Re: Press Release - Mailing list pgsql-advocacy

From scott.marlowe
Subject Re: Press Release
Date
Msg-id Pine.LNX.4.33.0310291604390.22178-100000@css120.ihs.com
Whole thread Raw
In response to Re: Press Release  ("Joshua D. Drake" <jd@commandprompt.com>)
Responses Re: Press Release
List pgsql-advocacy
On Wed, 29 Oct 2003, Joshua D. Drake wrote:

> First of all, be aware that we have already collected half the translations
>
> >for the press kit.  So at this point, we can only cut paragraphs and not
> >edit.  These comments would have been more timely a month ago ....
> >
> >
> >
> If only I had a nickel :)
>
> >>I believe is false. As long as you have to vacuum the above is not true.
> >>
> >>
> >
> >How?  Vacuuming does not require the database to be offline.  Vacuum full
> >does, but that can be eliminated with proper tuning.
> >
> >
> >
> No but vacuum will cause your machine to grind to a crawl. Try telling a
> customer
> that is pushing 240,000 transactions an hour, 24 hours a day to run a
> Vacuum.
> They are not pleased.

The autovacuum daemon makes this situation as good as it's likely to ever
get without some form of "vacuum this tuple at your leisure when you've
got free bandwidth" setting for an update/delete.  It only vacuums the
tables that have actually been changing, and you can set the period and
what not so that it runs often enough that it never has too much to do.
Keep in mind, the garbage collection HAS to happen sometime, so even if it
were rolled 100% efficiently into the individual update/delete itself, it
would still cost, just with slower transactions.


I agree though that the paragraph may make it sound like automatic
vacuuming is a default part of the install when it isn't.


pgsql-advocacy by date:

Previous
From: "scott.marlowe"
Date:
Subject: Re: Press Release
Next
From: Josh Berkus
Date:
Subject: Re: Press Release