Horton

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

Michael Paquier <michael.paquier@gmail.com>

View thread View on postgr.es
From: Michael Paquier <michael.paquier@gmail.com>
To: Andres Freund <andres@anarazel.de>
Cc: Alvaro Herrera <alvherre@alvh.no-ip.org>, "Wood, Dan" <hexpert@amazon.com>, Peter Geoghegan <pg@bowt.ie>, pgsql-committers@postgresql.org, "Wong, Yi Wen" <yiwong@amazon.com>, Robert Haas <robertmhaas@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>, PostgreSQL mailing lists <pgsql-hackers@postgresql.org>
Date: 2017-11-22T01:11:17Z
Lists: pgsql-hackers 
On Tue, Nov 21, 2017 at 4:18 AM, Andres Freund <andres@anarazel.de> wrote:
> On 2017-11-13 19:03:41 -0800, Andres Freund wrote:
>> On 2017-11-03 07:53:30 -0700, Andres Freund wrote:
>> > Here's that patch.  I've stared at this some, and Robert did too. Robert
>> > mentioned that the commit message might need some polish and I'm not
>> > 100% sure about the error message texts yet.
>> >
>> > I'm not yet convinced that the new elog in vacuumlazy can never trigger
>> > - but I also don't think we want to actually freeze the tuple in that
>> > case.
>>
>> I'm fairly sure it could be triggered, therefore I've rewritten that.
>>
>> I've played around quite some with the attached patch. So far, after
>> applying the second patch, neither VACUUM nor VACUUM FULL / CLUSTER make
>> the situation worse for already existing corruption. HOT pruning can
>> change the exact appearance of existing corruption a bit, but I don't
>> think it can make the corruption meaningfully worse.  It's a bit
>> annoying and scary to add so many checks to backbranches but it kinda
>> seems required.  The error message texts aren't perfect, but these are
>> "should never be hit" type elog()s so I'm not too worried about that.
>>
>> Please review!
>
> Ping? Alvaro, it'd be good to get some input here.

Note that I will be able to jump on the ship after being released from
commit fest duties. This is likely a multi-day task for testing and
looking at it, and I am not the most knowledgeable human being with
this code.
-- 
Michael

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.