pg_upgrade: Use COPY for large object metadata.

Reported-by: Hannu Krosing <hannuk@google.com>
Suggested-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAMT0RQSS-6qLH%2BzYsOeUbAYhop3wmQTkNmQpo5--QRDUR%2BqYmQ%40mail.gmail.com

Author: nathan-bossart

src/bin/pg_dump/pg_backup_archiver.c

@@ -31,6 +31,8 @@
 #endif
 
 #include "catalog/pg_class_d.h"
+#include "catalog/pg_largeobject_metadata.h"
+#include "catalog/pg_shdepend.h"
 #include "common/string.h"
 #include "compress_io.h"
 #include "dumputils.h"
@@ -2974,6 +2976,17 @@ _tocEntryRequired(TocEntry *te, teSection curSection, ArchiveHandle *AH)
 	int			res = REQ_SCHEMA | REQ_DATA;
 	RestoreOptions *ropt = AH->public.ropt;
 
+	/*
+	 * For binary upgrade mode, dump pg_largeobject_metadata and the
+	 * associated pg_shdepend rows. This is faster to restore than the
+	 * equivalent set of large object commands.
+	 */
+	if (ropt->binary_upgrade && AH->public.remoteVersion >= 120000 &&
+		strcmp(te->desc, "TABLE DATA") == 0 &&
+		(te->catalogId.oid == LargeObjectMetadataRelationId ||
+		 te->catalogId.oid == SharedDependRelationId))
+		return REQ_DATA;
+
 	/* These items are treated specially */
 	if (strcmp(te->desc, "ENCODING") == 0 ||
 		strcmp(te->desc, "STDSTRINGS") == 0 ||

src/bin/pg_dump/pg_dump.c

@@ -49,8 +49,10 @@
 #include "catalog/pg_class_d.h"
 #include "catalog/pg_default_acl_d.h"
 #include "catalog/pg_largeobject_d.h"
+#include "catalog/pg_largeobject_metadata_d.h"
 #include "catalog/pg_proc_d.h"
 #include "catalog/pg_publication_d.h"
+#include "catalog/pg_shdepend.h"
 #include "catalog/pg_subscription_d.h"
 #include "catalog/pg_type_d.h"
 #include "common/connect.h"
@@ -209,6 +211,12 @@ static int	nbinaryUpgradeClassOids = 0;
 static SequenceItem *sequences = NULL;
 static int	nsequences = 0;
 
+/*
+ * For binary upgrade, the dump ID of pg_largeobject_metadata is saved for use
+ * as a dependency for pg_shdepend and any large object comments/seclabels.
+ */
+static DumpId lo_metadata_dumpId;
+
 /* Maximum number of relations to fetch in a fetchAttributeStats() call. */
 #define MAX_ATTR_STATS_RELS 64
 
@@ -1083,6 +1091,34 @@ main(int argc, char **argv)
 	if (!dopt.dumpData && dopt.sequence_data)
 		getTableData(&dopt, tblinfo, numTables, RELKIND_SEQUENCE);
 
+	/*
+	 * For binary upgrade mode, dump pg_largeobject_metadata and the
+	 * associated pg_shdepend rows. This is faster to restore than the
+	 * equivalent set of large object commands.
+	 */
+	if (dopt.binary_upgrade && fout->remoteVersion >= 120000)
+	{
+		TableInfo  *lo_metadata = findTableByOid(LargeObjectMetadataRelationId);
+		TableInfo  *shdepend = findTableByOid(SharedDependRelationId);
+
+		makeTableDataInfo(&dopt, lo_metadata);
+		makeTableDataInfo(&dopt, shdepend);
+
+		/*
+		 * Save pg_largeobject_metadata's dump ID for use as a dependency on
+		 * pg_shdepend and any large object comments/seclabels.
+		 */
+		lo_metadata_dumpId = lo_metadata->dataObj->dobj.dumpId;
+		addObjectDependency(&shdepend->dataObj->dobj, lo_metadata_dumpId);
+
+		/*
+		 * Only dump large object shdepend rows for this database.
+		 */
+		shdepend->dataObj->filtercond = "WHERE classid = 'pg_largeobject'::regclass "
+			"AND dbid = (SELECT oid FROM pg_database "
+			"            WHERE datname = current_database())";
+	}
+
 	/*
 	 * In binary-upgrade mode, we do not have to worry about the actual LO
 	 * data or the associated metadata that resides in the pg_largeobject and
@@ -3922,10 +3958,29 @@ getLOs(Archive *fout)
 		 * as it will be copied by pg_upgrade, which simply copies the
 		 * pg_largeobject table. We *do* however dump out anything but the
 		 * data, as pg_upgrade copies just pg_largeobject, but not
-		 * pg_largeobject_metadata, after the dump is restored.
+		 * pg_largeobject_metadata, after the dump is restored.  In versions
+		 * before v12, this is done via proper large object commands.  In
+		 * newer versions, we dump the content of pg_largeobject_metadata and
+		 * any associated pg_shdepend rows, which is faster to restore.
 		 */
 		if (dopt->binary_upgrade)
-			loinfo->dobj.dump &= ~DUMP_COMPONENT_DATA;
+		{
+			if (fout->remoteVersion >= 120000)
+			{
+				loinfo->dobj.dump &= ~(DUMP_COMPONENT_DATA | DUMP_COMPONENT_ACL | DUMP_COMPONENT_DEFINITION);
+
+				/*
+				 * Mark the large object as dependent on
+				 * pg_largeobject_metadata so that any large object
+				 * comments/seclables are dumped after it.
+				 */
+				loinfo->dobj.dependencies = (DumpId *) pg_malloc(sizeof(DumpId));
+				loinfo->dobj.dependencies[0] = lo_metadata_dumpId;
+				loinfo->dobj.nDeps = loinfo->dobj.allocDeps = 1;
+			}
+			else
+				loinfo->dobj.dump &= ~DUMP_COMPONENT_DATA;
+		}
 
 		/*
 		 * Create a "BLOBS" data item for the group, too. This is just a
@@ -9034,8 +9089,18 @@ getTableAttrs(Archive *fout, TableInfo *tblinfo, int numTables)
 		if (tbinfo->relkind == RELKIND_SEQUENCE)
 			continue;
 
-		/* Don't bother with uninteresting tables, either */
-		if (!tbinfo->interesting)
+		/*
+		 * Don't bother with uninteresting tables, either.  For binary
+		 * upgrades, this is bypassed for pg_largeobject_metadata and
+		 * pg_shdepend so that the columns names are collected for the
+		 * corresponding COPY commands.  Restoring the data for those catalogs
+		 * is faster than restoring the equivalent set of large object
+		 * commands.
+		 */
+		if (!tbinfo->interesting &&
+			!(fout->dopt->binary_upgrade && fout->remoteVersion >= 120000 &&
+			  (tbinfo->dobj.catId.oid == LargeObjectMetadataRelationId ||
+			   tbinfo->dobj.catId.oid == SharedDependRelationId)))
 			continue;
 
 		/* OK, we need info for this table */
@@ -9232,7 +9297,10 @@ getTableAttrs(Archive *fout, TableInfo *tblinfo, int numTables)
 			pg_fatal("unrecognized table OID %u", attrelid);
 		/* cross-check that we only got requested tables */
 		if (tbinfo->relkind == RELKIND_SEQUENCE ||
-			!tbinfo->interesting)
+			(!tbinfo->interesting &&
+			 !(fout->dopt->binary_upgrade && fout->remoteVersion >= 120000 &&
+			   (tbinfo->dobj.catId.oid == LargeObjectMetadataRelationId ||
+				tbinfo->dobj.catId.oid == SharedDependRelationId))))
 			pg_fatal("unexpected column data for table \"%s\"",
 					 tbinfo->dobj.name);
 

src/bin/pg_dump/pg_dump_sort.c

@@ -76,10 +76,10 @@ enum dbObjectTypePriorities
 	PRIO_TABLE_ATTACH,
 	PRIO_DUMMY_TYPE,
 	PRIO_ATTRDEF,
-	PRIO_LARGE_OBJECT,
 	PRIO_PRE_DATA_BOUNDARY,		/* boundary! */
 	PRIO_TABLE_DATA,
 	PRIO_SEQUENCE_SET,
+	PRIO_LARGE_OBJECT,
 	PRIO_LARGE_OBJECT_DATA,
 	PRIO_STATISTICS_DATA_DATA,
 	PRIO_POST_DATA_BOUNDARY,	/* boundary! */

src/bin/pg_dump/t/002_pg_dump.pl

@@ -1029,6 +1029,7 @@
 			test_schema_plus_large_objects => 1,
 		},
 		unlike => {
+			binary_upgrade => 1,
 			no_large_objects => 1,
 			no_owner => 1,
 			schema_only => 1,
@@ -1517,6 +1518,7 @@
 			test_schema_plus_large_objects => 1,
 		},
 		unlike => {
+			binary_upgrade => 1,
 			schema_only => 1,
 			schema_only_with_statistics => 1,
 			no_large_objects => 1,
@@ -4524,9 +4526,9 @@
 			no_schema => 1,
 			section_data => 1,
 			test_schema_plus_large_objects => 1,
-			binary_upgrade => 1,
 		},
 		unlike => {
+			binary_upgrade => 1,
 			no_large_objects => 1,
 			no_privs => 1,
 			schema_only => 1,