To: "Bill Bartlett" <bbartlett@softwareanalytics.com>
cc: pgsql-admin@postgresql.org
Subject: Re: PG 8.0.4 - Does Dump + drop_db + reload reset XID to prevent
	wraparound?
In-reply-to: <001401c8d388$4b5d0720$e2171560$@com> 
References: <001401c8d388$4b5d0720$e2171560$@com>
Comments: In-reply-to "Bill Bartlett" <bbartlett@softwareanalytics.com>
	message dated "Sat, 21 Jun 2008 06:19:30 -0400"
Date: Sat, 21 Jun 2008 11:25:25 -0400
Message-ID: <18778.1214061925@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

"Bill Bartlett" <bbartlett@softwareanalytics.com> writes:
> In a PostgreSQL 8.0.4 environment, does dropping the database and
> reloading it completely reset the XID information that causes
> transaction ID wraparound?

I gather from the reference to "d:" that you're running on Windows.
You really, really, really need to put an urgent priority on getting
off 8.0.x.  (Or get off Windows, but I suppose I'm wasting my breath
suggesting that.)  The rename problem that you're hitting is fixed in
8.2 and up, as are a bunch of other Windows-specific problems that
will never be fixed in pre-8.2 branches, because we have abandoned
support for those branches on Windows.

As for the specific question, I think it'd work as long as template0
has never been unfrozen, but an initdb would be a lot more certain
--- and since you apparently have only one user database, there's
no difference in how much reload work you'd have to do.

Lastly, there is no need to use VACUUM FULL for wraparound protection;
plain VACUUM is sufficient.  The critical point though is that it has
to be a database-wide VACUUM (and done by a superuser); 8.0 doesn't
track this at a per-table grain, only per-database.

			regards, tom lane