Are you doing a lot of sorting with a small sort_area_size set?
That's what Note:33567.1 in Metalink seems to point to. Since this
involves the ST enqueue it is probably also a good idea to see if you
can switch to locally managed tablespaces. Switch temp over first.
You might want to run a
Jonathan,
Yes, please. I'd like the patch as well as the upgrade subscription path so I can
paste
'10' on as well. I'd also like installer insurance in case the patch hasn't been fully
QA'd and breaks two things while fixing one.
Seriously, I recommend the book because it explains a lot of
There are a lot of good items on your list. Harrison is the definitive work on Oracle
SQL tuning, though it is also worthwhile to pick up the O'Reilly pocket SQL tuning
book as well. Robert Freeman's books are uniformly excellent. Geoff Ingram's book
was a pleasure to read and contains a great
Ora-472 is a very non-specific PMON error. You've probably already worked
through this with support using a TAR, but if you still are having problems and
wanted to post the following it would help:
The stack trace from your pmon traces (they'll either be in bdump or udump, and by
stack
Another poor man's solution would be to unload the tables into flat files and attach
to them as needed using Oracle's external table feature from 9i. That solution
should hold for quite a while into the future since the external table function is
very
much like SQL*Loader, which is so
You could use VPD (virtual private database, AKA fine-grained
access control AKA row-level security) to control exactly which
rows a given user can delete based on a function using the
business rule which determines which rows they are authorized to
delete. This works from 8i onward, but there
There was some discussion recently (in the thread on Apache and
iAS) on version numbers of Oracle and how confusing they have
become.
Here is a document that explains the version numbers, at least for
the RDBMS. It lost it's authorship along the way, so forgive me for
not crediting it.
If
Here's a convincing stat for your architect:
There will be no RBO in a future version of Oracle. That version is
not far away (conjecture is perhaps as early as 10). As often
occurs, this is a case of do what we say, not what we do, since
RBO is still used in the data dictionary tables.
Patrice,
You write:
What is the meaning of relational in relational database
again?
Apparently not what you think. The relational in 'relational database'
comes from the term relation--a certain type of mathematical table
in set theory. It has nothing at all to do with relationships (the
View positives:
They simplify what the developer and user see, thus making SQL
logically simple
You can put hints on a view, thus forcing the optimizer to use the
view as you know best. This can sometimes be the only way to
optimize a third party application that puts all its SQL in a black
Take a look at Metalink note :1076835.6. You may be running out of
underlying OS user process memory. Changing some kernel parameters
may help.
One other thing to check: Is this query going parallel (i.e. is the degree of
any of the underlying objects 1 and you are seeing parallel query
Okay, I'll plug my own list of recommendations on Amazon. It
should come up from this URL:
http://www.amazon.com/exec/obidos/tg/listmania/list-browse/-
/VL8CI2YJANX1/qid=1013704150/sr=5-1/ref=sr_5_1/103-8633316-
6595843
If it doesn't, just enter 'Oracle performance tuning' in a search on
Srini,
If you have 400 concurrent users (2000 total users, with about 400
concurrent sessions typical for the busy time of the day), then you
should probably be using MTS instead of dedicated connections.
Using Microsoft pooling in COM may work out, but please check
with Oracle support on
While the script approach is excellent, and you have gotten several
good answers on how to use it, might I suggest looking into event
triggers? If you are running 8i or above (not 8.1.5, it's buggy on
this), you can create a database level trigger ON ERROR. If you
have Java turned on for your
There has been a version of Oracle for Solaris on Intel available for
some time. Unfortunately that time is coming to an end. There will
not be any Solaris Intel Oracle 9, so 8i is the end of the line for the
platform.
Regards,
Chris Gait
--
Please see the official ORACLE-L FAQ:
I'm afraid I can't agree with the statement that two different users
submitting the same SQL cause a hard parse. It is my
understanding of the process that part of the parsing occurs when
another user is submitting an identical SQL statement, but this
does not constitute a hard parse, just a
A client of mine is moving off some ancient hardware (with ancient
software) and needs for the short term to run SQL*Reportwriter 1.0
for DOS, then convert those files to version 1.1 of the same. If
anybody is an antique collector and still has either the install disks
or runtime of these
The key to the whole deal, as it turns out, being that the ANALYZE
ANY privilege must be granted explicitly in order to work in a stored
procedure. I tested this with a procedure for analyzing a table and
not even sys had sufficient privileges. But once I granted 'analyze
any' to a user they
If the production environment is already available, you could use
DBMS_STATS.EXPORT_TABLE_STATS on the production table
and import those stats into your development instance.
Regards,
Chris Gait
On 18 Apr 2001, at 11:00, Murali Vallath wrote:
Thanks for the feedback, I am coming from the
As others have pointed out on this thread, I was wrong, and you
can in fact pin packages in a startup trigger that have not been
loaded into memory. I was confusing this with functions and
procedures (and cursors) which have to be in memory before they
are pinned and will raise an error if
I ran into something like this on a client's system recently. They
had a LONG in the table, and that won't come over with the COPY
command. If that's the case here, then you need to use the
TO_LOB() function to take it from LONG to CLOB on the fly.
As sometimes happens, the error message can
I would need to double-check on your error, but the usual problem
with pinning packages at startup is that they haven't been loaded
into memory (standard and some of SYS's other packages load
almost at the initial moment of startup). Try to put some code into
the trigger to call the packages
22 matches
Mail list logo
