Thanks for the input, Michael.
You are correct in your error number assumption. It is
the 2064 error.
We did set the 2064 event in a logon trigger and last
week and got a trace, not much help there either. I
tried to get the analyst to choose between oradebug
and dbms_system.set_ev so that I could set it
independently of the session, but we finally just put
a good old alter session in our load script.
We ran a 10053 trace since the problem seems that it
might be some sort of parse error on the source side.
We wanted the 10046 trace to see exactly where the
error was occuring since it is in a stored procedure
and what it was waiting on.
In working the theory that it might be
object/stats/parse related I rebuilt the tables and
recreated the indexes, no effect.
The strange thing is that this error just started
showing up two weeks ago, in two stored procedures
(both doing the same thing, different tables) having
the exact same frequency, scope, and duration of
error. There must be a connection, but unfortunately
this is a very hot problem for us and I don't have the
window to figure it out. It is too bad, it is a very
good puzzle. Almost hate to solve it with an upgrade.
The main reason that I am upgrading is that Oracle
support has tentatively identified a possible bug
(although it is internal, so no details) that will be
fixed by 8.1.7.4. Their position is that unless I
upgrade they won't pursue this further. So, upgrading
I am, even though I do not feel that they have really
nailed the problem, but they won't give further
support. Reasonable on their part, since they might
find that we have a bug that is already solved.
jack
- Michael P Sale <Michael.Sale_at_oracle.com> wrote:
> Ahh yes, the elusive "Why". If you are already
> working with support then
> I would set an errorstack on the ora-2064 error
> number you are getting(I
> suspect that is what it is, you never say). This
> will give them
> extensive information as to the state of the call at
> the time of the
> error.
>
> If you are running a high enough version of the db
> this statement will
> work:
> alter system set events '2064 trace name errorstack
> level 3';
>
> Otherwise you can set it in the init.ora:
> event = "2064 trace name errorstack level 3"
> Where 2064 is the error number you are getting.
>
> You can also set it at the session level with the
> "alter session syntax"
> and on another session with oradebug. This is
> obviously the much more
> painful route.
>
> The 10053 would only be relevant if you were running
> into a CBO issue,
> not likely unless you are running with certain
> events and/or underscore
> parameters in place, and the 10046 should show the
> sql being run (if
> taken at the proper level) but not why the error
> message, only that this
> particular sql was run and resulted in this error.
>
> An upgrade to address this problem directly might
> not solve it
> **UNLESS** you are running into bug 2140287 "CREATE
> TABLE AS SELECT WITH
> DECODE CORRUPTING DATA" (see the 8.1.7.4 patchset
> readme or metalink
> note 120613.1). In which case you could possibly
> have serious issues
> that would require you move to this patchset!!!
>
> Either way, the errorstack will give you the "why"
> even though only
> Oracle support or development would likely be able
> to interpret this for
> you.
>
>
> Regards,
>
> Michael Sale
> Author: Oracle9i for Windows(R) 2000 Tips &
> Techniques
> http://www.amazon.com/exec/obidos/ASIN/0072194626
>
>
> -----Original Message-----
> Silvey
> Sent: Friday, June 21, 2002 2:28 PM
> To: Multiple recipients of list ORACLE-L
>
>
> Oracle support is saying much the same thing. They
> are
> advising an upgrade to 8.1.7.4 (from 8.1.7.2).
> Doesn't
> give us the reason WHY it is happening, but it might
> fix it. We have run 10053 and 10046 traces and can
> see
> no obvious issues.
>
>
>
>
> --- Michael P Sale <Michael.Sale_at_oracle.com> wrote:
> > Without seeing more information, I have seen this
> > type of a response
> > where either the instance is not registered with
> the listener, or the
> > db is not accepting connections for a variety of
> > reasons (e.g. no more
> > memory all the way to the db is shut down).
> >
> > Regards,
> >
> > Michael Sale
> > Author: Oracle9i for Windows(R) 2000 Tips &
> > Techniques
> > http://www.amazon.com/exec/obidos/ASIN/0072194626
> >
> >
> > -----Original Message-----
> > Silvey
> > Sent: Friday, June 21, 2002 12:08 PM
> > To: Multiple recipients of list ORACLE-L
> >
> >
> > Listers,
> >
> > We have two stored procedures that are suddenly
> > throwing random "distributed operation not
> > supported"
> > errors.
> >
> > These pieces of code are doing an "insert into as
> > select" across a database link. The data is being
> > pushed, not pulled. They have been running for
> > months
> > in production without a problem. Suddenly, they
> > start
> > throwing these errors, but will often succeed on
> the
> > retry.
> >
> > Anyone have a similiar problem?
> >
> > thx,
> >
> > jack
> >
>
> --
> Please see the official ORACLE-L FAQ:
> http://www.orafaq.com
> --
> Author: Michael P Sale
> INET: Michael.Sale_at_oracle.com
>
> Fat City Network Services -- (858) 538-5051 FAX:
> (858) 538-5051
> San Diego, California -- Public Internet
> access / Mailing Lists
>
> To REMOVE yourself from this mailing list, send an
> E-Mail message
> to: ListGuru_at_fatcity.com (note EXACT spelling of
> 'ListGuru') and in
> the message BODY, include a line containing: UNSUB
> ORACLE-L
> (or the name of mailing list you want to be removed
> from). You may
> also send the HELP command for other information
> (like subscribing).
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: Jack Silvey
INET: jack_silvey_at_yahoo.com
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
Received on Sat Jun 22 2002 - 12:16:23 CDT
