Horton

Re: allow granting CLUSTER, REFRESH MATERIALIZED VIEW, and REINDEX

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Nathan Bossart <nathandbossart@gmail.com>
Cc: Jeff Davis <pgsql@j-davis.com>, Alvaro Herrera <alvherre@alvh.no-ip.org>, Pavel Luzanov <p.luzanov@postgrespro.ru>, pgsql-hackers@postgresql.org
Date: 2022-12-15T00:12:26Z
Lists: pgsql-hackers 
On Wed, Dec 14, 2022 at 03:29:39PM -0800, Nathan Bossart wrote:
> On Wed, Dec 14, 2022 at 11:05:13AM -0800, Jeff Davis wrote:
>> On Wed, 2022-12-14 at 10:16 -0800, Nathan Bossart wrote:
>>> Okay.  Should all the privileges governed by MAINTAIN apply to a
>>> relation's
>>> TOAST table as well?
>> 
>> Yes, I agree.
> 
> This might be tricky, because AFAICT you have to scan pg_class to find a
> TOAST table's main relation.

Ugh, yeah.  Are we talking about a case where we know the toast
information but need to look back at some information of its parent to
do a decision?  I don't recall a case where we do that.  CLUSTER,
REINDEX and VACUUM lock first the parent when working on it, and no
AEL is taken on the parent if doing directly a VACUUM or a REINDEX on
the toast table, so that could lead to deadlock scenarios.  Shouldn't
MAINTAIN be sent down to the toast table as well if that's not done
this way?

FWIW, I have briefly poked at that here:
https://www.postgresql.org/message-id/YZI+aNEnnpBASxNU@paquier.xyz
--
Michael

Commits

  1. Improve privilege documentation for maintenance commands.

  2. Fix cache lookup hazards introduced by ff9618e82a.

  3. Move bool parameter for vacuum_rel() to option bits.

  4. Fix MAINTAIN privileges for toast tables and partitions.

  5. Add grantable MAINTAIN privilege and pg_maintain role.