On Sat, Jul 23, 2022 at 12:12:38PM -0700, David G. Johnston wrote:
> I concur, suggesting the following:
>
> "Specifically, updates result in multiple rows versions (tuples) existing on
> the table."
See my reply to Peter G. about tuples.
> "There is sufficient free space on the page containing the old tuple for the
> updated tuple."
Same.
> "Old tuples can be completely removed..."
It is really old row versions, not old tuples since tuples, in my mind,
don't have a chain of versions. There are cannot-be-viewed-by-anyone
tuples, but that is not the point here.
> Overall, I think that this is suitable to commit, and I don't want to
> make too much of a fuss. It's great that we're doing this.
>
>
>
> Agreed. The other suggestion listed are not clear-cut winners in my mind.
>
> The following, though, seems to just come out of nowhere. It would be better
> setup as a "(See <link> for why this is possible.)" instead of dropping "page
> item identifiers" on the reader.
>
> + This removal is possible because indexes
> + do not reference their <link linkend="storage-page-layout">page
> + item identifiers</link>.
I added parentheses around that.
>
> As a related thought, this has done a great job of being usable for a DBA
> operating at a high-level of system knowledge and interaction. I don't think
> burying it in storage.sgml is desirable, Maybe "Performance Tips" under
> "Avoid
> Unnecessary Indexes" (yes, a bit of a stretch, but nothing else seems to fit
> better, except maybe in concurrency control since we are discussing overcoming
> the limitation of our concurrency control choice.
Uh, not sure. Anyone else have an opinion?
> Summary paragraph:
> "can only happen if" => "can only be created if"
Yes, good point.
Updated patch attached.
--
Bruce Momjian <br...@momjian.us> https://momjian.us
EDB https://enterprisedb.com
Indecision is a decision. Inaction is an action. Mark Batterson
diff --git a/doc/src/sgml/acronyms.sgml b/doc/src/sgml/acronyms.sgml
index 9ed148ab84..2df6559acc 100644
--- a/doc/src/sgml/acronyms.sgml
+++ b/doc/src/sgml/acronyms.sgml
@@ -299,9 +299,7 @@
<term><acronym>HOT</acronym></term>
<listitem>
<para>
- <ulink
- url="https://git.postgresql.org/gitweb/?p=postgresql.git;a=blob;f=src/backend/access/heap/README.HOT;hb=HEAD";>Heap-Only
- Tuples</ulink>
+ <link linkend="storage-hot">Heap-Only Tuples</link>
</para>
</listitem>
</varlistentry>
diff --git a/doc/src/sgml/btree.sgml b/doc/src/sgml/btree.sgml
index a9200ee52e..6f608a14bf 100644
--- a/doc/src/sgml/btree.sgml
+++ b/doc/src/sgml/btree.sgml
@@ -639,7 +639,8 @@ options(<replaceable>relopts</replaceable> <type>local_relopts *</type>) returns
accumulate and adversely affect query latency and throughput. This
typically occurs with <command>UPDATE</command>-heavy workloads
where most individual updates cannot apply the
- <acronym>HOT</acronym> optimization. Changing the value of only
+ <link linkend="storage-hot"><acronym>HOT</acronym> optimization.</link>
+ Changing the value of only
one column covered by one index during an <command>UPDATE</command>
<emphasis>always</emphasis> necessitates a new set of index tuples
— one for <emphasis>each and every</emphasis> index on the
diff --git a/doc/src/sgml/catalogs.sgml b/doc/src/sgml/catalogs.sgml
index a186e35f00..248dbc0e26 100644
--- a/doc/src/sgml/catalogs.sgml
+++ b/doc/src/sgml/catalogs.sgml
@@ -4381,7 +4381,7 @@ SCRAM-SHA-256$<replaceable><iteration count></replaceable>:<replaceable>&l
<para>
If true, queries must not use the index until the <structfield>xmin</structfield>
of this <structname>pg_index</structname> row is below their <symbol>TransactionXmin</symbol>
- event horizon, because the table may contain broken HOT chains with
+ event horizon, because the table may contain broken <link linkend="storage-hot">HOT chains</link> with
incompatible rows that they can see
</para></entry>
</row>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index e2d728e0c4..e5a84ed76d 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -4482,7 +4482,8 @@ ANY <replaceable class="parameter">num_sync</replaceable> ( <replaceable class="
<listitem>
<para>
Specifies the number of transactions by which <command>VACUUM</command> and
- <acronym>HOT</acronym> updates will defer cleanup of dead row versions. The
+ <link linkend="storage-hot"><acronym>HOT</acronym> updates</link>
+ will defer cleanup of dead row versions. The
default is zero transactions, meaning that dead row versions can be
removed as soon as possible, that is, as soon as they are no longer
visible to any open transaction. You may wish to set this to a
diff --git a/doc/src/sgml/indexam.sgml b/doc/src/sgml/indexam.sgml
index cf359fa9ff..4f83970c85 100644
--- a/doc/src/sgml/indexam.sgml
+++ b/doc/src/sgml/indexam.sgml
@@ -45,7 +45,8 @@
extant versions of the same logical row; to an index, each tuple is
an independent object that needs its own index entry. Thus, an
update of a row always creates all-new index entries for the row, even if
- the key values did not change. (HOT tuples are an exception to this
+ the key values did not change. (<link linkend="storage-hot">HOT
+ tuples</link> are an exception to this
statement; but indexes do not deal with those, either.) Index entries for
dead tuples are reclaimed (by vacuuming) when the dead tuples themselves
are reclaimed.
diff --git a/doc/src/sgml/indices.sgml b/doc/src/sgml/indices.sgml
index 023157d888..42e1e86c8a 100644
--- a/doc/src/sgml/indices.sgml
+++ b/doc/src/sgml/indices.sgml
@@ -749,7 +749,7 @@ CREATE INDEX people_names ON people ((first_name || ' ' || last_name));
<para>
Index expressions are relatively expensive to maintain, because the
derived expression(s) must be computed for each row insertion
- and non-HOT update. However, the index expressions are
+ and <link linkend="storage-hot">non-HOT update.</link> However, the index expressions are
<emphasis>not</emphasis> recomputed during an indexed search, since they are
already stored in the index. In both examples above, the system
sees the query as just <literal>WHERE indexedcolumn = 'constant'</literal>
diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml
index 7dbbab6f5c..6408d28c5d 100644
--- a/doc/src/sgml/monitoring.sgml
+++ b/doc/src/sgml/monitoring.sgml
@@ -4426,7 +4426,7 @@ SELECT pid, wait_event_type, wait_event FROM pg_stat_activity WHERE wait_event i
<structfield>n_tup_upd</structfield> <type>bigint</type>
</para>
<para>
- Number of rows updated (includes HOT updated rows)
+ Number of rows updated (includes <link linkend="storage-hot">HOT updated rows</link>)
</para></entry>
</row>
diff --git a/doc/src/sgml/ref/create_table.sgml b/doc/src/sgml/ref/create_table.sgml
index 6bbf15ed1a..c14b2010d8 100644
--- a/doc/src/sgml/ref/create_table.sgml
+++ b/doc/src/sgml/ref/create_table.sgml
@@ -1435,7 +1435,9 @@ WITH ( MODULUS <replaceable class="parameter">numeric_literal</replaceable>, REM
to the indicated percentage; the remaining space on each page is
reserved for updating rows on that page. This gives <command>UPDATE</command>
a chance to place the updated copy of a row on the same page as the
- original, which is more efficient than placing it on a different page.
+ original, which is more efficient than placing it on a different
+ page, and makes <link linkend="storage-hot">heap-only tuple
+ updates</link> more likely.
For a table whose entries are never updated, complete packing is the
best choice, but in heavily updated tables smaller fillfactors are
appropriate. This parameter cannot be set for TOAST tables.
diff --git a/doc/src/sgml/storage.sgml b/doc/src/sgml/storage.sgml
index f4b9f66589..e5b9f3f1ff 100644
--- a/doc/src/sgml/storage.sgml
+++ b/doc/src/sgml/storage.sgml
@@ -1075,4 +1075,74 @@ data. Empty in ordinary tables.</entry>
</sect2>
</sect1>
+<sect1 id="storage-hot">
+
+ <title>Heap-Only Tuples (<acronym>HOT</acronym>)</title>
+
+ <para>
+ To allow for high concurrency, <productname>PostgreSQL</productname>
+ uses <link linkend="mvcc-intro">multiversion concurrency
+ control</link> (<acronym>MVCC</acronym>) to store rows. However,
+ <acronym>MVCC</acronym> has some downsides for update queries.
+ Specifically, updates require new versions of rows to be added to
+ tables. This can also require new index entries for each updated row,
+ and removal of old versions of rows and their index entries can be
+ expensive.
+ </para>
+
+ <para>
+ To help reduce the overhead of updates,
+ <productname>PostgreSQL</productname> has an optimization called
+ heap-only tuples (<acronym>HOT</acronym>). This optimization is
+ possible when:
+
+ <itemizedlist>
+ <listitem>
+ <para>
+ The update does not modify any columns referenced by the table's
+ indexes, including expression and partial indexes.
+ </para>
+ </listitem>
+ <listitem>
+ <para>
+ There is sufficient free space on the page containing the old row
+ for the updated row.
+ </para>
+ </listitem>
+ </itemizedlist>
+
+ In such cases, heap-only tuples provide two optimizations:
+
+ <itemizedlist>
+ <listitem>
+ <para>
+ New index entries are not needed to represent updated rows.
+ </para>
+ </listitem>
+ <listitem>
+ <para>
+ Old versions of updated rows can be completely removed during normal
+ operation, including <command>SELECT</command>s, instead of requiring
+ periodic vacuum operations. (This is possible because indexes
+ do not reference their <link linkend="storage-page-layout">page
+ item identifiers</link>.)
+ </para>
+ </listitem>
+ </itemizedlist>
+ </para>
+
+ <para>
+ In summary, heap-only tuple updates can only be created
+ if columns used by indexes are not updated. You can
+ increase the likelihood of sufficient page space for
+ <acronym>HOT</acronym> updates by decreasing a table's <link
+ linkend="sql-createtable"><literal>fillfactor</literal></link>.
+ If you don't, <acronym>HOT</acronym> updates will still happen because
+ new rows will naturally migrate to new pages and existing pages with
+ sufficient free space for new row versions. The system view <link
+ linkend="monitoring-pg-stat-all-tables-view">pg_stat_all_tables</link>
+ allows monitoring of the occurrence of HOT and non-HOT updates.
+ </para>
+</sect1>
+
</chapter>
