Re: Index Rebuilds

  • From: Stalin <stalinsk@xxxxxxxxx>
  • To: "Christian Antognini" <Christian.Antognini@xxxxxxxxxxxx>
  • Date: Sun, 30 Jul 2006 21:54:17 -0700

Thanks Christain and Alex. I didn't know index_ffs would scan all the
blocks in freelists. Now it makes sense.

Yes, regular coalesce after rebuild would solve the problem, but is
there a permanent solution for these ever growing index other than
regular coalesce.

Thanks again.

On 7/30/06, Christian Antognini <Christian.Antognini@xxxxxxxxxxxx> wrote:
Stalin

> LBLKS: 66653
> BlVL: 3
> CF: 679426
> NROWS: 829734
> AVG_DATA_BLK/KEY: 2
>
> Coalescing the index got the lblks to 11k and the desired plan
> however, the performance is still under water.

Coalesce reuse the index structure and puts the freed block in the free
list. I.e. I guess you have 55k blocks on the free list (a dump of the
segment can be used to confirm this point). Since the FFS reads all the
blocks up to the high water mark, also the free blocks are read.

> Any suggestions as to what i should be looking further. I really don't
> want to rebuild the index :)

To decrease the index size you have to rebuild it. In the future, if the
number of leaf blocks increase constantly (in the way you described),
regular coalesce may "solve" the problem, i.e. let reuse the already
allocated index blocks instead of allocating new ones.


HTH Chris

--
//www.freelists.org/webpage/oracle-l


Other related posts: