Re: Optimizer problems 7.0.16.4, Solaris
Date: Wed, 16 Nov 1994 08:38:04 GMT
Message-ID: <CzCqnG.B91_at_baan.nl>
In article <3ab2va$mc3_at_athos.cc.bellcore.com>, mjk_at_spanky.cc.bellcore.com (25994-kelly) writes:
|> Has anyone else had problems with the optimizer on 7.0.16.4?
|> I get completely different plans for the same query on the same data
|> seemingly at random. The data and the table statistics do not change but the plan does. One plan takes 1 minute. The other takes 10 minutes and involves
|> reading both involved tables with full scans 3 times each. It is the worst
|> imaginable way to satisfy the query. I have opened a problem with Oracle
|> about this but they seem to think I should just change to 7.1 because
|> they put fixes into 7.1. I can't do this easily because our product is
|> built on a workstation factory environment and I have to get those folks
|> to test/approve it for many applications before I can use it.
|>
|>
We have also problems with the optimizer.
Special on table with fragmentation.
The optimizer chooses a wrong execution path.
it goes for sorting, while hint FIRST_ROWS is given and index can be used.
Sometimes re-creation of the table or re-generation of statistics 'solves' the problem,
but the behaviour is unpredictable.
Oracle 7.1.3 has the same bug.
Oracle tells us that it should be solved in Oracle 7.2. (????)
+-------------------------+------------------+------------------------+
| Hans van Leeuwen | B.v. Nagellstr 89| Phone: +31 3420 28888 |
| Porting & Benchmarking | P.O. Box 143 | Fax : +31 3420 28606 |
| Baan Development B.V. | 3770 AC Barneveld| Email: hansvl_at_baan.nl |
+-------------------------+ The Netherlands +------------------------+
+------------------+
Received on Wed Nov 16 1994 - 09:38:04 CET