Horton

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

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Peter Geoghegan <pg@bowt.ie>
Cc: Alvaro Herrera <alvherre@alvh.no-ip.org>, "Wood, Dan" <hexpert@amazon.com>, pgsql-committers <pgsql-committers@postgresql.org>, "Wong, Yi Wen" <yiwong@amazon.com>, Michael Paquier <michael.paquier@gmail.com>, Robert Haas <robertmhaas@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2017-11-10T00:53:09Z
Lists: pgsql-hackers 
On 2017-11-09 16:45:07 -0800, Peter Geoghegan wrote:
> On Thu, Nov 9, 2017 at 4:17 PM, Andres Freund <andres@anarazel.de> wrote:
> >> Actually, on second thought, I take that back -- I don't think that
> >> REINDEXing will even finish once a HOT chain is broken by the bug.
> >> IndexBuildHeapScan() actually does quite a good job of making sure
> >> that HOT chains are sane, which is how the enhanced amcheck notices
> >> the bug here in practice.
> >
> > I think that's too optimistic.
> 
> Why? Because the "find the TID of the root" logic in
> IndexBuildHeapScan()/heap_get_root_tuples() won't reliably find the
> actual root (it might be some other HOT chain root following TID
> recycling by VACUUM)?

Primarily because it's not an anti-corruption tool. I'd be surprised if
there weren't ways to corrupt the page using these corruptions that
aren't detected by it.  But even if it were, I don't think there's
enough information to do so in the general case.  You very well can end
up with pages where subsequent hot pruning has removed a good bit of the
direct evidence of this bug.

But I'm not really sure why the error detection capabilities of matter
much for the principal point I raised, which is how much work we need to
do to not further worsen the corruption.

Greetings,

Andres Freund

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.