Comments on: A gotcha with parallel index builds, parallel degree and query plans

By: Index rebuilds in parallel change the parallel DEGREE « DBA n00b

Index rebuilds in parallel change the parallel DEGREE « DBA n00b — Thu, 04 Mar 2010 20:11:59 +0000

[...] 4, 2010 by tylerv As many of the comments to this handy article say, we got stung by this one today. Definitely check out the article. After a large data [...]

By: Parallel Reasoning « Charles Hooper's Oracle Notes

Parallel Reasoning « Charles Hooper's Oracle Notes — Wed, 17 Feb 2010 06:04:17 +0000

[...] A Gotcha with Parallel Index Builds, Parallel Degree and Query Plans I haven’t enabled parallelism for my table! How can Oracle go parallel without my consent? Parallel index (re)build will persistently set the index parallel degree in data dictionary to the value used during build! [...]

By: Tanel Poder

Tanel Poder — Mon, 27 Apr 2009 19:23:26 +0000

Did you gather optimizer stats on the table and index using dbms_stats? which command exactly did you use?

By: kwanyoung oh

kwanyoung oh — Mon, 27 Apr 2009 17:37:46 +0000

Hi,

I ran the scripts up there.
But I found one strange thing in the execution plan.

How come we have a lot different value in the Estimated rows and actual rows?

SQL> set linesize 132
SQL> select /*+ gather_plan_statistics */ sum(object_id) from t1 where object_id > 60000;

SUM(OBJECT_ID)
————–
65869754

SQL> @x

PLAN_TABLE_OUTPUT
————————————————————————————————————————————
SQL_ID 6tk45guj3gkcg, child number 0
————————————-
select /*+ gather_plan_statistics */ sum(object_id) from t1 where object_id > 60000

Plan hash value: 2596547647

————————————————————————————————
| Id | Operation | Name | Starts | E-Rows | A-Rows | A-Time | Buffers |
————————————————————————————————
| 1 | SORT AGGREGATE | | 1 | 1 | 1 |00:00:00.01 | 3 |
| 2 | PX COORDINATOR | | 1 | | 4 |00:00:00.01 | 3 |

PLAN_TABLE_OUTPUT
————————————————————————————————————————————
| 3 | PX SEND QC (RANDOM) | :TQ10000 | 0 | 1 | 0 |00:00:00.01 | 0 |
| 4 | SORT AGGREGATE | | 0 | 1 | 0 |00:00:00.01 | 0 |
| 5 | PX BLOCK ITERATOR | | 0 | 19660 | 0 |00:00:00.01 | 0 |
|* 6 | INDEX FAST FULL SCAN| I1 | 0 | 19660 | 0 |00:00:00.01 | 0 |
————————————————————————————————

Predicate Information (identified by operation id):
—————————————————

6 – access(:Z>=:Z AND :Z60000)

PLAN_TABLE_OUTPUT
————————————————————————————————————————————

23 rows selected.

By: Avi

Avi — Tue, 16 Dec 2008 09:03:13 +0000

Hi,
How value for DEFAULT is being calculated, can you please tell me?

By: Paul Leland

Paul Leland — Mon, 25 Aug 2008 18:50:50 +0000

We seem to be hit by the same problem. Suddenly everything is ending with ORA-04031:Unable To allocate shared memory (“large pool”,”Unknown object”,”Large Pool”,”PX msg Pool”).

By: Paul Leland

Paul Leland — Mon, 25 Aug 2008 18:50:02 +0000

We seem to be hit by the same problem. Suddenly everything is ending with “Unable To allocate shared memory (“large pool”,”Unknown object”,”Large Pool”,”PX msg Pool”).

By: paulk

paulk — Thu, 23 Aug 2007 16:37:27 +0000

We had some related problems last night, your post was invaluable.

From a message I sent to about 75 coworkers last night:

“Allegedly some types of DDL when done in parallel will change the degree of the object. ( google oracle poder parallel. Poder is very credible, pretty much my
rule of thumb is to read anything that he writes. )”

I know, I shouldn’t give away my trade secrets…

By: prodlife

prodlife — Sat, 23 Jun 2007 13:56:24 +0000

Oh, I’ve been bitten so hard by this issue.

We had a table that was really not very well designed and several times a month the number of objects changed from 3 to 30,000,000. So every once in a while we rebuilt indexes. One day we decided to save time and rebuild in parallel. We forgot the little note in the admin guide that said that rebuilding in parallel will change the degree of the index. A day later we knew this, and we also knew that having queries run in parallel can cause very severe performance issues.

Thanks for this article. It has good information, great examples and it also reassuring to know that I’m not the only DBA to make this mistake.