Re: Partition Indexes

Richard Foote wrote:

> Are you sure that the first query isn't run in parallel as well ?

Yes. 100% Not a single PQ is running when the hint PARALLEL_INDEX is not specified.

Awe damn..just remembered. I must update the stats and try again. Slipped my mind. One of the things that I still need to try is to see if updated/correct stats solve the PQ problem. There are usually issues around PQ when stats is missing or old, with Oracle not knowing how to dish the work out evenly.

But this still does not explain the discrepancy in performance - see the SQL1 and SQL2 samples I posted in this thread when replying to Jon. Something just does not seem to work right with partition indexes, when dealing with the partition tables as oppose to an individual partition (which in turns makes even less sense as it should be doing a simple partition iteration when processing the complete table, and actually deal with one partition at time).

> There are some bugs regarding reading tables in parallel with local
> indexes that have been discussed recently here (or in
> comp.databases.oracle.misc).

Drats. I must have missed it. I will hit Google Groups again to see what I can dig up.

Can you recall if these were dinkum Oracle bugs, or user/dba introduced ones?

--
Billy