Horton

Re: Emit fewer vacuum records by reaping removable tuples during pruning

Robert Haas <robertmhaas@gmail.com>

View thread View on postgr.es
From: Robert Haas <robertmhaas@gmail.com>
To: Melanie Plageman <melanieplageman@gmail.com>
Cc: Michael Paquier <michael@paquier.xyz>, Peter Geoghegan <pg@bowt.ie>, Pg Hackers <pgsql-hackers@postgresql.org>, Andres Freund <andres@anarazel.de>
Date: 2024-01-09T20:40:36Z
Lists: pgsql-hackers 
On Tue, Jan 9, 2024 at 3:13 PM Melanie Plageman
<melanieplageman@gmail.com> wrote:
> I had already written the patch for immediate reaping addressing the
> below feedback before I saw the emails that said everyone is happy
> with using hastup in lazy_scan_[no]prune() in a preliminary patch. Let
> me know if you have a strong preference for reordering. Otherwise, I
> will write the three subsequent patches on top of this one.

I don't know if it rises to the level of a strong preference. It's
just a preference.

> Ah, I like this a lot. Attached patch does this. I've added a modified
> version of the comment you suggested. My only question is if we are
> losing something without this sentence (from the old comment):
>
> -         * ... They don't need to be left in place as LP_DEAD items
> until VACUUM gets
> -         * around to doing index vacuuming.
>
> I don't feel like it adds a lot, but it is absent from the new
> comment, so thought I would check.

I agree that we can leave that out. It wouldn't be bad to include it
if someone had a nice way of doing that, but it doesn't seem critical,
and if forcing it in there makes the comment less clear overall, it's
a net loss IMHO.

> Hmm. Yes. I suppose I was trying to find something to validate. Is it
> worth checking that the line pointer is not already LP_UNUSED? Or is
> that a bit ridiculous?

I think that's worthwhile (hence my proposed wording).

-- 
Robert Haas
EDB: http://www.enterprisedb.com

Commits

  1. Combine FSM updates for prune and no-prune cases.

  2. Remove LVPagePruneState.

  3. Move VM update code from lazy_scan_heap() to lazy_scan_prune().

  4. Optimize vacuuming of relations with no indexes.

  5. Be more consistent about whether to update the FSM while vacuuming.

  6. Remove hastup from LVPagePruneState.

  7. Use scanned_pages to decide when to failsafe check.

  8. Simplify lazy_scan_heap's handling of scanned pages.

  9. While vacuuming a large table, update upper-level FSM data every so often.