From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Andres Freund <andres@anarazel.de>
Cc: Tomas Vondra <tomas.vondra@2ndquadrant.com>
Cc: Andrey Salnikov <andrey.salnikov@dataegret.com>
Cc: pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #16125: Crash of PostgreSQL's wal sender during logical replication
Date: Wed, 27 Nov 2019 11:22:04 -0300
Message-ID: <20191127142204.GA13686@alvherre.pgsql> (raw)
In-Reply-To: <20191119232849.qhtuhq4udrqrz4he@alap3.anarazel.de>
On 2019-Nov-19, Andres Freund wrote:
> > > 2525:rmgr: Heap len (rec/tot): 1671/ 1671, tx: 1667601530, lsn: 25EE/D66F5508, prev 25EE/D66F54C8, desc: INSERT off 2, blkref #0: rel 1663/13018/88964792 blk 18475
> > > 2532:rmgr: Heap len (rec/tot): 54/ 54, tx: 1667601530, lsn: 25EE/D66F7EF0, prev 25EE/D66F7EA0, desc: DELETE off 2 , blkref #0: rel 1663/13018/88964792 blk 18475
> Hm. I don't think it necessarily means that. compute_new_xmax_infomask(), when
> called from heap_delete(), will e.g. set HEAP_KEYS_UPDATED when
>
> if (old_infomask & HEAP_XMAX_INVALID)
>
> is true. Which is the most common case. I actually find it *more*
> curious to *not* see that for a delete. I've not traced fully through
> it, yet.
>
>
> Alvaro, it's not clear to me whether the the multixact codepaths in
> compute_new_xmax_infomask() actually work as expected for deletes.
Hmm ... I think you're complaining about failing to set the
HEAP_KEYS_UPDATED bit when the delete transaction is the same
as the insert transaction.
Grepping the new waldump file the OP sent for relfilenode 88964792
I saw this:
rmgr: Heap len (rec/tot): 1671/ 1671, tx: 1667601529, lsn: 25EE/D66F1A78, prev 25EE/D66F1A38, desc: INSERT off 5, blkref #0: rel 1663/13018/88964792 blk 18460
rmgr: Heap len (rec/tot): 1671/ 1671, tx: 1667601530, lsn: 25EE/D66F5508, prev 25EE/D66F54C8, desc: INSERT off 2, blkref #0: rel 1663/13018/88964792 blk 18475
rmgr: Heap len (rec/tot): 48/ 48, tx: 1667601529, lsn: 25EE/D66F7E10, prev 25EE/D66F6778, desc: HEAP_CONFIRM off 5, blkref #0: rel 1663/13018/88964792 blk 18460
rmgr: Heap len (rec/tot): 54/ 54, tx: 1667601530, lsn: 25EE/D66F7EF0, prev 25EE/D66F7EA0, desc: DELETE off 2 , blkref #0: rel 1663/13018/88964792 blk 18475
rmgr: Heap len (rec/tot): 54/ 54, tx: 1667601536, lsn: 25EE/D6BA3BC0, prev 25EE/D6BA3B70, desc: LOCK off 5: xid 1667601536: flags 0 LOCK_ONLY EXCL_LOCK , blkref #0: rel 1663/13018/88964792 blk 20806
rmgr: Heap len (rec/tot): 3955/ 3955, tx: 1667601536, lsn: 25EE/D6BA3D00, prev 25EE/D6BA3CB0, desc: UPDATE off 5 xmax 1667601536 ; new off 2 xmax 0, blkref #0: rel 1663/13018/88964792 blk 18491, blkref #1: rel 1663/13018/88964792 blk 20806
rmgr: Heap len (rec/tot): 54/ 54, tx: 1667601538, lsn: 25EE/D6DE53D0, prev 25EE/D6DE33B0, desc: LOCK off 5: xid 1667601538: flags 0 LOCK_ONLY EXCL_LOCK , blkref #0: rel 1663/13018/88964792 blk 18460
rmgr: Heap len (rec/tot): 6563/ 6563, tx: 1667601538, lsn: 25EE/D6DE5408, prev 25EE/D6DE53D0, desc: UPDATE off 5 xmax 1667601538 ; new off 3 xmax 0, blkref #0: rel 1663/13018/88964792 blk 18499, blkref #1: rel 1663/13018/88964792 blk 18460
I wonder about the HEAP_CONFIRM there ... it's in a completely different
transaction of course, so it's not related, but note that the
corresponding insert (25EE/D66F1A78) looks like a regular INSERT, it
doesn't indicate that it's a speculative insert, which it must have
been or there wouldn't be a confirm. So I wonder if our insert just
below is *also* a speculative insert ... which perhaps could cause the
compute_new_xmax_infomask code to misbehave.
I think pg_waldump should print out the speculativeness of an insertion.
--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients from your mail client; the link fills in
the headers below: reply via email
Some clients, webmail especially, drop In-Reply-To and References from
a link, and the reply then starts a new thread: use the mbox below.
To: pgsql-bugs@postgresql.org
Cc: alvherre@2ndquadrant.com, andres@anarazel.de, tomas.vondra@2ndquadrant.com, andrey.salnikov@dataegret.com
Subject: Re: BUG #16125: Crash of PostgreSQL's wal sender during logical replication
In-Reply-To: <20191127142204.GA13686@alvherre.pgsql>
References: <20191119232849.qhtuhq4udrqrz4he@alap3.anarazel.de> <20191127142204.GA13686@alvherre.pgsql>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox