Patchset 10.2.0.5 for Windows x86/x64
Last night Oracle finally released patchset 4 (10.2.0.5) for Oracle 10g R2 on Windows x64 (my download is going on right now :-).
In the last couple of days I was busy installing and testing 32-bit version of the patchset for Windows. Didn’t experience any difficulties so far, but it’s too early to push the upgrade to the client machines. There are some known issues related to ODP.NET for which I’m not sure that I did it right (need a confirmation from .net developers).
The rest of the week I’ll be installing and testing x64 version of the patchset on Windows 2003/2008 servers and if I find something worth sharing with the Oracle community, I’ll publish it here.
Posted on 28.07.2010, in Oracle and tagged oracle. Bookmark the permalink. 7 Comments.
Points of concern for 10.2.0.5
Mandatory reading (immediately) after checking patchset readme are the following MOS notes:
Among new bugs listed as introduced by 10.2.0.5 the following one sounds particularly scary to me:
To avoid any surprises we should grep our PL/SQL code for XMLTYPE() usage, such as v_xml := XMLTYPE (v_result_clob); and test the function on 10.2.0.5 with the example shown in MOS support note:
Bug 9871430: ORA-31011: LPX-00230: INVALID CHARACTER 0 (U+0000) FOUND IN A NAME OR NMTOKEN
In patchset readme, section 8.1.7 Installing 10.2.0.5.0 Version of ODAC for .NET Framework 2.0, is mentioned that the issue with ODP.NET is tracked under bug 8420421. Search on MOS didn’t show such document, it’s obviously a non-public bug. We’re using 10.2.0.3 + ODAC 10.2.0.2.21 (the latest and greatest ODAC release for 10g available on OTN), after reading the section in patchset readme I’m not 100% positive that our 10.2.0.5 patched ODP.NET client is trouble free. It looks like but need to confirm this with someone from app. team.
[Update, November 2010 – Bugs discovered so far after 10.2.0.5 + Patch 2 installation]
XML Database FAQ
Three (test) installations of patchset 10.2.0.5 on Windows 2003 x64 are behind me. All three went without a hitch.
The only thing that really bothers me is bug introduced in XmlType() function. This function is heavily used in one of our core applications with tones of PL/SQL code. I could not reproduce the bug with the simple test case shown in Bug 9871430. It must be something in data itself that triggers the error, of course customer sample data is only available to Oracle support. While searching the MOS for related bugs I stumbled on document: 416132.1 XML Database FAQ, with interesting recommendation:
For example if on 10.2.0.x we currently recommend the 10.2.0.5 patch set as soon as it is available it has over 100 XDB bug fixes in it alone.
Hmm…it throws totally different perspective on my concerns regarding xmltype() function — suddenly I feell lucky. At the moment lucky in the light of legendary question: “Do you feel lucky, punk? Well do ya…”
No wonder that regression hit back while nailing those 100+ little green ones… :-)
ORA-4043 during catupgrd.sql
During test installation of patchset 10.2.0.5 I greped the following error from the log:
ORA-04043: object XDB_DATASTORE_PROC does not exist
This error is a result of Oracle Text not being installed in database and can be ignored according to MOS note:
360907.1 Problem : catupgrd.sql gives error on XDB_DATASTORE_PROC ORA-4043
After 10.2.0.5 installation I noticed “noisy” trace files in udump directory with the contents like this:
This is a known bug in 10.2.0.5, some helpful MOS references:
At the time of writing this note interim patches exists for Linux x86-64, AIX and Solaris. No interim patch for Windows x64.
I didn’t find much technical info about severity of this bug, neither on MOS nor internet. I would say it’s more of an annoyance archiving those trace files than some serious bug, otherwise I would expect this bug to be listed in Known Issues section of the readme. Nevertheless, I postponed applying 10.2.0.5 on Windows x64 in production environment until interim patch is available or until Oracle support reassure me that this bug is a benign one.
SR opened for bug 9772888
On August 5th 2010 I opened Service request asking support to provide me interim patch for the bug for Windows x64. Right now discussion is going on (usual fun to pass tier 1 personnel), if async I/O is really turned on and if it is, could we turn it off as a workaround…then providing them some metalink notes, …as I said we have fun ;-). I hope I’ll be back with some good news. Fingers crossed.
Fix for bug 9772888
Oracle support notified me that fix for the bug 9772888 will be included in Patch 1 for 10.2.0.5. At the time of this writing the release date for patch 1 is not yet known, according to MOS note 161549.1 Bulletin : Oracle Database Server and Networking Patches for Microsoft Platforms.
10.2.0.5 Patch 1
Yesterday Oracle updated MOS note 161549.1 with the target release date for 10.2.0.5 Patch 1 for Windows x64, which is in mid September 2010.
Update [September 18, 2010]: Patch 1 for 10.2.0.5 is available for download for both, 32-bit [Patch#10058290] and 64-bit [Patch#10099855] Windows.