Comments on: Test Case: Oracle Event 10224 – index block split/delete trace http://oraclue.com/2008/10/28/test-case-oracle-event-10224-index-block-splitdelete-trace/ Oracle internals, debugging and undocumented features Thu, 27 Oct 2011 15:05:30 +0000 hourly 1 http://wordpress.com/ By: oraclue http://oraclue.com/2008/10/28/test-case-oracle-event-10224-index-block-splitdelete-trace/#comment-49 Mon, 03 Nov 2008 15:34:06 +0000 http://oraclue.wordpress.com/?p=294#comment-49 Hi Andrew,

Thanks for comment.This open new topic.

To prove it we can use oracle event 10081 – segment HWM advanced and dump the index with oradebug dump treedump command.

Regards,

Miladin

]]> By: Andre http://oraclue.com/2008/10/28/test-case-oracle-event-10224-index-block-splitdelete-trace/#comment-48 Sun, 02 Nov 2008 23:44:54 +0000 http://oraclue.wordpress.com/?p=294#comment-48 Hi, Miladin,

UNUSED_BLOCKS indicates how many blocks are above the high-watermark (hwm), i.e., have never contained data.

When the rows are deleted some blocks are put in the free-list because they now have space for new rows. However, as they have already been used, and therefore are below the hwm, the don’t count as unused_blocks.

When you rebuild the index, the data is reorganised and the index recreated from scratch, and then you have more pristine blocks (above the hwm) that appear as UNUSED.

Cheers,
Andre

]]>