Re: [PATCH] Unremovable tuple monitoring - Mailing list pgsql-hackers

From Robert Haas
Subject Re: [PATCH] Unremovable tuple monitoring
Date
Msg-id CA+TgmoZktbfNLM9OU=-oecRvHdL_mmCmPEWKd9GMDVfh7AYZxA@mail.gmail.com
Whole thread Raw
In response to Re: [PATCH] Unremovable tuple monitoring  (Royce Ausburn <royce.ml@inomial.com>)
List pgsql-hackers
On Tue, Nov 15, 2011 at 10:02 PM, Royce Ausburn <royce.ml@inomial.com> wrote:
> Fair enough -- someone knowledgable could set that up if they wanted.  My goal was mostly to have something helpful
inthe logs.  If that's not something postgres wants/needs Ill drop it =) 

IMHO, it's generally not desirable to provided hard-coded
functionality that could easily be duplicated in user-space.  We can't
really know whether the user wants this warning at all, and if so what
the cut-off ought to be for a "too old" transaction, and how often the
warning should be emitted.  It's far more flexible for the user to set
it up themselves.  Log clutter is a major problem for some users, and
we shouldn't add to it without a fairly compelling reason.

Just to take an example of something that *couldn't* easily be done in
userspace, suppose VACUUM emitted a NOTICE when the number of
recently-dead tuples was more than a certain (configurable?)
percentage of the table.  That's something that VACUUM could calculate
(or at least estimate) while scanning the table, but it wouldn't be so
easy for the user to get that same data, at least not cheaply.  Now
I'm not sure we need that particular thing; it's just an example.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company


pgsql-hackers by date:

Previous
From: Robert Haas
Date:
Subject: Re: [PATCH] Unremovable tuple monitoring
Next
From: Robert Haas
Date:
Subject: Re: Adding Node support in outfuncs.c and readfuncs.c