From: Tom Lane <tgl@sss.pgh.pa.us>
To: Peter Geoghegan <pg@bowt.ie>
cc: pgsql-committers <pgsql-committers@lists.postgresql.org>
Subject: Re: pgsql: Make heap TID a tiebreaker nbtree index column.
In-reply-to: 
 <CAH2-Wzny8NVzgvx=n6nBE9xNEf-W5o=ri2Dq3kspYmgDHtrHDw@mail.gmail.com>
References: <E1h6eep-0001Mw-Vd@gemulon.postgresql.org>
 <CAH2-WzmFdW-e9+uG5BqzuW7YWBdL9=Vrwgmt5p6vndCeKJ5Z4w@mail.gmail.com>
 <13999.1553105283@sss.pgh.pa.us>
 <CAH2-Wznu+i765yrRcpKC0m9N2LEFWg=stjF73U7XmakSXWDDAg@mail.gmail.com>
 <CAH2-WzkCQ_MtKeOpzozj7QhhgP1unXsK8o9DMAFvDqQFEPpkYQ@mail.gmail.com>
 <24277.1553114810@sss.pgh.pa.us> <16667.1553116435@sss.pgh.pa.us>
 <17604.1553118255@sss.pgh.pa.us>
 <CAH2-WzmfxXs3bSX6T3DP1wTEWbyhkTiyPZ6o9t2DmcHbBtsCNA@mail.gmail.com>
 <17845.1553118645@sss.pgh.pa.us> <8748.1553119737@sss.pgh.pa.us>
 <CAH2-Wz=U8tcaP1wx4t2n8-grz-9ar2fPJeYu9MExSyOu_NVmiA@mail.gmail.com>
 <CAH2-Wzm0nQyV3x06_XEuH62tda8X6op-4r29uj8VA0XambsUpw@mail.gmail.com>
 <21983.1553138224@sss.pgh.pa.us>
 <CAH2-WzmMakut=D8=TnmzhGtB+ELbb6k84LYNW28C8MZ2D-ZoJw@mail.gmail.com>
 <5446.1553272377@sss.pgh.pa.us> <6226.1553274783@sss.pgh.pa.us>
 <CAH2-Wzny8NVzgvx=n6nBE9xNEf-W5o=ri2Dq3kspYmgDHtrHDw@mail.gmail.com>
Comments: In-reply-to Peter Geoghegan <pg@bowt.ie>
	message dated "Fri, 22 Mar 2019 10:23:58 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <4641.1553276304.1@sss.pgh.pa.us>
Date: Fri, 22 Mar 2019 13:38:24 -0400
Message-ID: <4642.1553276304@sss.pgh.pa.us>
Precedence: bulk

Peter Geoghegan <pg@bowt.ie> writes:
> On Fri, Mar 22, 2019 at 10:13 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> 1a seems a bit ambitious but maybe it's worth doing, considering
>> that right now there's a non-negligible chance of OOM if you
>> try to drop a role that owns a huge number of objects, or just
>> plain failure due to the stringinfo buffer for the message getting
>> past MaxAllocSize.  Sending a gigabyte-sized message to the server
>> log could be pretty unfriendly in some contexts, too.

> All of these options seem acceptable. However, the problem is unlikely
> to get any worse, so going to the trouble of option 1 or 1a might not
> be the best use of time.

Yeah.  After thinking a bit more, the OOM hazard is probably pretty
far-fetched --- even a role owning hundreds of thousands of objects
wouldn't accumulate more than a small number of megabytes of DETAIL.
You could still argue that sending such a message to the server log
is a bad idea, perhaps, but I'm willing to let it go until we see
actual field complaints about it.

So that means that the de-revert is probably the best option for
now.  Will you do the honors?

			regards, tom lane