Re: Query ASH for Undo Blocks

This last discrepancy could be explained by the having some variation in the number of rows per ID. If there are a couple of IDs with N times the typical number rows then (potentially) there will be N times as many queries for each of those value, and each one has to work through N rows, giving an N^2 impact on the buffer gets. What does "select id, count(*)" look like ?

Regards
Jonathan Lewis

On Mon, 20 Jun 2022 at 11:24, Lothar Flatz <l.flatz_at_bluewin.ch> wrote:

>
> As you can see we have 54 buffer/execution.
>

> Still during the batch run there seems to be something not explained.
> Even if just the first column in the index is used we have 112885 Leaf
> blocks devided by 58 distinct keys = 2015 + 4 =2019 buffer on average and
> not 3700 as indicated below.
> Thus, it looks like things were different on production and undo chain
> seems a logical candidate.
>
>

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