On Tue, Aug 26, 2014 at 5:44 PM, Keith Fiske <ke...@omniti.com> wrote:
> So I recently wrote this script to help us in monitoring for table/index > bloat > > https://github.com/keithf4/pg_bloat_check > > I based my query off of the one contained in check_postgres, since I > thought it seemed fairly accurate with what I was expecting to get back. > > > https://github.com/bucardo/check_postgres/blob/master/check_postgres.pl#L3512 > > A client however, showed me that running the bloat check from heroku > brings back some additional results. The additional results are actually > tables we were kind of expecting to see that the check_postgres query does > not return. > > I've tried decoding these queries to understand what they're actually > doing, but they're both quite a bit over my head. I did notice one thing in > the check_postgres one: > > CASE WHEN v ~ 'mingw32' OR v ~ '64-bit' THEN 8 ELSE 4 END AS ma > > Is that a typo? It seems to be looking for the 32bit version of mingw but > the 64-bit version when compiled elsewhere. The heroku version just sets > this value to 4 outright (since they know what they're running). I tried > changing this value to see if it made a difference, but it doesn't seem to. > > Anyone that's better able to evaluate which one of these two bloat check > queries is more accurate able to provide some assistance? > > -- > Keith Fiske > Database Administrator > OmniTI Computer Consulting, Inc. > http://www.keithf4.com > Helps if I include the query from Heroku https://github.com/heroku/heroku-pg-extras/blob/master/lib/heroku/command/pg.rb#L438-L504 -- Keith Fiske Database Administrator OmniTI Computer Consulting, Inc. http://www.keithf4.com
