Re: Inserts and row chaining

zgeist_at_ee.ee.net (Unknown) wrote:

>
>
>Greetings all -
>
>I am aware of PCTFREE's function in regards to saving space for UPDATEs,
>but I am unable to determine via Oracle docs or O'Reilly's performance
>book what exactly happens with INSERTs.
>
>A co-worker insists that Oracle just keeps inserting bytes until PCTFREE
>is reached, at which point it continues writing the row into another data
>block. I thought Oracle might check to see if the row could be INSERTed
>and, if not, take that block off the free list. Does it then attempt
>to find a less-populated block that can hold the entire row ? I realize
>with LONGs and LONG RAWs, it would never find such a case.
>
>However, I would think that just inserting bytes until PCTFREE was
>reached (and then continuing in another block) would result in a ton of
>chained rows. What am I missing ?
>
>I am totally clear on UPDATEs, just not INSERTs.
>
>Thanks in advance for any assistance.
>
>Brett Cunningham
>zgeist_at_ee.net
>

According to the instructor in the DBA-1 class that I took, it won't spread a row across multiple blocks unless it has too. If the block has not reached the pctfree threshold, it'll insert it into another block. (i.e. It'll grab the free block in the extent, or allocate a new extent.)

--
Chuck Hamilton
chuckh_at_dvol.com

This message delivered by electronic sled dogs. WOOF!