I wrote:
> [ scratches head... ] That example fails to misbehave for me in 8.3.6,
Oh, wait, you omitted a step from the example: the ALTER has to be done
as somebody other than the table owner (eg a superuser, else you'd not
have the needed permissions).
regression=# create user a;
CREATE ROLE
r
Cott Lang writes:
> I fiddled around a while, found the problem, and I can repeat it at
> will. It's actually related to type changes we've made over the months
> since the restore.
[ scratches head... ] That example fails to misbehave for me in 8.3.6,
and I'm pretty certain there have been no r
On Fri, 2009-02-20 at 16:35 -0500, Tom Lane wrote:
> Can you see any pattern or common characteristic to the tables whose
> toast pg_type rows failed to change owner? I'm not sure what to look
> for exactly, but similarities in the column contents might be a
> possibility. Also, can you tell if
Cott Lang writes:
> I found a handful of other pg_types with an unusual owner and in every
> case, the toast type is owned by the user that created the database via
> full pg_restore some months ago.
> pg_restore should have created the table as the user running it, and
> immediately done an ALT
Tom,
Thanks for the inspiration - I've fixed them manually.
I spent a few minutes trying to recreate the obvious test case, and it
all works as designed. I reviewed our logs from the user removal
yesterday, and the tables linked to these toast tables did not have
ownership changed yesterday, so
On Fri, 2009-02-20 at 11:25 -0800, Joshua D. Drake wrote:
> >
>
> You can use alter type to change the owner of the type to a valid user
> but see above. Something is wrong.
That's what I thought too, but we tried that first with these results:
# alter type pg_toast.pg_toast_80075 OWNER TO post
Cott Lang writes:
> The owner of the actual table and index is correct, only the type has an
> invalid owner. I have thus far avoided the temptation to try a manual
> update...
That's probably your best bet.
> Since Postgres now prevents you from dropping users owning objects, is
> this a bug, o
On Fri, 2009-02-20 at 12:01 -0700, Cott Lang wrote:
> We're running 8.3.5 on RHEL4 x86_64.
>
> We removed a user yesterday and were greeted with warnings from pg_dump
> this morning. :)
>
> pg_dump: WARNING: owner of data type "pg_toast_80075" appears to be
> invalid
>
> The usual archives and
We're running 8.3.5 on RHEL4 x86_64.
We removed a user yesterday and were greeted with warnings from pg_dump
this morning. :)
pg_dump: WARNING: owner of data type "pg_toast_80075" appears to be
invalid
The usual archives and google searches produced mainly 8.0 and earlier
incidents and suggeste
