Fix split schedule

src/test/regress/citus_tests/run_test.py

@@ -73,6 +73,8 @@ elif "isolation" in test_schedule:
     test_schedule = 'base_isolation_schedule'
 elif "failure" in test_schedule:
     test_schedule = 'failure_base_schedule'
+elif "split" in test_schedule:
+    test_schedule = 'minimal_schedule'
 elif "mx" in test_schedule:
     if use_base_schedule:
         test_schedule = 'mx_base_schedule'

src/test/regress/enterprise_split_schedule

@@ -1,10 +0,0 @@
-# Split Shard tests.
-# Include tests from 'minimal_schedule' for setup.
-test: multi_test_helpers multi_test_helpers_superuser columnar_test_helpers
-test: multi_cluster_management
-test: multi_test_catalog_views
-test: tablespace
-# Split tests go here.
-test: citus_split_shard_by_split_points_negative
-test: citus_split_shard_by_split_points
-test: citus_split_shard_by_split_points_deferred_drop

src/test/regress/expected/citus_non_blocking_shard_split_cleanup.out

@@ -1,123 +0,0 @@
-/*
-Citus Shard Split Test.The test is model similar to 'shard_move_constraints'.
-Here is a high level overview of test plan:
- 1. Create a table 'sensors' (ShardCount = 2) to be split. Add indexes and statistics on this table.
- 2. Create two other tables: 'reference_table' and 'colocated_dist_table', co-located with sensors.
- 3. Create Foreign key constraints between the two co-located distributed tables.
- 4. Load data into the three tables.
- 5. Move one of the shards for 'sensors' to test ShardMove -> Split.
- 6. Trigger Split on both shards of 'sensors'. This will also split co-located tables.
- 7. Move one of the split shard to test Split -> ShardMove.
- 8. Split an already split shard second time on a different schema.
-*/
-CREATE SCHEMA "citus_split_test_schema";
-CREATE ROLE test_split_role WITH LOGIN;
-GRANT USAGE, CREATE ON SCHEMA "citus_split_test_schema" TO test_split_role;
-SET ROLE test_split_role;
-SET search_path TO "citus_split_test_schema";
-SET citus.next_shard_id TO 8981000;
-SET citus.next_placement_id TO 8610000;
-SET citus.shard_count TO 2;
-SET citus.shard_replication_factor TO 1;
--- BEGIN: Create table to split, along with other co-located tables. Add indexes, statistics etc.
-CREATE TABLE sensors(
-    measureid               integer,
-    eventdatetime           date,
-    measure_data            jsonb,
-	meaure_quantity         decimal(15, 2),
-    measure_status          char(1),
-	measure_comment         varchar(44),
-    PRIMARY KEY (measureid, eventdatetime, measure_data));
-SELECT create_distributed_table('sensors', 'measureid', colocate_with:='none');
- create_distributed_table
----------------------------------------------------------------------
-
-(1 row)
-
-INSERT INTO sensors SELECT i, '2020-01-05', '{}', 11011.10, 'A', 'I <3 Citus' FROM generate_series(0,1000)i;
--- END: Create table to split, along with other co-located tables. Add indexes, statistics etc.
--- BEGIN : Move one shard before we split it.
-\c - postgres - :master_port
-SET ROLE test_split_role;
-SET search_path TO "citus_split_test_schema";
-SET citus.next_shard_id TO 8981007;
-SELECT citus_move_shard_placement(8981000, 'localhost', :worker_1_port, 'localhost', :worker_2_port, shard_transfer_mode:='force_logical');
- citus_move_shard_placement
----------------------------------------------------------------------
-
-(1 row)
-
--- END : Move one shard before we split it.
--- BEGIN : Set node id variables
-SELECT nodeid AS worker_1_node FROM pg_dist_node WHERE nodeport=:worker_1_port \gset
-SELECT nodeid AS worker_2_node FROM pg_dist_node WHERE nodeport=:worker_2_port \gset
--- END   : Set node id variables
--- BEGIN : Split two shards : One with move and One without move.
--- Perform 2 way split
-SELECT * FROM citus_shards WHERE nodeport IN (:worker_1_port, :worker_2_port);
- table_name | shardid |               shard_name                | citus_table_type | colocation_id | nodename  | nodeport | shard_size
----------------------------------------------------------------------
- sensors    | 8981000 | citus_split_test_schema.sensors_8981000 | distributed      |       1390009 | localhost |    57638 |      40960
- sensors    | 8981001 | citus_split_test_schema.sensors_8981001 | distributed      |       1390009 | localhost |    57638 |      40960
-(2 rows)
-
-SELECT pg_catalog.citus_split_shard_by_split_points(
-    8981000,
-    ARRAY['-1073741824'],
-    ARRAY[:worker_2_node, :worker_2_node],
-    'force_logical');
-WARNING:  replication slot "citus_shard_split_template_slot_8981000" does not exist
-CONTEXT:  while executing command on localhost:xxxxx
- citus_split_shard_by_split_points
----------------------------------------------------------------------
-
-(1 row)
-
-SELECT * FROM citus_shards WHERE nodeport IN (:worker_1_port, :worker_2_port);
- table_name | shardid |               shard_name                | citus_table_type | colocation_id | nodename  | nodeport | shard_size
----------------------------------------------------------------------
- sensors    | 8981001 | citus_split_test_schema.sensors_8981001 | distributed      |       1390009 | localhost |    57638 |      40960
- sensors    | 8981007 | citus_split_test_schema.sensors_8981007 | distributed      |       1390009 | localhost |    57638 |      24576
- sensors    | 8981008 | citus_split_test_schema.sensors_8981008 | distributed      |       1390009 | localhost |    57638 |      24576
-(3 rows)
-
-\c - - - :worker_2_port
-SELECT slot_name FROM pg_replication_slots;
-                slot_name
----------------------------------------------------------------------
- citus_shard_split_template_slot_8981000
- citus_shard_split_18_20648
-(2 rows)
-
-\c - - - :master_port
-SELECT pg_catalog.citus_split_shard_by_split_points(
-    8981001,
-    ARRAY['536870911', '1610612735'],
-    ARRAY[:worker_1_node, :worker_1_node, :worker_2_node],
-    'force_logical');
-WARNING:  replication slot "citus_shard_split_template_slot_8981001" does not exist
-CONTEXT:  while executing command on localhost:xxxxx
- citus_split_shard_by_split_points
----------------------------------------------------------------------
-
-(1 row)
-
-SELECT * FROM citus_shards WHERE nodeport IN (:worker_1_port, :worker_2_port);
-           table_name            | shardid |               shard_name                | citus_table_type | colocation_id | nodename  | nodeport | shard_size
----------------------------------------------------------------------
- citus_split_test_schema.sensors |  102042 | citus_split_test_schema.sensors_102042  | distributed      |       1390009 | localhost |    57637 |       8192
- citus_split_test_schema.sensors |  102043 | citus_split_test_schema.sensors_102043  | distributed      |       1390009 | localhost |    57637 |      16384
- citus_split_test_schema.sensors |  102044 | citus_split_test_schema.sensors_102044  | distributed      |       1390009 | localhost |    57638 |      16384
- citus_split_test_schema.sensors | 8981007 | citus_split_test_schema.sensors_8981007 | distributed      |       1390009 | localhost |    57638 |      24576
- citus_split_test_schema.sensors | 8981008 | citus_split_test_schema.sensors_8981008 | distributed      |       1390009 | localhost |    57638 |      24576
-(5 rows)
-
-\c - - - :worker_2_port
-SELECT slot_name FROM pg_replication_slots;
-                slot_name
----------------------------------------------------------------------
- citus_shard_split_template_slot_8981001
- citus_shard_split_16_20648
- citus_shard_split_18_20648
-(3 rows)
-

src/test/regress/expected/citus_split_shard_by_split_points.out

@@ -498,4 +498,5 @@ DETAIL:  drop cascades to table citus_split_test_schema.sensors
 drop cascades to table citus_split_test_schema.reference_table
 drop cascades to table citus_split_test_schema.colocated_dist_table
 drop cascades to table citus_split_test_schema.table_with_index_rep_identity
+DROP USER test_split_role;
 --END : Cleanup

src/test/regress/expected/citus_split_shard_by_split_points_deferred_drop.out

@@ -13,9 +13,9 @@ SELECT * from pg_dist_cleanup;
 ---------------------------------------------------------------------
 (0 rows)
 
--- Set a very long(10mins) time interval to stop auto cleanup in case of deferred drop.
+-- Disable Deferred drop auto cleanup to avoid flaky tests.
 \c - postgres - :master_port
-ALTER SYSTEM SET citus.defer_shard_delete_interval TO 600000;
+ALTER SYSTEM SET citus.defer_shard_delete_interval TO -1;
 SELECT pg_reload_conf();
  pg_reload_conf
 ---------------------------------------------------------------------
@@ -28,8 +28,9 @@ SET citus.next_placement_id TO 8610000;
 SET citus.shard_count TO 2;
 SET citus.shard_replication_factor TO 1;
 SET citus.next_operation_id TO 777;
-SET citus.next_cleanup_record_id TO 11;
+SET citus.next_cleanup_record_id TO 511;
 SET ROLE test_split_role;
+SET search_path TO "citus_split_shard_by_split_points_deferred_schema";
 CREATE TABLE table_to_split(id int PRIMARY KEY, int_data int, data text);
 SELECT create_distributed_table('table_to_split', 'id');
  create_distributed_table
@@ -61,26 +62,25 @@ SELECT pg_catalog.citus_split_shard_by_split_points(
 
 (1 row)
 
--- The original shards are marked for deferred drop with policy_type = 2.
+-- The original shard is marked for deferred drop with policy_type = 2.
+-- The previous shard should be dropped at the beginning of the second split call
 SELECT * from pg_dist_cleanup;
  record_id | operation_id | object_type |                               object_name                                | node_group_id | policy_type
 ---------------------------------------------------------------------
-        11 |          777 |           1 | public.table_to_split_8981000 |            14 |           2
+       512 |          778 |           1 | citus_split_shard_by_split_points_deferred_schema.table_to_split_8981001 |            16 |           2
-        12 |          778 |           1 | public.table_to_split_8981001 |            16 |           2
+(1 row)
-(2 rows)
 
--- The physical shards should not be deleted.
+-- One of the physical shards should not be deleted, the other one should.
 \c - - - :worker_1_port
-SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r';
+SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r' ORDER BY relname;
         relname
 ---------------------------------------------------------------------
- table_to_split_8981000
  table_to_split_9999000
  table_to_split_9999002
-(3 rows)
+(2 rows)
 
 \c - - - :worker_2_port
-SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r';
+SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r' ORDER BY relname;
         relname
 ---------------------------------------------------------------------
  table_to_split_8981001
@@ -88,30 +88,19 @@ SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind
  table_to_split_9999003
 (3 rows)
 
--- Set a very short(1ms) time interval to force deferred drop cleanup.
+-- Perform deferred drop cleanup.
 \c - postgres - :master_port
-ALTER SYSTEM SET citus.defer_shard_delete_interval TO 1;
+CALL citus_cleanup_orphaned_resources();
-SELECT pg_reload_conf();
+NOTICE:  cleaned up 1 orphaned resources
- pg_reload_conf
----------------------------------------------------------------------
- t
-(1 row)
-
--- Give enough time for the deferred drop cleanup to run.
-SELECT pg_sleep(2);
- pg_sleep
----------------------------------------------------------------------
-
-(1 row)
-
 -- Clenaup has been done.
 SELECT * from pg_dist_cleanup;
  record_id | operation_id | object_type | object_name | node_group_id | policy_type
 ---------------------------------------------------------------------
 (0 rows)
 
+-- Verify that the shard to be dropped is dropped
 \c - - - :worker_1_port
-SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r';
+SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r' ORDER BY relname;
         relname
 ---------------------------------------------------------------------
  table_to_split_9999000
@@ -119,7 +108,7 @@ SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind
 (2 rows)
 
 \c - - - :worker_2_port
-SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r';
+SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r' ORDER BY relname;
         relname
 ---------------------------------------------------------------------
  table_to_split_9999001
@@ -128,5 +117,6 @@ SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind
 
 -- Test Cleanup
 \c - postgres - :master_port
+SET client_min_messages TO WARNING;
 DROP SCHEMA "citus_split_shard_by_split_points_deferred_schema" CASCADE;
-NOTICE:  drop cascades to table citus_split_shard_by_split_points_deferred_schema.temp_table
+DROP USER test_split_role;

src/test/regress/expected/split_shard_release_dsm.out

@@ -43,3 +43,8 @@ SELECT count(*) FROM pg_catalog.worker_split_shard_replication_setup(ARRAY[
      1
 (1 row)
 
+-- cleanup, we are done with these manually created test tables
+DROP TABLE table_to_split_1, table_to_split_2, table_to_split_3;
+\c - - - :worker_2_port
+SET search_path TO split_shard_replication_setup_schema;
+DROP TABLE table_to_split_1, table_to_split_2, table_to_split_3;

src/test/regress/split_schedule

@@ -17,6 +17,7 @@ test: worker_split_binary_copy_test
 test: worker_split_text_copy_test
 test: citus_split_shard_by_split_points_negative
 test: citus_split_shard_by_split_points
+test: citus_split_shard_by_split_points_deferred_drop
 test: citus_split_shard_by_split_points_failure
 # Name citus_split_shard_by_split_points_columnar_partitioned was too long and being truncated.
 # use citus_split_shard_columnar_partitioned instead.

src/test/regress/sql/citus_split_shard_by_split_points.sql

@@ -256,4 +256,5 @@ SELECT COUNT(*) FROM colocated_dist_table;
 ALTER SYSTEM RESET citus.defer_shard_delete_interval;
 SELECT pg_reload_conf();
 DROP SCHEMA "citus_split_test_schema" CASCADE;
+DROP USER test_split_role;
 --END : Cleanup

src/test/regress/sql/citus_split_shard_by_split_points_deferred_drop.sql

@@ -24,8 +24,9 @@ SET citus.next_placement_id TO 8610000;
 SET citus.shard_count TO 2;
 SET citus.shard_replication_factor TO 1;
 SET citus.next_operation_id TO 777;
-SET citus.next_cleanup_record_id TO 11;
+SET citus.next_cleanup_record_id TO 511;
 SET ROLE test_split_role;
+SET search_path TO "citus_split_shard_by_split_points_deferred_schema";
 
 CREATE TABLE table_to_split(id int PRIMARY KEY, int_data int, data text);
 SELECT create_distributed_table('table_to_split', 'id');
@@ -48,29 +49,33 @@ SELECT pg_catalog.citus_split_shard_by_split_points(
     ARRAY[:worker_1_node, :worker_2_node],
     'force_logical');
 
--- The original shards are marked for deferred drop with policy_type = 2.
+-- The original shard is marked for deferred drop with policy_type = 2.
+-- The previous shard should be dropped at the beginning of the second split call
 SELECT * from pg_dist_cleanup;
 
--- The physical shards should not be deleted.
+-- One of the physical shards should not be deleted, the other one should.
 \c - - - :worker_1_port
-SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r';
+SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r' ORDER BY relname;
 
 \c - - - :worker_2_port
-SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r';
+SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r' ORDER BY relname;
 
 -- Perform deferred drop cleanup.
 \c - postgres - :master_port
-CALL citus_cleanup_orphaned_shards();
+CALL citus_cleanup_orphaned_resources();
 
 -- Clenaup has been done.
 SELECT * from pg_dist_cleanup;
 
+-- Verify that the shard to be dropped is dropped
 \c - - - :worker_1_port
-SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r';
+SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r' ORDER BY relname;
 
 \c - - - :worker_2_port
-SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r';
+SELECT relname FROM pg_class where relname LIKE '%table_to_split_%' AND relkind = 'r' ORDER BY relname;
 
 -- Test Cleanup
 \c - postgres - :master_port
+SET client_min_messages TO WARNING;
 DROP SCHEMA "citus_split_shard_by_split_points_deferred_schema" CASCADE;
+DROP USER test_split_role;

src/test/regress/sql/split_shard_release_dsm.sql

@@ -26,3 +26,10 @@ SELECT count(*) FROM pg_catalog.worker_split_shard_replication_setup(ARRAY[
     ROW(1, 'id', 2, '-2147483648', '-1', :worker_1_node)::pg_catalog.split_shard_info,
     ROW(1, 'id', 3, '0', '2147483647', :worker_1_node)::pg_catalog.split_shard_info
     ], 0);
+
+-- cleanup, we are done with these manually created test tables
+DROP TABLE table_to_split_1, table_to_split_2, table_to_split_3;
+
+\c - - - :worker_2_port
+SET search_path TO split_shard_replication_setup_schema;
+DROP TABLE table_to_split_1, table_to_split_2, table_to_split_3;