citus/src/test/regress/sql/single_shard_table_udfs.sql

337 lines
15 KiB
PL/PgSQL

CREATE SCHEMA null_dist_key_udfs;
SET search_path TO null_dist_key_udfs;
SET citus.next_shard_id TO 1820000;
SET citus.shard_count TO 32;
SET citus.shard_replication_factor TO 1;
ALTER SEQUENCE pg_catalog.pg_dist_colocationid_seq RESTART 198000;
SET client_min_messages TO ERROR;
SELECT 1 FROM citus_add_node('localhost', :master_port, groupid=>0);
RESET client_min_messages;
-- test some other udf's with single shard tables
CREATE TABLE null_dist_key_table(a int);
SELECT create_distributed_table('null_dist_key_table', null, colocate_with=>'none', distribution_type=>null);
SELECT truncate_local_data_after_distributing_table('null_dist_key_table');
-- should work --
-- insert some data & create an index for table size udf's
INSERT INTO null_dist_key_table VALUES (1), (2), (3);
CREATE INDEX null_dist_key_idx ON null_dist_key_table(a);
SELECT citus_table_size('null_dist_key_table');
SELECT citus_total_relation_size('null_dist_key_table');
SELECT citus_relation_size('null_dist_key_table');
SELECT * FROM pg_catalog.citus_shard_sizes() WHERE table_name LIKE '%null_dist_key_table%';
BEGIN;
SELECT lock_relation_if_exists('null_dist_key_table', 'ACCESS SHARE');
SELECT count(*) FROM pg_locks where relation='null_dist_key_table'::regclass;
COMMIT;
SELECT partmethod, repmodel FROM pg_dist_partition WHERE logicalrelid = 'null_dist_key_table'::regclass;
SELECT master_get_table_ddl_events('null_dist_key_table');
SELECT column_to_column_name(logicalrelid, partkey)
FROM pg_dist_partition WHERE logicalrelid = 'null_dist_key_table'::regclass;
SELECT column_name_to_column('null_dist_key_table', 'a');
SELECT master_update_shard_statistics(shardid)
FROM (SELECT shardid FROM pg_dist_shard WHERE logicalrelid='null_dist_key_table'::regclass) as shardid;
SELECT truncate_local_data_after_distributing_table('null_dist_key_table');
-- should return a single element array that only includes its own shard id
SELECT shardid=unnest(get_colocated_shard_array(shardid))
FROM (SELECT shardid FROM pg_dist_shard WHERE logicalrelid='null_dist_key_table'::regclass) as shardid;
BEGIN;
SELECT master_remove_partition_metadata('null_dist_key_table'::regclass::oid, 'null_dist_key_udfs', 'null_dist_key_table');
-- should print 0
select count(*) from pg_dist_partition where logicalrelid='null_dist_key_table'::regclass;
ROLLBACK;
SELECT master_create_empty_shard('null_dist_key_table');
-- return true
SELECT citus_table_is_visible('null_dist_key_table'::regclass::oid);
-- return false
SELECT relation_is_a_known_shard('null_dist_key_table');
-- return | false | true |
SELECT citus_table_is_visible(tableName::regclass::oid), relation_is_a_known_shard(tableName::regclass)
FROM (SELECT tableName FROM pg_catalog.pg_tables WHERE tablename LIKE 'null_dist_key_table%') as tableName;
-- should fail, maybe support in the future
SELECT create_reference_table('null_dist_key_table');
SELECT create_distributed_table('null_dist_key_table', 'a');
SELECT create_distributed_table_concurrently('null_dist_key_table', 'a');
SELECT citus_add_local_table_to_metadata('null_dist_key_table');
-- test altering distribution column, fails for single shard tables
SELECT alter_distributed_table('null_dist_key_table', distribution_column := 'a');
-- test altering shard count, fails for single shard tables
SELECT alter_distributed_table('null_dist_key_table', shard_count := 6);
-- test shard splitting udf, fails for single shard tables
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
SELECT citus_split_shard_by_split_points(
1820000,
ARRAY['-1073741826'],
ARRAY[:worker_1_node, :worker_2_node],
'block_writes');
SELECT colocationid FROM pg_dist_partition WHERE logicalrelid::text LIKE '%null_dist_key_table%';
-- test alter_table_set_access_method and verify it doesn't change the colocation id
SELECT alter_table_set_access_method('null_dist_key_table', 'columnar');
SELECT colocationid FROM pg_dist_partition WHERE logicalrelid::text LIKE '%null_dist_key_table%';
-- undistribute
SELECT undistribute_table('null_dist_key_table');
-- verify that the metadata is gone
SELECT COUNT(*) = 0 FROM pg_dist_partition WHERE logicalrelid::text LIKE '%null_dist_key_table%';
SELECT COUNT(*) = 0 FROM pg_dist_placement WHERE shardid IN (SELECT shardid FROM pg_dist_shard WHERE logicalrelid::text LIKE '%null_dist_key_table%');
SELECT COUNT(*) = 0 FROM pg_dist_shard WHERE logicalrelid::text LIKE '%null_dist_key_table%';
-- create 7 single shard tables, 3 of them are colocated, for testing shard moves / rebalance on them
CREATE TABLE single_shard_table_col1_1 (a INT PRIMARY KEY);
CREATE TABLE single_shard_table_col1_2 (a TEXT PRIMARY KEY);
CREATE TABLE single_shard_table_col1_3 (a TIMESTAMP PRIMARY KEY);
CREATE TABLE single_shard_table_col2_1 (a INT PRIMARY KEY);
CREATE TABLE single_shard_table_col3_1 (a INT PRIMARY KEY);
CREATE TABLE single_shard_table_col4_1 (a INT PRIMARY KEY);
CREATE TABLE single_shard_table_col5_1 (a INT PRIMARY KEY);
SELECT create_distributed_table('single_shard_table_col1_1', null, colocate_with=>'none');
SELECT create_distributed_table('single_shard_table_col1_2', null, colocate_with=>'single_shard_table_col1_1');
SELECT create_distributed_table('single_shard_table_col1_3', null, colocate_with=>'single_shard_table_col1_2');
SELECT create_distributed_table('single_shard_table_col2_1', null, colocate_with=>'none');
SELECT create_distributed_table('single_shard_table_col3_1', null, colocate_with=>'none');
SELECT create_distributed_table('single_shard_table_col4_1', null, colocate_with=>'none');
SELECT create_distributed_table('single_shard_table_col5_1', null, colocate_with=>'none');
-- initial status
SELECT shardid, nodeport FROM pg_dist_shard_placement WHERE shardid > 1820000 ORDER BY shardid;
-- errors out because streaming replicated
SELECT citus_copy_shard_placement(1820005, 'localhost', :worker_1_port, 'localhost', :worker_2_port);
SELECT master_copy_shard_placement(1820005, 'localhost', :worker_1_port, 'localhost', :worker_2_port);
SELECT citus_copy_shard_placement(1820005, :worker_1_node, :worker_2_node);
-- no changes because it's already balanced
SELECT rebalance_table_shards();
-- same placements
SELECT shardid, nodeport FROM pg_dist_shard_placement WHERE shardid > 1820000 ORDER BY shardid;
-- manually move 2 shard from 2 colocation groups to make the cluster unbalanced
SELECT citus_move_shard_placement(1820005, 'localhost', :worker_1_port, 'localhost', :worker_2_port);
SELECT citus_move_shard_placement(1820007, :worker_1_node, :worker_2_node);
-- all placements are located on worker 2
SELECT shardid, nodeport FROM pg_dist_shard_placement WHERE shardid > 1820000 ORDER BY shardid;
-- move some of them to worker 1 to balance the cluster
SELECT rebalance_table_shards();
-- the final status, balanced
SELECT shardid, nodeport FROM pg_dist_shard_placement WHERE shardid > 1820000 ORDER BY shardid;
-- verify we didn't break any colocations
SELECT logicalrelid, colocationid FROM pg_dist_partition WHERE logicalrelid::text LIKE '%single_shard_table_col%' ORDER BY colocationid;
-- again, manually move 2 shard from 2 colocation groups to make the cluster unbalanced
-- consider using citus_drain_node when the issue is fixed: https://github.com/citusdata/citus/issues/6948
SELECT citus_move_shard_placement(1820005, 'localhost', :worker_1_port, 'localhost', :worker_2_port);
SELECT citus_move_shard_placement(1820003, :worker_1_node, :worker_2_node);
-- see the plan for moving 4 shards, 3 of them are in the same colocation group
SELECT * FROM get_rebalance_table_shards_plan();
-- move some of them to worker 1 to balance the cluster
SELECT 1 FROM citus_rebalance_start();
-- stop it
SELECT * FROM citus_rebalance_stop();
-- show rebalance status, see the cancelled job for two moves
SELECT state, details FROM citus_rebalance_status();
-- start again
SELECT 1 FROM citus_rebalance_start();
-- show rebalance status, scheduled a job for two moves
SELECT state, details FROM citus_rebalance_status();
-- wait for rebalance to be completed
SELECT * FROM citus_rebalance_wait();
-- the final status, balanced
SELECT shardid, nodeport FROM pg_dist_shard_placement WHERE shardid > 1820000 ORDER BY shardid;
-- test update_distributed_table_colocation
CREATE TABLE update_col_1 (a INT);
CREATE TABLE update_col_2 (a INT);
CREATE TABLE update_col_3 (a INT);
-- create colocated single shard distributed tables, so the shards will be
-- in the same worker node
SELECT create_distributed_table ('update_col_1', null, colocate_with:='none');
SELECT create_distributed_table ('update_col_2', null, colocate_with:='update_col_1');
-- now create a third single shard distributed table that is not colocated,
-- with the new colocation id the new table will be in the other worker node
SELECT create_distributed_table ('update_col_3', null, colocate_with:='none');
-- make sure nodes are correct
SELECT c1.nodeport = c2.nodeport AS same_node
FROM citus_shards c1, citus_shards c2, pg_dist_node p1, pg_dist_node p2
WHERE c1.table_name::text = 'update_col_1' AND c2.table_name::text = 'update_col_2' AND
p1.nodeport = c1.nodeport AND p2.nodeport = c2.nodeport AND
p1.noderole = 'primary' AND p2.noderole = 'primary';
SELECT c1.nodeport = c2.nodeport AS same_node
FROM citus_shards c1, citus_shards c2, pg_dist_node p1, pg_dist_node p2
WHERE c1.table_name::text = 'update_col_1' AND c2.table_name::text = 'update_col_3' AND
p1.nodeport = c1.nodeport AND p2.nodeport = c2.nodeport AND
p1.noderole = 'primary' AND p2.noderole = 'primary';
-- and the update_col_1 and update_col_2 are colocated
SELECT c1.colocation_id = c2.colocation_id AS colocated
FROM public.citus_tables c1, public.citus_tables c2
WHERE c1.table_name::text = 'update_col_1' AND c2.table_name::text = 'update_col_2';
-- break the colocation
SELECT update_distributed_table_colocation('update_col_2', colocate_with:='none');
SELECT c1.colocation_id = c2.colocation_id AS colocated
FROM public.citus_tables c1, public.citus_tables c2
WHERE c1.table_name::text = 'update_col_1' AND c2.table_name::text = 'update_col_2';
-- re-colocate, the shards were already in the same node
SELECT update_distributed_table_colocation('update_col_2', colocate_with:='update_col_1');
SELECT c1.colocation_id = c2.colocation_id AS colocated
FROM public.citus_tables c1, public.citus_tables c2
WHERE c1.table_name::text = 'update_col_1' AND c2.table_name::text = 'update_col_2';
-- update_col_1 and update_col_3 are not colocated, because they are not in the some node
SELECT c1.colocation_id = c2.colocation_id AS colocated
FROM public.citus_tables c1, public.citus_tables c2
WHERE c1.table_name::text = 'update_col_1' AND c2.table_name::text = 'update_col_3';
-- they should not be able to be colocated since the shards are in different nodes
SELECT update_distributed_table_colocation('update_col_3', colocate_with:='update_col_1');
SELECT c1.colocation_id = c2.colocation_id AS colocated
FROM public.citus_tables c1, public.citus_tables c2
WHERE c1.table_name::text = 'update_col_1' AND c2.table_name::text = 'update_col_3';
-- hash distributed and single shard distributed tables cannot be colocated
CREATE TABLE update_col_4 (a INT);
SELECT create_distributed_table ('update_col_4', 'a', colocate_with:='none');
SELECT update_distributed_table_colocation('update_col_1', colocate_with:='update_col_4');
SELECT update_distributed_table_colocation('update_col_4', colocate_with:='update_col_1');
-- test columnar UDFs
CREATE TABLE columnar_tbl (a INT) USING COLUMNAR;
SELECT create_distributed_table('columnar_tbl', NULL, colocate_with:='none');
SELECT * FROM columnar.options WHERE relation = 'columnar_tbl'::regclass;
SELECT alter_columnar_table_set('columnar_tbl', compression_level => 2);
SELECT * FROM columnar.options WHERE relation = 'columnar_tbl'::regclass;
SELECT alter_columnar_table_reset('columnar_tbl', compression_level => true);
SELECT * FROM columnar.options WHERE relation = 'columnar_tbl'::regclass;
SELECT columnar_internal.upgrade_columnar_storage(c.oid)
FROM pg_class c, pg_am a
WHERE c.relam = a.oid AND amname = 'columnar' AND relname = 'columnar_tbl';
SELECT columnar_internal.downgrade_columnar_storage(c.oid)
FROM pg_class c, pg_am a
WHERE c.relam = a.oid AND amname = 'columnar' AND relname = 'columnar_tbl';
CREATE OR REPLACE FUNCTION columnar_storage_info(
rel regclass,
version_major OUT int4,
version_minor OUT int4,
storage_id OUT int8,
reserved_stripe_id OUT int8,
reserved_row_number OUT int8,
reserved_offset OUT int8)
STRICT
LANGUAGE c AS 'citus', $$columnar_storage_info$$;
SELECT version_major, version_minor, reserved_stripe_id, reserved_row_number, reserved_offset FROM columnar_storage_info('columnar_tbl');
SELECT columnar.get_storage_id(oid) = storage_id FROM pg_class, columnar_storage_info('columnar_tbl') WHERE relname = 'columnar_tbl';
-- test time series functions
CREATE TABLE part_tbl (a DATE) PARTITION BY RANGE (a);
CREATE TABLE part_tbl_1 PARTITION OF part_tbl FOR VALUES FROM ('2000-01-01') TO ('2010-01-01');
CREATE TABLE part_tbl_2 PARTITION OF part_tbl FOR VALUES FROM ('2020-01-01') TO ('2030-01-01');
SELECT create_distributed_table('part_tbl', NULL, colocate_with:='none');
SELECT * FROM time_partitions WHERE parent_table::text = 'part_tbl';
SELECT time_partition_range('part_tbl_2');
SELECT get_missing_time_partition_ranges('part_tbl', INTERVAL '10 years', '2050-01-01', '2000-01-01');
SELECT create_time_partitions('part_tbl', INTERVAL '10 years', '2050-01-01', '2000-01-01');
CALL drop_old_time_partitions('part_tbl', '2030-01-01');
SELECT * FROM time_partitions WHERE parent_table::text = 'part_tbl';
-- test locking shards
CREATE TABLE lock_tbl_1 (a INT);
SELECT create_distributed_table('lock_tbl_1', NULL, colocate_with:='none');
CREATE TABLE lock_tbl_2 (a INT);
SELECT create_distributed_table('lock_tbl_2', NULL, colocate_with:='none');
BEGIN;
SELECT lock_shard_metadata(3, array_agg(distinct(shardid)))
FROM citus_shards WHERE table_name::text = 'lock_tbl_1';
SELECT lock_shard_metadata(5, array_agg(distinct(shardid)))
FROM citus_shards WHERE table_name::text LIKE 'lock\_tbl\__';
SELECT table_name, classid, mode, granted
FROM pg_locks, public.citus_tables
WHERE
locktype = 'advisory' AND
table_name::text LIKE 'lock\_tbl\__' AND
objid = colocation_id
ORDER BY 1, 3;
END;
BEGIN;
SELECT lock_shard_resources(3, array_agg(distinct(shardid)))
FROM citus_shards WHERE table_name::text = 'lock_tbl_1';
SELECT lock_shard_resources(5, array_agg(distinct(shardid)))
FROM citus_shards WHERE table_name::text LIKE 'lock\_tbl\__';
SELECT locktype, table_name, mode, granted
FROM pg_locks, citus_shards, pg_dist_node
WHERE
objid = shardid AND
table_name::text LIKE 'lock\_tbl\__' AND
citus_shards.nodeport = pg_dist_node.nodeport AND
noderole = 'primary'
ORDER BY 2, 3;
END;
SET client_min_messages TO WARNING;
DROP SCHEMA null_dist_key_udfs CASCADE;