Log message of across shard queries according to the log level

src/backend/distributed/executor/multi_server_executor.c

@@ -29,6 +29,7 @@
 int RemoteTaskCheckInterval = 100; /* per cycle sleep interval in millisecs */
 int TaskExecutorType = MULTI_EXECUTOR_REAL_TIME; /* distributed executor type */
 bool BinaryMasterCopyFormat = false; /* copy data from workers in binary format */
+int MultiTaskQueryLogLevel = MULTI_TASK_QUERY_INFO_OFF; /* multi-task query log level */
 
 
 /*
@@ -58,6 +59,15 @@ JobExecutorType(MultiPlan *multiPlan)
 		return MULTI_EXECUTOR_ROUTER;
 	}
 
+	/* if it is not a router executable plan, inform user according to the log level */
+	if (MultiTaskQueryLogLevel != MULTI_TASK_QUERY_INFO_OFF)
+	{
+		ereport(MultiTaskQueryLogLevel, (errmsg("multi-task query about to be executed"),
+										 errhint("Queries are split to multiple tasks "
+												 "if they have to be split into several"
+												 " queries on the workers.")));
+	}
+
 	if (executorType == MULTI_EXECUTOR_REAL_TIME)
 	{
 		double reasonableConnectionCount = 0;

src/backend/distributed/planner/multi_router_planner.c

@@ -34,6 +34,7 @@
 #include "distributed/multi_logical_optimizer.h"
 #include "distributed/multi_physical_planner.h"
 #include "distributed/multi_router_planner.h"
+#include "distributed/multi_server_executor.h"
 #include "distributed/listutils.h"
 #include "distributed/citus_ruleutils.h"
 #include "distributed/relation_restriction_equivalence.h"
@@ -343,6 +344,15 @@ CreateInsertSelectRouterPlan(Query *originalQuery,
 		++taskIdIndex;
 	}
 
+	if (MultiTaskQueryLogLevel != MULTI_TASK_QUERY_INFO_OFF &&
+		list_length(sqlTaskList) > 1)
+	{
+		ereport(MultiTaskQueryLogLevel, (errmsg("multi-task query about to be executed"),
+										 errhint("Queries are split to multiple tasks "
+												 "if they have to be split into several"
+												 " queries on the workers.")));
+	}
+
 	/* Create the worker job */
 	workerJob = CitusMakeNode(Job);
 	workerJob->taskList = sqlTaskList;

src/backend/distributed/shared_library_init.c

@@ -92,6 +92,16 @@ static const struct config_enum_entry multi_shard_commit_protocol_options[] = {
 	{ NULL, 0, false }
 };
 
+static const struct config_enum_entry multi_task_query_log_level_options[] = {
+	{ "off", MULTI_TASK_QUERY_INFO_OFF, false },
+	{ "debug", DEBUG2, false },
+	{ "log", LOG, false },
+	{ "notice", NOTICE, false },
+	{ "warning", WARNING, false },
+	{ "error", ERROR, false },
+	{ NULL, 0, false }
+};
+
 /* *INDENT-ON* */
 
 
@@ -614,6 +624,16 @@ RegisterCitusConfigVariables(void)
 		0,
 		NULL, NULL, NULL);
 
+	DefineCustomEnumVariable(
+		"citus.multi_task_query_log_level",
+		gettext_noop("Sets the level of multi task query execution log messages"),
+		NULL,
+		&MultiTaskQueryLogLevel,
+		MULTI_TASK_QUERY_INFO_OFF, multi_task_query_log_level_options,
+		PGC_USERSET,
+		0,
+		NULL, NULL, NULL);
+
 	DefineCustomStringVariable(
 		"citus.version",
 		gettext_noop("Shows the Citus library version"),

src/include/distributed/multi_server_executor.h

@@ -36,6 +36,8 @@
 #define JOB_CLEANUP_QUERY "SELECT task_tracker_cleanup_job("UINT64_FORMAT ")"
 #define JOB_CLEANUP_TASK_ID INT_MAX
 
+#define MULTI_TASK_QUERY_INFO_OFF 0  /* do not log multi-task queries */
+
 
 /* Enumeration to track one task's execution status */
 typedef enum
@@ -184,6 +186,7 @@ extern int RemoteTaskCheckInterval;
 extern int MaxAssignTaskBatchSize;
 extern int TaskExecutorType;
 extern bool BinaryMasterCopyFormat;
+extern int MultiTaskQueryLogLevel;
 
 
 /* Function declarations for distributed execution */

src/test/regress/expected/multi_cross_shard.out

@@ -0,0 +1,191 @@
+--
+-- MULTI_CROSS_SHARD
+--
+-- Tests to log cross shard queries according to error log level
+--
+-- Create a distributed table and add data to it
+CREATE TABLE multi_task_table
+(
+	id int, 
+	name varchar(20)
+);
+SELECT create_distributed_table('multi_task_table', 'id');
+ create_distributed_table 
+--------------------------
+ 
+(1 row)
+
+INSERT INTO multi_task_table VALUES(1, 'elem_1');
+INSERT INTO multi_task_table VALUES(2, 'elem_2');
+INSERT INTO multi_task_table VALUES(3, 'elem_3');
+-- Shouldn't log anything when the log level is 'off'
+SHOW citus.multi_task_query_log_level;
+ citus.multi_task_query_log_level 
+----------------------------------
+ off
+(1 row)
+
+SELECT * FROM multi_task_table;
+ id |  name  
+----+--------
+  1 | elem_1
+  3 | elem_3
+  2 | elem_2
+(3 rows)
+
+-- Get messages with the log level 'notice'
+SET citus.multi_task_query_log_level TO notice;
+SELECT * FROM multi_task_table;
+NOTICE:  multi-task query about to be executed
+HINT:  Queries are split to multiple tasks if they have to be split into several queries on the workers.
+ id |  name  
+----+--------
+  1 | elem_1
+  3 | elem_3
+  2 | elem_2
+(3 rows)
+
+SELECT AVG(id) AS avg_id FROM multi_task_table;
+NOTICE:  multi-task query about to be executed
+HINT:  Queries are split to multiple tasks if they have to be split into several queries on the workers.
+       avg_id       
+--------------------
+ 2.0000000000000000
+(1 row)
+
+-- Get messages with the log level 'error'
+SET citus.multi_task_query_log_level TO error;
+SELECT * FROM multi_task_table;
+ERROR:  multi-task query about to be executed
+HINT:  Queries are split to multiple tasks if they have to be split into several queries on the workers.
+-- Check the log message with INSERT INTO ... SELECT 
+CREATE TABLE raw_table
+(
+    id int,
+    order_count int
+);
+CREATE TABLE summary_table
+(
+    id int,
+    order_sum BIGINT
+);
+SELECT create_distributed_table('raw_table', 'id');
+ create_distributed_table 
+--------------------------
+ 
+(1 row)
+
+SELECT create_distributed_table('summary_table', 'id');
+ create_distributed_table 
+--------------------------
+ 
+(1 row)
+
+INSERT INTO raw_table VALUES(1, '15');
+INSERT INTO raw_table VALUES(2, '15');
+INSERT INTO raw_table VALUES(3, '15');
+INSERT INTO raw_table VALUES(1, '20');
+INSERT INTO raw_table VALUES(2, '25');
+INSERT INTO raw_table VALUES(3, '35');
+--  Should notice user that the query is multi-task one
+SET citus.multi_task_query_log_level TO notice;
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table GROUP BY id;
+NOTICE:  multi-task query about to be executed
+HINT:  Queries are split to multiple tasks if they have to be split into several queries on the workers.
+-- Should error out since the query is multi-task one
+SET citus.multi_task_query_log_level TO error;
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table GROUP BY id;
+ERROR:  multi-task query about to be executed
+HINT:  Queries are split to multiple tasks if they have to be split into several queries on the workers.
+-- Shouldn't error out since it is a single task insert-into select query
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table WHERE id = 1 GROUP BY id;
+-- Should have four rows (three rows from the query without where and the one from with where)
+SET citus.multi_task_query_log_level to DEFAULT;
+SELECT * FROM summary_table;
+ id | order_sum 
+----+-----------
+  1 |        35
+  1 |        35
+  3 |        50
+  2 |        40
+(4 rows)
+
+-- Set log-level to different levels inside the transaction
+BEGIN;
+--  Should notice user that the query is multi-task one
+SET citus.multi_task_query_log_level TO notice;
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table GROUP BY id;
+NOTICE:  multi-task query about to be executed
+HINT:  Queries are split to multiple tasks if they have to be split into several queries on the workers.
+-- Should error out since the query is multi-task one
+SET citus.multi_task_query_log_level TO error;
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table GROUP BY id;
+ERROR:  multi-task query about to be executed
+HINT:  Queries are split to multiple tasks if they have to be split into several queries on the workers.
+ROLLBACK;
+-- Should have only four rows since the transaction is rollbacked.
+SET citus.multi_task_query_log_level to DEFAULT;
+SELECT * FROM summary_table;
+ id | order_sum 
+----+-----------
+  1 |        35
+  1 |        35
+  3 |        50
+  2 |        40
+(4 rows)
+
+-- Test router-select query
+SET citus.multi_task_query_log_level TO notice;
+-- Shouldn't log since it is a router select query
+SELECT * FROM raw_table WHERE ID = 1;
+ id | order_count 
+----+-------------
+  1 |          15
+  1 |          20
+(2 rows)
+
+-- Task tracker query test
+CREATE TABLE tt1
+(
+	id int,
+	name varchar(20)
+);
+CREATE TABLE tt2
+(
+	id int,
+	name varchar(20),
+	count bigint
+);
+SELECT create_distributed_table('tt1', 'id');
+ create_distributed_table 
+--------------------------
+ 
+(1 row)
+
+SELECT create_distributed_table('tt2', 'name');
+ create_distributed_table 
+--------------------------
+ 
+(1 row)
+
+INSERT INTO tt1 VALUES(1, 'Ahmet');
+INSERT INTO tt1 VALUES(2, 'Mehmet');
+INSERT INTO tt2 VALUES(1, 'Ahmet', 5);
+INSERT INTO tt2 VALUES(2, 'Mehmet', 15);
+-- Should notice since it is a task-tracker query
+SET citus.task_executor_type to "task-tracker";
+SELECT tt1.id, tt2.count from tt1,tt2 where tt1.id = tt2.id;
+NOTICE:  multi-task query about to be executed
+HINT:  Queries are split to multiple tasks if they have to be split into several queries on the workers.
+ id | count 
+----+-------
+  1 |     5
+  2 |    15
+(2 rows)
+
+SET citus.task_executor_type to DEFAULT;
+DROP TABLE tt2;
+DROP TABLE tt1;
+DROP TABLE multi_task_table;
+DROP TABLE raw_table;
+DROP TABLE summary_table;

src/test/regress/multi_schedule

@@ -52,7 +52,7 @@ test: multi_partition_pruning
 test: multi_join_pruning multi_hash_pruning
 test: multi_null_minmax_value_pruning
 test: multi_query_directory_cleanup
-test: multi_task_assignment_policy
+test: multi_task_assignment_policy multi_cross_shard
 test: multi_utility_statements
 test: multi_dropped_column_aliases
 test: multi_binary_master_copy_format

src/test/regress/sql/multi_cross_shard.sql

@@ -0,0 +1,124 @@
+--
+-- MULTI_CROSS_SHARD
+--
+-- Tests to log cross shard queries according to error log level
+--
+
+-- Create a distributed table and add data to it
+CREATE TABLE multi_task_table
+(
+	id int, 
+	name varchar(20)
+);
+SELECT create_distributed_table('multi_task_table', 'id');
+
+INSERT INTO multi_task_table VALUES(1, 'elem_1');
+INSERT INTO multi_task_table VALUES(2, 'elem_2');
+INSERT INTO multi_task_table VALUES(3, 'elem_3');
+
+-- Shouldn't log anything when the log level is 'off'
+SHOW citus.multi_task_query_log_level;
+SELECT * FROM multi_task_table;
+
+-- Get messages with the log level 'notice'
+SET citus.multi_task_query_log_level TO notice;
+SELECT * FROM multi_task_table;
+SELECT AVG(id) AS avg_id FROM multi_task_table;
+
+-- Get messages with the log level 'error'
+SET citus.multi_task_query_log_level TO error;
+SELECT * FROM multi_task_table;
+
+-- Check the log message with INSERT INTO ... SELECT 
+CREATE TABLE raw_table
+(
+    id int,
+    order_count int
+);
+
+CREATE TABLE summary_table
+(
+    id int,
+    order_sum BIGINT
+);
+
+SELECT create_distributed_table('raw_table', 'id');
+SELECT create_distributed_table('summary_table', 'id');
+
+INSERT INTO raw_table VALUES(1, '15');
+INSERT INTO raw_table VALUES(2, '15');
+INSERT INTO raw_table VALUES(3, '15');
+INSERT INTO raw_table VALUES(1, '20');
+INSERT INTO raw_table VALUES(2, '25');
+INSERT INTO raw_table VALUES(3, '35');
+
+--  Should notice user that the query is multi-task one
+SET citus.multi_task_query_log_level TO notice;
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table GROUP BY id;
+
+-- Should error out since the query is multi-task one
+SET citus.multi_task_query_log_level TO error;
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table GROUP BY id;
+
+-- Shouldn't error out since it is a single task insert-into select query
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table WHERE id = 1 GROUP BY id;
+
+-- Should have four rows (three rows from the query without where and the one from with where)
+SET citus.multi_task_query_log_level to DEFAULT;
+SELECT * FROM summary_table;
+
+-- Set log-level to different levels inside the transaction
+BEGIN;
+--  Should notice user that the query is multi-task one
+SET citus.multi_task_query_log_level TO notice;
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table GROUP BY id;
+
+-- Should error out since the query is multi-task one
+SET citus.multi_task_query_log_level TO error;
+INSERT INTO summary_table SELECT id, SUM(order_count) FROM raw_table GROUP BY id;
+ROLLBACK;
+
+-- Should have only four rows since the transaction is rollbacked.
+SET citus.multi_task_query_log_level to DEFAULT;
+SELECT * FROM summary_table;
+
+-- Test router-select query
+SET citus.multi_task_query_log_level TO notice;
+
+-- Shouldn't log since it is a router select query
+SELECT * FROM raw_table WHERE ID = 1;
+
+-- Task tracker query test
+CREATE TABLE tt1
+(
+	id int,
+	name varchar(20)
+);
+
+CREATE TABLE tt2
+(
+	id int,
+	name varchar(20),
+	count bigint
+);
+
+SELECT create_distributed_table('tt1', 'id');
+SELECT create_distributed_table('tt2', 'name');
+
+INSERT INTO tt1 VALUES(1, 'Ahmet');
+INSERT INTO tt1 VALUES(2, 'Mehmet');
+
+INSERT INTO tt2 VALUES(1, 'Ahmet', 5);
+INSERT INTO tt2 VALUES(2, 'Mehmet', 15);
+
+-- Should notice since it is a task-tracker query
+SET citus.task_executor_type to "task-tracker";
+SELECT tt1.id, tt2.count from tt1,tt2 where tt1.id = tt2.id;
+
+SET citus.task_executor_type to DEFAULT;
+
+DROP TABLE tt2;
+DROP TABLE tt1;
+DROP TABLE multi_task_table;
+DROP TABLE raw_table;
+DROP TABLE summary_table;