Re: "Orphaned" files after initdb

From: Andres Freund <andres(at)2ndquadrant(dot)com>
To: Magnus Hagander <magnus(at)hagander(dot)net>
Cc: PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: "Orphaned" files after initdb
Date: 2013-04-01 20:53:21
Message-ID: 20130401205321.GB7244@alap2.anarazel.de
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On 2013-04-01 21:24:06 +0200, Magnus Hagander wrote:
> On Mon, Apr 1, 2013 at 9:09 PM, Andres Freund <andres(at)2ndquadrant(dot)com> wrote:
> > Hi,
> >
> > On 2013-04-01 20:43:36 +0200, Magnus Hagander wrote:
> >> Maybe my thinking is just missing something obvious here. But looking
> >> at a fresh initdb, I'm seeing a number of files in the data directory
> >> for this database that don't have corresponding entries in pg_class. I
> >> looked for entries based on relfilenode in pg_class - am I missing
> >> some other way we put files in there?
> >>
> >> I'm seeing the same on both head and 9.1 as an example, but different oids.
> >>
> >> I feel I'm either missing something obvious, or we have a problem
> >> here? And somehow the first one seems more likely..
> >>
> >> The query I ran (yeah, i'm sure it can be written prettier, but this
> >> was quick-n-dirty):
> >>
> >> with t as (
> >> select * from pg_ls_dir('<datadir>/base/1') l(f)
> >> where f not in (
> >> select relfilenode::text from pg_class
> >> union all select relfilenode::text || '_vm' from pg_class
> >> union all select relfilenode::text || '_fsm' from pg_class)
> >> )
> >> select f,
> >> size,
> >> modification
> >> from t, pg_stat_file('<datadir>/base/1/' || f)
> >> order by 1;
> >
> > You're missing nailed tables which don't have a valid relfilenode (but
> > InvalidOid/0) in pg_class.relfilenode. You need to query the refilenode like
> > SELECT pg_relation_filenode(pg_class.oid) FROM pg_class.
>
> Ha. Of course. I knew it was me getting it wrong.
>
> You'd get extra happy if only you knew how many pretty senior pg folks
> i pestered about that one on irc before posting here :D

Most of those probably didn't spend too much time on developing wal->logical
changes transformations ;)

For some real reason to send this email: Even if you add pg_relation_filenode()
to the above query, its still not correct. You also need to disambiguate by
tablespace. Just for the case youre using it for something more interesting
than a freshly initdbed cluster.

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Magnus Hagander 2013-04-01 20:54:13 Re: "Orphaned" files after initdb
Previous Message Bruce Momjian 2013-04-01 20:32:49 Re: Page replacement algorithm in buffer cache