Re: Postgresql performance in production environment

On 19/08/07, Magnus Hagander <magnus@hagander.net> wrote:
> Phoenix Kiula wrote:

.....snipped....

> > I can merrily increase the "max_fsm_pages" directive, but the manual
> > also caveats that with "this can use more system V memory than
> > available on your system". My full verbose vacuum info below includes
> > the line:
>
> Do you actually run VACUUM FULL, or do you just mean you run VACUUM over
> the full database? If you run VACUUM FULL, you need to stop doing that
> :-) However, you will need to run it at least once over the whole
> database once you've fixed your max_fsm_pages setting.



No we only do a "vacuum analyze" when we do something manually.
Otherwise, it's all "autovacuum". Never done  a "vacuum full" --
should we do one now given that we've overrun our max_fsm_pages?



> > [-------------
> > INFO:  "traders": scanned 3000 of 199396 pages, containing 40775 live
> > rows and 0 dead rows; 3000 rows in sample, 2710124
> > estimated total rows
> > -------------]
>
> There should be a line like this at the end of a "VACUUM VERBOSE" command:
> INFO:  free space map contains 33 pages in 74 relations
> DETAIL:  A total of 1184 page slots are in use (including overhead).
> 1184 page slots are required to track all free space.
> Current limits are:  153600 page slots, 1000 relations, using 965 kB.
> VACUUM
>


Nope, there's no line that includes such useful info. The entire
verbose output was included in my note. I did not see the words
"Current limits are". Do I need to enable something in the conf file
to get more verbose output? "debug2", "debug3" -- kind of stuff?


TIA