ora-600导致SMON异常终止，实例重启

Fung920

主要的错误：
ORA-00600: internal error code, arguments: [kglGetMutex]
ORA-00600: internal error code, arguments: [kglpnlt]
附件为alert日志及trace信息。

平台：Linux
数据库版本：11.2.0.3.0 单实例

xifenfei

APPLIES TO:Oracle Database - Enterprise Edition - Version 11.2.0.3 and later
Information in this document applies to any platform.
SYMPTOMS

As per alert.log, below errors reported in their environment:

    ORA-00600: internal error code, arguments: [kglGetMutex], [], [], [], [], [], [], [], [], [], [], []
    ORA-00600: internal error code, arguments: [kglpnlt], [], [], [], [], [], [], [], [], [], [], []
    ORA-07445: exception encountered: core dump [kglhdgn()+282] [SIGSEGV] [ADDR:0x0] [PC:0x93908AA] [SI_KERNEL(general_protection)] []
    ORA-07445: exception encountered: core dump [kglGetMutex()+80] [SIGBUS] [ADDR:0xF67D5310] [PC:0x9392516] [Non-existent physical address] []

Call stack from ORA-00600 trace file is reported as:

    skdstdst <- ksedst1 <- ksedst <- dbkedDefDump <- ksedmp
       <- ssexhd <- sighandler <- kglGetMutex <- kglScanHandles4Unke <- ktmmon
        <- ep <- ktmSmonMain <- ksbrdp <- opirip <- opidrv
         <- sou2o <- opimai_real <- ssthrdmain <- main <- libc_start_main
          <- start

Note: Here the call stack is similar for ORA-00600 [kglGetMutex] and ORA-07445 [kglGetMutex()+80].

During our analysis, we refereed the Doc ID 13550185.8 and found that CURSOR_SHARING=EXACT along with SYS_Bnnn binds in SQL statements.

This was confirmed from incident trace file and by executing below SQL statement in customer environment-

    SQL> select count(*) from v$sql_bind_capture where name like ':SYS%';

CAUSE

This issue is matching with Bug 17059810, which was closed as duplicate of Bug 13550185 .
This was confirmed through a new Bug 20348053

SOLUTION

The suggestion from development is to follow the below action plan:

1.    Download and apply patch 13550185 available for your platform on top of 11.2.0.3.
2.    It should resolve the issue but you can monitor and keep the following set:
   
        SQL > alter system set max_dump_file_size=unlimited;
        SQL> alter system set events '7445 trace name errorstack level 3;name library_cache level 90';
      
If the issue is seen again then provide:

a.    Full alert log and the tracefile/incident file for the latest crash.
b.    Opatch output from:

          opatch lsinventory -detail

And create a SR along with these details, so that Oracle support can file a new Bug.

Note: To disable the 7445 event:

    SQL> alter system set events '7445 trace name errorstack off;name library_cache off';