analyze-in-stages post upgrade questions

On Fri, Jun 27, 2025 at 9:35 AM Zechman, Derek S <Derek.S.Zechman@snapon.com>
wrote:

We recently performed an upgrade from pg14 (14.18) to pg16 (16.9) and
performed the analyze-in-stages post upgrade. It has been noticed that
some plans changed to use hash joins instead of nested loops. Further
investigation found it was because the parent table of partitioned tables
did not have stats. After running an ANALYZE on the parent tables we got
similar plan an execution times as before.

I have two questions

1 - Why does analyze-in-stages not analyze the parent tables?

2 – What happens if we do not run analyze-in-stages post upgrade and just
run an analyze?

It takes more time, and you don't have *any* statistics on a given table
until the ANALYZE on that table completes.

How long did "vacuumdb --analyze-only --jobs=$mumble your_db" take?

--
Death to <Redacted>, and butter sauce.
Don't boil me, I'm still alive.
<Redacted> lobster!

On 6/27/25 06:35, Zechman, Derek S wrote:

We recently performed an upgrade from pg14 (14.18) to pg16 (16.9) and
performed the analyze-in-stages post upgrade.  It has been noticed that
some plans changed to use hash joins instead of nested loops.  Further
investigation found it was because the parent table of partitioned
tables did not have stats.  After running an ANALYZE on the parent
tables we got similar plan an execution times as before.

I have two questions

1 - Why does analyze-in-stages not analyze the parent tables?

2 – What happens if we do not run analyze-in-stages post upgrade and
just run an analyze?

It is spelled out in the docs:

https://www.postgresql.org/docs/current/pgupgrade.html

Emphasis added

"Using vacuumdb --all --analyze-only can efficiently generate such
statistics, and the use of --jobs can speed it up. Option
--analyze-in-stages can be used to generate **minimal statistics**
quickly. If vacuum_cost_delay is set to a non-zero value, this can be
overridden to speed up statistics generation using PGOPTIONS, e.g.,
PGOPTIONS='-c vacuum_cost_delay=0' vacuumdb ...."

and from here:

https://www.postgresql.org/docs/current/app-vacuumdb.html

"--analyze-in-stages

Only calculate statistics for use by the optimizer (no vacuum),
like --analyze-only. Run three stages of analyze; the first stage uses
the lowest possible statistics target (see default_statistics_target) to
produce usable statistics faster, and subsequent stages build the full
statistics.

This option is only useful to analyze a database that currently has
no statistics or has wholly incorrect ones, such as if it is newly
populated from a restored dump or by pg_upgrade. Be aware that running
with this option in a database with existing statistics may cause the
query optimizer choices to become transiently worse due to the low
statistics targets of the early stages.
"

Thanks,

Sean

--
Adrian Klaver
adrian.klaver@aklaver.com

On Fri, 2025-06-27 at 08:31 -0700, Adrian Klaver wrote:

On 6/27/25 06:35, Zechman, Derek S wrote:

We recently performed an upgrade from pg14 (14.18) to pg16 (16.9) and
performed the analyze-in-stages post upgrade.  It has been noticed that
some plans changed to use hash joins instead of nested loops.  Further
investigation found it was because the parent table of partitioned
tables did not have stats.  After running an ANALYZE on the parent
tables we got similar plan an execution times as before.

I have two questions

1 - Why does analyze-in-stages not analyze the parent tables?

2 – What happens if we do not run analyze-in-stages post upgrade and
just run an analyze?

It is spelled out in the docs:

https://www.postgresql.org/docs/current/pgupgrade.html

Emphasis added

"Using vacuumdb --all --analyze-only can efficiently generate such
statistics, and the use of --jobs can speed it up. Option
--analyze-in-stages can be used to generate **minimal statistics**
quickly. If vacuum_cost_delay is set to a non-zero value, this can be
overridden to speed up statistics generation using PGOPTIONS, e.g.,
PGOPTIONS='-c vacuum_cost_delay=0' vacuumdb ...."

and from here:

https://www.postgresql.org/docs/current/app-vacuumdb.html

"--analyze-in-stages

Only calculate statistics for use by the optimizer (no vacuum),
like --analyze-only. Run three stages of analyze; the first stage uses
the lowest possible statistics target (see default_statistics_target) to
produce usable statistics faster, and subsequent stages build the full
statistics.

This option is only useful to analyze a database that currently has
no statistics or has wholly incorrect ones, such as if it is newly
populated from a restored dump or by pg_upgrade. Be aware that running
with this option in a database with existing statistics may cause the
query optimizer choices to become transiently worse due to the low
statistics targets of the early stages.

Well, that wouldn't explain why it doesn't work on partitioned tables.
I am under the impression that it should.

Derek, can cou share the pg_stats entries for the partitioned table?

Yours,
Laurenz Albe

We recently performed an upgrade from pg14 (14.18) to pg16 (16.9) and performed the analyze-in-stages post upgrade. It has been noticed that some plans changed to use hash joins instead of nested loops. Further investigation found it was because the parent table of partitioned tables did not have stats. After running an ANALYZE on the parent tables we got similar plan an execution times as before.

I have two questions
1 - Why does analyze-in-stages not analyze the parent tables?
2 – What happens if we do not run analyze-in-stages post upgrade and just run an analyze?

“It takes more time, and you don't have any statistics on a given table until the ANALYZE on that table completes.

How long did "vacuumdb --analyze-only --jobs=$mumble your_db" take?”

Thanks – that makes sense. I understand what analyze in stages does just wish it would include parent tables.

"vacuumdb --all --analyze-only --jobs=7" took about 75 minutes where the analyze-in-stages after upgrade took 115 minutes. Neither of these activities analyzed the parent tables.
Reading more and it seems vacuumdb doesn’t analyze parent tables and a manual analyze on those is needed if we want better planner statistics.

We recently performed an upgrade from pg14 (14.18) to pg16 (16.9) and

performed the analyze-in-stages post upgrade. It has been noticed that

some plans changed to use hash joins instead of nested loops. Further

investigation found it was because the parent table of partitioned

tables did not have stats. After running an ANALYZE on the parent

tables we got similar plan an execution times as before.

I have two questions

1 - Why does analyze-in-stages not analyze the parent tables?

2 – What happens if we do not run analyze-in-stages post upgrade and

just run an analyze?

It is spelled out in the docs:

https://urldefense.com/v3/__https://www.postgresql.org/docs/current/pgupgrade.html__;!!Lf_9VycLqA!jU5QgA1re0Txg_h2dD1N3XvK_l8hYdyMvpcxrLL5GDyQ5qN4aGQcxmDE8qmaV_p5telTzYmOL6S3fR8eRc0s_8UvnFFR6Ws$&lt;https://urldefense.com/v3/__https:/www.postgresql.org/docs/current/pgupgrade.html__;!!Lf_9VycLqA!jU5QgA1re0Txg_h2dD1N3XvK_l8hYdyMvpcxrLL5GDyQ5qN4aGQcxmDE8qmaV_p5telTzYmOL6S3fR8eRc0s_8UvnFFR6Ws$&gt;

Emphasis added

"Using vacuumdb --all --analyze-only can efficiently generate such

statistics, and the use of --jobs can speed it up. Option

--analyze-in-stages can be used to generate **minimal statistics**

quickly. If vacuum_cost_delay is set to a non-zero value, this can be

overridden to speed up statistics generation using PGOPTIONS, e.g.,

PGOPTIONS='-c vacuum_cost_delay=0' vacuumdb ...."

and from here:

https://urldefense.com/v3/__https://www.postgresql.org/docs/current/app-vacuumdb.html__;!!Lf_9VycLqA!jU5QgA1re0Txg_h2dD1N3XvK_l8hYdyMvpcxrLL5GDyQ5qN4aGQcxmDE8qmaV_p5telTzYmOL6S3fR8eRc0s_8UvvC7gfd0$&lt;https://urldefense.com/v3/__https:/www.postgresql.org/docs/current/app-vacuumdb.html__;!!Lf_9VycLqA!jU5QgA1re0Txg_h2dD1N3XvK_l8hYdyMvpcxrLL5GDyQ5qN4aGQcxmDE8qmaV_p5telTzYmOL6S3fR8eRc0s_8UvvC7gfd0$&gt;

"--analyze-in-stages

Only calculate statistics for use by the optimizer (no vacuum),

like --analyze-only. Run three stages of analyze; the first stage uses

the lowest possible statistics target (see default_statistics_target) to

produce usable statistics faster, and subsequent stages build the full

statistics.

This option is only useful to analyze a database that currently has

no statistics or has wholly incorrect ones, such as if it is newly

populated from a restored dump or by pg_upgrade. Be aware that running

with this option in a database with existing statistics may cause the

query optimizer choices to become transiently worse due to the low

statistics targets of the early stages.

Well, that wouldn't explain why it doesn't work on partitioned tables.

I am under the impression that it should.

Derek, can cou share the pg_stats entries for the partitioned table?

Yours,

Laurenz Albe

There are no entries in pg_stats for the parent table until after I manually run an analyze on it – Example below

=> select relname, reltuples, relkind from pg_class where relname ~ '^chapter_[0-9]+$' or relname='chapter' order by 1;

relname | reltuples | relkind

-------------+-----------+---------

chapter | -1 | p

chapter_1 | 4 | r

chapter_10 | 4 | r

chapter_100 | 30 | r

chapter_101 | 15 | r

chapter_102 | 15 | r

…

=> select count(*) from pg_stats where tablename='chapter';

count

-------

0

(1 row)

=> analyze chapter;

ANALYZE

=> select relname, reltuples, relkind from pg_class where relkind ='p' and relname='chapter';

relname | reltuples | relkind

---------+-----------+---------

chapter | 7589 | p

(1 row)

=> select count(*) from pg_stats where tablename='chapter';

count

-------

49

(1 row)

toy_epc_stg_1_db=>

On Sat, 2025-06-28 at 01:23 +0000, Zechman, Derek S wrote:

Well, that wouldn't explain why it doesn't work on partitioned tables.
I am under the impression that it should.

Derek, can cou share the pg_stats entries for the partitioned table?

There are no entries in pg_stats for the parent table until after I manually run an analyze on it – Example below

You are right. I looked at the code, and "vacuumdb" does not process
partitiond tables, even if --analyze-only is specified. I find that
surprising, aince the SQL command ANALYZE (without a table name) will
also collect statistics for partitioned tables.

I think that it would be a good idea to change that behavior.
In particular, it makes a lot of sense to collect statistics for
partitioned tables after a "pg_upgrade".

Attached is a patch to make "vacuumdb --analyze-only" consider
partitioned tables as well.

Yours,
Laurenz Albe

From 07cbd491011ff5da1243b117c111d51531293782 Mon Sep 17 00:00:00 2001
From: Laurenz Albe <laurenz.albe@cybertec.at>
Date: Sat, 28 Jun 2025 07:20:02 +0200
Subject: [PATCH v1] Make vacuumdb -Z process partitioned tables

Autoanalyze won't process partitioned tables, but a manual ANALYZE
of the whole database does.  So it was surprising that partitioned
tables were not processed by "vacuumdb --analyze-only".

In addition, "vacuumdb --analyze-only" is what you run to collect
missing statistics after a "pg_upgrade", and it makes a lot of sense
to collect statistics for partitioned tables at that point.

However, running VACUUM on partitioned tables adds no benefit over
vacuuming the partitions, and VACUUM is more expensive than ANALYZE,
so we won't treat partitioned tables unless --analyze-only is given.
Otherwise, we'd end up vacuuming the partitions twice, which would
be a waste of resources.
---
 doc/src/sgml/ref/vacuumdb.sgml |  4 ++++
 src/bin/scripts/vacuumdb.c     | 23 +++++++++++++++++++----
 2 files changed, 23 insertions(+), 4 deletions(-)

diff --git a/doc/src/sgml/ref/vacuumdb.sgml b/doc/src/sgml/ref/vacuumdb.sgml
index b0680a61814..6a4c8d51b7f 100644
--- a/doc/src/sgml/ref/vacuumdb.sgml
+++ b/doc/src/sgml/ref/vacuumdb.sgml
@@ -440,6 +440,10 @@ PostgreSQL documentation
       <listitem>
        <para>
         Only calculate statistics for use by the optimizer (no vacuum).
+        If that option is specified, <command>vacuumdb</command> will also
+        process partitioned tables.  Without that option, only the partitions
+        will be considered, unless a partitioned table is explicitly specified
+        with the <option>--table</option> option.
        </para>
       </listitem>
      </varlistentry>
diff --git a/src/bin/scripts/vacuumdb.c b/src/bin/scripts/vacuumdb.c
index 79b1096eb08..d7ca8300f43 100644
--- a/src/bin/scripts/vacuumdb.c
+++ b/src/bin/scripts/vacuumdb.c
@@ -911,10 +911,25 @@ retrieve_objects(PGconn *conn, vacuumingOptions *vacopts,
 	 */
 	if ((objfilter & OBJFILTER_TABLE) == 0)
 	{
-		appendPQExpBufferStr(&catalog_query,
-							 " AND c.relkind OPERATOR(pg_catalog.=) ANY (array["
-							 CppAsString2(RELKIND_RELATION) ", "
-							 CppAsString2(RELKIND_MATVIEW) "])\n");
+		/*
+		 * VACUUMing partitioned tables would be unreasonably expensive, since
+		 * that entails processing the partitions twice (once as part of the
+		 * partitioned table, once as tables in their own right) for no
+		 * benefit. But if we only ANALYZE, collecting statistics for
+		 * partitioned tables is worth the effort.
+		 */
+		if (vacopts->analyze_only)
+			appendPQExpBufferStr(&catalog_query,
+								 " AND c.relkind OPERATOR(pg_catalog.=) ANY (array["
+								 CppAsString2(RELKIND_RELATION) ", "
+								 CppAsString2(RELKIND_MATVIEW) ", "
+								 CppAsString2(RELKIND_PARTITIONED_TABLE) "])\n");
+		else
+			appendPQExpBufferStr(&catalog_query,
+								 " AND c.relkind OPERATOR(pg_catalog.=) ANY (array["
+								 CppAsString2(RELKIND_RELATION) ", "
+								 CppAsString2(RELKIND_MATVIEW) "])\n");
+
 	}
 
 	/*
-- 
2.50.0

The following review has been posted through the commitfest application:
make installcheck-world: tested, passed
Implements feature: tested, passed
Spec compliant: tested, passed
Documentation: tested, passed

Hi Laurenz,

Nice focused patch.

Applied successfully on commit 62a17a92 from master.

Documentation is updated and there is an explanatory comment for the code, as well as a descriptive commit message.

To check the patch, I added the following test at the end of src/bin/scripts/t/100_vacuumdb.pl in both master (as experiment) and your patch. In master it does not pass, but with your patch applied it does.

You can consider adding it to your patch, or I could also do that.

$node->safe_psql('postgres',
"CREATE TABLE parent_table (a INT) PARTITION BY LIST (a);\n"
. "CREATE TABLE child_table PARTITION OF parent_table FOR VALUES IN (1);\n"
. "INSERT INTO parent_table VALUES (1);\n");
$node->issues_sql_like(
[
'vacuumdb', '--analyze-only', 'postgres'
],
qr/statement:\s+ANALYZE\s+public\.parent_table/s,
'--analyze_only updates statistics for partitioned tables');

Kind regards,
Mircea Cadariu

The new status of this patch is: Waiting on Author

Well, that wouldn't explain why it doesn't work on partitioned tables.

I am under the impression that it should.

Derek, can cou share the pg_stats entries for the partitioned table?

There are no entries in pg_stats for the parent table until after I manually run an analyze on it – Example below

You are right. I looked at the code, and "vacuumdb" does not process

partitiond tables, even if --analyze-only is specified. I find that

surprising, aince the SQL command ANALYZE (without a table name) will

also collect statistics for partitioned tables.

I think that it would be a good idea to change that behavior.

In particular, it makes a lot of sense to collect statistics for

partitioned tables after a "pg_upgrade".

Attached is a patch to make "vacuumdb --analyze-only" consider

partitioned tables as well.

Yours,

Laurenz Albe

Is there a plan to include this patch in future releases/patches of postgres?

Thanks,

(Derek) Sean

On Wed, 2025-07-09 at 11:30 +0000, Zechman, Derek S wrote:

There are no entries in pg_stats for the parent table until after I manually run an analyze on it – Example below

You are right.  I looked at the code, and "vacuumdb" does not process
partitiond tables, even if --analyze-only is specified.  I find that
surprising, aince the SQL command ANALYZE (without a table name) will
also collect statistics for partitioned tables.

I think that it would be a good idea to change that behavior.
In particular, it makes a lot of sense to collect statistics for
partitioned tables after a "pg_upgrade".

Attached is a patch to make "vacuumdb --analyze-only" consider
partitioned tables as well.

Is there a plan to include this patch in future releases/patches of postgres?

I have added the patch to the current commitfest:
https://commitfest.postgresql.org/patch/5871/

So far, it has not got any peer review. So yes, I'd like to include
the patch, but I cannot make it happen by myself.
Essentially, patches get applied if
a) they get peer review and
b) a committer applies them

If you want this to happen, the best thing you could do would be
to review the patch and see if it works for you, if it does what you
need and so on:
https://wiki.postgresql.org/wiki/Reviewing_a_Patch

Yours,
Laurenz Albe

Hi Laurenz,

On 09/07/2025 16:26, Laurenz Albe wrote:

I have added the patch to the current commitfest:
https://commitfest.postgresql.org/patch/5871/

Just to let you know that I have added a review through the commitfest app.

You can see it here:
/messages/by-id/175189219162.2200286.3306593311375985296.pgcf@coridan.postgresql.org

Kind regards,

Mircea Cadariu

On Wed, 2025-07-09 at 17:37 +0100, Mircea Cadariu wrote:

Just to let you know that I have added a review through the commitfest app. 

Thanks!

The patch is still in state "needs review".
If there is something that I should change, you should set it to
"waiting on author". If you think that the patch is ready to go
as it is, please set it to "ready for committer".

Yours,
Laurenz Albe

Hi Laurenz,

Got it. I have only one suggestion for the patch. Consider adding a
corresponding test in src/bin/scripts/t/100_vacuumdb.pl.

Proposal (I used this to check the patch):

$node->safe_psql('postgres',
    "CREATE TABLE parent_table (a INT) PARTITION BY LIST (a);\n"
      . "CREATE TABLE child_table PARTITION OF parent_table FOR VALUES
IN (1);\n"
      . "INSERT INTO parent_table VALUES (1);\n");
$node->issues_sql_like(
    [
        'vacuumdb', '--analyze-only', 'postgres'
    ],
    qr/statement:\s+ANALYZE\s+public\.parent_table/s,
    '--analyze_only updates statistics for partitioned tables');

Kind regards,

Mircea Cadariu

[moving to pgsql-hackers]
On Thu, 2025-07-10 at 17:20 +0100, Mircea Cadariu wrote:

I have only one suggestion for the patch. Consider adding a
corresponding test in src/bin/scripts/t/100_vacuumdb.pl.

Proposal (I used this to check the patch):

$node->safe_psql('postgres',
    "CREATE TABLE parent_table (a INT) PARTITION BY LIST (a);\n"
      . "CREATE TABLE child_table PARTITION OF parent_table FOR VALUES
IN (1);\n"
      . "INSERT INTO parent_table VALUES (1);\n");
$node->issues_sql_like(
    [
        'vacuumdb', '--analyze-only', 'postgres'
    ],
    qr/statement:\s+ANALYZE\s+public\.parent_table/s,
    '--analyze_only updates statistics for partitioned tables');

Good idea; done in the attached version 2 of the patch.

Yours,
Laurenz Albe

From 4bf5048da6ffd250ee204fff419cf975aa7a6548 Mon Sep 17 00:00:00 2001
From: Laurenz Albe <laurenz.albe@cybertec.at>
Date: Fri, 11 Jul 2025 11:46:55 +0200
Subject: [PATCH v2] Make vacuumdb -Z process partitioned tables

Autoanalyze won't process partitioned tables, but a manual ANALYZE
of the whole database does.  So it was surprising that partitioned
tables were not processed by "vacuumdb --analyze-only".

In addition, "vacuumdb --analyze-only" is what you run to collect
missing statistics after a "pg_upgrade", and it makes a lot of sense
to collect statistics for partitioned tables at that point.

However, running VACUUM on partitioned tables adds no benefit over
vacuuming the partitions, and VACUUM is more expensive than ANALYZE,
so we won't treat partitioned tables unless --analyze-only is given.
Otherwise, we'd end up vacuuming the partitions twice, which would
be a waste of resources.

Author: Laurenz Albe <laurenz.albe@cybertec.at>
Author: Mircea Cadariu <cadariu.mircea@gmail.com>
Discussion: https://postgr.es/m/CO1PR04MB8281387B9AD9DE30976966BBC045A%40CO1PR04MB8281.namprd04.prod.outlook.com
---
 doc/src/sgml/ref/vacuumdb.sgml    |  4 ++++
 src/bin/scripts/t/100_vacuumdb.pl | 12 ++++++++++++
 src/bin/scripts/vacuumdb.c        | 23 +++++++++++++++++++----
 3 files changed, 35 insertions(+), 4 deletions(-)

diff --git a/doc/src/sgml/ref/vacuumdb.sgml b/doc/src/sgml/ref/vacuumdb.sgml
index b0680a61814..6a4c8d51b7f 100644
--- a/doc/src/sgml/ref/vacuumdb.sgml
+++ b/doc/src/sgml/ref/vacuumdb.sgml
@@ -440,6 +440,10 @@ PostgreSQL documentation
       <listitem>
        <para>
         Only calculate statistics for use by the optimizer (no vacuum).
+        If that option is specified, <command>vacuumdb</command> will also
+        process partitioned tables.  Without that option, only the partitions
+        will be considered, unless a partitioned table is explicitly specified
+        with the <option>--table</option> option.
        </para>
       </listitem>
      </varlistentry>
diff --git a/src/bin/scripts/t/100_vacuumdb.pl b/src/bin/scripts/t/100_vacuumdb.pl
index ff56a13b46b..9ab42c39a94 100644
--- a/src/bin/scripts/t/100_vacuumdb.pl
+++ b/src/bin/scripts/t/100_vacuumdb.pl
@@ -340,4 +340,16 @@ $node->issues_sql_unlike(
 	qr/statement:\ ANALYZE/sx,
 	'--missing-stats-only with no missing partition stats');
 
+$node->safe_psql('postgres',
+	"CREATE TABLE parent_table (a INT) PARTITION BY LIST (a);\n"
+	  . "CREATE TABLE child_table PARTITION OF parent_table FOR VALUES
+IN (1);\n"
+	  . "INSERT INTO parent_table VALUES (1);\n");
+$node->issues_sql_like(
+	[
+		'vacuumdb', '--analyze-only', 'postgres'
+	],
+	qr/statement:\s+ANALYZE\s+public\.parent_table/s,
+	'--analyze_only updates statistics for partitioned tables');
+
 done_testing();
diff --git a/src/bin/scripts/vacuumdb.c b/src/bin/scripts/vacuumdb.c
index 79b1096eb08..d7ca8300f43 100644
--- a/src/bin/scripts/vacuumdb.c
+++ b/src/bin/scripts/vacuumdb.c
@@ -911,10 +911,25 @@ retrieve_objects(PGconn *conn, vacuumingOptions *vacopts,
 	 */
 	if ((objfilter & OBJFILTER_TABLE) == 0)
 	{
-		appendPQExpBufferStr(&catalog_query,
-							 " AND c.relkind OPERATOR(pg_catalog.=) ANY (array["
-							 CppAsString2(RELKIND_RELATION) ", "
-							 CppAsString2(RELKIND_MATVIEW) "])\n");
+		/*
+		 * VACUUMing partitioned tables would be unreasonably expensive, since
+		 * that entails processing the partitions twice (once as part of the
+		 * partitioned table, once as tables in their own right) for no
+		 * benefit. But if we only ANALYZE, collecting statistics for
+		 * partitioned tables is worth the effort.
+		 */
+		if (vacopts->analyze_only)
+			appendPQExpBufferStr(&catalog_query,
+								 " AND c.relkind OPERATOR(pg_catalog.=) ANY (array["
+								 CppAsString2(RELKIND_RELATION) ", "
+								 CppAsString2(RELKIND_MATVIEW) ", "
+								 CppAsString2(RELKIND_PARTITIONED_TABLE) "])\n");
+		else
+			appendPQExpBufferStr(&catalog_query,
+								 " AND c.relkind OPERATOR(pg_catalog.=) ANY (array["
+								 CppAsString2(RELKIND_RELATION) ", "
+								 CppAsString2(RELKIND_MATVIEW) "])\n");
+
 	}
 
 	/*
-- 
2.50.0

On 11/07/2025 10:51, Laurenz Albe wrote:

Good idea; done in the attached version 2 of the patch.

Thanks! Looks good. I have set the status of the Commitfest entry to
"Ready for Committer".

Kind regards,

Mircea Cadariu

On Fri, Jul 11, 2025 at 7:42 PM Mircea Cadariu <cadariu.mircea@gmail.com> wrote:

On 11/07/2025 10:51, Laurenz Albe wrote:

Good idea; done in the attached version 2 of the patch.

Thanks! Looks good. I have set the status of the Commitfest entry to
"Ready for Committer".

I've started reviewing the patch since it's marked as ready for committer.

Overall, I like the change. But I have one question: should this be treated as
a bug fix that we back-patch to supported branches, or is it more of
an improvement that should only go into master?

         Only calculate statistics for use by the optimizer (no vacuum).
+        If that option is specified, <command>vacuumdb</command> will also
+        process partitioned tables.  Without that option, only the partitions
+        will be considered, unless a partitioned table is explicitly specified
+        with the <option>--table</option> option.

This wording seems a bit out of place in the --analyze-only section,
since it also describes the default behavior of vacuumdb without that option.
Wouldn't it make more sense to move that explanation in the --table section?

For example, we could add something like:

------------------
If no tables are specified with the --table option, vacuumdb will
clean all regular tables and materialized views in the connected
database. If --analyze-only or --analyze-in-stages is also specified,
it will analyze all regular tables, partitioned tables, and
materialized views (but not foreign tables).
------------------

+ /*
+ * VACUUMing partitioned tables would be unreasonably expensive, since
+ * that entails processing the partitions twice (once as part of the
+ * partitioned table, once as tables in their own right) for no
+ * benefit. But if we only ANALYZE, collecting statistics for
+ * partitioned tables is worth the effort.
+ */

This is probably true. But isn't the main reason more about aligning with
the behavior of the underlying VACUUM and ANALYZE commands? As the vacuumdb
docs says, "There is no effective difference between vacuuming and analyzing
databases via this utility and via other methods for accessing the server.",
so its default target objects should match: VACUUM skips partitioned tables
by default, while ANALYZE includes them. If that's the case, maybe the comment
should reflect that instead.

+ qr/statement:\s+ANALYZE\s+public\.parent_table/s,
+ '--analyze_only updates statistics for partitioned tables');

A plain space might be sufficient instead of \s+.
Also, I don't think the backslash before ".parent_table" is necessary.

Regards,

--
Fujii Masao

Hi,

On 30/07/2025 12:49, Fujii Masao wrote:

I've started reviewing the patch since it's marked as ready for committer.

Thanks!

Overall, I like the change. But I have one question: should this be treated as
a bug fix that we back-patch to supported branches, or is it more of
an improvement that should only go into master?

I reckon it might make sense to back-patch it to previous versions, as
users might not upgrade always to the latest version.

Only calculate statistics for use by the optimizer (no vacuum).
+        If that option is specified, <command>vacuumdb</command> will also
+        process partitioned tables.  Without that option, only the partitions
+        will be considered, unless a partitioned table is explicitly specified
+        with the <option>--table</option> option.

This wording seems a bit out of place in the --analyze-only section,
since it also describes the default behavior of vacuumdb without that option.
Wouldn't it make more sense to move that explanation in the --table section?

For example, we could add something like:

------------------
If no tables are specified with the --table option, vacuumdb will
clean all regular tables and materialized views in the connected
database. If --analyze-only or --analyze-in-stages is also specified,
it will analyze all regular tables, partitioned tables, and
materialized views (but not foreign tables).
------------------

Yes, agreed.

+ /*
+ * VACUUMing partitioned tables would be unreasonably expensive, since
+ * that entails processing the partitions twice (once as part of the
+ * partitioned table, once as tables in their own right) for no
+ * benefit. But if we only ANALYZE, collecting statistics for
+ * partitioned tables is worth the effort.
+ */

This is probably true. But isn't the main reason more about aligning with
the behavior of the underlying VACUUM and ANALYZE commands? As the vacuumdb
docs says, "There is no effective difference between vacuuming and analyzing
databases via this utility and via other methods for accessing the server.",
so its default target objects should match: VACUUM skips partitioned tables
by default, while ANALYZE includes them. If that's the case, maybe the comment
should reflect that instead.

I see what you mean. From that perspective, I wonder if we even need a
comment there at all.

+ qr/statement:\s+ANALYZE\s+public\.parent_table/s,
+ '--analyze_only updates statistics for partitioned tables');

A plain space might be sufficient instead of \s+.
Also, I don't think the backslash before ".parent_table" is necessary.

Good catch! Indeed let's simplify that to contain strictly only what's
necessary.

Kind regards,

Mircea Cadariu

On Wed, Aug 6, 2025 at 1:01 PM Mircea Cadariu <cadariu.mircea@gmail.com> wrote:

Overall, I like the change. But I have one question: should this be treated as
a bug fix that we back-patch to supported branches, or is it more of
an improvement that should only go into master?

I reckon it might make sense to back-patch it to previous versions, as users might not upgrade always to the latest version.

I understand your point. But on second thought, since the patch changes
behavior, I'm leaning toward treating it as an improvement, so it should
only go to master...

+ /*
+ * VACUUMing partitioned tables would be unreasonably expensive, since
+ * that entails processing the partitions twice (once as part of the
+ * partitioned table, once as tables in their own right) for no
+ * benefit. But if we only ANALYZE, collecting statistics for
+ * partitioned tables is worth the effort.
+ */

This is probably true. But isn't the main reason more about aligning with
the behavior of the underlying VACUUM and ANALYZE commands? As the vacuumdb
docs says, "There is no effective difference between vacuuming and analyzing
databases via this utility and via other methods for accessing the server.",
so its default target objects should match: VACUUM skips partitioned tables
by default, while ANALYZE includes them. If that's the case, maybe the comment
should reflect that instead.

I see what you mean. From that perspective, I wonder if we even need a comment there at all.

Or, if we keep it, though, I'd like to update it to something like
the following:

--------------------
vacuumdb should generally follow the behavior of the underlying
VACUUM and ANALYZE commands. If analyze_only is true, process
regular tables, materialized views, and partitioned tables, just like
ANALYZE (with no specific target tables) does. Otherwise, process
only regular tables and materialized views, since VACUUM skips
partitioned tables when no target tables are specified.
--------------------

Regards,

--
Fujii Masao

On Wed, Aug 06, 2025 at 11:25:53PM +0900, Fujii Masao wrote:

On Wed, Aug 6, 2025 at 1:01 PM Mircea Cadariu <cadariu.mircea@gmail.com> wrote:

Overall, I like the change. But I have one question: should this be treated as
a bug fix that we back-patch to supported branches, or is it more of
an improvement that should only go into master?

I reckon it might make sense to back-patch it to previous versions, as users might not upgrade always to the latest version.

I understand your point. But on second thought, since the patch changes
behavior, I'm leaning toward treating it as an improvement, so it should
only go to master...

I also am leaning towards treating this as v19 material. It's a nontrivial
behavior change, and this option is useful for major version upgrades,
which is an area that we really don't want to surprise users too much.
Furthermore, auto-analyze doesn't process partitioned tables, either, so
this introduces a bit of divergence. (I'd love to see that project picked
up again someday. Perhaps I will take a gander...)

--
nathan

On Wed, 2025-08-06 at 23:25 +0900, Fujii Masao wrote:

On Wed, Aug 6, 2025 at 1:01 PM Mircea Cadariu <cadariu.mircea@gmail.com> wrote:

Overall, I like the change. But I have one question: should this be treated as
a bug fix that we back-patch to supported branches, or is it more of
an improvement that should only go into master?

I reckon it might make sense to back-patch it to previous versions, as users might not upgrade always to the latest version.

I understand your point. But on second thought, since the patch changes
behavior, I'm leaning toward treating it as an improvement, so it should
only go to master...

I agree that this behavior change should not be backpatched.
That is not a bugfix.

+ /*
+ * VACUUMing partitioned tables would be unreasonably expensive, since
+ * that entails processing the partitions twice (once as part of the
+ * partitioned table, once as tables in their own right) for no
+ * benefit. But if we only ANALYZE, collecting statistics for
+ * partitioned tables is worth the effort.
+ */

This is probably true. But isn't the main reason more about aligning with
the behavior of the underlying VACUUM and ANALYZE commands? As the vacuumdb
docs says, "There is no effective difference between vacuuming and analyzing
databases via this utility and via other methods for accessing the server.",
so its default target objects should match: VACUUM skips partitioned tables
by default, while ANALYZE includes them. If that's the case, maybe the comment
should reflect that instead.

I see what you mean. From that perspective, I wonder if we even need a comment there at all.

Or, if we keep it, though, I'd like to update it to something like
the following:

--------------------
vacuumdb should generally follow the behavior of the underlying
VACUUM and ANALYZE commands. If analyze_only is true, process
regular tables, materialized views, and partitioned tables, just like
ANALYZE (with no specific target tables) does. Otherwise, process
only regular tables and materialized views, since VACUUM skips
partitioned tables when no target tables are specified.
--------------------

I am fine with that suggestion.

Alternatively, my original comment could be amended with

Besides, ANALYZE (without an option) processes partitioned tables, and
"vacuumdb -Z" should behave like ANALYZE.

Yours,
Laurenz Albe

On Thu, Aug 7, 2025 at 3:14 AM Nathan Bossart <nathandbossart@gmail.com> wrote:

I also am leaning towards treating this as v19 material. It's a nontrivial
behavior change, and this option is useful for major version upgrades,
which is an area that we really don't want to surprise users too much.

+1

Furthermore, auto-analyze doesn't process partitioned tables, either, so
this introduces a bit of divergence. (I'd love to see that project picked
up again someday. Perhaps I will take a gander...)

Sounds good!

Regards,

--
Fujii Masao

On Thu, Aug 7, 2025 at 5:52 AM Laurenz Albe <laurenz.albe@cybertec.at> wrote:

I understand your point. But on second thought, since the patch changes
behavior, I'm leaning toward treating it as an improvement, so it should
only go to master...

I agree that this behavior change should not be backpatched.
That is not a bugfix.

+1

--------------------
vacuumdb should generally follow the behavior of the underlying
VACUUM and ANALYZE commands. If analyze_only is true, process
regular tables, materialized views, and partitioned tables, just like
ANALYZE (with no specific target tables) does. Otherwise, process
only regular tables and materialized views, since VACUUM skips
partitioned tables when no target tables are specified.
--------------------

I am fine with that suggestion.

Thanks! So I've updated the patch based on my earlier comments.
Unless there are objections, I'll commit the attached version to master only.

Regards,

--
Fujii Masao

From 5c969379514aed6de612165f39fa709740a56933 Mon Sep 17 00:00:00 2001
From: Fujii Masao <fujii@postgresql.org>
Date: Mon, 18 Aug 2025 10:41:15 +0900
Subject: [PATCH v3] vacuumdb: Make vacuumdb --analyze-only process partitioned
 tables.

vacuumdb should follow the behavior of the underlying VACUUM and ANALYZE
commands. When --analyze-only is used, it ought to analyze regular tables,
materialized views, and partitioned tables, just as ANALYZE (with no explicit
target tables) does. Otherwise, it should only process regular tables and
materialized views, since VACUUM skips partitioned tables when no targets
are given.

Previously, vacuumdb --analyze-only skipped partitioned tables. This was
inconsistent, and also inconvenient after pg_upgrade, where --analyze-only
is typically used to gather missing statistics.

This commit fixes the behavior so that vacuumdb --analyze-only also processes
partitioned tables. As a result, both vacuumdb --analyze-only and
ANALYZE (with no explicit targets) now analyze regular tables,
partitioned tables, and materialized views, but not foreign tables.

Because this is a nontrivial behavior change, it is applied only to master.

Reported-by: Zechman, Derek S <Derek.S.Zechman@snapon.com>
Author: Laurenz Albe <laurenz.albe@cybertec.at>
Co-authored-by: Mircea Cadariu <cadariu.mircea@gmail.com>
Reviewed-by: Fujii Masao <masao.fujii@gmail.com>
Discussion: https://postgr.es/m/CO1PR04MB8281387B9AD9DE30976966BBC045A%40CO1PR04MB8281.namprd04.prod.outlook.com
---
 doc/src/sgml/ref/vacuumdb.sgml    |  9 +++++++++
 src/bin/scripts/t/100_vacuumdb.pl | 11 +++++++++++
 src/bin/scripts/vacuumdb.c        | 24 ++++++++++++++++++++----
 3 files changed, 40 insertions(+), 4 deletions(-)

diff --git a/doc/src/sgml/ref/vacuumdb.sgml b/doc/src/sgml/ref/vacuumdb.sgml
index c7d9dca17b8..53147480515 100644
--- a/doc/src/sgml/ref/vacuumdb.sgml
+++ b/doc/src/sgml/ref/vacuumdb.sgml
@@ -397,6 +397,15 @@ PostgreSQL documentation
         Multiple tables can be vacuumed by writing multiple
         <option>-t</option> switches.
        </para>
+       <para>
+        If no tables are specified with the <option>--table</option> option,
+        <application>vacuumdb</application> will clean all regular tables
+        and materialized views in the connected database.
+        If <option>--analyze-only</option> or
+        <option>--analyze-in-stages</option> is also specified,
+        it will analyze all regular tables, partitioned tables,
+        and materialized views (but not foreign tables).
+       </para>
        <tip>
         <para>
          If you specify columns, you probably have to escape the parentheses
diff --git a/src/bin/scripts/t/100_vacuumdb.pl b/src/bin/scripts/t/100_vacuumdb.pl
index ff56a13b46b..240f0fdd3e5 100644
--- a/src/bin/scripts/t/100_vacuumdb.pl
+++ b/src/bin/scripts/t/100_vacuumdb.pl
@@ -340,4 +340,15 @@ $node->issues_sql_unlike(
 	qr/statement:\ ANALYZE/sx,
 	'--missing-stats-only with no missing partition stats');
 
+$node->safe_psql('postgres',
+	"CREATE TABLE parent_table (a INT) PARTITION BY LIST (a);\n"
+	  . "CREATE TABLE child_table PARTITION OF parent_table FOR VALUES IN (1);\n"
+	  . "INSERT INTO parent_table VALUES (1);\n");
+$node->issues_sql_like(
+	[
+		'vacuumdb', '--analyze-only', 'postgres'
+	],
+	qr/statement: ANALYZE public.parent_table/s,
+	'--analyze-only updates statistics for partitioned tables');
+
 done_testing();
diff --git a/src/bin/scripts/vacuumdb.c b/src/bin/scripts/vacuumdb.c
index 79b1096eb08..22093e50aa5 100644
--- a/src/bin/scripts/vacuumdb.c
+++ b/src/bin/scripts/vacuumdb.c
@@ -911,10 +911,26 @@ retrieve_objects(PGconn *conn, vacuumingOptions *vacopts,
 	 */
 	if ((objfilter & OBJFILTER_TABLE) == 0)
 	{
-		appendPQExpBufferStr(&catalog_query,
-							 " AND c.relkind OPERATOR(pg_catalog.=) ANY (array["
-							 CppAsString2(RELKIND_RELATION) ", "
-							 CppAsString2(RELKIND_MATVIEW) "])\n");
+		/*
+		 * vacuumdb should generally follow the behavior of the underlying
+		 * VACUUM and ANALYZE commands. If analyze_only is true, process
+		 * regular tables, materialized views, and partitioned tables, just
+		 * like ANALYZE (with no specific target tables) does. Otherwise,
+		 * process only regular tables and materialized views, since VACUUM
+		 * skips partitioned tables when no target tables are specified.
+		 */
+		if (vacopts->analyze_only)
+			appendPQExpBufferStr(&catalog_query,
+								 " AND c.relkind OPERATOR(pg_catalog.=) ANY (array["
+								 CppAsString2(RELKIND_RELATION) ", "
+								 CppAsString2(RELKIND_MATVIEW) ", "
+								 CppAsString2(RELKIND_PARTITIONED_TABLE) "])\n");
+		else
+			appendPQExpBufferStr(&catalog_query,
+								 " AND c.relkind OPERATOR(pg_catalog.=) ANY (array["
+								 CppAsString2(RELKIND_RELATION) ", "
+								 CppAsString2(RELKIND_MATVIEW) "])\n");
+
 	}
 
 	/*
-- 
2.50.1

On Mon, 2025-08-18 at 11:38 +0900, Fujii Masao wrote:

Thanks! So I've updated the patch based on my earlier comments.
Unless there are objections, I'll commit the attached version to master only.

I am fine with your patch.

One suggestion:

--- a/doc/src/sgml/ref/vacuumdb.sgml
+++ b/doc/src/sgml/ref/vacuumdb.sgml
@@ -397,6 +397,15 @@ PostgreSQL documentation
Multiple tables can be vacuumed by writing multiple
<option>-t</option> switches.
</para>
+       <para>
+        If no tables are specified with the <option>--table</option> option,
+        <application>vacuumdb</application> will clean all regular tables
+        and materialized views in the connected database.
+        If <option>--analyze-only</option> or
+        <option>--analyze-in-stages</option> is also specified,
+        it will analyze all regular tables, partitioned tables,
+        and materialized views (but not foreign tables).
+       </para>

I suggest replacing "clean" with "process", since VACUUM does so much more than
clean up dead tuples.

Concerning backpatching, I voted against, but I suggest that this be backpatched
to v18. I don't feel very strongly about it though.

Yours,
Laurenz Albe

On Mon, Aug 18, 2025 at 3:40 PM Laurenz Albe <laurenz.albe@cybertec.at> wrote:

On Mon, 2025-08-18 at 11:38 +0900, Fujii Masao wrote:

Thanks! So I've updated the patch based on my earlier comments.
Unless there are objections, I'll commit the attached version to master only.

I am fine with your patch.

Thanks for the review!

One suggestion:

--- a/doc/src/sgml/ref/vacuumdb.sgml
+++ b/doc/src/sgml/ref/vacuumdb.sgml
@@ -397,6 +397,15 @@ PostgreSQL documentation
Multiple tables can be vacuumed by writing multiple
<option>-t</option> switches.
</para>
+       <para>
+        If no tables are specified with the <option>--table</option> option,
+        <application>vacuumdb</application> will clean all regular tables
+        and materialized views in the connected database.
+        If <option>--analyze-only</option> or
+        <option>--analyze-in-stages</option> is also specified,
+        it will analyze all regular tables, partitioned tables,
+        and materialized views (but not foreign tables).
+       </para>

I suggest replacing "clean" with "process", since VACUUM does so much more than
clean up dead tuples.

I see your point. However, since the vacuumdb docs already use "clean"
in several places, I think it's better to keep using "clean" here
for consistency. Thought?

Concerning backpatching, I voted against, but I suggest that this be backpatched
to v18. I don't feel very strongly about it though.

As for back-patching, I failed to find a strong reason to apply this change
to v18 over the many other patches that could not be committed before
the feature freeze... Of course if there's broad support for back-patching,
we can certainly revisit it. But for now I'm thinking to commit the patch
to master.

Regards,

--
Fujii Masao

On Tue, 2025-08-19 at 23:40 +0900, Fujii Masao wrote:

--- a/doc/src/sgml/ref/vacuumdb.sgml
+++ b/doc/src/sgml/ref/vacuumdb.sgml
@@ -397,6 +397,15 @@ PostgreSQL documentation
          Multiple tables can be vacuumed by writing multiple
          <option>-t</option> switches.
         </para>
+       <para>
+        If no tables are specified with the <option>--table</option> option,
+        <application>vacuumdb</application> will clean all regular tables
+        and materialized views in the connected database.
+        If <option>--analyze-only</option> or
+        <option>--analyze-in-stages</option> is also specified,
+        it will analyze all regular tables, partitioned tables,
+        and materialized views (but not foreign tables).
+       </para>

I suggest replacing "clean" with "process", since VACUUM does so much more than
clean up dead tuples.

I see your point. However, since the vacuumdb docs already use "clean"
in several places, I think it's better to keep using "clean" here
for consistency. Thought?

Works for me; I didn't consider that.

Concerning backpatching, I voted against, but I suggest that this be backpatched
to v18.  I don't feel very strongly about it though.

As for back-patching, I failed to find a strong reason to apply this change
to v18 over the many other patches that could not be committed before
the feature freeze... Of course if there's broad support for back-patching,
we can certainly revisit it. But for now I'm thinking to commit the patch
to master.

I don't have a strong reason either - my reasoning was that the change is small
and unlikely to introduce a bug, and that it would be nice to get more accurate
statistics on partitioned tables after "pg_upgrade" a year earlier.

But I won't object if the patch is only in v19.

Yours,
Laurenz Albe

On Wed, Aug 20, 2025 at 12:16 AM Laurenz Albe <laurenz.albe@cybertec.at> wrote:

On Tue, 2025-08-19 at 23:40 +0900, Fujii Masao wrote:

--- a/doc/src/sgml/ref/vacuumdb.sgml
+++ b/doc/src/sgml/ref/vacuumdb.sgml
@@ -397,6 +397,15 @@ PostgreSQL documentation
Multiple tables can be vacuumed by writing multiple
<option>-t</option> switches.
</para>
+       <para>
+        If no tables are specified with the <option>--table</option> option,
+        <application>vacuumdb</application> will clean all regular tables
+        and materialized views in the connected database.
+        If <option>--analyze-only</option> or
+        <option>--analyze-in-stages</option> is also specified,
+        it will analyze all regular tables, partitioned tables,
+        and materialized views (but not foreign tables).
+       </para>

I suggest replacing "clean" with "process", since VACUUM does so much more than
clean up dead tuples.

I see your point. However, since the vacuumdb docs already use "clean"
in several places, I think it's better to keep using "clean" here
for consistency. Thought?

Works for me; I didn't consider that.

Concerning backpatching, I voted against, but I suggest that this be backpatched
to v18. I don't feel very strongly about it though.

As for back-patching, I failed to find a strong reason to apply this change
to v18 over the many other patches that could not be committed before
the feature freeze... Of course if there's broad support for back-patching,
we can certainly revisit it. But for now I'm thinking to commit the patch
to master.

I don't have a strong reason either - my reasoning was that the change is small
and unlikely to introduce a bug, and that it would be nice to get more accurate
statistics on partitioned tables after "pg_upgrade" a year earlier.

But I won't object if the patch is only in v19.

OK, so for now I've pushed the patch to master. Thanks!

Regards,

--
Fujii Masao

On Wed, 2025-08-20 at 13:31 +0900, Fujii Masao wrote:

OK, so for now I've pushed the patch to master. Thanks!

Thank you for working on that!

Yours,
Laurenz Albe

When analyzing a partitioned table, I think you should use ANALYZE ONLY,
or otherwise avoid processing the children twice.

Thanks for handling this. I was recently suprised to learn that
vacuumdb doesn't process parents.

--
Justin