Horton

Re: Is this a problem in GenericXLogFinish()?

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Robert Haas <robertmhaas@gmail.com>
Cc: Jeff Davis <pgsql@j-davis.com>, Alexander Lakhin <exclusion@gmail.com>, Heikki Linnakangas <hlinnaka@iki.fi>, pgsql-hackers@postgresql.org
Date: 2023-10-27T00:15:19Z
Lists: pgsql-hackers 
On Thu, Oct 26, 2023 at 09:40:09AM -0400, Robert Haas wrote:
> Because of this, it is possible for bucketbuf, prevbuf, and wbuf to be
> the same (your first scenario) but the second scenario you mention
> (nextbuf  == wbuf) should be impossible.

Okay..

> It seems to me that maybe we shouldn't even be registering wbuf or
> doing anything at all to it if there are no tuples that need moving.
> That would also require some adjustment of the redo routine.

Hmm.  So my question is: do we need the cleanup lock on the write
buffer even if there are no tuples, and even if primary bucket and the
write bucket are the same?  I'd like to think that what you say is OK,
still I am not completely sure after reading the lock assumptions in
the hash README or 6d46f4783efe.  A simpler thing would be to mark
buffer 1 with REGBUF_NO_CHANGE when the primary and write buffers are
the same if we expect the lock to always be taken, I guess..

I've noticed that the replay paths for XLOG_HASH_MOVE_PAGE_CONTENTS
and XLOG_HASH_SQUEEZE_PAGE are similar with their page handlings (some
copy-pastes?).  A MOVE record should never have zero tuples, still the
replay path assumes that this can be possible, so it could be
simplified.
--
Michael

Commits

  1. Fix inconsistency with replay of hash squeeze record for clean buffers

  2. Set LSN for wbuf in _hash_freeovflpage() iff wbuf is modified.

  3. Fix an uninitialized access in hash_xlog_squeeze_page().

  4. Use REGBUF_NO_CHANGE at one more place in the hash index.

  5. Assert that buffers are marked dirty before XLogRegisterBuffer().

  6. Fix bug in GenericXLogFinish().

  7. Improve hash index bucket split behavior.