Kevin Grittner <Kevin(dot)Grittner(at)wicourts(dot)gov>
Cc:
Peter Eisentraut <peter_e(at)gmx(dot)net>, Daniel Farina <daniel(at)heroku(dot)com>,Greg Stark <stark(at)mit(dot)edu>,pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>,Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
On Thu, Mar 15, 2012 at 10:20:02AM -0500, Kevin Grittner wrote:
> Bruce Momjian <bruce(at)momjian(dot)us> wrote:
>
> > I think we have two choices --- either migrate the statistics, or
> > adopt my approach to generating incremental statistics quickly.
> > Does anyone see any other options?
>
> Would it make any sense to modify the incremental approach to do a
> first pass of any tables with target overrides, using the default
> GUC setting, and then proceed through the passes you describe for
> all tables *except* those? I'm thinking that any overrides were
> probably set because the columns are particularly important in terms
> of accurate statistics, and that running with different GUC settings
> will just be a waste of time for those tables -- if they have a high
> setting for any column, they will sample more blocks for every run,
> right?
I just added text telling users they might want to remove and re-add
those per-table statistics. I could try coding up something to store
and reset those values, but it is going to be complex, partly because
they are in different databases. I would need to create a pg_upgrade
schema in every database, store and reset the targets, and restore them
once complete.
I don't think it makes sense to do the custom targets first because it
would likely delay all tables from getting any statistics.
--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +
