Re: Statistics mismatch between n_live_tup and actual row count

On 12/06/2012 06:45 PM, tim_wilson wrote:
>    <http://postgresql.1045698.n5.nabble.com/file/n5735593/pg_drift.png>
>
>
>
> 86% of the updates are HOT updates!
>
> The difference between the first and second line of this image above is that
> 366 updates happened of which 299 where HOT. And a vacuum on the table was
> run.
> Look at the retuples number change after the vacuum!
>
> Open transactions... how does this impact those numbers?
>

The very simple version:
Because of MVCC a particular row may have many tuple versions
associated with it. If those tuple versions are visible to open
transactions they may not be marked for removal until those transactions
are completed.

Others on this list are more capable than I of filling in the fine detail.


--
Adrian Klaver
adrian.klaver@gmail.com