Re: Catbundle causes sessions using text index to throw ORA-4068 on RAC
- From: Yong Huang <yong321@xxxxxxxxx>
- To: michaelw436@xxxxxxxxx
- Date: Tue, 30 Mar 2010 10:50:08 -0700 (PDT)
> The instructions for the CPU don't require you to run catbundle before
> the database is opened.
A good point if I understand you correctly. I opened an SR. The analyst
says "If the script [catbundle.sql] is not being run in the database, it
is as good as not being patched with the CPU patch." So there's no risk
in postponing running this script till whatever time is convenient (other
than the security risk the CPU patch handles). It dismissed my worry that
patching at OS level (replacing various shared libraries) without running
catbundle would cause problems to the database.
> Also, the package body in question doesn't need a recompile after the
> catbundle,
You misunderstood. That package (driutl) does not become invalid. The
driutl.plb script provided by CPU is doing "create or replace package
body", because it has new code. Replacing utlrp.sql with manual
compilation of invalid objects is irrelevant.
The title of my message mentions "RAC", because it's only a problem to
watch for if you apply CPU a single node at a time, run catbundle on the
last node, and never bounce the entire database. If it were non-RAC, you
would bounce the database anyway, so there's no possibility of
ORA-4068,-4061, etc.
Yong Huang
--
//www.freelists.org/webpage/oracle-l
Other related posts:
