From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
cc: Michael Paquier <michael@paquier.xyz>,
        Postgres hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: REINDEX INDEX results in a crash for an index of pg_class since
 9.6
In-reply-to: <20190502145045.mdrxilmm5bhq55ti@alap3.anarazel.de>
References: <20190430182741.s6o7mo5ebmgxxi5x@alap3.anarazel.de>
 <16277.1556649660@sss.pgh.pa.us>
 <20190430190308.kga2z5oz7vab7w22@alap3.anarazel.de>
 <17996.1556651503@sss.pgh.pa.us>
 <20190430221045.6slpi2vm2noda5jb@alap3.anarazel.de>
 <28313.1556663080@sss.pgh.pa.us> <28926.1556664156@sss.pgh.pa.us>
 <20190430225307.lvfvcf2776iywhw4@alap3.anarazel.de>
 <20190501013638.7u7fwr4zktbyyng5@alap3.anarazel.de>
 <8749.1556685814@sss.pgh.pa.us>
 <20190502145045.mdrxilmm5bhq55ti@alap3.anarazel.de>
Comments: In-reply-to Andres Freund <andres@anarazel.de>
	message dated "Thu, 02 May 2019 07:50:45 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <26299.1556809723.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Thu, 02 May 2019 11:08:43 -0400
Message-ID: <26300.1556809723@sss.pgh.pa.us>
Precedence: bulk

Andres Freund <andres@anarazel.de> writes:
> On 2019-05-01 00:43:34 -0400, Tom Lane wrote:
>> ...  What it's really currently doing at the
>> moment of the deadlock is cleaning out its temporary schema after the
>> client disconnected.

> I'm inclined to remove the tests from the backbranches, once we've
> committed a fix for the actual REINDEX issue, and most of the farm has
> been through a cycle or three. I don't think we'll figure out how to
> make them robust in time for next week's release.

Yeah, as I just said in my other message, I see no other alternative for
next week's releases.  We can leave the test in place in HEAD a bit
longer, but I don't really want it there for the beta either, unless we
can think of some better plan.

> I don't think we can really rely on the post-disconnect phase completing
> in a particularly deterministic time.

Exactly :-(

>> Another fairly interesting thing is that this log includes the telltale
>> 2019-05-01 05:24:48.887 CEST [97694:7] pg_regress/reindex_catalog CONTE=
XT:  while checking uniqueness of tuple (12,71) in relation "pg_class"
>> Why did I have to dig to find that information in HEAD?  Have we lost
>> some useful context reporting?  (Note this run is in the v10 branch.)

FWIW, as best I can reconstruct the sequence of events, I might just
not've looked.  I got an error and just assumed it was the same as what
we'd seen in the buildfarm; but now we realize that there were multiple
ways to get deadlocks, and only some of them would have shown this.
For the moment I'm willing to assume this isn't a real issue.

			regards, tom lane