Merge pull request #2673 from citusdata/fix_multishard_transactions

Fix savepoint rollback after multi-shard modify/copy failure.

src/backend/distributed/commands/multi_copy.c

@@ -154,6 +154,8 @@ static void CopySendInt32(CopyOutState outputState, int32 val);
 static void CopySendInt16(CopyOutState outputState, int16 val);
 static void CopyAttributeOutText(CopyOutState outputState, char *string);
 static inline void CopyFlushOutput(CopyOutState outputState, char *start, char *pointer);
+static bool CitusSendTupleToPlacements(TupleTableSlot *slot,
+									   CitusCopyDestReceiver *copyDest);
 
 /* CitusCopyDestReceiver functions */
 static void CitusCopyDestReceiverStartup(DestReceiver *copyDest, int operation,
@@ -2251,8 +2253,37 @@ CitusCopyDestReceiverStartup(DestReceiver *dest, int operation,
 static bool
 CitusCopyDestReceiverReceive(TupleTableSlot *slot, DestReceiver *dest)
 {
+	bool result = false;
 	CitusCopyDestReceiver *copyDest = (CitusCopyDestReceiver *) dest;
 
+	PG_TRY();
+	{
+		result = CitusSendTupleToPlacements(slot, copyDest);
+	}
+	PG_CATCH();
+	{
+		/*
+		 * We might be able to recover from errors with ROLLBACK TO SAVEPOINT,
+		 * so unclaim the connections before throwing errors.
+		 */
+		HTAB *shardConnectionHash = copyDest->shardConnectionHash;
+		UnclaimAllShardConnections(shardConnectionHash);
+
+		PG_RE_THROW();
+	}
+	PG_END_TRY();
+
+	return result;
+}
+
+
+/*
+ * CitusSendTupleToPlacements sends the given TupleTableSlot to the appropriate
+ * shard placement(s).
+ */
+static bool
+CitusSendTupleToPlacements(TupleTableSlot *slot, CitusCopyDestReceiver *copyDest)
+{
 	int partitionColumnIndex = copyDest->partitionColumnIndex;
 	TupleDesc tupleDescriptor = copyDest->tupleDescriptor;
 	CopyStmt *copyStatement = copyDest->copyStatement;
@@ -2409,6 +2440,9 @@ CitusCopyDestReceiverShutdown(DestReceiver *destReceiver)
 	Relation distributedRelation = copyDest->distributedRelation;
 
 	shardConnectionsList = ShardConnectionList(shardConnectionHash);
+
+	PG_TRY();
+	{
 		foreach(shardConnectionsCell, shardConnectionsList)
 		{
 			ShardConnections *shardConnections = (ShardConnections *) lfirst(
@@ -2422,8 +2456,21 @@ CitusCopyDestReceiverShutdown(DestReceiver *destReceiver)
 			}
 
 			/* close the COPY input on all shard placements */
-		EndRemoteCopy(shardConnections->shardId, shardConnections->connectionList, true);
+			EndRemoteCopy(shardConnections->shardId, shardConnections->connectionList,
+						  true);
 		}
+	}
+	PG_CATCH();
+	{
+		/*
+		 * We might be able to recover from errors with ROLLBACK TO SAVEPOINT,
+		 * so unclaim the connections before throwing errors.
+		 */
+		UnclaimAllShardConnections(shardConnectionHash);
+
+		PG_RE_THROW();
+	}
+	PG_END_TRY();
 
 	heap_close(distributedRelation, NoLock);
 }

src/backend/distributed/connection/connection_management.c

@@ -464,7 +464,7 @@ ShutdownConnection(MultiConnection *connection)
 	if (PQstatus(connection->pgConn) == CONNECTION_OK &&
 		PQtransactionStatus(connection->pgConn) == PQTRANS_ACTIVE)
 	{
-		char errorMessage[256] = { 0 };
+		char errorMessage[ERROR_BUFFER_SIZE] = { 0 };
 		PGcancel *cancel = PQgetCancel(connection->pgConn);
 
 		if (!PQcancel(cancel, errorMessage, sizeof(errorMessage)))

src/backend/distributed/connection/remote_commands.c

@@ -29,6 +29,8 @@
 bool LogRemoteCommands = false;
 
 
+static bool ClearResultsInternal(MultiConnection *connection, bool raiseErrors,
+								 bool discardWarnings);
 static bool FinishConnectionIO(MultiConnection *connection, bool raiseInterrupts);
 static WaitEventSet * BuildWaitEventSet(MultiConnection **allConnections,
 										int totalConnectionCount,
@@ -71,7 +73,7 @@ ForgetResults(MultiConnection *connection)
 
 
 /*
- * ClearResults clears a connection from pending activity,
+ * ClearResultsInternal clears a connection from pending activity,
  * returns true if all pending commands return success. It raises
  * error if raiseErrors flag is set, any command fails and transaction
  * is marked critical.
@@ -81,6 +83,27 @@ ForgetResults(MultiConnection *connection)
  */
 bool
 ClearResults(MultiConnection *connection, bool raiseErrors)
+{
+	return ClearResultsInternal(connection, raiseErrors, false);
+}
+
+
+/*
+ * ClearResultsDiscardWarnings does the same thing as ClearResults, but doesn't
+ * emit warnings.
+ */
+bool
+ClearResultsDiscardWarnings(MultiConnection *connection, bool raiseErrors)
+{
+	return ClearResultsInternal(connection, raiseErrors, true);
+}
+
+
+/*
+ * ClearResultsInternal is used by ClearResults and ClearResultsDiscardWarnings.
+ */
+static bool
+ClearResultsInternal(MultiConnection *connection, bool raiseErrors, bool discardWarnings)
 {
 	bool success = true;
 
@@ -102,8 +125,12 @@ ClearResults(MultiConnection *connection, bool raiseErrors)
 		}
 
 		if (!IsResponseOK(result))
+		{
+			if (!discardWarnings)
 			{
 				ReportResultError(connection, result, WARNING);
+			}
+
 			MarkRemoteTransactionFailed(connection, raiseErrors);
 
 			success = false;
@@ -1008,3 +1035,26 @@ BuildWaitEventSet(MultiConnection **allConnections, int totalConnectionCount,
 
 	return waitEventSet;
 }
+
+
+/*
+ * SendCancelationRequest sends a cancelation request on the given connection.
+ * Return value indicates whether the cancelation request was sent successfully.
+ */
+bool
+SendCancelationRequest(MultiConnection *connection)
+{
+	char errorBuffer[ERROR_BUFFER_SIZE] = { 0 };
+	PGcancel *cancelObject = PQgetCancel(connection->pgConn);
+
+	bool cancelSent = PQcancel(cancelObject, errorBuffer, sizeof(errorBuffer));
+	if (!cancelSent)
+	{
+		ereport(WARNING, (errmsg("could not issue cancel request"),
+						  errdetail("Client error: %s", errorBuffer)));
+	}
+
+	PQfreeCancel(cancelObject);
+
+	return cancelSent;
+}

src/backend/distributed/executor/multi_client_executor.c

@@ -443,27 +443,13 @@ bool
 MultiClientCancel(int32 connectionId)
 {
 	MultiConnection *connection = NULL;
-	PGcancel *cancelObject = NULL;
-	int cancelSent = 0;
 	bool canceled = true;
-	char errorBuffer[STRING_BUFFER_SIZE];
 
 	Assert(connectionId != INVALID_CONNECTION_ID);
 	connection = ClientConnectionArray[connectionId];
 	Assert(connection != NULL);
 
-	cancelObject = PQgetCancel(connection->pgConn);
-
-	cancelSent = PQcancel(cancelObject, errorBuffer, sizeof(errorBuffer));
-	if (cancelSent == 0)
-	{
-		ereport(WARNING, (errmsg("could not issue cancel request"),
-						  errdetail("Client error: %s", errorBuffer)));
-
-		canceled = false;
-	}
-
-	PQfreeCancel(cancelObject);
+	canceled = SendCancelationRequest(connection);
 
 	return canceled;
 }

src/backend/distributed/executor/multi_router_executor.c

@@ -1574,6 +1574,7 @@ ExecuteModifyTasks(List *taskList, bool expectResults, ParamListInfo paramListIn
 			queryOK = SendQueryInSingleRowMode(connection, queryString, paramListInfo);
 			if (!queryOK)
 			{
+				UnclaimAllShardConnections(shardConnectionHash);
 				ReportConnectionError(connection, ERROR);
 			}
 		}
@@ -1616,6 +1617,8 @@ ExecuteModifyTasks(List *taskList, bool expectResults, ParamListInfo paramListIn
 				SetCitusNoticeLevel(INFO);
 			}
 
+			PG_TRY();
+			{
 				/*
 				 * If caller is interested, store query results the first time
 				 * through. The output of the query's execution on other shards is
@@ -1634,10 +1637,22 @@ ExecuteModifyTasks(List *taskList, bool expectResults, ParamListInfo paramListIn
 					queryOK = ConsumeQueryResult(connection, alwaysThrowErrorOnFailure,
 												 &currentAffectedTupleCount);
 				}
+			}
+			PG_CATCH();
+			{
+				/*
+				 * We might be able to recover from errors with ROLLBACK TO SAVEPOINT,
+				 * so unclaim the connections before throwing errors.
+				 */
+				UnclaimAllShardConnections(shardConnectionHash);
+				PG_RE_THROW();
+			}
+			PG_END_TRY();
 
 			/* We error out if the worker fails to return a result for the query. */
 			if (!queryOK)
 			{
+				UnclaimAllShardConnections(shardConnectionHash);
 				ReportConnectionError(connection, ERROR);
 			}
 

src/backend/distributed/transaction/remote_transaction.c

@@ -1068,6 +1068,13 @@ CoordinatedRemoteTransactionsSavepointRollback(SubTransactionId subId)
 		MultiConnection *connection = dlist_container(MultiConnection, transactionNode,
 													  iter.cur);
 		RemoteTransaction *transaction = &connection->remoteTransaction;
+
+		/* cancel any ongoing queries before issuing rollback */
+		SendCancelationRequest(connection);
+
+		/* clear results, but don't show cancelation warning messages from workers. */
+		ClearResultsDiscardWarnings(connection, raiseInterrupts);
+
 		if (transaction->transactionFailed)
 		{
 			if (transaction->lastSuccessfulSubXact <= subId)

src/include/distributed/connection_management.h

@@ -21,6 +21,9 @@
 /* maximum (textual) lengths of hostname and port */
 #define MAX_NODE_LENGTH 255 /* includes 0 byte */
 
+/* used for libpq commands that get an error buffer. Postgres docs recommend 256. */
+#define ERROR_BUFFER_SIZE 256
+
 /* default notice level */
 #define DEFAULT_CITUS_NOTICE_LEVEL DEBUG1
 

src/include/distributed/remote_commands.h

@@ -27,6 +27,7 @@ extern bool LogRemoteCommands;
 extern bool IsResponseOK(struct pg_result *result);
 extern void ForgetResults(MultiConnection *connection);
 extern bool ClearResults(MultiConnection *connection, bool raiseErrors);
+extern bool ClearResultsDiscardWarnings(MultiConnection *connection, bool raiseErrors);
 extern bool ClearResultsIfReady(MultiConnection *connection);
 extern bool SqlStateMatchesCategory(char *sqlStateString, int category);
 
@@ -57,5 +58,6 @@ extern bool PutRemoteCopyEnd(MultiConnection *connection, const char *errormsg);
 /* waiting for multiple command results */
 extern void WaitForAllConnections(List *connectionList, bool raiseInterrupts);
 
+extern bool SendCancelationRequest(MultiConnection *connection);
 
 #endif /* REMOTE_COMMAND_H */

src/test/regress/expected/failure_savepoints.out

@@ -38,6 +38,8 @@ CONTEXT:  while executing command on localhost:9060
 WARNING:  connection not open
 CONTEXT:  while executing command on localhost:9060
 DELETE FROM artists WHERE id=4;
+WARNING:  could not issue cancel request
+DETAIL:  Client error: PQcancel() -- no cancel object supplied
 WARNING:  connection not open
 CONTEXT:  while executing command on localhost:9060
 WARNING:  connection error: localhost:9060
@@ -195,6 +197,7 @@ INSERT INTO artists VALUES (7, 'Emily Carr');
 ROLLBACK TO SAVEPOINT s1;
 WARNING:  connection not open
 WARNING:  connection not open
+WARNING:  could not issue cancel request
 COMMIT;
 ERROR:  could not make changes to shard 100950 on any node
 SELECT * FROM artists WHERE id=6;
@@ -230,6 +233,7 @@ WARNING:  connection not open
 WARNING:  connection not open
 INSERT INTO researchers VALUES (8, 4, 'Alonzo Church');
 ROLLBACK TO s1;
+WARNING:  could not issue cancel request
 WARNING:  connection not open
 WARNING:  connection error: localhost:9060
 WARNING:  connection not open

src/test/regress/expected/multi_subtransactions.out

@@ -129,6 +129,140 @@ ERROR:  current transaction is aborted, commands ignored until end of transactio
 ROLLBACK TO SAVEPOINT s3;
 ERROR:  savepoint "s3" does not exist
 COMMIT;
+-- Recover from multi-shard modify errors
+BEGIN;
+INSERT INTO artists VALUES (8, 'Sogand');
+SAVEPOINT s1;
+UPDATE artists SET name = NULL;
+ERROR:  null value in column "name" violates not-null constraint
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (9, 'Mohsen Namjoo');
+COMMIT;
+SELECT * FROM artists WHERE id IN (7, 8, 9) ORDER BY id;
+ id |     name      
+----+---------------
+  8 | Sogand
+  9 | Mohsen Namjoo
+(2 rows)
+
+-- Recover from multi-shard copy shutdown failure.
+-- Constraint check for non-partition columns happen only at copy shutdown.
+BEGIN;
+DELETE FROM artists;
+SAVEPOINT s1;
+INSERT INTO artists SELECT i, NULL FROM generate_series(1, 5) i;
+ERROR:  null value in column "name" violates not-null constraint
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (10, 'Mahmoud Farshchian');
+COMMIT;
+SELECT * FROM artists WHERE id IN (9, 10) ORDER BY id;
+ id |        name        
+----+--------------------
+ 10 | Mahmoud Farshchian
+(1 row)
+
+-- Recover from multi-shard copy send failure.
+-- Constraint check for partition column happens at copy send.
+BEGIN;
+DELETE FROM artists;
+SAVEPOINT s1;
+INSERT INTO artists SELECT NULL, NULL FROM generate_series(1, 5) i;
+ERROR:  the partition column of table public.artists cannot be NULL
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (11, 'Egon Schiele');
+COMMIT;
+SELECT * FROM artists WHERE id IN (10, 11) ORDER BY id;
+ id |     name     
+----+--------------
+ 11 | Egon Schiele
+(1 row)
+
+-- Recover from multi-shard copy startup failure.
+-- Check for existence of a value for partition columnn happens at copy startup.
+BEGIN;
+DELETE FROM artists;
+SAVEPOINT s1;
+INSERT INTO artists(name) SELECT 'a' FROM generate_series(1, 5) i;
+ERROR:  the partition column of table public.artists should have a value
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (12, 'Marc Chagall');
+COMMIT;
+SELECT * FROM artists WHERE id IN (11, 12) ORDER BY id;
+ id |     name     
+----+--------------
+ 12 | Marc Chagall
+(1 row)
+
+-- Recover from multi-shard CTE modify failures
+create table t1(a int, b int);
+create table t2(a int, b int CHECK(b > 0));
+ALTER SEQUENCE pg_catalog.pg_dist_shardid_seq RESTART 1190000;
+select create_distributed_table('t1', 'a'),
+       create_distributed_table('t2', 'a');
+ create_distributed_table | create_distributed_table 
+--------------------------+--------------------------
+                          | 
+(1 row)
+
+begin;
+insert into t2 select i, i+1 from generate_series(1, 3) i;
+with r AS (
+    update t2 set b = b + 1
+    returning *
+) insert into t1 select * from r;
+savepoint s1;
+with r AS (
+    update t1 set b = b - 10
+    returning *
+) insert into t2 select * from r;
+ERROR:  new row for relation "t2_1190005" violates check constraint "t2_b_check"
+rollback to savepoint s1;
+savepoint s2;
+with r AS (
+    update t2 set b = b - 10
+    returning *
+) insert into t1 select * from r;
+ERROR:  new row for relation "t2_1190004" violates check constraint "t2_b_check"
+rollback to savepoint s2;
+savepoint s3;
+with r AS (
+    insert into t2 select i, i+1 from generate_series(-10,-5) i
+    returning *
+) insert into t1 select * from r;
+ERROR:  new row for relation "t2_1190004" violates check constraint "t2_b_check"
+rollback to savepoint s3;
+savepoint s4;
+with r AS (
+    insert into t1 select i, i+1 from generate_series(-10,-5) i
+    returning *
+) insert into t2 select * from r;
+ERROR:  new row for relation "t2_1190005" violates check constraint "t2_b_check"
+rollback to savepoint s4;
+with r AS (
+    update t2 set b = b + 1
+    returning *
+) insert into t1 select * from r;
+commit;
+select * from t2 order by a, b;
+ a | b 
+---+---
+ 1 | 4
+ 2 | 5
+ 3 | 6
+(3 rows)
+
+select * from t1 order by a, b;
+ a | b 
+---+---
+ 1 | 3
+ 1 | 4
+ 2 | 4
+ 2 | 5
+ 3 | 5
+ 3 | 6
+(6 rows)
+
+drop table t1, t2;
 -- ===================================================================
 -- Tests for replication factor > 1
 -- ===================================================================

src/test/regress/expected/multi_subtransactions_0.out

@@ -129,6 +129,140 @@ ERROR:  current transaction is aborted, commands ignored until end of transactio
 ROLLBACK TO SAVEPOINT s3;
 ERROR:  no such savepoint
 COMMIT;
+-- Recover from multi-shard modify errors
+BEGIN;
+INSERT INTO artists VALUES (8, 'Sogand');
+SAVEPOINT s1;
+UPDATE artists SET name = NULL;
+ERROR:  null value in column "name" violates not-null constraint
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (9, 'Mohsen Namjoo');
+COMMIT;
+SELECT * FROM artists WHERE id IN (7, 8, 9) ORDER BY id;
+ id |     name      
+----+---------------
+  8 | Sogand
+  9 | Mohsen Namjoo
+(2 rows)
+
+-- Recover from multi-shard copy shutdown failure.
+-- Constraint check for non-partition columns happen only at copy shutdown.
+BEGIN;
+DELETE FROM artists;
+SAVEPOINT s1;
+INSERT INTO artists SELECT i, NULL FROM generate_series(1, 5) i;
+ERROR:  null value in column "name" violates not-null constraint
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (10, 'Mahmoud Farshchian');
+COMMIT;
+SELECT * FROM artists WHERE id IN (9, 10) ORDER BY id;
+ id |        name        
+----+--------------------
+ 10 | Mahmoud Farshchian
+(1 row)
+
+-- Recover from multi-shard copy send failure.
+-- Constraint check for partition column happens at copy send.
+BEGIN;
+DELETE FROM artists;
+SAVEPOINT s1;
+INSERT INTO artists SELECT NULL, NULL FROM generate_series(1, 5) i;
+ERROR:  the partition column of table public.artists cannot be NULL
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (11, 'Egon Schiele');
+COMMIT;
+SELECT * FROM artists WHERE id IN (10, 11) ORDER BY id;
+ id |     name     
+----+--------------
+ 11 | Egon Schiele
+(1 row)
+
+-- Recover from multi-shard copy startup failure.
+-- Check for existence of a value for partition columnn happens at copy startup.
+BEGIN;
+DELETE FROM artists;
+SAVEPOINT s1;
+INSERT INTO artists(name) SELECT 'a' FROM generate_series(1, 5) i;
+ERROR:  the partition column of table public.artists should have a value
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (12, 'Marc Chagall');
+COMMIT;
+SELECT * FROM artists WHERE id IN (11, 12) ORDER BY id;
+ id |     name     
+----+--------------
+ 12 | Marc Chagall
+(1 row)
+
+-- Recover from multi-shard CTE modify failures
+create table t1(a int, b int);
+create table t2(a int, b int CHECK(b > 0));
+ALTER SEQUENCE pg_catalog.pg_dist_shardid_seq RESTART 1190000;
+select create_distributed_table('t1', 'a'),
+       create_distributed_table('t2', 'a');
+ create_distributed_table | create_distributed_table 
+--------------------------+--------------------------
+                          | 
+(1 row)
+
+begin;
+insert into t2 select i, i+1 from generate_series(1, 3) i;
+with r AS (
+    update t2 set b = b + 1
+    returning *
+) insert into t1 select * from r;
+savepoint s1;
+with r AS (
+    update t1 set b = b - 10
+    returning *
+) insert into t2 select * from r;
+ERROR:  new row for relation "t2_1190005" violates check constraint "t2_b_check"
+rollback to savepoint s1;
+savepoint s2;
+with r AS (
+    update t2 set b = b - 10
+    returning *
+) insert into t1 select * from r;
+ERROR:  new row for relation "t2_1190004" violates check constraint "t2_b_check"
+rollback to savepoint s2;
+savepoint s3;
+with r AS (
+    insert into t2 select i, i+1 from generate_series(-10,-5) i
+    returning *
+) insert into t1 select * from r;
+ERROR:  new row for relation "t2_1190004" violates check constraint "t2_b_check"
+rollback to savepoint s3;
+savepoint s4;
+with r AS (
+    insert into t1 select i, i+1 from generate_series(-10,-5) i
+    returning *
+) insert into t2 select * from r;
+ERROR:  new row for relation "t2_1190005" violates check constraint "t2_b_check"
+rollback to savepoint s4;
+with r AS (
+    update t2 set b = b + 1
+    returning *
+) insert into t1 select * from r;
+commit;
+select * from t2 order by a, b;
+ a | b 
+---+---
+ 1 | 4
+ 2 | 5
+ 3 | 6
+(3 rows)
+
+select * from t1 order by a, b;
+ a | b 
+---+---
+ 1 | 3
+ 1 | 4
+ 2 | 4
+ 2 | 5
+ 3 | 5
+ 3 | 6
+(6 rows)
+
+drop table t1, t2;
 -- ===================================================================
 -- Tests for replication factor > 1
 -- ===================================================================

src/test/regress/sql/multi_subtransactions.sql

@@ -105,6 +105,103 @@ SAVEPOINT s3;
 ROLLBACK TO SAVEPOINT s3;
 COMMIT;
 
+-- Recover from multi-shard modify errors
+BEGIN;
+INSERT INTO artists VALUES (8, 'Sogand');
+SAVEPOINT s1;
+UPDATE artists SET name = NULL;
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (9, 'Mohsen Namjoo');
+COMMIT;
+
+SELECT * FROM artists WHERE id IN (7, 8, 9) ORDER BY id;
+
+-- Recover from multi-shard copy shutdown failure.
+-- Constraint check for non-partition columns happen only at copy shutdown.
+BEGIN;
+DELETE FROM artists;
+SAVEPOINT s1;
+INSERT INTO artists SELECT i, NULL FROM generate_series(1, 5) i;
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (10, 'Mahmoud Farshchian');
+COMMIT;
+
+SELECT * FROM artists WHERE id IN (9, 10) ORDER BY id;
+
+-- Recover from multi-shard copy send failure.
+-- Constraint check for partition column happens at copy send.
+BEGIN;
+DELETE FROM artists;
+SAVEPOINT s1;
+INSERT INTO artists SELECT NULL, NULL FROM generate_series(1, 5) i;
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (11, 'Egon Schiele');
+COMMIT;
+
+SELECT * FROM artists WHERE id IN (10, 11) ORDER BY id;
+
+-- Recover from multi-shard copy startup failure.
+-- Check for existence of a value for partition columnn happens at copy startup.
+BEGIN;
+DELETE FROM artists;
+SAVEPOINT s1;
+INSERT INTO artists(name) SELECT 'a' FROM generate_series(1, 5) i;
+ROLLBACK TO s1;
+INSERT INTO artists VALUES (12, 'Marc Chagall');
+COMMIT;
+
+SELECT * FROM artists WHERE id IN (11, 12) ORDER BY id;
+
+-- Recover from multi-shard CTE modify failures
+create table t1(a int, b int);
+create table t2(a int, b int CHECK(b > 0));
+
+ALTER SEQUENCE pg_catalog.pg_dist_shardid_seq RESTART 1190000;
+
+select create_distributed_table('t1', 'a'),
+       create_distributed_table('t2', 'a');
+
+begin;
+insert into t2 select i, i+1 from generate_series(1, 3) i;
+with r AS (
+    update t2 set b = b + 1
+    returning *
+) insert into t1 select * from r;
+savepoint s1;
+with r AS (
+    update t1 set b = b - 10
+    returning *
+) insert into t2 select * from r;
+rollback to savepoint s1;
+savepoint s2;
+with r AS (
+    update t2 set b = b - 10
+    returning *
+) insert into t1 select * from r;
+rollback to savepoint s2;
+savepoint s3;
+with r AS (
+    insert into t2 select i, i+1 from generate_series(-10,-5) i
+    returning *
+) insert into t1 select * from r;
+rollback to savepoint s3;
+savepoint s4;
+with r AS (
+    insert into t1 select i, i+1 from generate_series(-10,-5) i
+    returning *
+) insert into t2 select * from r;
+rollback to savepoint s4;
+with r AS (
+    update t2 set b = b + 1
+    returning *
+) insert into t1 select * from r;
+commit;
+
+select * from t2 order by a, b;
+select * from t1 order by a, b;
+
+drop table t1, t2;
+
 -- ===================================================================
 -- Tests for replication factor > 1
 -- ===================================================================
@@ -216,3 +313,4 @@ SELECT * FROM researchers WHERE lab_id=10;
 -- Clean-up
 DROP TABLE artists;
 DROP TABLE researchers;
+