasewillnotstartup,andwillgiveadditionalinformation. 
ReferencesBug5411629-Ora-00600:Arguments:[Ktte_Append_File_Info-1],[12], 
-------------------------------------------------------------------------------- 
Helpusimproveourservice.Pleaseemailusyourcommentsforthisdocument.. 
三、解决步骤 
1. 备份表空间XCZH(export或者其它方式) 
2.crs_stat-t 查看rac状态 
3. 关闭所有服务与库 
4. 仅启动第一个实例与库 
5. 在Oracle数据库10g第2版中，可以只删除数据文件。(1数据文件必须为空才能删除。无法删除表空间中的最后一个数据文件；必须删除表空间本身。同时，表空间必须处于联机和读写状态。) 
altertablespaceXCZHdropdatafile'/dev/md/oraset/rdsk/d104'; 
6. 重新添加数据文件至XCZH表空间 
altertablespacexczhadddatafile'/dev/md/oraset/rdsk/d114'size29998M; 
四、总结： 
安装初期， 
/dev/md/oraset/dsk/d103 
/dev/md/oraset/rdsk/d103 
30000 
XCZH 
修船综合 
/dev/md/oraset/dsk/d104 
/dev/md/oraset/rdsk/d104 
30000
oracle 用户仅在fujitsu1实例机上有对d104裸设备操作的权限，fujitsu2实例机上没有权限操作而引起的ora-600错误。 
09 年2月12日 
一、alert_FUJITSU2.log 
WedFeb1108:00:562009 
MemoryNotification:LibraryCacheObjectloadedintoSGA 
Heapsize5086Kexceedsnotificationthreshold(2048K) 
KGLobjectname:select*from(selectnulltable_qualifier,decode(o1.owner,'PUBLIC',chr(0),o1.owner)table_owner,o1.object_nametable_name,decode(o1.owner,'SYS',decode(o1.object_type,'TABLE','SYSTEMTABLE','VIEW','SYSTEMVIEW',o1.object_type),'SYSTEM',decode(o1.object_type,'TABLE','SYSTEMTABLE','VIEW','SYSTEMVIEW',o1.object_type),o1.object_type)table_type,nullremarksfromall_objectso1whereo1.object_typein('TABLE','VIEW')unionselectnulltable_qualifier,decode(o2.owner,'PUBLIC',chr(0),o2.owner)tab 
WedFeb1108:02:472009 
MemoryNotification:LibraryCacheObjectloadedintoSGA 
Heapsize5054Kexceedsnotificationthreshold(2048K) 
KGLobjectname:select*from(selectnulltable_qualifier,decode(o1.owner,'PUBLIC',chr(0),o1.owner)table_owner,o1.object_nametable_name,decode(o1.owner,'SYS',decode(o1.object_type,'TABLE','SYSTEMTABLE','VIEW','SYSTEMVIEW',o1.object_type),'SYSTEM',decode(o1.object_type,'TABLE','SYSTEMTABLE','VIEW','SYSTEMVIEW',o1.object_type),o1.object_type)table_type,nullremarksfromall_objectso1whereo1.object_typein('TABLE','VIEW')unionselectnulltable_qualifier,decode(o2.owner,'PUBLIC',chr(0),o2.owner)tab 
二、fujitsu2实例机频繁的写trace文件。查看问题 
这是个警告，属于一个bug，在10.2.0.2中被fixed。
问题原因：
进程花费大量时间在内存中查找剩余空间，没找到，进行了扩展，有可能是内存碎片比较严重。
解决办法：
10g 中有隐藏参数_kgl_large_heap_warning_threshold，将其扩大即可. 在10.2.0.2中，这个值被增加到了50MB.
SQL>altersystemset"_kgl_large_heap_warning_threshold"=52428800scope=spfile;
停库，启库
SQL>showparameter_kgl_large_heap_warning_threshold
NAMETYPEVALUE
-----------------------------------------------------------------------------
_kgl_large_heap_warning_thresholdinteger52428800
总结：除了10.2.0.1的bug外，其它问题都是安装初期没有跟系统工程师沟通完全所致。 

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/11780477/viewspace-705090/，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/11780477/viewspace-705090/