Note in-line.
--
Regards
Jonathan Lewis
http://www.jlcomp.demon.co.uk
The educated person is not the person
who can answer the questions, but the
person who can question the answers -- T. Schick Jr
Next public appearance2:
March 2004 Hotsos Symposium - Keynote
March 2004 Charlotte NC - OUG Tutorial
April 2004 Iceland
One-day tutorials:
http://www.jlcomp.demon.co.uk/tutorial.html
Three-day seminar:
see http://www.jlcomp.demon.co.uk/seminar.html
____UK___February
The Co-operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/faq/ind_faq.html
"Richard Kuhler" <noone_at_nowhere.com> wrote in message
news:w0jLb.45890$C87.11580_at_twister.socal.rr.com...
> We believe we've found the source of the problem although not really the
> explanation. At the time the process was running, a DBA kicked of a
> DBMS_STATS analysis for this instance. We killed our process and the
> DBMS_STATS session and retried our process. This time it finished
> without any problems in the expected amount of time. To confirm this
> interaction, we ran our process again and started DBMS_STATS while it
> was running again. The same repeated rollback and retry behavior was
> observed. Something about the DBMS_STATS run is causing the issues with
> the MERGE operation.
>
Just bouncing ideas around - but when you generate stats
you invalidate any cursors that may exist on the table.
In simple case, if you are running in parallel, the parallel slaves
get called about 13 times each to process a statement that
addresses part of the data. But what they are given to process appears
as an SQL statement with a few input parameters. If these cursors
become invalid between calls, perhaps Oracle has to go into
defensive mode and rollback the parallel update because reparsing
the statement may change its execution path, which means one pass
of the query MIGHT not identify the section of data that was intended
when the QC wrote the SQL and this could therefore allow some data
to be updated twice, and some data to be missed.
(This seems a bit unlikely - but it's the only hypothesis I
can come up with at the moment).
Received on Fri Jan 09 2004 - 02:59:59 CST
