Installing 11g on Centos 4.2 – getting ORA600 [kghssgai2]

This is my second Oracle11g (11.1.0.6) installation on Linux. The first one was a successful one, on Dell PowerEdge 6400, running Centos 4.5. Installation on Dell server went well, without any errors.

Now, I’m strugling installing 11g on home based test machine (AMD 1Ghz, 1GB RAM), running Centos 4.2. All checks done by OUI passed the test – all kernel parameters are set and necessary packages installed with yum. All is going well until OUI comes to the point when JVM is added to the database. In the middle of JVM installation OUI throws ORA-600 error.

From the alert.log (note that trace files and alert.log location changed in 11g, now it’s under directory $ORACLE_BASE/diag/rdbms/$DBNAME/trace):

ORA-00600: internal error code, arguments: [kghssgai2], [1], [36], [], [], [], [], [] Incident 
details in: /oracle/diag/rdbms/ora11/incident/incdir_835/ora11_m000_12152_i835.trc

An interesting part from trace file /oracle/diag/rdbms/ora11/ora11_m000_12152.trc:

......
*** 2007-10-10 12:13:35.903
joez_compile_method: done compiling method oracle/aurora/zephyr/analysis/MIRSSA$HeapVariableMana
ger.copyHeapVariable(0xb6e3caa7).
joez_compile_method beginning compile of method java/util/Stack.push

*** 2007-10-10 12:13:36.973
joez_compile_method: done compiling method java/util/Stack.push(0x331d2c85).
joez_compile_method beginning compile of method java/text/MessageFormat.applyPattern

*** 2007-10-10 12:13:39.794
Incident 835 created, dump file: /oracle/diag/rdbms/ora11/ora11/incident/incdir_835/ora11_m000_1
2152_i835.trc
ORA-00600: internal error code, arguments: [kghssgai2], [1], [36], [], [], [], [], []

joxjitexe0 caught: 600; Aborting
joxjitexe1 caught: 600;
joxjitexe exiting
KEBM: MMON slave action policy violation. joxjdp_; viol=0; err=600

*** 2007-10-10 12:14:41.787
ORA-00600: internal error code, arguments: [kghssgai2], [1], [36], [], [], [], [], []
OPIRIP: Uncaught error 447. Error stack:
ORA-00447: fatal error in background process

For now, I’m not sure what could be the cause for the error. Metalink search didn’t reveal much (interestingly ORA-600 search facility is not even ready yet for 11g, Oracle 10.2 is the last version that you can pick from the list of values in ORA-600 Search facility :-).
Perhaps the culprit of the problem is hidden in Centos version 4.2. I’ll update this thread with the solution (or workaround), if I’ll found one.

Posted on 10.10.2007, in Oracle and tagged . Bookmark the permalink. 1 Comment.

  1. Problem solved…
    It seems that the problem was related either to insufficient JAVA_POOL or SHARED_POOL.
    At first attempt I had a mere 50MB java_pool – 70MB was enough in second attempt, I also raised SHARED_POOL from 256MB to 300MB.
    In both cases I used (obviously) Manual Memory management.