ORA-00600([ksires_1])错误

在做压力测试的时候，客户数据库的alertlog 文件查询ORA-00600([ksires_1])错误

 

进一步查询跟踪文件：

 

[oracle@sxffdb1 ~]$ more /u01/oracle/admin/ffdb/udump/ffdb1_ora_202570.trc
/u01/oracle/admin/ffdb/udump/ffdb1_ora_202570.trc
Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bit Production
With the Partitioning, Real Application Clusters, OLAP, Data Mining
and Real Application Testing options
ORACLE_HOME = /u01/oracle/product/db10gr2
System name:    AIX
Node name:      sxffdb1
Release:        3
Version:        5
Machine:        00C24AC04C00
Instance name: ffdb1
Redo thread mounted by this instance: 1
Oracle process number: 215
Unix process pid: 202570, image:

*** ACTION NAME:(/rac/racSitemap) 2012-03-31 21:24:43.842
*** MODULE NAME:(OEM.DefaultPool) 2012-03-31 21:24:43.842
*** SERVICE NAME:(ffdb) 2012-03-31 21:24:43.842
*** CLIENT ID:( (compatible; MSIE 6.0; Windows N) 2012-03-31 21:24:43.842
*** SESSION ID:(1426.565) 2012-03-31 21:24:43.842
clscugblmini: NS user global alloc failed
2012-03-31 21:24:43.848: [ default]prom_init:9:could not initialize clscugblm 3
2012-03-31 21:24:43.848: [  OCRAPI]a_init:11!: Messaging init unsuccessful : [24]
clscugblmini: NS user global alloc failed
2012-03-31 21:24:43.851: [ CSSCLNT]clsssInitNative: clscugblmini failed, rc 3
kgxgncin: CLSS init failed with status 3
kjfmsgr: unable to connect to NM for reg in shared group
*** 2012-03-31 21:24:43.851
ksedmp: internal or fatal error
ORA-00600: internal error code, arguments: [ksires_1], [], [], [], [], [], [], []
Current SQL statement for this session:
SELECT oldest_flashback_time
         FROM v$flashback_database_log
----- PL/SQL Call Stack -----
  object      line  object
  handle    number  name
70000063b382b48      1581  package body SYSMAN.EMD_DATABASE
70000063b382b48      1903  package body SYSMAN.EMD_DATABASE

TRACE文件中可以看到导致这个问题的SQL是：

SELECT oldest_flashback_time  FROM v$flashback_database_log

 

这个SQL用于查询可以闪回到的最早时间点。

 

查询了一下MOS, 文档[ID 1311870.1] 和 这个错误最接近。

 

文档上说导致问题的原因是：CRS所有者用户周期性的删除目录/var/tmp下的所有CRS所有者的套接字文件。

The problem is caused by a cronjob which remove files owned by user oprdi (crs owner) under /var/tmp, which removed all CRS sockets files owned by user oprdi, this in turn causes problem at database and CRS layer.

 

ORACLE给出的解决方法是：

Remove the cronjob which periodically clean up /var/tmp for user oprdi.

 

不过客户的环境并没有发现有定期删除文件的脚本。

 

好歹这个错误并未发现对数据库又很大影响。