Horton

Re: HOT chain validation in verify_heapam()

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Robert Haas <robertmhaas@gmail.com>
Cc: Himanshu Upadhyaya <upadhyaya.himanshu@gmail.com>, Aleksander Alekseev <aleksander@timescale.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2022-11-15T19:50:05Z
Lists: pgsql-hackers 
Hi,

On 2022-11-15 11:36:21 -0500, Robert Haas wrote:
> On Mon, Nov 14, 2022 at 5:02 PM Andres Freund <andres@anarazel.de> wrote:
> > It seems like we should do a bit more validation within a chain of
> > tuples. E.g. that no live tuple can follow an !DidCommit xmin?
> 
> I think this check is already present in stronger form. If we see a
> !DidCommit xmin, the xmin of the next tuple in the chain not only can't be
> committed, but had better be the same.

As I think I mentioned before, I don't think the "better be the same" aspect
is correct, think subxacts. E.g.

off 0: xmin: top, xmax: child_1
off 1: xmin: child_1, xmax: invalid

If top hasn't committed yet, the current logic afaict will warn about this
situation, no? And I don't think we can generally the subxid parent at this
point, unfortunately (might have truncated subtrans).


Different aspect: Is it ok that we use TransactionIdDidCommit() without a
preceding IsInProgress() check?


I do think there's some potential for additional checks that don't run into
the above issue, e.g. checking that no in-progress xids follow an explicitly
aborted xact, that a committed xid can't follow an uncommitted xid etc.

Greetings,

Andres Freund

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.