Date: Tue, 3 Jan 2017 10:59:03 -0600
From: Justin Pryzby <pryzby@telsasoft.com>
To: Robert Haas <robertmhaas@gmail.com>
Cc: "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Subject: Re: ALTER TABLE .. ALTER COLUMN .. ERROR: attribute .. has
 wrong type
Message-ID: <20170103165903.GP32031@telsasoft.com>
References: 
 <CALcU3kLseKzasSpnAhA7zQgd6Oebhqmb3SS-dPWA2R-K=TbGgw@mail.gmail.com>
 <20161012152505.GD12747@telsasoft.com>
 <20170102225618.GA10071@telsasoft.com>
 <20170103003240.GM32031@telsasoft.com>
 <CA+TgmoY9W99m3cW9-4Kjnw4KO+PRqF8Lh_AqWh8Qoop9ZhnUdw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <CA+TgmoY9W99m3cW9-4Kjnw4KO+PRqF8Lh_AqWh8Qoop9ZhnUdw@mail.gmail.com>
User-Agent: Mutt/1.5.23 (2014-03-12)
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

On Tue, Jan 03, 2017 at 11:45:33AM -0500, Robert Haas wrote:
> > ts=# begin; drop view umts_eric_ch_switch_view, eric_umts_rnc_utrancell_view, umts_eric_cell_integrity_view; ALTER TABLE eric_umts_rnc_utrancell_metrics ALTER COLUMN PMSUMPACKETLATENCY_000 TYPE BIGINT USING PMSUMPACKETLATENCY_000::BIGINT;
> > BEGIN
> > DROP VIEW
> > ERROR:  attribute 424 has wrong type
> > DETAIL:  Table has type smallint, but query expects integer.
> > ts=#
> >
> > ts=# begin; drop view umts_eric_ch_switch_view, eric_umts_rnc_utrancell_view, umts_eric_cell_integrity_view; ALTER TABLE eric_umts_rnc_utrancell_metrics ALTER COLUMN PMSUMPACKETLATENCY_000 TYPE BIGINT ;
> > BEGIN
> > DROP VIEW
> > ALTER TABLE
> > ts=#
> >
> > Is it useful to send something from pg_attribute, or other clues ??
> 
> So, are these errors reproducible?  Like, if you create a brand new

I can cause the error at will on the existing table, but I wouldn't know how to
reproduce the problem on a new table/database.  I'm guessing it has something
to do with dropped columns or historic alters (which I mentioned are typically
done separately on child tables vs their parent).

Since it's happened 3 times now on this table, but not others on this database,
I would guess it's an "data issue", possibly related to pg_upgrades.  IOW it
may be impossible to get into this state from a fresh initdb from a current
version.

I considered that perhaps it only affected our oldest tables, and would stop
happening once they were dropped, but note this ALTER is only of a parent and
its 3 most recent children.  So only the empty parent could be described as
"old".

Justin


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers