Fix distributing tables owned by extensions

src/backend/distributed/commands/create_distributed_table.c

@@ -27,6 +27,7 @@
 #include "catalog/pg_attribute.h"
 #include "catalog/pg_enum.h"
 #include "catalog/pg_extension.h"
+#include "catalog/pg_namespace.h"
 #include "catalog/pg_opclass.h"
 #if PG_VERSION_NUM >= 12000
 #include "catalog/pg_proc.h"
@@ -156,7 +157,6 @@ master_create_distributed_table(PG_FUNCTION_ARGS)
 	bool viaDeprecatedAPI = true;
 	ObjectAddress tableAddress = { 0 };
 
-
 	/*
 	 * distributed tables might have dependencies on different objects, since we create
 	 * shards for a distributed table via multiple sessions these objects will be created
@@ -683,6 +683,13 @@ EnsureRelationCanBeDistributed(Oid relationId, Var *distributionColumn,
 	TupleDesc relationDesc = RelationGetDescr(relation);
 	char *relationName = RelationGetRelationName(relation);
 
+	if (relation->rd_rel->relnamespace == PG_CATALOG_NAMESPACE)
+	{
+		ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+						errmsg("cannot distribute catalog tables")));
+	}
+
+
 	if (!RelationUsesHeapAccessMethodOrNone(relation))
 	{
 		ereport(ERROR, (errmsg(

src/backend/distributed/metadata/dependency.c

@@ -631,6 +631,20 @@ SupportedDependencyByCitus(const ObjectAddress *address)
 }
 
 
+/*
+ * IsTableOwnedByExtension returns whether the table with the given relation ID is
+ * owned by an extension.
+ */
+bool
+IsTableOwnedByExtension(Oid relationId)
+{
+	ObjectAddress tableAddress = { 0 };
+	ObjectAddressSet(tableAddress, RelationRelationId, relationId);
+
+	return IsObjectAddressOwnedByExtension(&tableAddress, NULL);
+}
+
+
 /*
  * IsObjectAddressOwnedByExtension returns whether or not the object is owned by an
  * extension. It is assumed that an object having a dependency on an extension is created

src/backend/distributed/metadata/metadata_sync.c

@@ -36,6 +36,7 @@
 #include "distributed/coordinator_protocol.h"
 #include "distributed/metadata_cache.h"
 #include "distributed/metadata_sync.h"
+#include "distributed/metadata/distobject.h"
 #include "distributed/multi_join_order.h"
 #include "distributed/multi_partitioning_utils.h"
 #include "distributed/pg_dist_node.h"
@@ -54,6 +55,7 @@
 #include "utils/syscache.h"
 
 
+static List * GetDistributedTableDDLEvents(Oid relationId);
 static char * LocalGroupIdUpdateCommand(int32 groupId);
 static void UpdateDistNodeBoolAttr(const char *nodeName, int32 nodePort,
 								   int attrNum, bool value);
@@ -384,6 +386,12 @@ MetadataCreateCommands(void)
 		Oid relationId = cacheEntry->relationId;
 		ObjectAddress tableAddress = { 0 };
 
+		if (IsTableOwnedByExtension(relationId))
+		{
+			/* skip table creation when the Citus table is owned by an extension */
+			continue;
+		}
+
 		List *workerSequenceDDLCommands = SequenceDDLCommandsForTable(relationId);
 		List *ddlCommandList = GetTableDDLEvents(relationId, includeSequenceDefaults);
 		char *tableOwnerResetCommand = TableOwnerResetCommand(relationId);
@@ -407,8 +415,16 @@ MetadataCreateCommands(void)
 	/* construct the foreign key constraints after all tables are created */
 	foreach_ptr(cacheEntry, propagatedTableList)
 	{
+		Oid relationId = cacheEntry->relationId;
+
+		if (IsTableOwnedByExtension(relationId))
+		{
+			/* skip foreign key creation when the Citus table is owned by an extension */
+			continue;
+		}
+
 		List *foreignConstraintCommands =
-			GetReferencingForeignConstaintCommands(cacheEntry->relationId);
+			GetReferencingForeignConstaintCommands(relationId);
 
 		metadataSnapshotCommandList = list_concat(metadataSnapshotCommandList,
 												  foreignConstraintCommands);
@@ -417,10 +433,18 @@ MetadataCreateCommands(void)
 	/* construct partitioning hierarchy after all tables are created */
 	foreach_ptr(cacheEntry, propagatedTableList)
 	{
-		if (PartitionTable(cacheEntry->relationId))
+		Oid relationId = cacheEntry->relationId;
+
+		if (IsTableOwnedByExtension(relationId))
+		{
+			/* skip partition creation when the Citus table is owned by an extension */
+			continue;
+		}
+
+		if (PartitionTable(relationId))
 		{
 			char *alterTableAttachPartitionCommands =
-				GenerateAlterTableAttachPartitionCommand(cacheEntry->relationId);
+				GenerateAlterTableAttachPartitionCommand(relationId);
 
 			metadataSnapshotCommandList = lappend(metadataSnapshotCommandList,
 												  alterTableAttachPartitionCommands);
@@ -461,7 +485,7 @@ MetadataCreateCommands(void)
  * sequences, setting the owner of the table, inserting table and shard metadata,
  * setting the truncate trigger and foreign key constraints.
  */
-List *
+static List *
 GetDistributedTableDDLEvents(Oid relationId)
 {
 	CitusTableCacheEntry *cacheEntry = GetCitusTableCacheEntry(relationId);
@@ -469,6 +493,10 @@ GetDistributedTableDDLEvents(Oid relationId)
 	List *commandList = NIL;
 	bool includeSequenceDefaults = true;
 
+	/* if the table is owned by an extension we only propagate pg_dist_* records */
+	bool tableOwnedByExtension = IsTableOwnedByExtension(relationId);
+	if (!tableOwnedByExtension)
+	{
 		/* commands to create sequences */
 		List *sequenceDDLCommands = SequenceDDLCommandsForTable(relationId);
 		commandList = list_concat(commandList, sequenceDDLCommands);
@@ -480,6 +508,7 @@ GetDistributedTableDDLEvents(Oid relationId)
 		/* command to reset the table owner */
 		char *tableOwnerResetCommand = TableOwnerResetCommand(relationId);
 		commandList = lappend(commandList, tableOwnerResetCommand);
+	}
 
 	/* command to insert pg_dist_partition entry */
 	char *metadataCommand = DistributionCreateCommand(cacheEntry);
@@ -494,6 +523,8 @@ GetDistributedTableDDLEvents(Oid relationId)
 	List *shardMetadataInsertCommandList = ShardListInsertCommand(shardIntervalList);
 	commandList = list_concat(commandList, shardMetadataInsertCommandList);
 
+	if (!tableOwnedByExtension)
+	{
 		/* commands to create foreign key constraints */
 		List *foreignConstraintCommands =
 			GetReferencingForeignConstaintCommands(relationId);
@@ -506,6 +537,7 @@ GetDistributedTableDDLEvents(Oid relationId)
 				GenerateAlterTableAttachPartitionCommand(relationId);
 			commandList = lappend(commandList, alterTableAttachPartitionCommands);
 		}
+	}
 
 	return commandList;
 }

src/backend/distributed/worker/worker_drop_protocol.c

@@ -24,6 +24,7 @@
 #include "distributed/metadata_utility.h"
 #include "distributed/coordinator_protocol.h"
 #include "distributed/metadata_cache.h"
+#include "distributed/metadata/distobject.h"
 #include "foreign/foreign.h"
 #include "utils/builtins.h"
 #include "utils/fmgroids.h"
@@ -100,9 +101,15 @@ worker_drop_distributed_table(PG_FUNCTION_ARGS)
 		performMultipleDeletions(objects, DROP_RESTRICT,
 								 PERFORM_DELETION_INTERNAL);
 	}
-	else
+	else if (!IsObjectAddressOwnedByExtension(&distributedTableObject, NULL))
 	{
-		/* drop the table with cascade since other tables may be referring to it */
+		/*
+		 * If the table is owned by an extension, we cannot drop it, nor should we
+		 * until the user runs DROP EXTENSION. Therefore, we skip dropping the
+		 * table and only delete the metadata.
+		 *
+		 * We drop the table with cascade since other tables may be referring to it.
+		 */
 		performDeletion(&distributedTableObject, DROP_CASCADE,
 						PERFORM_DELETION_INTERNAL);
 	}

src/include/distributed/metadata/distobject.h

@@ -22,6 +22,7 @@ extern bool IsObjectDistributed(const ObjectAddress *address);
 extern bool ClusterHasDistributedFunctionWithDistArgument(void);
 extern void MarkObjectDistributed(const ObjectAddress *distAddress);
 extern void UnmarkObjectDistributed(const ObjectAddress *address);
+extern bool IsTableOwnedByExtension(Oid relationId);
 extern bool IsObjectAddressOwnedByExtension(const ObjectAddress *target,
 											ObjectAddress *extensionAddress);
 

src/include/distributed/metadata_sync.h

@@ -32,7 +32,6 @@ extern void StartMetadataSyncToNode(const char *nodeNameString, int32 nodePort);
 extern bool ClusterHasKnownMetadataWorkers(void);
 extern bool ShouldSyncTableMetadata(Oid relationId);
 extern List * MetadataCreateCommands(void);
-extern List * GetDistributedTableDDLEvents(Oid relationId);
 extern List * MetadataDropCommands(void);
 extern char * DistributionCreateCommand(CitusTableCacheEntry *cacheEntry);
 extern char * DistributionDeleteCommand(const char *schemaName,

src/test/regress/expected/multi_mx_ddl.out

@@ -243,3 +243,65 @@ SELECT :worker_1_lastval = :worker_2_lastval;
 -- the type of sequences can't be changed
 ALTER TABLE mx_sequence ALTER value TYPE BIGINT;
 ALTER TABLE mx_sequence ALTER value TYPE INT;
+-- test distributed tables owned by extension
+CREATE TABLE seg_test (x int);
+INSERT INTO seg_test VALUES (42);
+-- pretend this table belongs to an extension
+CREATE EXTENSION seg;
+ALTER EXTENSION seg ADD TABLE seg_test;
+NOTICE:  Citus does not propagate adding/dropping member objects
+HINT:  You can add/drop the member objects on the workers as well.
+\c - - - :worker_1_port
+-- pretend the extension created the table on the worker as well
+CREATE TABLE seg_test (x int);
+ALTER EXTENSION seg ADD TABLE seg_test;
+NOTICE:  Citus does not propagate adding/dropping member objects
+HINT:  You can add/drop the member objects on the workers as well.
+\c - - - :worker_2_port
+-- pretend the extension created the table on the worker as well
+CREATE TABLE seg_test (x int);
+ALTER EXTENSION seg ADD TABLE seg_test;
+NOTICE:  Citus does not propagate adding/dropping member objects
+HINT:  You can add/drop the member objects on the workers as well.
+\c - - - :master_port
+-- sync table metadata, but skip CREATE TABLE
+SET citus.shard_replication_factor TO 1;
+SET citus.shard_count TO 4;
+SET citus.replication_model TO streaming;
+SELECT create_distributed_table('seg_test', 'x');
+NOTICE:  Copying data from local table...
+NOTICE:  copying the data has completed
+DETAIL:  The local data in the table is no longer visible, but is still on disk.
+HINT:  To remove the local data, run: SELECT truncate_local_data_after_distributing_table($$public.seg_test$$)
+ create_distributed_table
+---------------------------------------------------------------------
+
+(1 row)
+
+\c - - - :worker_1_port
+-- should be able to see contents from worker
+SELECT * FROM seg_test;
+ x
+---------------------------------------------------------------------
+ 42
+(1 row)
+
+\c - - - :master_port
+-- test metadata sync in the presence of an extension-owned table
+SELECT start_metadata_sync_to_node('localhost', :worker_1_port);
+ start_metadata_sync_to_node
+---------------------------------------------------------------------
+
+(1 row)
+
+\c - - - :worker_1_port
+-- should be able to see contents from worker
+SELECT * FROM seg_test;
+ x
+---------------------------------------------------------------------
+ 42
+(1 row)
+
+\c - - - :master_port
+-- also drops table on both worker and master
+DROP EXTENSION seg CASCADE;

src/test/regress/sql/multi_mx_ddl.sql

@@ -129,3 +129,50 @@ SELECT :worker_1_lastval = :worker_2_lastval;
 ALTER TABLE mx_sequence ALTER value TYPE BIGINT;
 ALTER TABLE mx_sequence ALTER value TYPE INT;
 
+-- test distributed tables owned by extension
+CREATE TABLE seg_test (x int);
+INSERT INTO seg_test VALUES (42);
+
+-- pretend this table belongs to an extension
+CREATE EXTENSION seg;
+ALTER EXTENSION seg ADD TABLE seg_test;
+
+\c - - - :worker_1_port
+
+-- pretend the extension created the table on the worker as well
+CREATE TABLE seg_test (x int);
+ALTER EXTENSION seg ADD TABLE seg_test;
+
+\c - - - :worker_2_port
+
+-- pretend the extension created the table on the worker as well
+CREATE TABLE seg_test (x int);
+ALTER EXTENSION seg ADD TABLE seg_test;
+
+\c - - - :master_port
+
+-- sync table metadata, but skip CREATE TABLE
+SET citus.shard_replication_factor TO 1;
+SET citus.shard_count TO 4;
+SET citus.replication_model TO streaming;
+SELECT create_distributed_table('seg_test', 'x');
+
+\c - - - :worker_1_port
+
+-- should be able to see contents from worker
+SELECT * FROM seg_test;
+
+\c - - - :master_port
+
+-- test metadata sync in the presence of an extension-owned table
+SELECT start_metadata_sync_to_node('localhost', :worker_1_port);
+
+\c - - - :worker_1_port
+
+-- should be able to see contents from worker
+SELECT * FROM seg_test;
+
+\c - - - :master_port
+
+-- also drops table on both worker and master
+DROP EXTENSION seg CASCADE;