From: Tom Lane <tgl@sss.pgh.pa.us>
To: Stephen Frost <sfrost@snowman.net>
cc: pgsql-hackers@lists.postgresql.org
Subject: Re: More parallel pg_dump bogosities
In-reply-to: <21154.1535392691@sss.pgh.pa.us>
References: <19784.1535390902@sss.pgh.pa.us>
 <20180827173932.GY3326@tamriel.snowman.net> <21154.1535392691@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Mon, 27 Aug 2018 13:58:11 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <11449.1535483506.1@sss.pgh.pa.us>
Date: Tue, 28 Aug 2018 15:11:46 -0400
Message-ID: <11450.1535483506@sss.pgh.pa.us>
Precedence: bulk

... just when you thought it was safe to go back in the water ...

Doesn't pg_backup_archiver.c's identify_locking_dependencies() need to
treat POLICY and ROW SECURITY items as requiring exclusive lock on
the referenced table?  Those commands definitely acquire
AccessExclusiveLock in a quick test.

I haven't looked hard, but I'm suspicious that other recently-added
dump object types may have been missed here too, and even more
suspicious that we'll forget this again in future.  I wonder if we
shouldn't invert the logic, so that instead of a blacklist of object
types that we assume need exclusive lock, we keep a whitelist of
object types that are known not to (which might be just INDEX,
not sure).  That way, we'd at least be failing in a safe direction.

			regards, tom lane