Re: Oracle XE Corruption
- From: Niall Litchfield <niall.litchfield@xxxxxxxxx>
- To: daniel.fink@xxxxxxxxxxxxxx
- Date: Thu, 26 Aug 2010 08:59:00 +0100
Dan, I entirely agree, but the whole unusable index behaviour is a bugbear
for me (in the default setup I.e without skip_unusable_indexes set) imo
marking an index as unusable *should* exactly invalidate cursors and lead to
a reoptimisation without the indexed access path being considered. However
the default is (or at least was) to throw the error "index blah or partition
of such index is unusable". I'd expect smon to barf on that. This does of
course involve several levels of speculation many of which could be tested.
I might try and get some time to do this before Alex G reads this and
correctly points out BAAG.
On 25 Aug 2010 20:09, "Daniel Fink" <daniel.fink@xxxxxxxxxxxxxx> wrote:
To Andy's point about not being marked as unusable, I agree. Throwing
serious errors on an object but not deciding to mark that object as unusable
strikes me as not wise. Were the statements failing or just using bad data?
To Niall's point - I really want to know if my metadata is hosed. Bad
metadata could mean bad data being returned without any indication. If it
requires an instance down, so be it. If it had not been for a file system
space issue, this might not have been caught...that I don't like.
Of course, I did not write smon code, nor any Oracle code (I flail away at
SQL and write PL/SQL at the same level as an unusually dim hedgehog). But as
a DBA, bad metadata causes way too many problems (been there, done that in
Oracle7) to be left to a simple trace file.
------- Original Message -------
On 8/25/2010 5:51 PM Niall Litchfield wrote:
Methinks If it was marked unusable then the instance w...
--
//www.freelists.org/webpage/oracle-l
Other related posts:
