citus/src/test/regress/sql/background_rebalance_parall...

407 lines
17 KiB
PL/PgSQL

--
-- BACKGROUND_REBALANCE_PARALLEL
--
-- Test to check if the background tasks scheduled by the background rebalancer
-- have the correct dependencies
--
-- Test to verify that we do not allow parallel rebalancer moves involving a
-- particular node (either as source or target) more than
-- citus.max_background_task_executors_per_node, and that we can change the GUC on
-- the fly, and that will affect the ongoing balance as it should
--
-- Test to verify that there's a hard dependency when a specific node is first being
-- used as a source for a move, and then later as a target.
--
CREATE SCHEMA background_rebalance_parallel;
SET search_path TO background_rebalance_parallel;
SET citus.next_shard_id TO 85674000;
SET citus.shard_replication_factor TO 1;
SET client_min_messages TO ERROR;
ALTER SEQUENCE pg_dist_background_job_job_id_seq RESTART 17777;
ALTER SEQUENCE pg_dist_background_task_task_id_seq RESTART 1000;
ALTER SEQUENCE pg_catalog.pg_dist_colocationid_seq RESTART 50050;
SELECT nextval('pg_catalog.pg_dist_groupid_seq') AS last_group_id_cls \gset
SELECT nextval('pg_catalog.pg_dist_node_nodeid_seq') AS last_node_id_cls \gset
ALTER SEQUENCE pg_catalog.pg_dist_groupid_seq RESTART 50;
ALTER SEQUENCE pg_catalog.pg_dist_node_nodeid_seq RESTART 50;
SELECT 1 FROM master_remove_node('localhost', :worker_1_port);
SELECT 1 FROM master_remove_node('localhost', :worker_2_port);
SELECT 1 FROM master_add_node('localhost', :worker_1_port);
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
ALTER SYSTEM SET citus.background_task_queue_interval TO '1s';
SELECT pg_reload_conf();
-- Colocation group 1: create two tables table1_colg1, table2_colg1 and in a colocation group
CREATE TABLE table1_colg1 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg1', 'a', shard_count => 4, colocate_with => 'none');
CREATE TABLE table2_colg1 (b int PRIMARY KEY);
SELECT create_distributed_table('table2_colg1', 'b', colocate_with => 'table1_colg1');
-- Colocation group 2: create two tables table1_colg2, table2_colg2 and in a colocation group
CREATE TABLE table1_colg2 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg2', 'a', shard_count => 4, colocate_with => 'none');
CREATE TABLE table2_colg2 (b int primary key);
SELECT create_distributed_table('table2_colg2', 'b', colocate_with => 'table1_colg2');
-- Colocation group 3: create two tables table1_colg3, table2_colg3 and in a colocation group
CREATE TABLE table1_colg3 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg3', 'a', shard_count => 4, colocate_with => 'none');
CREATE TABLE table2_colg3 (b int primary key);
SELECT create_distributed_table('table2_colg3', 'b', colocate_with => 'table1_colg3');
-- Add two new nodes so that we can rebalance
SELECT 1 FROM citus_add_node('localhost', :worker_3_port);
SELECT 1 FROM citus_add_node('localhost', :worker_4_port);
SELECT * FROM get_rebalance_table_shards_plan() ORDER BY shardid;
SELECT * FROM citus_rebalance_start();
SELECT citus_rebalance_wait();
-- PART 1
-- Test to check if the background tasks scheduled by the background rebalancer
-- have the correct dependencies
-- Check that a move is dependent on
-- any other move scheduled earlier in its colocation group.
SELECT S.shardid, P.colocationid
FROM pg_dist_shard S, pg_dist_partition P
WHERE S.logicalrelid = P.logicalrelid ORDER BY S.shardid ASC;
SELECT D.task_id,
(SELECT T.command FROM pg_dist_background_task T WHERE T.task_id = D.task_id),
D.depends_on,
(SELECT T.command FROM pg_dist_background_task T WHERE T.task_id = D.depends_on)
FROM pg_dist_background_task_depend D WHERE job_id = 17777 ORDER BY D.task_id, D.depends_on ASC;
-- Check that if there is a reference table that needs to be synched to a node,
-- any move without a dependency must depend on the move task for reference table.
SELECT 1 FROM citus_drain_node('localhost',:worker_4_port);
SELECT public.wait_for_resource_cleanup();
SELECT 1 FROM citus_disable_node('localhost', :worker_4_port, synchronous:=true);
-- Drain worker_3 so that we can move only one colocation group to worker_3
-- to create an unbalance that would cause parallel rebalancing.
SELECT 1 FROM citus_drain_node('localhost',:worker_3_port);
SELECT citus_set_node_property('localhost', :worker_3_port, 'shouldhaveshards', true);
CALL citus_cleanup_orphaned_resources();
CREATE TABLE ref_table(a int PRIMARY KEY);
SELECT create_reference_table('ref_table');
-- Move all the shards of Colocation group 3 to worker_3.
SELECT
master_move_shard_placement(shardid, 'localhost', nodeport, 'localhost', :worker_3_port, 'block_writes')
FROM
pg_dist_shard NATURAL JOIN pg_dist_shard_placement
WHERE
logicalrelid = 'table1_colg3'::regclass AND nodeport <> :worker_3_port
ORDER BY
shardid;
CALL citus_cleanup_orphaned_resources();
-- Activate and new nodes so that we can rebalance.
SELECT 1 FROM citus_activate_node('localhost', :worker_4_port);
SELECT citus_set_node_property('localhost', :worker_4_port, 'shouldhaveshards', true);
SELECT 1 FROM citus_add_node('localhost', :worker_5_port);
SELECT 1 FROM citus_add_node('localhost', :worker_6_port);
SELECT * FROM citus_rebalance_start();
SELECT citus_rebalance_wait();
SELECT S.shardid, P.colocationid
FROM pg_dist_shard S, pg_dist_partition P
WHERE S.logicalrelid = P.logicalrelid ORDER BY S.shardid ASC;
SELECT D.task_id,
(SELECT T.command FROM pg_dist_background_task T WHERE T.task_id = D.task_id),
D.depends_on,
(SELECT T.command FROM pg_dist_background_task T WHERE T.task_id = D.depends_on)
FROM pg_dist_background_task_depend D WHERE job_id = 17778 ORDER BY D.task_id, D.depends_on ASC;
-- PART 2
-- Test to verify that we do not allow parallel rebalancer moves involving a
-- particular node (either as source or target)
-- more than citus.max_background_task_executors_per_node
-- and that we can change the GUC on the fly
-- citus_task_wait calls are used to ensure consistent pg_dist_background_task query
-- output i.e. to avoid flakiness
-- First let's restart the scenario
DROP SCHEMA background_rebalance_parallel CASCADE;
TRUNCATE pg_dist_background_job CASCADE;
TRUNCATE pg_dist_background_task CASCADE;
TRUNCATE pg_dist_background_task_depend;
SELECT public.wait_for_resource_cleanup();
select citus_remove_node('localhost', :worker_2_port);
select citus_remove_node('localhost', :worker_3_port);
select citus_remove_node('localhost', :worker_4_port);
select citus_remove_node('localhost', :worker_5_port);
select citus_remove_node('localhost', :worker_6_port);
CREATE SCHEMA background_rebalance_parallel;
SET search_path TO background_rebalance_parallel;
-- Create 8 tables in 4 colocation groups, and populate them
CREATE TABLE table1_colg1 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg1', 'a', shard_count => 3, colocate_with => 'none');
INSERT INTO table1_colg1 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table2_colg1 (b int PRIMARY KEY);
SELECT create_distributed_table('table2_colg1', 'b', colocate_with => 'table1_colg1');
INSERT INTO table2_colg1 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table1_colg2 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg2', 'a', shard_count => 3, colocate_with => 'none');
INSERT INTO table1_colg2 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table2_colg2 (b int PRIMARY KEY);
SELECT create_distributed_table('table2_colg2', 'b', colocate_with => 'table1_colg2');
INSERT INTO table2_colg2 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table1_colg3 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg3', 'a', shard_count => 3, colocate_with => 'none');
INSERT INTO table1_colg3 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table2_colg3 (b int primary key);
SELECT create_distributed_table('table2_colg3', 'b', colocate_with => 'table1_colg3');
INSERT INTO table2_colg3 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table1_colg4 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg4', 'a', shard_count => 3, colocate_with => 'none');
INSERT INTO table1_colg4 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table2_colg4 (b int PRIMARY KEY);
SELECT create_distributed_table('table2_colg4', 'b', colocate_with => 'table1_colg4');
INSERT INTO table2_colg4 SELECT i FROM generate_series(0, 100)i;
-- Add nodes so that we can rebalance
SELECT citus_add_node('localhost', :worker_2_port);
SELECT citus_add_node('localhost', :worker_3_port);
SELECT citus_rebalance_start AS job_id from citus_rebalance_start() \gset
-- see dependent tasks to understand which tasks remain runnable because of
-- citus.max_background_task_executors_per_node
-- and which tasks are actually blocked from colocation group dependencies
SELECT task_id, depends_on
FROM pg_dist_background_task_depend
WHERE job_id in (:job_id)
ORDER BY 1, 2 ASC;
-- default citus.max_background_task_executors_per_node is 1
-- show that first exactly one task per node is running
-- among the tasks that are not blocked
SELECT citus_task_wait(1013, desired_status => 'running');
SELECT job_id, task_id, status, nodes_involved
FROM pg_dist_background_task WHERE job_id in (:job_id) ORDER BY task_id;
-- increase citus.max_background_task_executors_per_node
SELECT citus_task_wait(1013, desired_status => 'done');
ALTER SYSTEM SET citus.max_background_task_executors_per_node = 2;
SELECT pg_reload_conf();
SELECT citus_task_wait(1014, desired_status => 'running');
SELECT citus_task_wait(1015, desired_status => 'running');
-- show that at most 2 tasks per node are running
-- among the tasks that are not blocked
SELECT job_id, task_id, status, nodes_involved
FROM pg_dist_background_task WHERE job_id in (:job_id) ORDER BY task_id;
-- decrease to default (1)
ALTER SYSTEM RESET citus.max_background_task_executors_per_node;
SELECT pg_reload_conf();
SELECT citus_task_wait(1015, desired_status => 'done');
SELECT citus_task_wait(1014, desired_status => 'done');
SELECT citus_task_wait(1016, desired_status => 'running');
-- show that exactly one task per node is running
-- among the tasks that are not blocked
SELECT job_id, task_id, status, nodes_involved
FROM pg_dist_background_task WHERE job_id in (:job_id) ORDER BY task_id;
SELECT citus_rebalance_stop();
-- PART 3
-- Test to verify that there's a hard dependency when A specific node is first being used as a
-- source for a move, and then later as a target.
-- First let's restart the scenario
DROP SCHEMA background_rebalance_parallel CASCADE;
TRUNCATE pg_dist_background_job CASCADE;
TRUNCATE pg_dist_background_task CASCADE;
TRUNCATE pg_dist_background_task_depend;
SELECT public.wait_for_resource_cleanup();
select citus_remove_node('localhost', :worker_1_port);
select citus_remove_node('localhost', :worker_2_port);
select citus_remove_node('localhost', :worker_3_port);
CREATE SCHEMA background_rebalance_parallel;
SET search_path TO background_rebalance_parallel;
SET citus.next_shard_id TO 85674051;
ALTER SEQUENCE pg_catalog.pg_dist_node_nodeid_seq RESTART 61;
-- add the first node
-- nodeid here is 61
select citus_add_node('localhost', :worker_1_port);
-- create, populate and distribute 6 tables, each with 1 shard, none colocated with each other
CREATE TABLE table1_colg1 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg1', 'a', shard_count => 1, colocate_with => 'none');
INSERT INTO table1_colg1 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table1_colg2 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg2', 'a', shard_count => 1, colocate_with => 'none');
INSERT INTO table1_colg2 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table1_colg3 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg3', 'a', shard_count => 1, colocate_with => 'none');
INSERT INTO table1_colg3 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table1_colg4 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg4', 'a', shard_count => 1, colocate_with => 'none');
INSERT INTO table1_colg4 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table1_colg5 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg5', 'a', shard_count => 1, colocate_with => 'none');
INSERT INTO table1_colg5 SELECT i FROM generate_series(0, 100)i;
CREATE TABLE table1_colg6 (a int PRIMARY KEY);
SELECT create_distributed_table('table1_colg6', 'a', shard_count => 1, colocate_with => 'none');
INSERT INTO table1_colg6 SELECT i FROM generate_series(0, 100)i;
-- add two other nodes
-- nodeid here is 62
select citus_add_node('localhost', :worker_2_port);
-- nodeid here is 63
select citus_add_node('localhost', :worker_3_port);
CREATE OR REPLACE FUNCTION shard_placement_rebalance_array(
worker_node_list json[],
shard_placement_list json[],
threshold float4 DEFAULT 0,
max_shard_moves int DEFAULT 1000000,
drain_only bool DEFAULT false,
improvement_threshold float4 DEFAULT 0.5
)
RETURNS json[]
AS 'citus'
LANGUAGE C STRICT VOLATILE;
-- we are simulating the following from shard_rebalancer_unit.sql
-- the following steps are all according to this scenario
-- where the third move should be dependent of the first two
-- because the third move's target is the source of the first two
SELECT unnest(shard_placement_rebalance_array(
ARRAY['{"node_name": "hostname1", "disallowed_shards": "1,2,3,5,6"}',
'{"node_name": "hostname2", "disallowed_shards": "4"}',
'{"node_name": "hostname3", "disallowed_shards": "4"}'
]::json[],
ARRAY['{"shardid":1, "nodename":"hostname1"}',
'{"shardid":2, "nodename":"hostname1"}',
'{"shardid":3, "nodename":"hostname2"}',
'{"shardid":4, "nodename":"hostname2"}',
'{"shardid":5, "nodename":"hostname3"}',
'{"shardid":6, "nodename":"hostname3"}'
]::json[]
));
-- manually balance the cluster such that we have
-- a balanced cluster like above with 1,2,3,4,5,6 and hostname1/2/3
-- shardid 85674051 (1) nodeid 61 (hostname1)
-- shardid 85674052 (2) nodeid 61 (hostname1)
-- shardid 85674053 (3) nodeid 62 (hostname2)
-- shardid 85674054 (4) nodeid 62 (hostname2)
-- shardid 85674055 (5) nodeid 63 (hostname3)
-- shardid 85674056 (6) nodeid 63 (hostname3)
SELECT pg_catalog.citus_move_shard_placement(85674053,61,62,'auto');
SELECT pg_catalog.citus_move_shard_placement(85674054,61,62,'auto');
SELECT pg_catalog.citus_move_shard_placement(85674055,61,63,'auto');
SELECT pg_catalog.citus_move_shard_placement(85674056,61,63,'auto');
-- now create another rebalance strategy in order to simulate moves
-- which use as target a node that has been previously used as source
CREATE OR REPLACE FUNCTION test_shard_allowed_on_node(shardid bigint, nodeid int)
RETURNS boolean AS
$$
-- analogous to '{"node_name": "hostname1", "disallowed_shards": "1,2,3,5,6"}'
select case when (shardid != 85674054 and nodeid = 61)
then false
-- analogous to '{"node_name": "hostname2", "disallowed_shards": "4"}'
-- AND '{"node_name": "hostname2", "disallowed_shards": "4"}'
when (shardid = 85674054 and nodeid != 61)
then false
else true
end;
$$ LANGUAGE sql;
-- insert the new test rebalance strategy
INSERT INTO
pg_catalog.pg_dist_rebalance_strategy(
name,
default_strategy,
shard_cost_function,
node_capacity_function,
shard_allowed_on_node_function,
default_threshold,
minimum_threshold,
improvement_threshold
) VALUES (
'test_source_then_target',
false,
'citus_shard_cost_1',
'citus_node_capacity_1',
'background_rebalance_parallel.test_shard_allowed_on_node',
0,
0,
0
);
SELECT * FROM get_rebalance_table_shards_plan(rebalance_strategy := 'test_source_then_target');
SELECT citus_rebalance_start AS job_id from citus_rebalance_start(rebalance_strategy := 'test_source_then_target') \gset
-- check that the third move is blocked and depends on the first two
SELECT job_id, task_id, status, nodes_involved
FROM pg_dist_background_task WHERE job_id in (:job_id) ORDER BY task_id;
SELECT D.task_id,
(SELECT T.command FROM pg_dist_background_task T WHERE T.task_id = D.task_id),
D.depends_on,
(SELECT T.command FROM pg_dist_background_task T WHERE T.task_id = D.depends_on)
FROM pg_dist_background_task_depend D WHERE job_id in (:job_id) ORDER BY D.task_id, D.depends_on ASC;
SELECT citus_rebalance_stop();
DELETE FROM pg_catalog.pg_dist_rebalance_strategy WHERE name='test_source_then_target';
DROP SCHEMA background_rebalance_parallel CASCADE;
TRUNCATE pg_dist_background_job CASCADE;
TRUNCATE pg_dist_background_task CASCADE;
TRUNCATE pg_dist_background_task_depend;
SELECT public.wait_for_resource_cleanup();
select citus_remove_node('localhost', :worker_3_port);
-- keep the rest of the tests inact that depends node/group ids
ALTER SEQUENCE pg_catalog.pg_dist_groupid_seq RESTART :last_group_id_cls;
ALTER SEQUENCE pg_catalog.pg_dist_node_nodeid_seq RESTART :last_node_id_cls;