Horton

Re: [COMMITTERS] pgsql: Fix freezing of a dead HOT-updated tuple

Stephen Frost <sfrost@snowman.net>

View thread View on postgr.es
From: Stephen Frost <sfrost@snowman.net>
To: Alvaro Herrera <alvherre@alvh.no-ip.org>
Cc: Michael Paquier <michael.paquier@gmail.com>, "Wood, Dan" <hexpert@amazon.com>, "Wong, Yi Wen" <yiwong@amazon.com>, Peter Geoghegan <pg@bowt.ie>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>, Robert Haas <robertmhaas@gmail.com>
Date: 2017-10-06T14:11:40Z
Lists: pgsql-hackers 
Robert,

* Alvaro Herrera (alvherre@alvh.no-ip.org) wrote:
> Michael Paquier wrote:
> > On Fri, Oct 6, 2017 at 10:18 PM, Alvaro Herrera <alvherre@alvh.no-ip.org> wrote:
> > > Wood, Dan wrote:
> > >> Yes, I’ve been testing 9.6.  I’ll try Alvaro’s patch today.
> > >>
> > >> I would prefer to focus on either latest 9X or 11dev.
> > >
> > > I tested my patch on 9.4 and 9.5 today and it seems to close the problem
> > > (with the patch, I waited 10x as many iterations as it took for the
> > > problem to occur ~10 times without the patch), but I can reproduce a
> > > problem in 9.6 with my patch installed.  There must be something new in
> > > 9.6 that is causing the problem to reappear.
> > 
> > The freeze visibility map has been introduced in 9.6... There could be
> > interactions on this side.
> 
> Ah, thanks for the tip.  I hope the authors of that can do the gruntwork
> of researching this problem there, then.  I'll go commit what I have
> now.

I don't doubt you're watching this thread too, but just to be 110% sure
that we don't end up with the November releases still having this issue,
I'm adding you to the CC on this thread as the one who did the freeze
visibility map work.  Depending on hope here is a bit too squishy for me
when we're talking about corruption issues.

Thanks!

Stephen

Commits

  1. Backport addition of rs_old_rel to rewriteheap's state.

  2. Perform a lot more sanity checks when freezing tuples.

  3. Revert bogus fixes of HOT-freezing bug

  4. Fix traversal of half-frozen update chains

  5. Fix freezing of a dead HOT-updated tuple

  6. During index build, check and elog (not just Assert) for broken HOT chain.

  7. Fix WAL replay of locking an updated tuple

  8. Change the way we mark tuples as frozen.