Index Usage and Monitoring...

Thought of blogging about this long ago but never took a step and now is the time to put it up here instead of keep digging my repository for queries.

Creating an Index is the first solution for every developer to overcome the issue :)
But once its created its way too hard to decide whether to keep it or drop it after days go by...

How do we know if the particular index is being used or not? Big Question...

Well, one can turn on monitoring on index:

alter index index_name monitoring usage;

Once done, you can check v$object_usage to see if the index is being used or not.
Great. But how often? Sorry view wont have answer for that.

Below query helps in finding how many times the index is being used (and one can digg thru more to find the actual query also by using sql_id):

select di.table_name "Table Name", di.index_name "Index Name",
       round(sum((ds.bytes)/1024/1024),2) "Index Size -MB",
       dhsp.options "Index Operation", count(dhsp.options) "No# of Executions"
  from dba_hist_sql_plan dhsp, dba_indexes di, dba_segments ds
 where di.owner = user
   and dhsp.object_name(+) = di.index_name
   and dhsp.object_owner(+) = di.owner
   and di.index_name = ds.segment_name
   and di.owner = ds.owner
   and not exists (select 1 from dba_constraints dc where dc.constraint_name = di.index_name and dc.constraint_type = 'P')
   --and di.table_name in ('TABLE1', 'TABLE2')
 group by di.table_name, di.index_name, dhsp.options
 having count(dhsp.options) = 0
 order by 1,2,3;

As you can see, the above query is filtering for count = 0 meaning, I am only interested in showing indexes which were never used.

Note: Recently executed queries wont show up in DBA_HIST_SQL_PLAN so replace with V$SQL_PLAN. Also check the min(timestamp) from dba_hist_sql_plan to see the retention of the history. 

ORA-02050: transaction 11.9.23629 rolled back, some remote DBs may be in-doubt is local tran 11.9.23629 (hex=0b.09.5c4d)

Full stack of error message:
ORA-02050: transaction 11.9.23629 rolled back, some remote DBs may be in-doubt
  is local tran 11.9.23629 (hex=0b.09.5c4d)
ORA-02051: another session or branch in same transaction failed or finalized
  insert pending collecting tran, scn=7859252260641 (hex=725.dff06f21)

Developer reported this error while he is trying to perform an update.

First thought it might be due to exceeded connections over db links.

Later found it was due to Materialized View Log that was created on the base table on which the update was being performed. This log was not being refreshed due to some reasons and thus causing this failure.

Check the alert log for any additional/following errors for this ORA-02050 messages:

Resetting HWM (High Water Mark) on a Table...

The high-water mark is the level at which blocks have never been formatted to receive data.

Prior to 10g, only option to reset HWM (High Water Mark) was to either truncate data and re-insert or export import.
With 10g, there is an option that can be used to reset HWM on a table using "shrink space".

eg.,
alter table table_name shrink space; --can be used cascade along with it (Read the manual)

Here is an example showing how HWM takes place and how this option helps to reset it back:

create table hm_test as select *from all_tables where 1=2;

select blocks, extents
  from dba_segments
 where segment_name='HM_TEST';

--no records (read this to see why)

insert into hm_test select *from all_tables where rownum < 2;

commit;

select blocks, extents
  from dba_segments
 where segment_name='HM_TEST';

--blocks  extents
--8       1

analyze table hm_test compute statistics;

select blocks,empty_blocks,num_rows from dba_tables where table_name='HM_TEST';

--Querying dba_tables for 'Blocks' and 'Empty_blocks' should give the high water mark.
--
--Blocks -- > Number blocks that has been formatted to recieve data
--Empty_blocks ---> Among the allocated blocks, the blocks that were never used

--BLOCKS  EMPTY_BLOCKS    NUM_ROWS
--5                3                                 1

insert into hm_test select *from all_tables;

commit;

select blocks, extents
  from dba_segments
 where segment_name='HM_TEST';

--blocks  extents
--88        11

analyze table hm_test compute statistics;

select blocks,empty_blocks,num_rows from dba_tables where table_name='HM_TEST';

--BLOCKS  EMPTY_BLOCKS    NUM_ROWS
--80              8                                 2285

--now delete the records to see if empty_blocks number changes

delete hm_test;

commit;

select blocks,empty_blocks,num_rows from dba_tables where table_name='HM_TEST';

--BLOCKS  EMPTY_BLOCKS    NUM_ROWS
--80              8                                 2285

analyze table hm_test compute statistics;

select blocks,empty_blocks,num_rows from dba_tables where table_name='HM_TEST';
--BLOCKS  EMPTY_BLOCKS    NUM_ROWS
--80              8                                 0

--as you see empty_blocks is still 8

--To determine the exact number of blocks that contain data i.e. space used by table below the high water mark, query rowid and get the first and last used block from the rowid.
select count (distinct dbms_rowid.rowid_block_number(rowid)) "used blocks" from hm_test;

--0

--This works fine if only one file is used for the segment.
--If more files, we need to include the file number in some way, for instance:

select count (distinct dbms_rowid.rowid_block_number(rowid)||'-'||dbms_rowid.rowid_relative_fno (rowid)) "used blocks" from hm_test;

--o

--resetting HWM

alter table hm_test shrink space compact cascade;

--ORA-10636: ROW MOVEMENT is not enabled

alter table hm_test enable row movement;

alter table hm_test shrink space compact cascade;

select blocks,empty_blocks,num_rows from dba_tables where table_name='HM_TEST';

--BLOCKS  EMPTY_BLOCKS    NUM_ROWS
--80              8                                  0

analyze table hm_test compute statistics;

select blocks,empty_blocks,num_rows from dba_tables where table_name='HM_TEST';

--BLOCKS  EMPTY_BLOCKS    NUM_ROWS
--80              8                                  0

--still no change

--now lets to shrink space cascade

alter table hm_test shrink space cascade;

select blocks,empty_blocks,num_rows from dba_tables where table_name='HM_TEST';

--BLOCKS  EMPTY_BLOCKS    NUM_ROWS
--80              8                                  0

analyze table hm_test compute statistics;

select blocks,empty_blocks,num_rows from dba_tables where table_name='HM_TEST';

--BLOCKS  EMPTY_BLOCKS    NUM_ROWS
--1                7                                 0