agora inbox for [email protected]  

help / color / mirror / Atom feed

[PATCH] fix "Preserve relfilenodes" for pg_largeobject and its index
499+ messages / 3 participants
[nested] [flat]

* [PATCH] fix "Preserve relfilenodes" for pg_largeobject and its index
@ 2022-07-02 05:48  Justin Pryzby <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Justin Pryzby @ 2022-07-02 05:48 UTC (permalink / raw)

An empty table is created by initdb, and the filenode was still pointing to
that, so pg_largeobject was empty after binary upgrades.

Test like:
| ./tmp_install/usr/local/pgsql/bin/initdb --no-sync -D pg15b1.dat -k
| ./tmp_install/usr/local/pgsql/bin/postgres -D pg15b1.dat -c logging_collector=no -p 5678 -k /tmp&
| psql -h /tmp postgres -p 5678 -c '\lo_import /etc/shells' -c 'VACUUM FULL pg_largeobject'
| rm -fr pg15b2.dat && ./tmp_install/usr/local/pgsql/bin/initdb --no-sync -k -D pg15b2.dat && ./tmp_install/usr/local/pgsql/bin/pg_upgrade -d pg15b1.dat -D pg15b2.dat -b ./tmp_install/usr/local/pgsql/bin
| ./tmp_install/usr/local/pgsql/bin/postgres -D pg15b2.dat -c logging_collector=no -p 5678 -k /tmp&

See also:
9a974cbcba005256a19991203583a94b4f9a21a9
https://www.postgresql.org/message-id/20210601140949.GC22012%40momjian.us
https://www.postgresql.org/message-id/20220701231413.GI13040%40telsasoft.com
---
 src/bin/pg_dump/pg_dump.c              | 37 ++++++++------
 src/bin/pg_upgrade/relfilenumber.c     | 68 +++++++++++++++++++++++---
 src/bin/pg_upgrade/t/002_pg_upgrade.pl |  2 +
 3 files changed, 86 insertions(+), 21 deletions(-)

diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index 227cdca2e2d..06d436f2f34 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -3135,7 +3135,7 @@ dumpDatabase(Archive *fout)
 
 	/*
 	 * pg_largeobject comes from the old system intact, so set its
-	 * relfrozenxids and relminmxids.
+	 * relfrozenxids, relminmxids and relfilenode.
 	 */
 	if (dopt->binary_upgrade)
 	{
@@ -3143,34 +3143,41 @@ dumpDatabase(Archive *fout)
 		PQExpBuffer loFrozenQry = createPQExpBuffer();
 		PQExpBuffer loOutQry = createPQExpBuffer();
 		int			i_relfrozenxid,
+					i_relfilenode,
+					i_oid,
 					i_relminmxid;
 
 		/*
 		 * pg_largeobject
 		 */
 		if (fout->remoteVersion >= 90300)
-			appendPQExpBuffer(loFrozenQry, "SELECT relfrozenxid, relminmxid\n"
+			appendPQExpBuffer(loFrozenQry, "SELECT relfrozenxid, relminmxid, relfilenode, oid\n"
 							  "FROM pg_catalog.pg_class\n"
-							  "WHERE oid = %u;\n",
-							  LargeObjectRelationId);
+							  "WHERE oid IN (%u, %u);\n",
+							  LargeObjectRelationId, LargeObjectLOidPNIndexId);
 		else
-			appendPQExpBuffer(loFrozenQry, "SELECT relfrozenxid, 0 AS relminmxid\n"
+			appendPQExpBuffer(loFrozenQry, "SELECT relfrozenxid, 0 AS relminmxid, relfilenode, oid\n"
 							  "FROM pg_catalog.pg_class\n"
-							  "WHERE oid = %u;\n",
-							  LargeObjectRelationId);
+							  "WHERE oid IN (%u, %u);\n",
+							  LargeObjectRelationId, LargeObjectLOidPNIndexId);
 
-		lo_res = ExecuteSqlQueryForSingleRow(fout, loFrozenQry->data);
+		lo_res = ExecuteSqlQuery(fout, loFrozenQry->data, PGRES_TUPLES_OK);
 
 		i_relfrozenxid = PQfnumber(lo_res, "relfrozenxid");
 		i_relminmxid = PQfnumber(lo_res, "relminmxid");
+		i_relfilenode = PQfnumber(lo_res, "relfilenode");
+		i_oid = PQfnumber(lo_res, "oid");
+
+		appendPQExpBufferStr(loOutQry, "\n-- For binary upgrade, preserve values for pg_largeobject and its index\n");
+		for (int i = 0; i < PQntuples(lo_res); ++i)
+			appendPQExpBuffer(loOutQry, "UPDATE pg_catalog.pg_class\n"
+							  "SET relfrozenxid = '%u', relminmxid = '%u', relfilenode = '%u'\n"
+							  "WHERE oid = %u;\n",
+							  atooid(PQgetvalue(lo_res, i, i_relfrozenxid)),
+							  atooid(PQgetvalue(lo_res, i, i_relminmxid)),
+							  atooid(PQgetvalue(lo_res, i, i_relfilenode)),
+							  atooid(PQgetvalue(lo_res, i, i_oid)));
 
-		appendPQExpBufferStr(loOutQry, "\n-- For binary upgrade, set pg_largeobject relfrozenxid and relminmxid\n");
-		appendPQExpBuffer(loOutQry, "UPDATE pg_catalog.pg_class\n"
-						  "SET relfrozenxid = '%u', relminmxid = '%u'\n"
-						  "WHERE oid = %u;\n",
-						  atooid(PQgetvalue(lo_res, 0, i_relfrozenxid)),
-						  atooid(PQgetvalue(lo_res, 0, i_relminmxid)),
-						  LargeObjectRelationId);
 		ArchiveEntry(fout, nilCatalogId, createDumpId(),
 					 ARCHIVE_OPTS(.tag = "pg_largeobject",
 								  .description = "pg_largeobject",
diff --git a/src/bin/pg_upgrade/relfilenumber.c b/src/bin/pg_upgrade/relfilenumber.c
index b3ad8209eca..0dd649aa0f5 100644
--- a/src/bin/pg_upgrade/relfilenumber.c
+++ b/src/bin/pg_upgrade/relfilenumber.c
@@ -13,9 +13,13 @@
 
 #include "access/transam.h"
 #include "catalog/pg_class_d.h"
+#include "catalog/pg_largeobject_d.h"
 #include "pg_upgrade.h"
 
-static void transfer_single_new_db(FileNameMap *maps, int size, char *old_tablespace);
+static void transfer_single_new_db(FileNameMap *maps, int size, char *old_tablespace,
+	bool *has_lotable, bool *has_loindex);
+static void remove_stray_largeobject_files(DbInfo *new_db, bool has_lotable,
+		bool has_loindex);
 static void transfer_relfile(FileNameMap *map, const char *suffix, bool vm_must_add_frozenbit);
 
 
@@ -98,6 +102,8 @@ transfer_all_new_dbs(DbInfoArr *old_db_arr, DbInfoArr *new_db_arr,
 				   *new_db = NULL;
 		FileNameMap *mappings;
 		int			n_maps;
+		bool		has_lotable = false;
+		bool		has_loindex = false;
 
 		/*
 		 * Advance past any databases that exist in the new cluster but not in
@@ -117,22 +123,66 @@ transfer_all_new_dbs(DbInfoArr *old_db_arr, DbInfoArr *new_db_arr,
 
 		mappings = gen_db_file_maps(old_db, new_db, &n_maps, old_pgdata,
 									new_pgdata);
-		if (n_maps)
-		{
-			transfer_single_new_db(mappings, n_maps, old_tablespace);
-		}
+		transfer_single_new_db(mappings, n_maps, old_tablespace,
+			&has_lotable, &has_loindex);
+
+		/*
+		 * Remove the original pg_largeobject files created by initdb if they
+		 * exist in the new cluster with different relfilenodes.  This avoids
+		 * leaving behind stray files unassociated with any relation.
+		 */
+		remove_stray_largeobject_files(new_db, has_lotable, has_loindex);
+
 		/* We allocate something even for n_maps == 0 */
 		pg_free(mappings);
 	}
 }
 
+/*
+ * If there's no relation in the cluster whose relfilenode is the OID of the
+ * largeobject table/index, then remove the stray file.  It ought to have a
+ * single segment and fork.
+ */
+static void
+remove_stray_largeobject_files(DbInfo *new_db, bool has_lotable,
+		bool has_loindex)
+{
+	char		prefix[MAXPGPATH];
+	char		largeobj_file[MAXPGPATH];
+
+	if (new_db->db_tablespace[0] != '\0')
+		snprintf(prefix, sizeof(prefix), "%s/%s/%u",
+				new_db->db_tablespace, new_cluster.tablespace_suffix,
+				new_db->db_oid);
+	else
+		snprintf(prefix, sizeof(prefix), "%s/base/%u",
+				new_cluster.pgdata, new_db->db_oid);
+
+	if (!has_lotable)
+	{
+		snprintf(largeobj_file, sizeof(largeobj_file), "%s/%u",
+				prefix, LargeObjectRelationId);
+		if (unlink(largeobj_file) != 0)
+			pg_fatal("Unable to remove %s.\n", largeobj_file);
+	}
+
+	if (!has_loindex)
+	{
+		snprintf(largeobj_file, sizeof(largeobj_file), "%s/%u",
+				prefix, LargeObjectLOidPNIndexId);
+		if (unlink(largeobj_file) != 0)
+			pg_fatal("Unable to remove %s.\n", largeobj_file);
+	}
+}
+
 /*
  * transfer_single_new_db()
  *
  * create links for mappings stored in "maps" array.
  */
 static void
-transfer_single_new_db(FileNameMap *maps, int size, char *old_tablespace)
+transfer_single_new_db(FileNameMap *maps, int size, char *old_tablespace,
+	bool *has_lotable, bool *has_loindex)
 {
 	int			mapnum;
 	bool		vm_must_add_frozenbit = false;
@@ -146,6 +196,12 @@ transfer_single_new_db(FileNameMap *maps, int size, char *old_tablespace)
 
 	for (mapnum = 0; mapnum < size; mapnum++)
 	{
+		/* Keep track of whether a filenode matches the OID */
+		if (maps[mapnum].relfilenumber == LargeObjectRelationId)
+			*has_lotable = true;
+		if (maps[mapnum].relfilenumber == LargeObjectLOidPNIndexId)
+			*has_loindex = true;
+
 		if (old_tablespace == NULL ||
 			strcmp(maps[mapnum].old_tablespace, old_tablespace) == 0)
 		{
diff --git a/src/bin/pg_upgrade/t/002_pg_upgrade.pl b/src/bin/pg_upgrade/t/002_pg_upgrade.pl
index 2f9b13bf0ae..9ed48c4e06a 100644
--- a/src/bin/pg_upgrade/t/002_pg_upgrade.pl
+++ b/src/bin/pg_upgrade/t/002_pg_upgrade.pl
@@ -208,6 +208,8 @@ if (defined($ENV{oldinstall}))
 	}
 }
 
+$oldnode->safe_psql("regression", "VACUUM FULL pg_largeobject;");
+
 # In a VPATH build, we'll be started in the source directory, but we want
 # to run pg_upgrade in the build directory so that any files generated finish
 # in it, like delete_old_cluster.{sh,bat}.
-- 
2.17.1


--m+utxuhC6KVTvgNz--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH v8 1/7] Row pattern recognition patch for raw parser.
@ 2023-09-25 05:01  Tatsuo Ishii <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Tatsuo Ishii @ 2023-09-25 05:01 UTC (permalink / raw)

---
 src/backend/parser/gram.y       | 222 ++++++++++++++++++++++++++++++--
 src/include/nodes/parsenodes.h  |  56 ++++++++
 src/include/parser/kwlist.h     |   8 ++
 src/include/parser/parse_node.h |   1 +
 4 files changed, 273 insertions(+), 14 deletions(-)

diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
index 7d2032885e..74c2069050 100644
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -251,6 +251,8 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	DefElem	   *defelt;
 	SortBy	   *sortby;
 	WindowDef  *windef;
+	RPCommonSyntax	*rpcom;
+	RPSubsetItem	*rpsubset;
 	JoinExpr   *jexpr;
 	IndexElem  *ielem;
 	StatsElem  *selem;
@@ -278,6 +280,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	MergeWhenClause *mergewhen;
 	struct KeyActions *keyactions;
 	struct KeyAction *keyaction;
+	RPSkipTo	skipto;
 }
 
 %type <node>	stmt toplevel_stmt schema_stmt routine_body_stmt
@@ -453,8 +456,12 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 				TriggerTransitions TriggerReferencing
 				vacuum_relation_list opt_vacuum_relation_list
 				drop_option_list pub_obj_list
-
-%type <node>	opt_routine_body
+				row_pattern_measure_list row_pattern_definition_list
+				opt_row_pattern_subset_clause
+				row_pattern_subset_list row_pattern_subset_rhs
+				row_pattern
+%type <rpsubset>	 row_pattern_subset_item
+%type <node>	opt_routine_body row_pattern_term
 %type <groupclause> group_clause
 %type <list>	group_by_list
 %type <node>	group_by_item empty_grouping_set rollup_clause cube_clause
@@ -551,6 +558,8 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 %type <range>	relation_expr_opt_alias
 %type <node>	tablesample_clause opt_repeatable_clause
 %type <target>	target_el set_target insert_column_item
+				row_pattern_measure_item row_pattern_definition
+%type <skipto>	first_or_last
 
 %type <str>		generic_option_name
 %type <node>	generic_option_arg
@@ -633,6 +642,9 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 %type <list>	window_clause window_definition_list opt_partition_clause
 %type <windef>	window_definition over_clause window_specification
 				opt_frame_clause frame_extent frame_bound
+%type <rpcom>	opt_row_pattern_common_syntax opt_row_pattern_skip_to
+%type <boolean>	opt_row_pattern_initial_or_seek
+%type <list>	opt_row_pattern_measures
 %type <ival>	opt_window_exclusion_clause
 %type <str>		opt_existing_window_name
 %type <boolean> opt_if_not_exists
@@ -659,7 +671,6 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 				json_object_constructor_null_clause_opt
 				json_array_constructor_null_clause_opt
 
-
 /*
  * Non-keyword token types.  These are hard-wired into the "flex" lexer.
  * They must be listed first so that their numeric codes do not depend on
@@ -702,7 +713,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	CURRENT_TIME CURRENT_TIMESTAMP CURRENT_USER CURSOR CYCLE
 
 	DATA_P DATABASE DAY_P DEALLOCATE DEC DECIMAL_P DECLARE DEFAULT DEFAULTS
-	DEFERRABLE DEFERRED DEFINER DELETE_P DELIMITER DELIMITERS DEPENDS DEPTH DESC
+	DEFERRABLE DEFERRED DEFINE DEFINER DELETE_P DELIMITER DELIMITERS DEPENDS DEPTH DESC
 	DETACH DICTIONARY DISABLE_P DISCARD DISTINCT DO DOCUMENT_P DOMAIN_P
 	DOUBLE_P DROP
 
@@ -718,7 +729,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	HANDLER HAVING HEADER_P HOLD HOUR_P
 
 	IDENTITY_P IF_P ILIKE IMMEDIATE IMMUTABLE IMPLICIT_P IMPORT_P IN_P INCLUDE
-	INCLUDING INCREMENT INDENT INDEX INDEXES INHERIT INHERITS INITIALLY INLINE_P
+	INCLUDING INCREMENT INDENT INDEX INDEXES INHERIT INHERITS INITIAL INITIALLY INLINE_P
 	INNER_P INOUT INPUT_P INSENSITIVE INSERT INSTEAD INT_P INTEGER
 	INTERSECT INTERVAL INTO INVOKER IS ISNULL ISOLATION
 
@@ -731,7 +742,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MEASURES MERGE METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -743,8 +754,8 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	ORDER ORDINALITY OTHERS OUT_P OUTER_P
 	OVER OVERLAPS OVERLAY OVERRIDING OWNED OWNER
 
-	PARALLEL PARAMETER PARSER PARTIAL PARTITION PASSING PASSWORD
-	PLACING PLANS POLICY
+	PARALLEL PARAMETER PARSER PARTIAL PARTITION PASSING PASSWORD PAST
+	PATTERN_P PERMUTE PLACING PLANS POLICY
 	POSITION PRECEDING PRECISION PRESERVE PREPARE PREPARED PRIMARY
 	PRIOR PRIVILEGES PROCEDURAL PROCEDURE PROCEDURES PROGRAM PUBLICATION
 
@@ -755,12 +766,13 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 	RESET RESTART RESTRICT RETURN RETURNING RETURNS REVOKE RIGHT ROLE ROLLBACK ROLLUP
 	ROUTINE ROUTINES ROW ROWS RULE
 
-	SAVEPOINT SCALAR SCHEMA SCHEMAS SCROLL SEARCH SECOND_P SECURITY SELECT
+	SAVEPOINT SCALAR SCHEMA SCHEMAS SCROLL SEARCH SECOND_P SECURITY SEEK SELECT
 	SEQUENCE SEQUENCES
+
 	SERIALIZABLE SERVER SESSION SESSION_USER SET SETS SETOF SHARE SHOW
 	SIMILAR SIMPLE SKIP SMALLINT SNAPSHOT SOME SQL_P STABLE STANDALONE_P
 	START STATEMENT STATISTICS STDIN STDOUT STORAGE STORED STRICT_P STRIP_P
-	SUBSCRIPTION SUBSTRING SUPPORT SYMMETRIC SYSID SYSTEM_P SYSTEM_USER
+	SUBSCRIPTION SUBSET SUBSTRING SUPPORT SYMMETRIC SYSID SYSTEM_P SYSTEM_USER
 
 	TABLE TABLES TABLESAMPLE TABLESPACE TEMP TEMPLATE TEMPORARY TEXT_P THEN
 	TIES TIME TIMESTAMP TO TRAILING TRANSACTION TRANSFORM
@@ -853,6 +865,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
  */
 %nonassoc	UNBOUNDED		/* ideally would have same precedence as IDENT */
 %nonassoc	IDENT PARTITION RANGE ROWS GROUPS PRECEDING FOLLOWING CUBE ROLLUP
+%nonassoc	MEASURES AFTER INITIAL SEEK PATTERN_P
 %left		Op OPERATOR		/* multi-character ops and user-defined operators */
 %left		'+' '-'
 %left		'*' '/' '%'
@@ -15857,7 +15870,8 @@ over_clause: OVER window_specification
 		;
 
 window_specification: '(' opt_existing_window_name opt_partition_clause
-						opt_sort_clause opt_frame_clause ')'
+						opt_sort_clause opt_row_pattern_measures opt_frame_clause
+						opt_row_pattern_common_syntax ')'
 				{
 					WindowDef  *n = makeNode(WindowDef);
 
@@ -15865,10 +15879,12 @@ window_specification: '(' opt_existing_window_name opt_partition_clause
 					n->refname = $2;
 					n->partitionClause = $3;
 					n->orderClause = $4;
+					n->rowPatternMeasures = $5;
 					/* copy relevant fields of opt_frame_clause */
-					n->frameOptions = $5->frameOptions;
-					n->startOffset = $5->startOffset;
-					n->endOffset = $5->endOffset;
+					n->frameOptions = $6->frameOptions;
+					n->startOffset = $6->startOffset;
+					n->endOffset = $6->endOffset;
+					n->rpCommonSyntax = $7;
 					n->location = @1;
 					$$ = n;
 				}
@@ -15892,6 +15908,31 @@ opt_partition_clause: PARTITION BY expr_list		{ $$ = $3; }
 			| /*EMPTY*/								{ $$ = NIL; }
 		;
 
+/*
+ * ROW PATTERN_P MEASURES
+ */
+opt_row_pattern_measures: MEASURES row_pattern_measure_list	{ $$ = $2; }
+			| /*EMPTY*/								{ $$ = NIL; }
+		;
+
+row_pattern_measure_list:
+			row_pattern_measure_item
+					{ $$ = list_make1($1); }
+			| row_pattern_measure_list ',' row_pattern_measure_item
+					{ $$ = lappend($1, $3); }
+		;
+
+row_pattern_measure_item:
+			a_expr AS ColLabel
+				{
+					$$ = makeNode(ResTarget);
+					$$->name = $3;
+					$$->indirection = NIL;
+					$$->val = (Node *) $1;
+					$$->location = @1;
+				}
+		;
+
 /*
  * For frame clauses, we return a WindowDef, but only some fields are used:
  * frameOptions, startOffset, and endOffset.
@@ -16051,6 +16092,143 @@ opt_window_exclusion_clause:
 			| /*EMPTY*/				{ $$ = 0; }
 		;
 
+opt_row_pattern_common_syntax:
+opt_row_pattern_skip_to opt_row_pattern_initial_or_seek
+				PATTERN_P '(' row_pattern ')'
+				opt_row_pattern_subset_clause
+				DEFINE row_pattern_definition_list
+			{
+				RPCommonSyntax *n = makeNode(RPCommonSyntax);
+				n->rpSkipTo = $1->rpSkipTo;
+				n->rpSkipVariable = $1->rpSkipVariable;
+				n->initial = $2;
+				n->rpPatterns = $5;
+				n->rpSubsetClause = $7;
+				n->rpDefs = $9;
+				$$ = n;
+			}
+			| /*EMPTY*/		{ $$ = NULL; }
+	;
+
+opt_row_pattern_skip_to:
+			AFTER MATCH SKIP TO NEXT ROW
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = ST_NEXT_ROW;
+					n->rpSkipVariable = NULL;
+					$$ = n;
+			}
+			| AFTER MATCH SKIP PAST LAST_P ROW
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = ST_PAST_LAST_ROW;
+					n->rpSkipVariable = NULL;
+					$$ = n;
+				}
+			| AFTER MATCH SKIP TO first_or_last ColId
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = $5;
+					n->rpSkipVariable = $6;
+					$$ = n;
+				}
+/*
+			| AFTER MATCH SKIP TO LAST_P ColId		%prec LAST_P
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = ST_LAST_VARIABLE;
+					n->rpSkipVariable = $6;
+					$$ = n;
+				}
+			| AFTER MATCH SKIP TO ColId
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					n->rpSkipTo = ST_VARIABLE;
+					n->rpSkipVariable = $5;
+					$$ = n;
+				}
+*/
+			| /*EMPTY*/
+				{
+					RPCommonSyntax *n = makeNode(RPCommonSyntax);
+					/* temporary set default to ST_NEXT_ROW */
+					n->rpSkipTo = ST_PAST_LAST_ROW;
+					n->rpSkipVariable = NULL;
+					$$ = n;
+				}
+	;
+
+first_or_last:
+			FIRST_P		{ $$ = ST_FIRST_VARIABLE; }
+			| LAST_P	{ $$ = ST_LAST_VARIABLE; }
+	;
+
+opt_row_pattern_initial_or_seek:
+			INITIAL			{ $$ = true; }
+			| SEEK
+				{
+					ereport(ERROR,
+							(errcode(ERRCODE_SYNTAX_ERROR),
+							 errmsg("SEEK is not supported"),
+							 errhint("Use INITIAL."),
+							 parser_errposition(@1)));
+				}
+			| /*EMPTY*/		{ $$ = true; }
+		;
+
+row_pattern:
+			row_pattern_term							{ $$ = list_make1($1); }
+			| row_pattern row_pattern_term				{ $$ = lappend($1, $2); }
+		;
+
+row_pattern_term:
+			ColId	{ $$ = (Node *) makeSimpleA_Expr(AEXPR_OP, "", (Node *)makeString($1), NULL, @1); }
+			| ColId '*'	{ $$ = (Node *) makeSimpleA_Expr(AEXPR_OP, "*", (Node *)makeString($1), NULL, @1); }
+			| ColId '+'	{ $$ = (Node *) makeSimpleA_Expr(AEXPR_OP, "+", (Node *)makeString($1), NULL, @1); }
+			| ColId '?'	{ $$ = (Node *) makeSimpleA_Expr(AEXPR_OP, "?", (Node *)makeString($1), NULL, @1); }
+		;
+
+opt_row_pattern_subset_clause:
+			SUBSET row_pattern_subset_list	{ $$ = $2; }
+			| /*EMPTY*/												{ $$ = NIL; }
+		;
+
+row_pattern_subset_list:
+			row_pattern_subset_item									{ $$ = list_make1($1); }
+			| row_pattern_subset_list ',' row_pattern_subset_item	{ $$ = lappend($1, $3); }
+			| /*EMPTY*/												{ $$ = NIL; }
+		;
+
+row_pattern_subset_item: ColId '=' '(' row_pattern_subset_rhs ')'
+			{
+				RPSubsetItem *n = makeNode(RPSubsetItem);
+				n->name = $1;
+				n->rhsVariable = $4;
+				$$ = n;
+			}
+		;
+
+row_pattern_subset_rhs:
+			ColId								{ $$ = list_make1(makeStringConst($1, @1)); }
+			| row_pattern_subset_rhs ',' ColId	{ $$ = lappend($1, makeStringConst($3, @1)); }
+			| /*EMPTY*/							{ $$ = NIL; }
+		;
+
+row_pattern_definition_list:
+			row_pattern_definition										{ $$ = list_make1($1); }
+			| row_pattern_definition_list ',' row_pattern_definition	{ $$ = lappend($1, $3); }
+		;
+
+row_pattern_definition:
+			ColId AS a_expr
+				{
+					$$ = makeNode(ResTarget);
+					$$->name = $1;
+					$$->indirection = NIL;
+					$$->val = (Node *) $3;
+					$$->location = @1;
+				}
+		;
 
 /*
  * Supporting nonterminals for expressions.
@@ -17146,6 +17324,7 @@ unreserved_keyword:
 			| INDEXES
 			| INHERIT
 			| INHERITS
+			| INITIAL
 			| INLINE_P
 			| INPUT_P
 			| INSENSITIVE
@@ -17173,6 +17352,7 @@ unreserved_keyword:
 			| MATCHED
 			| MATERIALIZED
 			| MAXVALUE
+			| MEASURES
 			| MERGE
 			| METHOD
 			| MINUTE_P
@@ -17215,6 +17395,9 @@ unreserved_keyword:
 			| PARTITION
 			| PASSING
 			| PASSWORD
+			| PAST
+			| PATTERN_P
+			| PERMUTE
 			| PLANS
 			| POLICY
 			| PRECEDING
@@ -17265,6 +17448,7 @@ unreserved_keyword:
 			| SEARCH
 			| SECOND_P
 			| SECURITY
+			| SEEK
 			| SEQUENCE
 			| SEQUENCES
 			| SERIALIZABLE
@@ -17290,6 +17474,7 @@ unreserved_keyword:
 			| STRICT_P
 			| STRIP_P
 			| SUBSCRIPTION
+			| SUBSET
 			| SUPPORT
 			| SYSID
 			| SYSTEM_P
@@ -17477,6 +17662,7 @@ reserved_keyword:
 			| CURRENT_USER
 			| DEFAULT
 			| DEFERRABLE
+			| DEFINE
 			| DESC
 			| DISTINCT
 			| DO
@@ -17639,6 +17825,7 @@ bare_label_keyword:
 			| DEFAULTS
 			| DEFERRABLE
 			| DEFERRED
+			| DEFINE
 			| DEFINER
 			| DELETE_P
 			| DELIMITER
@@ -17714,6 +17901,7 @@ bare_label_keyword:
 			| INDEXES
 			| INHERIT
 			| INHERITS
+			| INITIAL
 			| INITIALLY
 			| INLINE_P
 			| INNER_P
@@ -17763,6 +17951,7 @@ bare_label_keyword:
 			| MATCHED
 			| MATERIALIZED
 			| MAXVALUE
+			| MEASURES
 			| MERGE
 			| METHOD
 			| MINVALUE
@@ -17816,6 +18005,9 @@ bare_label_keyword:
 			| PARTITION
 			| PASSING
 			| PASSWORD
+			| PAST
+			| PATTERN_P
+			| PERMUTE
 			| PLACING
 			| PLANS
 			| POLICY
@@ -17872,6 +18064,7 @@ bare_label_keyword:
 			| SCROLL
 			| SEARCH
 			| SECURITY
+			| SEEK
 			| SELECT
 			| SEQUENCE
 			| SEQUENCES
@@ -17903,6 +18096,7 @@ bare_label_keyword:
 			| STRICT_P
 			| STRIP_P
 			| SUBSCRIPTION
+			| SUBSET
 			| SUBSTRING
 			| SUPPORT
 			| SYMMETRIC
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
index fef4c714b8..657651df1d 100644
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -547,6 +547,44 @@ typedef struct SortBy
 	int			location;		/* operator location, or -1 if none/unknown */
 } SortBy;
 
+/*
+ * AFTER MATCH row pattern skip to types in row pattern common syntax
+ */
+typedef enum RPSkipTo
+{
+	ST_NONE,					/* AFTER MATCH omitted */
+	ST_NEXT_ROW,				/* SKIP TO NEXT ROW */
+	ST_PAST_LAST_ROW,			/* SKIP TO PAST LAST ROW */
+	ST_FIRST_VARIABLE,			/* SKIP TO FIRST variable name */
+	ST_LAST_VARIABLE,			/* SKIP TO LAST variable name */
+	ST_VARIABLE					/* SKIP TO variable name */
+} RPSkipTo;
+
+/*
+ * Row Pattern SUBSET clause item
+ */
+typedef struct RPSubsetItem
+{
+	NodeTag		type;
+	char	   *name;			/* Row Pattern SUBSET clause variable name */
+	List	   *rhsVariable;	/* Row Pattern SUBSET rhs variables (list of char *string) */
+} RPSubsetItem;
+
+/*
+ * RowPatternCommonSyntax - raw representation of row pattern common syntax
+ *
+ */
+typedef struct RPCommonSyntax
+{
+	NodeTag		type;
+	RPSkipTo	rpSkipTo;		/* Row Pattern AFTER MATCH SKIP type */
+	char	   *rpSkipVariable;	/* Row Pattern Skip To variable name, if any */
+	bool		initial;		/* true if <row pattern initial or seek> is initial */
+	List	   *rpPatterns;		/* PATTERN variables (list of A_Expr) */
+	List	   *rpSubsetClause;	/* row pattern subset clause (list of RPSubsetItem), if any */
+	List	   *rpDefs;			/* row pattern definitions clause (list of ResTarget) */
+} RPCommonSyntax;
+
 /*
  * WindowDef - raw representation of WINDOW and OVER clauses
  *
@@ -562,6 +600,8 @@ typedef struct WindowDef
 	char	   *refname;		/* referenced window name, if any */
 	List	   *partitionClause;	/* PARTITION BY expression list */
 	List	   *orderClause;	/* ORDER BY (list of SortBy) */
+	List	   *rowPatternMeasures;	/* row pattern measures (list of ResTarget) */
+	RPCommonSyntax *rpCommonSyntax;	/* row pattern common syntax */
 	int			frameOptions;	/* frame_clause options, see below */
 	Node	   *startOffset;	/* expression for starting bound, if any */
 	Node	   *endOffset;		/* expression for ending bound, if any */
@@ -1483,6 +1523,11 @@ typedef struct GroupingSet
  * the orderClause might or might not be copied (see copiedOrder); the framing
  * options are never copied, per spec.
  *
+ * "defineClause" is Row Pattern Recognition DEFINE clause (list of
+ * TargetEntry). TargetEntry.resname represents row pattern definition
+ * variable name. "patternVariable" and "patternRegexp" represents PATTERN
+ * clause.
+ *
  * The information relevant for the query jumbling is the partition clause
  * type and its bounds.
  */
@@ -1514,6 +1559,17 @@ typedef struct WindowClause
 	Index		winref;			/* ID referenced by window functions */
 	/* did we copy orderClause from refname? */
 	bool		copiedOrder pg_node_attr(query_jumble_ignore);
+	/* Row Pattern AFTER MACH SKIP clause */
+	RPSkipTo	rpSkipTo;		/* Row Pattern Skip To type */
+	bool		initial;		/* true if <row pattern initial or seek> is initial */
+	/* Row Pattern DEFINE clause (list of TargetEntry) */
+	List		*defineClause;
+	/* Row Pattern DEFINE variable initial names (list of String) */
+	List		*defineInitial;
+	/* Row Pattern PATTERN variable name (list of String) */
+	List		*patternVariable;
+	/* Row Pattern PATTERN regular expression quantifier ('+' or ''. list of String) */
+	List		*patternRegexp;
 } WindowClause;
 
 /*
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
index 5984dcfa4b..2804333b53 100644
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -128,6 +128,7 @@ PG_KEYWORD("default", DEFAULT, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("defaults", DEFAULTS, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("deferrable", DEFERRABLE, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("deferred", DEFERRED, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("define", DEFINE, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("definer", DEFINER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("delete", DELETE_P, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("delimiter", DELIMITER, UNRESERVED_KEYWORD, BARE_LABEL)
@@ -212,6 +213,7 @@ PG_KEYWORD("index", INDEX, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("indexes", INDEXES, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("inherit", INHERIT, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("inherits", INHERITS, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("initial", INITIAL, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("initially", INITIALLY, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("inline", INLINE_P, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("inner", INNER_P, TYPE_FUNC_NAME_KEYWORD, BARE_LABEL)
@@ -265,6 +267,7 @@ PG_KEYWORD("match", MATCH, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("matched", MATCHED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("measures", MEASURES, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
@@ -326,6 +329,9 @@ PG_KEYWORD("partial", PARTIAL, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("partition", PARTITION, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("passing", PASSING, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("password", PASSWORD, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("past", PAST, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("pattern", PATTERN_P, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("permute", PERMUTE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("placing", PLACING, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("plans", PLANS, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("policy", POLICY, UNRESERVED_KEYWORD, BARE_LABEL)
@@ -385,6 +391,7 @@ PG_KEYWORD("scroll", SCROLL, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("search", SEARCH, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("second", SECOND_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("security", SECURITY, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("seek", SEEK, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("select", SELECT, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("sequence", SEQUENCE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("sequences", SEQUENCES, UNRESERVED_KEYWORD, BARE_LABEL)
@@ -416,6 +423,7 @@ PG_KEYWORD("stored", STORED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("strict", STRICT_P, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("strip", STRIP_P, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("subscription", SUBSCRIPTION, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("subset", SUBSET, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("substring", SUBSTRING, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("support", SUPPORT, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("symmetric", SYMMETRIC, RESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
index f589112d5e..6640090910 100644
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -51,6 +51,7 @@ typedef enum ParseExprKind
 	EXPR_KIND_WINDOW_FRAME_RANGE,	/* window frame clause with RANGE */
 	EXPR_KIND_WINDOW_FRAME_ROWS,	/* window frame clause with ROWS */
 	EXPR_KIND_WINDOW_FRAME_GROUPS,	/* window frame clause with GROUPS */
+	EXPR_KIND_RPR_DEFINE,		/* DEFINE */
 	EXPR_KIND_SELECT_TARGET,	/* SELECT target list item */
 	EXPR_KIND_INSERT_TARGET,	/* INSERT target list item */
 	EXPR_KIND_UPDATE_SOURCE,	/* UPDATE assignment source item */
-- 
2.25.1


----Next_Part(Mon_Sep_25_14_26_30_2023_752)--
Content-Type: Text/X-Patch; charset=us-ascii
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename="v8-0002-Row-pattern-recognition-patch-parse-analysis.patch"



^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread

* [PATCH 5/5] Serialize decoded tuples without flattening.
@ 2026-02-27 18:01  Antonin Houska <[email protected]>
  0 siblings, 0 replies; 499+ messages in thread

From: Antonin Houska @ 2026-02-27 18:01 UTC (permalink / raw)

So far we used to flatten the tuples produced by logical decoding in order to
write them to a temporary file. That means that all the TOASTed attributes
ended up in a single chunk of memory. The problem is that even if individual
attributes do fit into the maximum chunk size (currently 1 GB), the whole
tuple is not guaranteed to fit.

The flattening might be a problem in the future anyway, since there's an
effort to restrict the impact of REPACK (CONCURRENTLY) on the VACUUM xmin
horizon. However, as the flattening assumes scans of the TOAST relaation, the
less we restrict VACUUM, the more likely it is that the TOAST attributes
needed for the flattening won't be available at decoding time.

(This patch also removes unnecessary field 'tupdesc' from RepackDecodingState
and allocates it in the appropriate memory context.)
---
 src/backend/commands/cluster.c                | 260 +++++++++++++++---
 .../pgoutput_repack/pgoutput_repack.c         | 114 ++++++--
 src/include/commands/cluster.h                |   5 +-
 .../expected/repack_toast.out                 |   1 +
 .../injection_points/specs/repack_toast.spec  |   7 +
 5 files changed, 318 insertions(+), 69 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index db710259eeb..e4de410e733 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -35,6 +35,7 @@
 #include "postgres.h"
 
 #include "access/amapi.h"
+#include "access/detoast.h"
 #include "access/heapam.h"
 #include "access/multixact.h"
 #include "access/relscan.h"
@@ -285,6 +286,10 @@ static void apply_concurrent_update(Relation rel, HeapTuple tup,
 									IndexInsertState *iistate,
 									TupleTableSlot *index_slot);
 static void apply_concurrent_delete(Relation rel, HeapTuple tup_target);
+static HeapTuple restore_tuple(BufFile *file, Relation relation,
+							   MemoryContext cxt);
+static HeapTuple adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+									   HeapTuple tup_src, MemoryContext cxt);
 static HeapTuple find_target_tuple(Relation rel, ChangeDest *dest,
 								   HeapTuple tup_key,
 								   TupleTableSlot *ident_slot);
@@ -2574,6 +2579,7 @@ setup_logical_decoding(Oid relid)
 	LogicalDecodingContext *ctx;
 	NameData	slotname;
 	RepackDecodingState *dstate;
+	MemoryContext oldcxt;
 
 	/*
 	 * REPACK CONCURRENTLY is not allowed in a transaction block, so this
@@ -2624,14 +2630,6 @@ setup_logical_decoding(Oid relid)
 	 */
 	ctx->reader->routine.page_read = read_local_xlog_page_no_wait;
 
-	/*
-	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
-	 * WAL.
-	 */
-	ctx->reader->private_data = MemoryContextAllocZero(ctx->context,
-													   sizeof(ReadLocalXLogPageNoWaitPrivate));
-
-
 	/* Some WAL records should have been read. */
 	Assert(ctx->reader->EndRecPtr != InvalidXLogRecPtr);
 
@@ -2642,18 +2640,23 @@ setup_logical_decoding(Oid relid)
 	XLByteToSeg(ctx->reader->EndRecPtr, repack_current_segment,
 				wal_segment_size);
 
-	dstate = palloc0_object(RepackDecodingState);
-	dstate->relid = relid;
+	/* Our private state belongs to the decoding context. */
+	oldcxt = MemoryContextSwitchTo(ctx->context);
 
 	/*
-	 * Tuple descriptor may be needed to flatten a tuple before we write it to
-	 * a file. A copy is needed because the decoding worker invalidates system
-	 * caches before it starts to do the actual work.
+	 * read_local_xlog_page_no_wait() needs to be able to indicate the end of
+	 * WAL.
 	 */
-	rel = table_open(relid, AccessShareLock);
-	dstate->tupdesc = CreateTupleDescCopy(RelationGetDescr(rel));
+	ctx->reader->private_data = palloc0_object(ReadLocalXLogPageNoWaitPrivate);
+	dstate = palloc0_object(RepackDecodingState);
+	MemoryContextSwitchTo(oldcxt);
+
+#ifdef	USE_ASSERT_CHECKING
+	dstate->relid = relid;
+#endif
 
 	/* Avoid logical decoding of other relations. */
+	rel = table_open(relid, AccessShareLock);
 	repacked_rel_locator = rel->rd_locator;
 	toastrelid = rel->rd_rel->reltoastrelid;
 	if (OidIsValid(toastrelid))
@@ -2822,10 +2825,10 @@ static void
 apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 {
 	char		kind;
-	uint32		t_len;
 	Relation	rel = dest->rel;
 	TupleTableSlot *index_slot,
 			   *ident_slot;
+	MemoryContext apply_cxt;
 	HeapTuple	tup_old = NULL;
 
 	/* TupleTableSlot is needed to pass the tuple to ExecInsertIndexTuples(). */
@@ -2835,6 +2838,14 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	/* A slot to fetch tuples from identity index. */
 	ident_slot = table_slot_create(rel, NULL);
 
+	/*
+	 * Use a separate memory context for the tuples and any memory needed to
+	 * process them.
+	 */
+	apply_cxt = AllocSetContextCreate(TopTransactionContext,
+									  "REPACK - apply",
+									  ALLOCSET_DEFAULT_SIZES);
+
 	while (true)
 	{
 		size_t		nread;
@@ -2849,14 +2860,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 			break;
 
 		/* Read the tuple. */
-		BufFileReadExact(file, &t_len, sizeof(t_len));
-		tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
-		tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
-		BufFileReadExact(file, tup->t_data, t_len);
-		tup->t_len = t_len;
-		ItemPointerSetInvalid(&tup->t_self);
-		tup->t_tableOid = RelationGetRelid(dest->rel);
-
+		tup = restore_tuple(file, rel, apply_cxt);
 		if (kind == CHANGE_UPDATE_OLD)
 		{
 			Assert(tup_old == NULL);
@@ -2868,7 +2872,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 
 			apply_concurrent_insert(rel, tup, dest->iistate, index_slot);
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else if (kind == CHANGE_UPDATE_NEW || kind == CHANGE_DELETE)
 		{
@@ -2892,18 +2896,28 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 				elog(ERROR, "failed to find target tuple");
 
 			if (kind == CHANGE_UPDATE_NEW)
+			{
+				/*
+				 * If 'tup' contains TOAST pointers, they point to the old
+				 * relation's toast. Copy the corresponding TOAST pointers for
+				 * the new relation from the existing tuple. (The fact that we
+				 * received a TOAST pointer here implies that the attribute
+				 * hasn't changed.)
+				 */
+				if (HeapTupleHasExternal(tup))
+					tup = adjust_toast_pointers(rel, tup, tup_exist,
+												apply_cxt);
+
 				apply_concurrent_update(rel, tup, tup_exist, dest->iistate,
 										index_slot);
+			}
 			else
 				apply_concurrent_delete(rel, tup_exist);
 
 			if (tup_old != NULL)
-			{
-				pfree(tup_old);
 				tup_old = NULL;
-			}
 
-			pfree(tup);
+			MemoryContextReset(apply_cxt);
 		}
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
@@ -2920,6 +2934,7 @@ apply_concurrent_changes(BufFile *file, ChangeDest *dest)
 	}
 
 	/* Cleanup. */
+	MemoryContextDelete(apply_cxt);
 	ExecDropSingleTupleTableSlot(index_slot);
 	ExecDropSingleTupleTableSlot(ident_slot);
 }
@@ -3030,6 +3045,186 @@ apply_concurrent_delete(Relation rel, HeapTuple tup_target)
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_DELETED, 1);
 }
 
+/*
+ * Read tuple from file and store it in given memory context.
+ *
+ * External attributes are stored in separate memory chunks, in order to avoid
+ * exceeding MaxAllocSize - that could happen if the individual attributes are
+ * smaller than MaxAllocSize but the whole tuple is bigger.
+ */
+static HeapTuple
+restore_tuple(BufFile *file, Relation relation, MemoryContext cxt)
+{
+	uint32		t_len;
+	HeapTuple	tup;
+	MemoryContext oldcxt;
+	uint32		natt_ext;
+	List	   *attrs_ext = NIL;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	/* First, read the external attributes if there are some. */
+	BufFileReadExact(file, &natt_ext, sizeof(natt_ext));
+	for (int i = 0; i < natt_ext; i++)
+	{
+		varlena		varhdr;
+		char	   *ext_val;
+		Size		ext_val_size;
+
+		/*
+		 * Read the header into properly aligned memory before accessing it.
+		 */
+		BufFileReadExact(file, &varhdr, VARHDRSZ);
+		ext_val_size = VARSIZE_ANY(&varhdr);
+		ext_val = palloc0(ext_val_size);
+		memcpy(ext_val, &varhdr, VARHDRSZ);
+		BufFileReadExact(file, ext_val + VARHDRSZ, ext_val_size - VARHDRSZ);
+
+		attrs_ext = lappend(attrs_ext, ext_val);
+	}
+
+	/* Read the tuple. */
+	BufFileReadExact(file, &t_len, sizeof(t_len));
+	tup = (HeapTuple) palloc(HEAPTUPLESIZE + t_len);
+	tup->t_data = (HeapTupleHeader) ((char *) tup + HEAPTUPLESIZE);
+	BufFileReadExact(file, tup->t_data, t_len);
+	tup->t_len = t_len;
+	ItemPointerSetInvalid(&tup->t_self);
+	tup->t_tableOid = RelationGetRelid(relation);
+
+	/*
+	 * If there are "external indirect" attributes, make the tuple reference
+	 * them.
+	 */
+	if (natt_ext > 0)
+	{
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+		ListCell   *lc;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tup, desc, attrs, isnull);
+
+		lc = list_head(attrs_ext);
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer,
+					   *new_datum;
+			varatt_indirect redirect_pointer;
+
+			/* Find the external attributes, like in store_change(). */
+			if (attr->attisdropped)
+				continue;
+			if (attr->attlen != -1)
+				continue;
+			if (isnull[i])
+				continue;
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+			if (!VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				continue;
+
+			/* Update the attribute. */
+			memset(&redirect_pointer, 0, sizeof(redirect_pointer));
+			redirect_pointer.pointer = lfirst(lc);
+
+			new_datum = (varlena *) palloc0(INDIRECT_POINTER_SIZE);
+			SET_VARTAG_EXTERNAL(new_datum, VARTAG_INDIRECT);
+			memcpy(VARDATA_EXTERNAL(new_datum), &redirect_pointer,
+				   sizeof(redirect_pointer));
+
+			attrs[i] = PointerGetDatum(new_datum);
+
+			lc = lnext(attrs_ext, lc);
+		}
+		Assert(lc == NULL);
+
+		tup = heap_form_tuple(desc, attrs, isnull);
+	}
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return tup;
+}
+
+/*
+ * Copy TOAST pointers from 'tup_src' to 'tup_dst' and return the modified
+ * 'tup_dst'. Use 'cxt' for all memory allocations.
+ */
+static HeapTuple
+adjust_toast_pointers(Relation relation, HeapTuple tup_dst,
+					  HeapTuple tup_src, MemoryContext cxt)
+{
+	TupleDesc	desc;
+	Datum	   *attrs_dst;
+	bool	   *isnull_dst;
+	Datum	   *attrs_src = NULL;
+	bool	   *isnull_src = NULL;
+	MemoryContext oldcxt;
+	HeapTuple	result;
+
+	oldcxt = MemoryContextSwitchTo(cxt);
+
+	desc = RelationGetDescr(relation);
+	attrs_dst = palloc0_array(Datum, desc->natts);
+	isnull_dst = palloc0_array(bool, desc->natts);
+
+	heap_deform_tuple(tup_dst, desc, attrs_dst, isnull_dst);
+
+	for (int i = 0; i < desc->natts; i++)
+	{
+		CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+		varlena    *varlena_dst,
+				   *varlena_src;
+
+		if (attr->attisdropped)
+			continue;
+		if (attr->attlen != -1)
+			continue;
+		if (isnull_dst[i])
+			continue;
+
+		varlena_dst = (varlena *) DatumGetPointer(attrs_dst[i]);
+		if (!VARATT_IS_EXTERNAL_ONDISK(varlena_dst))
+			continue;
+
+		/*
+		 * The first TOAST pointer? The caller can easily check if the tuple
+		 * has external attributes(s) in general, but is not supposed to check
+		 * specifically for TOAST.
+		 */
+		if (attrs_src == NULL)
+		{
+			attrs_src = palloc0_array(Datum, desc->natts);
+			isnull_src = palloc0_array(bool, desc->natts);
+
+			heap_deform_tuple(tup_src, desc, attrs_src, isnull_src);
+		}
+
+		varlena_src = (varlena *) DatumGetPointer(attrs_src[i]);
+
+		/*
+		 * Logical decoding only uses TOAST pointer for attributes that
+		 * haven't changed.
+		 */
+		Assert(!isnull_src[i]);
+		Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_src));
+
+		/* Copy the pointer. */
+		memcpy(VARDATA_EXTERNAL(varlena_dst), VARDATA_EXTERNAL(varlena_src),
+			   sizeof(varatt_external));
+	}
+	result = heap_form_tuple(desc, attrs_dst, isnull_dst);
+
+	MemoryContextSwitchTo(oldcxt);
+
+	return result;
+}
+
 /*
  * Find the tuple to be updated or deleted.
  *
@@ -3263,15 +3458,8 @@ free_index_insert_state(IndexInsertState *iistate)
 static void
 cleanup_logical_decoding(LogicalDecodingContext *ctx)
 {
-	RepackDecodingState *dstate;
-
-	dstate = (RepackDecodingState *) ctx->output_writer_private;
-
-	FreeTupleDesc(dstate->tupdesc);
 	FreeDecodingContext(ctx);
-
 	ReplicationSlotDropAcquired();
-	pfree(dstate);
 }
 
 /*
diff --git a/src/backend/replication/pgoutput_repack/pgoutput_repack.c b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
index d0c464a98d5..8e22aeb163b 100644
--- a/src/backend/replication/pgoutput_repack/pgoutput_repack.c
+++ b/src/backend/replication/pgoutput_repack/pgoutput_repack.c
@@ -12,7 +12,7 @@
  */
 #include "postgres.h"
 
-#include "access/heaptoast.h"
+#include "access/detoast.h"
 #include "commands/cluster.h"
 #include "replication/snapbuild.h"
 #include "utils/memutils.h"
@@ -28,7 +28,7 @@ static void plugin_commit_txn(LogicalDecodingContext *ctx,
 							  ReorderBufferTXN *txn, XLogRecPtr commit_lsn);
 static void plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 						  Relation rel, ReorderBufferChange *change);
-static void store_change(LogicalDecodingContext *ctx,
+static void store_change(LogicalDecodingContext *ctx, Relation relation,
 						 ConcurrentChangeKind kind, HeapTuple tuple);
 
 void
@@ -97,9 +97,8 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
 
-	/* Only interested in one particular relation. */
-	if (relation->rd_id != dstate->relid)
-		return;
+	/* Changes of other relation should not have been decoded. */
+	Assert(RelationGetRelid(relation) == dstate->relid);
 
 	/* Decode entry depending on its type */
 	switch (change->action)
@@ -117,7 +116,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (newtuple == NULL)
 					elog(ERROR, "incomplete insert info.");
 
-				store_change(ctx, CHANGE_INSERT, newtuple);
+				store_change(ctx, relation, CHANGE_INSERT, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_UPDATE:
@@ -132,9 +131,9 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 					elog(ERROR, "incomplete update info.");
 
 				if (oldtuple != NULL)
-					store_change(ctx, CHANGE_UPDATE_OLD, oldtuple);
+					store_change(ctx, relation, CHANGE_UPDATE_OLD, oldtuple);
 
-				store_change(ctx, CHANGE_UPDATE_NEW, newtuple);
+				store_change(ctx, relation, CHANGE_UPDATE_NEW, newtuple);
 			}
 			break;
 		case REORDER_BUFFER_CHANGE_DELETE:
@@ -146,7 +145,7 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 				if (oldtuple == NULL)
 					elog(ERROR, "incomplete delete info.");
 
-				store_change(ctx, CHANGE_DELETE, oldtuple);
+				store_change(ctx, relation, CHANGE_DELETE, oldtuple);
 			}
 			break;
 		default:
@@ -164,39 +163,94 @@ plugin_change(LogicalDecodingContext *ctx, ReorderBufferTXN *txn,
 
 /* Store concurrent data change. */
 static void
-store_change(LogicalDecodingContext *ctx, ConcurrentChangeKind kind,
-			 HeapTuple tuple)
+store_change(LogicalDecodingContext *ctx, Relation relation,
+			 ConcurrentChangeKind kind, HeapTuple tuple)
 {
 	RepackDecodingState *dstate;
+	BufFile    *file;
 	char		kind_byte = (char) kind;
-	bool		flattened = false;
+	List	   *attrs_ext = NIL;
+	uint32		natt_ext;
 
 	dstate = (RepackDecodingState *) ctx->output_writer_private;
+	file = dstate->file;
 
 	/* Store the change kind. */
-	BufFileWrite(dstate->file, &kind_byte, 1);
+	BufFileWrite(file, &kind_byte, 1);
 
 	/*
-	 * ReorderBufferCommit() stores the TOAST chunks in its private memory
-	 * context and frees them after having called apply_change().  Therefore
-	 * we need flat copy (including TOAST) that we eventually copy into the
-	 * memory context which is available to decode_concurrent_changes().
+	 * If the tuple contains "external indirect" attributes, we need to write
+	 * the contents to the file because we have no control over that memory.
 	 */
 	if (HeapTupleHasExternal(tuple))
 	{
-		/*
-		 * toast_flatten_tuple_to_datum() might be more convenient but we
-		 * don't want the decompression it does.
-		 */
-		tuple = toast_flatten_tuple(tuple, dstate->tupdesc);
-		flattened = true;
+		TupleDesc	desc;
+		Datum	   *attrs;
+		bool	   *isnull;
+
+		desc = RelationGetDescr(relation);
+		attrs = palloc0_array(Datum, desc->natts);
+		isnull = palloc0_array(bool, desc->natts);
+
+		heap_deform_tuple(tuple, desc, attrs, isnull);
+
+		/* First, gather and count the "external indirect" attributes. */
+		for (int i = 0; i < desc->natts; i++)
+		{
+			CompactAttribute *attr = TupleDescCompactAttr(desc, i);
+			varlena    *varlena_pointer;
+
+			if (attr->attisdropped)
+				continue;
+
+			/* not a varlena datatype */
+			if (attr->attlen != -1)
+				continue;
+
+			/* no data */
+			if (isnull[i])
+				continue;
+
+			/* ok, we know we have a toast datum */
+			varlena_pointer = (varlena *) DatumGetPointer(attrs[i]);
+
+			if (!VARATT_IS_EXTERNAL(varlena_pointer))
+				continue;
+
+			if (VARATT_IS_EXTERNAL_INDIRECT(varlena_pointer))
+				attrs_ext = lappend(attrs_ext, varlena_pointer);
+			else
+			{
+				/*
+				 * Logical decoding should not produce "external expanded"
+				 * attributes (those actually should never appear on disk), so
+				 * only TOASTed attribute can be seen here.
+				 */
+				Assert(VARATT_IS_EXTERNAL_ONDISK(varlena_pointer));
+			}
+		}
+		natt_ext = list_length(attrs_ext);
 	}
-	/* Store the tuple size ... */
-	BufFileWrite(dstate->file, &tuple->t_len, sizeof(tuple->t_len));
-	/* ... and the tuple itself. */
-	BufFileWrite(dstate->file, tuple->t_data, tuple->t_len);
+	else
+		natt_ext = 0;
 
-	/* Free the flat copy if created above. */
-	if (flattened)
-		pfree(tuple);
+	/* Write the number of external attributes. */
+	BufFileWrite(file, &natt_ext, sizeof(natt_ext));
+	/* ... and the attributes themselves, if there are some. */
+	foreach_ptr(varlena, attr_val, attrs_ext)
+	{
+		varlena    *ext_val;
+		Size		ext_val_size;
+
+		ext_val = detoast_external_attr(attr_val);
+		ext_val_size = VARSIZE_ANY(ext_val);
+		BufFileWrite(file, ext_val, ext_val_size);
+		pfree(ext_val);
+	}
+	list_free(attrs_ext);
+
+	/* Finally write the tuple size ... */
+	BufFileWrite(file, &tuple->t_len, sizeof(tuple->t_len));
+	/* ... and the tuple itself. */
+	BufFileWrite(file, tuple->t_data, tuple->t_len);
 }
diff --git a/src/include/commands/cluster.h b/src/include/commands/cluster.h
index 1b05d5d418b..a32813e8530 100644
--- a/src/include/commands/cluster.h
+++ b/src/include/commands/cluster.h
@@ -65,11 +65,10 @@ typedef enum
  */
 typedef struct RepackDecodingState
 {
+#ifdef	USE_ASSERT_CHECKING
 	/* The relation whose changes we're decoding. */
 	Oid			relid;
-
-	/* Tuple descriptor of the relation being processed. */
-	TupleDesc	tupdesc;
+#endif
 
 	/* The current output file. */
 	BufFile    *file;
diff --git a/src/test/modules/injection_points/expected/repack_toast.out b/src/test/modules/injection_points/expected/repack_toast.out
index 4f866a74e32..b56dde134f8 100644
--- a/src/test/modules/injection_points/expected/repack_toast.out
+++ b/src/test/modules/injection_points/expected/repack_toast.out
@@ -13,6 +13,7 @@ step change:
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 
 step check2: 
 	INSERT INTO relfilenodes(node)
diff --git a/src/test/modules/injection_points/specs/repack_toast.spec b/src/test/modules/injection_points/specs/repack_toast.spec
index b48abf21450..f84917daca2 100644
--- a/src/test/modules/injection_points/specs/repack_toast.spec
+++ b/src/test/modules/injection_points/specs/repack_toast.spec
@@ -74,10 +74,17 @@ teardown
 
 session s2
 step change
+# Separately test UPDATE of both plain ("i") and TOASTed ("j") attribute. In
+# the first case, the new tuple we get from reorderbuffer.c contains "j" as a
+# TOAST pointer, wich we need to update so it points to the new heap. In the
+# latter case, we receive "j" as "external indirect" value - here we test that
+# the decoding worker writes the tuple to a file correctly and that the
+# backend executing REPACK manages to restore it.
 {
 	UPDATE repack_test SET j=get_long_string() where i=2;
 	DELETE FROM repack_test WHERE i=3;
 	INSERT INTO repack_test(i, j) VALUES (4, get_long_string());
+	UPDATE repack_test SET i=3 where i=1;
 }
 # Check the table from the perspective of s2.
 step check2
-- 
2.47.3


--=-=-=--





^ permalink  raw  reply  [nested|flat] 499+ messages in thread