Horton

Re: HOT chain validation in verify_heapam()

Himanshu Upadhyaya <upadhyaya.himanshu@gmail.com>

View thread View on postgr.es
From: Himanshu Upadhyaya <upadhyaya.himanshu@gmail.com>
To: Andres Freund <andres@anarazel.de>
Cc: Robert Haas <robertmhaas@gmail.com>, Aleksander Alekseev <aleksander@timescale.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2022-11-17T16:03:17Z
Lists: pgsql-hackers 
On Tue, Nov 15, 2022 at 3:32 AM Andres Freund <andres@anarazel.de> wrote:

>
> > Furthermore, it is
> > possible that successor[x] = successor[x'] since the page might be
> corrupted
> > and we haven't checked otherwise.
> >
> > predecessor[y] = x means that successor[x] = y but in addition we've
> > checked that y is sane, and that x.xmax=y.xmin. If there are multiple
> > tuples for which these conditions hold, we've issued complaints about
> > all but one and entered the last into the predecessor array.
>
> As shown by the isolationtester test I just posted, this doesn't quite work
> right now. Probably fixable.
>
> I don't think we can follow non-HOT ctid chains if they're older than the
> xmin
> horizon, including all cases of xmin being frozen. There's just nothing
> guaranteeing that the tuples are actually "related".
>
> I understand the problem with frozen tuples but don't understand the
concern with non-HOT chains,
could you please help with some explanation around it?


-- 
Regards,
Himanshu Upadhyaya
EnterpriseDB: http://www.enterprisedb.com

Commits

  1. amcheck: Generalize one of the recently-added update chain checks.

  2. amcheck: Tighten up validation of redirect line pointers.

  3. amcheck: Fix verify_heapam for tuples where xmin or xmax is 0.

  4. amcheck: Fix a few bugs in new update chain validation.

  5. Fix new test case to work on (some?) big-endian architectures.

  6. Don't test HEAP_XMAX_INVALID when freezing xmax.