Send down SYNC message to a failed remote session that was running extended

query protocol.

While running extended query protocol, a backend that has thrown an error will
keep ignoring all messages until it sees a SYNC message. We now carefully track
the messages that we are sending to the remote node and remember if we must
send a SYNC message even before sending a ROLLBACK command.

While the regression was running fine even without this patch, this issue was
noticed as part of some other work and hence fixed

3 files changed, 43 insertions, 0 deletions

diff --git a/src/backend/pgxc/pool/execRemote.c b/src/backend/pgxc/pool/execRemote.c
index eb3807c70e..ab0a96986c 100644
--- a/src/backend/pgxc/pool/execRemote.c
+++ b/src/backend/pgxc/pool/execRemote.c
@@ -1718,6 +1718,15 @@ handle_response(PGXCNodeHandle *conn, ResponseCombiner *combiner)
 			case 'E':			/* ErrorResponse */
 				HandleError(combiner, msg, msg_len, conn);
 				add_error_message(conn, combiner->errorMessage);
+				/*
+				 * In case the remote node was running an extended query
+				 * protocol and reported an error, it will keep ignoring all
+				 * subsequent commands until it sees a SYNC message. So make
+				 * sure that we send down SYNC even before sending a ROLLBACK
+				 * command
+				 */
+				if (conn->in_extended_query)
+					conn->needSync = true;
 				return RESPONSE_ERROR;
 			case 'A':			/* NotificationResponse */
 			case 'N':			/* NoticeResponse */
@@ -2638,6 +2647,16 @@ pgxc_node_remote_abort(void)
 				BufferConnection(conn);
 
 			/*
+			 * If the remote session was running extended query protocol when
+			 * it failed, it will expect a SYNC message before it accepts any
+			 * other command
+			 */
+			if (conn->needSync)
+			{
+				pgxc_node_send_sync(conn);
+				pgxc_node_receive(1, &conn, NULL);
+			}
+			/*
 			 * Do not matter, is there committed or failed transaction,
 			 * just send down rollback to finish it.
 			 */
@@ -2664,6 +2683,12 @@ pgxc_node_remote_abort(void)
 
 		if (conn->transaction_status != 'I')
 		{
+			/* Send SYNC if the remote session is expecting one */
+			if (conn->needSync)
+			{
+				pgxc_node_send_sync(conn);
+				pgxc_node_receive(1, &conn, NULL);
+			}
 			/*
 			 * Do not matter, is there committed or failed transaction,
 			 * just send down rollback to finish it.
@@ -2902,6 +2927,8 @@ DataNodeCopyIn(char *data_row, int len, int conn_count, PGXCNodeHandle** copy_co
 			memcpy(handle->outBuffer + handle->outEnd, data_row, len);
 			handle->outEnd += len;
 			handle->outBuffer[handle->outEnd++] = '\n';
+
+			handle->in_extended_query = false;
 		}
 		else
 		{
@@ -3045,6 +3072,7 @@ DataNodeCopyEnd(PGXCNodeHandle *handle, bool is_error)
 	memcpy(handle->outBuffer + handle->outEnd, &nLen, 4);
 	handle->outEnd += 4;
 
+	handle->in_extended_query = false;
 	/* We need response right away, so send immediately */
 	if (pgxc_node_flush(handle) < 0)
 		return true;
diff --git a/src/backend/pgxc/pool/pgxcnode.c b/src/backend/pgxc/pool/pgxcnode.c
index 87a2f6167f..579cb4b30a 100644
--- a/src/backend/pgxc/pool/pgxcnode.c
+++ b/src/backend/pgxc/pool/pgxcnode.c
@@ -143,6 +143,7 @@ init_pgxc_handle(PGXCNodeHandle *pgxc_handle)
 	pgxc_handle->inEnd = 0;
 	pgxc_handle->inCursor = 0;
 	pgxc_handle->outEnd = 0;
+	pgxc_handle->needSync = false;
 
 	if (pgxc_handle->outBuffer == NULL || pgxc_handle->inBuffer == NULL)
 	{
@@ -415,6 +416,7 @@ pgxc_node_init(PGXCNodeHandle *handle, int sock, bool global_session, int pid)
 	handle->inStart = 0;
 	handle->inEnd = 0;
 	handle->inCursor = 0;
+	handle->needSync = false;
 	/*
 	 * We got a new connection, set on the remote node the session parameters
 	 * if defined. The transaction parameter should be sent after BEGIN
@@ -1325,6 +1327,7 @@ pgxc_node_send_plan(PGXCNodeHandle * handle, const char *statement,
 	}
 	pfree(paramTypes);
 
+	handle->in_extended_query = true;
  	return 0;
 }
 
@@ -1405,6 +1408,7 @@ pgxc_node_send_bind(PGXCNodeHandle * handle, const char *portal,
 	handle->outBuffer[handle->outEnd++] = 0;
 	handle->outBuffer[handle->outEnd++] = 0;
 
+	handle->in_extended_query = true;
  	return 0;
 }
 
@@ -1452,6 +1456,7 @@ pgxc_node_send_describe(PGXCNodeHandle * handle, bool is_statement,
 	else
 		handle->outBuffer[handle->outEnd++] = '\0';
 
+	handle->in_extended_query = true;
  	return 0;
 }
 
@@ -1492,6 +1497,7 @@ pgxc_node_send_close(PGXCNodeHandle * handle, bool is_statement,
 	else
 		handle->outBuffer[handle->outEnd++] = '\0';
 
+	handle->in_extended_query = true;
  	return 0;
 }
 
@@ -1535,6 +1541,7 @@ pgxc_node_send_execute(PGXCNodeHandle * handle, const char *portal, int fetch)
 
 	PGXCNodeSetConnectionState(handle, DN_CONNECTION_STATE_QUERY);
 
+	handle->in_extended_query = true;
 	return 0;
 }
 
@@ -1561,6 +1568,7 @@ pgxc_node_send_flush(PGXCNodeHandle * handle)
 	memcpy(handle->outBuffer + handle->outEnd, &msgLen, 4);
 	handle->outEnd += 4;
 
+	handle->in_extended_query = true;
 	return pgxc_node_flush(handle);
 }
 
@@ -1587,6 +1595,9 @@ pgxc_node_send_sync(PGXCNodeHandle * handle)
 	memcpy(handle->outBuffer + handle->outEnd, &msgLen, 4);
 	handle->outEnd += 4;
 
+	handle->in_extended_query = false;
+	handle->needSync = false;
+
 	return pgxc_node_flush(handle);
 }
 
@@ -1721,6 +1732,7 @@ pgxc_node_send_query_internal(PGXCNodeHandle * handle, const char *query,
 
 	PGXCNodeSetConnectionState(handle, DN_CONNECTION_STATE_QUERY);
 
+	handle->in_extended_query = false;
  	return pgxc_node_flush(handle);
 }
 
diff --git a/src/include/pgxc/pgxcnode.h b/src/include/pgxc/pgxcnode.h
index 64cf432ee8..93cc925c70 100644
--- a/src/include/pgxc/pgxcnode.h
+++ b/src/include/pgxc/pgxcnode.h
@@ -92,6 +92,9 @@ struct pgxc_node_handle
 	 * For details see comments of RESP_ROLLBACK
 	 */
 	bool		ck_resp_rollback;
+
+	bool		in_extended_query;
+	bool		needSync;
 };
 typedef struct pgxc_node_handle PGXCNodeHandle;