Horton

Re: Removing vacuum_cleanup_index_scale_factor

Masahiko Sawada <sawada.mshk@gmail.com>

View thread View on postgr.es
From: Masahiko Sawada <sawada.mshk@gmail.com>
To: Peter Geoghegan <pg@bowt.ie>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2021-03-09T06:21:13Z
Lists: pgsql-hackers 
On Tue, Mar 9, 2021 at 7:35 AM Peter Geoghegan <pg@bowt.ie> wrote:
>
> On Mon, Mar 8, 2021 at 1:38 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > As you say, the history here is a bit convoluted, but it seems like
> > a good principle to avoid interconnections between VACUUM and ANALYZE
> > as much as we can.  I haven't been paying enough attention to this
> > thread to have more insight than that.
>
> The attached patch does what I proposed earlier today: it teaches
> do_analyze_rel() to always set pg_class.reltuples for indexes when it
> would do the same thing for the heap/table relation already. It's now
> uniform in that sense.

Thank you for the patches. I looked at 0001 patch and have a comment:

+    * We don't report to the stats collector here because the stats collector
+    * only tracks per-table stats.  Reset the changes_since_analyze counter
+    * only if we analyzed all columns; otherwise, there is still work for
+    * auto-analyze to do.

I think the comment becomes clearer if we add "if doing inherited
stats" at top of the above paragraph since we actually report to the
stats collector in !inh case.

>
> Also included is a patch that removes the
> vacuum_cleanup_index_scale_factor mechanism for triggering an index
> scan during VACUUM -- that's what the second patch does (this depends
> on the first patch, really).

0002 patch looks good to me.

Regards,

--
Masahiko Sawada
EDB:  https://www.enterprisedb.com/

Commits

  1. VACUUM ANALYZE: Always update pg_class.reltuples.

  2. Don't consider newly inserted tuples in nbtree VACUUM.

  3. Fix nbtree cleanup-only VACUUM stats inaccuracies.

  4. Disable recheck_on_update optimization to avoid crashes.

  5. Fix VACUUM so that it always updates pg_class.reltuples/relpages.

  6. Clean up API for ambulkdelete/amvacuumcleanup as per today's discussion.

  7. First steps towards statistics on expressional (nee functional) indexes.