To: Matthew Wakeling <matthew@flymine.org>
cc: pgsql-performance@postgresql.org
Subject: Re: Disc space usage 
In-reply-to: <alpine.DEB.1.10.0810081443260.15851@aragorn.flymine.org> 
References: <alpine.DEB.1.10.0810081238380.15851@aragorn.flymine.org>
	<27956.1223471810@sss.pgh.pa.us>
	<alpine.DEB.1.10.0810081443260.15851@aragorn.flymine.org>
Comments: In-reply-to Matthew Wakeling <matthew@flymine.org>
	message dated "Wed, 08 Oct 2008 14:46:49 +0100"
Date: Wed, 08 Oct 2008 10:01:05 -0400
Message-ID: <28838.1223474465@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Matthew Wakeling <matthew@flymine.org> writes:
> On Wed, 8 Oct 2008, Tom Lane wrote:
>> Anything under $PGDATA/base that doesn't correspond to a live row in
>> pg_database is junk.

> So I can delete it? Might be safer to stop the db server while I do that 
> though.

In principle, at least, you shouldn't need to --- there shouldn't be any
buffers representing such files.

>> What PG version is this, anyway?

> Postgres 8.3.0

You should consider an update to 8.3.4.  A quick look in the post-8.3.0
CVS logs shows a couple of possibly relevant fixes:

2008-04-18 13:05  tgl

	* src/: backend/commands/dbcommands.c, include/port.h,
	port/dirmod.c (REL8_3_STABLE): Fix rmtree() so that it keeps going
	after failure to remove any individual file; the idea is that we
	should clean up as much as we can, even if there's some problem
	removing one file.  Make the error messages a bit less misleading,
	too.  In passing, const-ify function arguments.

2008-04-16 19:59  tgl

	* src/: backend/access/nbtree/nbtree.c,
	backend/access/nbtree/nbtutils.c, backend/access/transam/xlog.c,
	backend/commands/dbcommands.c, backend/port/ipc_test.c,
	backend/storage/ipc/ipc.c, include/access/nbtree.h,
	include/storage/ipc.h, include/utils/elog.h (REL8_3_STABLE): Repair
	two places where SIGTERM exit could leave shared memory state
	corrupted.  (Neither is very important if SIGTERM is used to shut
	down the whole database cluster together, but there's a problem if
	someone tries to SIGTERM individual backends.)	To do this,
	introduce new infrastructure macros
	PG_ENSURE_ERROR_CLEANUP/PG_END_ENSURE_ERROR_CLEANUP that take care
	of transiently pushing an on_shmem_exit cleanup hook.  Also use
	this method for createdb cleanup --- that wasn't a
	shared-memory-corruption problem, but SIGTERM abort of createdb
	could leave orphaned files lying around.

			regards, tom lane