Re: HOT - preliminary results - Mailing list pgsql-hackers

From Simon Riggs
Subject Re: HOT - preliminary results
Date
Msg-id 1172873613.3760.1702.camel@silverbirch.site
Whole thread Raw
In response to Re: HOT - preliminary results  (Alvaro Herrera <alvherre@commandprompt.com>)
Responses Re: HOT - preliminary results  (Alvaro Herrera <alvherre@commandprompt.com>)
List pgsql-hackers
On Fri, 2007-03-02 at 18:37 -0300, Alvaro Herrera wrote:
> Bruce Momjian escribió:
> > Simon Riggs wrote:
> > > CLUSTER does not need changes for HOT, as things stand currently, mainly
> > > because its MVCC behaviour is broken.
> >               ------------------------
> >
> > That's oddly discouraging.  :-)
>
> Apparently no one has been bothered enough to fix CLUSTER.

We learned from Csaba just the other day that this is a backdoor used on
production systems. I've not seen anyone admit it before, even though
I've seen it discussed.

People know that CLUSTER works better than VACUUM FULL and they use
that. The reason it hasn't been fixed is because its useful, I observe.

The oddly discouraging bit is that VACUUM FULL is only second best at
the thing its designed to achieve (compaction). Thats why I'm not
enthralled by the prospect of adding code to make VACUUM FULL work with
HOT - ISTM a good opportunity to make it work better.

--  Simon Riggs              EnterpriseDB   http://www.enterprisedb.com




pgsql-hackers by date:

Previous
From: Larry Rosenman
Date:
Subject: Re: Possible Bug: high CPU usage for stats collector in 8.2
Next
From: Bruno Wolff III
Date:
Subject: Re: Is there a way to run heap_insert() AFTER ExecInsertIndexTuples() ?