citus/src/test/regress/expected/multi_cluster_management.out

1153 lines
47 KiB
Plaintext

SET citus.next_shard_id TO 1220000;
ALTER SEQUENCE pg_catalog.pg_dist_colocationid_seq RESTART 1390000;
ALTER SEQUENCE pg_catalog.pg_dist_groupid_seq RESTART 1;
-- Tests functions related to cluster membership
-- add the nodes to the cluster
SELECT 1 FROM master_add_node('localhost', :worker_1_port);
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
-- I am coordinator
SELECT citus_is_coordinator();
citus_is_coordinator
---------------------------------------------------------------------
t
(1 row)
-- workers are not coordinator
SELECT result FROM run_command_on_workers('SELECT citus_is_coordinator()');
result
---------------------------------------------------------------------
f
f
(2 rows)
-- get the active nodes
SELECT master_get_active_worker_nodes();
master_get_active_worker_nodes
---------------------------------------------------------------------
(localhost,57638)
(localhost,57637)
(2 rows)
-- try to add a node that is already in the cluster
SELECT * FROM master_add_node('localhost', :worker_1_port);
master_add_node
---------------------------------------------------------------------
1
(1 row)
-- get the active nodes
SELECT master_get_active_worker_nodes();
master_get_active_worker_nodes
---------------------------------------------------------------------
(localhost,57638)
(localhost,57637)
(2 rows)
-- try to remove a node (with no placements)
SELECT master_remove_node('localhost', :worker_2_port);
master_remove_node
---------------------------------------------------------------------
(1 row)
-- verify that the node has been deleted
SELECT master_get_active_worker_nodes();
master_get_active_worker_nodes
---------------------------------------------------------------------
(localhost,57637)
(1 row)
-- try to disable a node with no placements see that node is s=removed
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT citus_disable_node('localhost', :worker_2_port);
citus_disable_node
---------------------------------------------------------------------
(1 row)
SELECT public.wait_until_metadata_sync();
wait_until_metadata_sync
---------------------------------------------------------------------
(1 row)
SELECT master_get_active_worker_nodes();
master_get_active_worker_nodes
---------------------------------------------------------------------
(localhost,57637)
(1 row)
-- add some shard placements to the cluster
SET citus.shard_count TO 16;
SET citus.shard_replication_factor TO 1;
-- test warnings on setting the deprecated guc for replication model
BEGIN;
SET citus.replication_model to 'statement';
NOTICE: Setting citus.replication_model has no effect. Please use citus.shard_replication_factor instead.
DETAIL: Citus determines the replication model based on the replication factor and the replication models of the colocated shards. If a colocated table is present, the replication model is inherited. Otherwise 'streaming' replication is preferred if supported by the replication factor.
ROLLBACK;
-- check that the rebalancer works even if there are no distributed tables
SELECT * FROM get_rebalance_table_shards_plan();
table_name | shardid | shard_size | sourcename | sourceport | targetname | targetport
---------------------------------------------------------------------
(0 rows)
SELECT * FROM rebalance_table_shards();
rebalance_table_shards
---------------------------------------------------------------------
(1 row)
-- TODO: Figure out why this is necessary, rebalance_table_shards shouldn't
-- insert stuff into pg_dist_colocation
TRUNCATE pg_dist_colocation;
SELECT run_command_on_workers('TRUNCATE pg_dist_colocation');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,"TRUNCATE TABLE")
(1 row)
ALTER SEQUENCE pg_catalog.pg_dist_colocationid_seq RESTART 1390000;
SELECT 1 FROM citus_activate_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
CREATE TABLE cluster_management_test (col_1 text, col_2 int);
SELECT create_distributed_table('cluster_management_test', 'col_1', 'hash');
create_distributed_table
---------------------------------------------------------------------
(1 row)
-- see that there are some active placements in the candidate node
SELECT shardid, shardstate, nodename, nodeport FROM pg_dist_shard_placement WHERE nodeport=:worker_2_port;
shardid | shardstate | nodename | nodeport
---------------------------------------------------------------------
1220001 | 1 | localhost | 57638
1220003 | 1 | localhost | 57638
1220005 | 1 | localhost | 57638
1220007 | 1 | localhost | 57638
1220009 | 1 | localhost | 57638
1220011 | 1 | localhost | 57638
1220013 | 1 | localhost | 57638
1220015 | 1 | localhost | 57638
(8 rows)
-- try to remove a node with active placements and see that node removal is failed
SELECT master_remove_node('localhost', :worker_2_port);
ERROR: cannot remove or disable the node localhost:xxxxx because because it contains the only shard placement for shard xxxxx
DETAIL: One of the table(s) that prevents the operation complete successfully is public.cluster_management_test
HINT: To proceed, either drop the tables or use undistribute_table() function to convert them to local tables
SELECT master_get_active_worker_nodes();
master_get_active_worker_nodes
---------------------------------------------------------------------
(localhost,57638)
(localhost,57637)
(2 rows)
-- insert a row so that citus_disable_node() exercises closing connections
CREATE TABLE test_reference_table (y int primary key, name text);
SELECT create_reference_table('test_reference_table');
create_reference_table
---------------------------------------------------------------------
(1 row)
INSERT INTO test_reference_table VALUES (1, '1');
-- try to remove a node with active placements and reference tables
SELECT citus_remove_node('localhost', :worker_2_port);
ERROR: cannot remove or disable the node localhost:xxxxx because because it contains the only shard placement for shard xxxxx
DETAIL: One of the table(s) that prevents the operation complete successfully is public.cluster_management_test
HINT: To proceed, either drop the tables or use undistribute_table() function to convert them to local tables
-- try to disable a node with active placements
-- which should fail because there are some placements
-- which are the only placements for a given shard
SELECT citus_disable_node('localhost', :worker_2_port);
ERROR: cannot remove or disable the node localhost:xxxxx because because it contains the only shard placement for shard xxxxx
DETAIL: One of the table(s) that prevents the operation complete successfully is public.cluster_management_test
HINT: To proceed, either drop the tables or use undistribute_table() function to convert them to local tables
SELECT master_get_active_worker_nodes();
master_get_active_worker_nodes
---------------------------------------------------------------------
(localhost,57638)
(localhost,57637)
(2 rows)
-- try to disable a node which does not exist and see that an error is thrown
SELECT citus_disable_node('localhost.noexist', 2345);
ERROR: node at "localhost.noexist:2345" does not exist
-- drop the table without leaving a shard placement behind (messes up other tests)
SELECT master_activate_node('localhost', :worker_2_port);
master_activate_node
---------------------------------------------------------------------
3
(1 row)
DROP TABLE test_reference_table, cluster_management_test;
-- create users like this so results of community and enterprise are same
SET client_min_messages TO ERROR;
CREATE USER non_super_user;
CREATE USER node_metadata_user;
SELECT 1 FROM run_command_on_workers('CREATE USER node_metadata_user');
?column?
---------------------------------------------------------------------
1
1
(2 rows)
RESET client_min_messages;
GRANT EXECUTE ON FUNCTION master_activate_node(text,int) TO node_metadata_user;
GRANT EXECUTE ON FUNCTION master_add_inactive_node(text,int,int,noderole,name) TO node_metadata_user;
GRANT EXECUTE ON FUNCTION master_add_node(text,int,int,noderole,name) TO node_metadata_user;
GRANT EXECUTE ON FUNCTION master_add_secondary_node(text,int,text,int,name) TO node_metadata_user;
GRANT EXECUTE ON FUNCTION citus_disable_node(text,int,bool) TO node_metadata_user;
GRANT EXECUTE ON FUNCTION citus_disable_node_and_wait(text,int,bool) TO node_metadata_user;
GRANT EXECUTE ON FUNCTION master_remove_node(text,int) TO node_metadata_user;
GRANT EXECUTE ON FUNCTION master_update_node(int,text,int,bool,int) TO node_metadata_user;
-- user needs permission for the pg_dist_node and pg_dist_local_group for metadata syncing
SELECT run_command_on_workers('GRANT ALL ON pg_dist_node TO node_metadata_user');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,GRANT)
(localhost,57638,t,GRANT)
(2 rows)
SELECT run_command_on_workers('GRANT ALL ON pg_dist_local_group TO node_metadata_user');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,GRANT)
(localhost,57638,t,GRANT)
(2 rows)
SELECT run_command_on_workers('GRANT ALL ON ALL TABLES IN SCHEMA citus TO node_metadata_user');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,GRANT)
(localhost,57638,t,GRANT)
(2 rows)
SELECT run_command_on_workers('GRANT ALL ON SCHEMA citus TO node_metadata_user');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,GRANT)
(localhost,57638,t,GRANT)
(2 rows)
SELECT master_remove_node('localhost', :worker_2_port);
master_remove_node
---------------------------------------------------------------------
(1 row)
-- Removing public schema from pg_dist_object because it breaks the next tests
DELETE FROM pg_catalog.pg_dist_object WHERE objid = 'public'::regnamespace::oid;
DELETE FROM pg_catalog.pg_dist_object WHERE objid = (SELECT oid FROM pg_extension WHERE extname = 'plpgsql');
-- try to manipulate node metadata via non-super user
SET ROLE non_super_user;
SELECT 1 FROM master_add_inactive_node('localhost', :worker_2_port + 1);
ERROR: permission denied for function master_add_inactive_node
SELECT 1 FROM master_activate_node('localhost', :worker_2_port + 1);
ERROR: permission denied for function master_activate_node
SELECT 1 FROM citus_disable_node('localhost', :worker_2_port + 1);
ERROR: permission denied for function citus_disable_node
SELECT 1 FROM master_remove_node('localhost', :worker_2_port + 1);
ERROR: permission denied for function master_remove_node
SELECT 1 FROM master_add_node('localhost', :worker_2_port + 1);
ERROR: permission denied for function master_add_node
SELECT 1 FROM master_add_secondary_node('localhost', :worker_2_port + 2, 'localhost', :worker_2_port);
ERROR: permission denied for function master_add_secondary_node
SELECT master_update_node(nodeid, 'localhost', :worker_2_port + 3) FROM pg_dist_node WHERE nodeport = :worker_2_port;
ERROR: permission denied for function master_update_node
-- try to manipulate node metadata via privileged user
SET ROLE node_metadata_user;
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
ERROR: operation is not allowed
HINT: Run the command with a superuser.
BEGIN;
SELECT 1 FROM master_add_inactive_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT 1 FROM master_remove_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT 1 FROM master_add_secondary_node('localhost', :worker_2_port + 2, 'localhost', :worker_1_port);
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT master_update_node(nodeid, 'localhost', :worker_2_port + 3) FROM pg_dist_node WHERE nodeport = :worker_2_port;
master_update_node
---------------------------------------------------------------------
(0 rows)
SELECT nodename, nodeport, noderole FROM pg_dist_node ORDER BY nodeport;
nodename | nodeport | noderole
---------------------------------------------------------------------
localhost | 57637 | primary
localhost | 57640 | secondary
(2 rows)
ABORT;
\c - postgres - :master_port
SET citus.next_shard_id TO 1220000;
SET citus.shard_count TO 16;
SET citus.shard_replication_factor TO 1;
SELECT master_get_active_worker_nodes();
master_get_active_worker_nodes
---------------------------------------------------------------------
(localhost,57637)
(1 row)
-- restore the node for next tests
SELECT * FROM master_add_node('localhost', :worker_2_port);
master_add_node
---------------------------------------------------------------------
7
(1 row)
ALTER SEQUENCE pg_dist_node_nodeid_seq RESTART WITH 7;
ALTER SEQUENCE pg_dist_groupid_seq RESTART WITH 6;
CREATE TABLE cluster_management_test (col_1 text, col_2 int);
SELECT create_distributed_table('cluster_management_test', 'col_1', 'hash');
create_distributed_table
---------------------------------------------------------------------
(1 row)
-- try to remove a node with active placements and see that node removal is failed
SELECT master_remove_node('localhost', :worker_2_port);
ERROR: cannot remove or disable the node localhost:xxxxx because because it contains the only shard placement for shard xxxxx
DETAIL: One of the table(s) that prevents the operation complete successfully is public.cluster_management_test
HINT: To proceed, either drop the tables or use undistribute_table() function to convert them to local tables
-- mark all placements in the candidate node as inactive
SELECT groupid AS worker_2_group FROM pg_dist_node WHERE nodeport=:worker_2_port \gset
UPDATE pg_dist_placement SET shardstate=3 WHERE groupid=:worker_2_group;
-- manual updates to pg_dist* tables are not automatically reflected to the workers, so we manually do that too
SELECT run_command_on_workers('UPDATE pg_dist_placement SET shardstate=3 WHERE groupid=' || :'worker_2_group');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,"UPDATE 8")
(localhost,57638,t,"UPDATE 8")
(2 rows)
SELECT shardid, shardstate, nodename, nodeport FROM pg_dist_shard_placement WHERE nodeport=:worker_2_port;
shardid | shardstate | nodename | nodeport
---------------------------------------------------------------------
1220001 | 3 | localhost | 57638
1220003 | 3 | localhost | 57638
1220005 | 3 | localhost | 57638
1220007 | 3 | localhost | 57638
1220009 | 3 | localhost | 57638
1220011 | 3 | localhost | 57638
1220013 | 3 | localhost | 57638
1220015 | 3 | localhost | 57638
(8 rows)
-- try to remove a node with only inactive placements and see that removal still fails
SELECT master_remove_node('localhost', :worker_2_port);
ERROR: cannot remove or disable the node localhost:xxxxx because because it contains the only shard placement for shard xxxxx
DETAIL: One of the table(s) that prevents the operation complete successfully is public.cluster_management_test
HINT: To proceed, either drop the tables or use undistribute_table() function to convert them to local tables
SELECT master_get_active_worker_nodes();
master_get_active_worker_nodes
---------------------------------------------------------------------
(localhost,57638)
(localhost,57637)
(2 rows)
-- mark all placements in the candidate node as to be deleted
UPDATE pg_dist_placement SET shardstate=4 WHERE groupid=:worker_2_group;
SELECT run_command_on_workers('UPDATE pg_dist_placement SET shardstate=4 WHERE groupid=' || :'worker_2_group');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,"UPDATE 8")
(localhost,57638,t,"UPDATE 8")
(2 rows)
SELECT shardid, shardstate, nodename, nodeport FROM pg_dist_shard_placement WHERE nodeport=:worker_2_port;
shardid | shardstate | nodename | nodeport
---------------------------------------------------------------------
1220001 | 4 | localhost | 57638
1220003 | 4 | localhost | 57638
1220005 | 4 | localhost | 57638
1220007 | 4 | localhost | 57638
1220009 | 4 | localhost | 57638
1220011 | 4 | localhost | 57638
1220013 | 4 | localhost | 57638
1220015 | 4 | localhost | 57638
(8 rows)
CREATE TABLE cluster_management_test_colocated (col_1 text, col_2 int);
-- Check that we warn the user about colocated shards that will not get created for shards that do not have active placements
SELECT create_distributed_table('cluster_management_test_colocated', 'col_1', 'hash', colocate_with => 'cluster_management_test');
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220019
WARNING: could not find any shard placements for shardId 1220021
WARNING: could not find any shard placements for shardId 1220023
WARNING: could not find any shard placements for shardId 1220025
WARNING: could not find any shard placements for shardId 1220027
WARNING: could not find any shard placements for shardId 1220029
WARNING: could not find any shard placements for shardId 1220031
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
WARNING: could not find any shard placements for shardId 1220017
create_distributed_table
---------------------------------------------------------------------
(1 row)
-- Check that colocated shards don't get created for shards that are to be deleted
SELECT logicalrelid, shardid, shardstate, nodename, nodeport FROM pg_dist_shard_placement NATURAL JOIN pg_dist_shard ORDER BY shardstate, shardid;
logicalrelid | shardid | shardstate | nodename | nodeport
---------------------------------------------------------------------
cluster_management_test | 1220000 | 1 | localhost | 57637
cluster_management_test | 1220002 | 1 | localhost | 57637
cluster_management_test | 1220004 | 1 | localhost | 57637
cluster_management_test | 1220006 | 1 | localhost | 57637
cluster_management_test | 1220008 | 1 | localhost | 57637
cluster_management_test | 1220010 | 1 | localhost | 57637
cluster_management_test | 1220012 | 1 | localhost | 57637
cluster_management_test | 1220014 | 1 | localhost | 57637
cluster_management_test_colocated | 1220016 | 1 | localhost | 57637
cluster_management_test_colocated | 1220018 | 1 | localhost | 57637
cluster_management_test_colocated | 1220020 | 1 | localhost | 57637
cluster_management_test_colocated | 1220022 | 1 | localhost | 57637
cluster_management_test_colocated | 1220024 | 1 | localhost | 57637
cluster_management_test_colocated | 1220026 | 1 | localhost | 57637
cluster_management_test_colocated | 1220028 | 1 | localhost | 57637
cluster_management_test_colocated | 1220030 | 1 | localhost | 57637
cluster_management_test | 1220001 | 4 | localhost | 57638
cluster_management_test | 1220003 | 4 | localhost | 57638
cluster_management_test | 1220005 | 4 | localhost | 57638
cluster_management_test | 1220007 | 4 | localhost | 57638
cluster_management_test | 1220009 | 4 | localhost | 57638
cluster_management_test | 1220011 | 4 | localhost | 57638
cluster_management_test | 1220013 | 4 | localhost | 57638
cluster_management_test | 1220015 | 4 | localhost | 57638
(24 rows)
-- clean-up
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
UPDATE pg_dist_placement SET shardstate=1 WHERE groupid=:worker_2_group;
SELECT run_command_on_workers('UPDATE pg_dist_placement SET shardstate=1 WHERE groupid=' || :'worker_2_group');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,"UPDATE 8")
(localhost,57638,t,"UPDATE 8")
(2 rows)
SET client_min_messages TO ERROR;
DROP TABLE cluster_management_test_colocated;
RESET client_min_messages;
-- when there is no primary we should get a pretty error
UPDATE pg_dist_node SET noderole = 'secondary' WHERE nodeport=:worker_2_port;
SELECT * FROM cluster_management_test;
ERROR: node group 6 does not have a primary node
-- when there is no node at all in the group we should get a different error
DELETE FROM pg_dist_node WHERE nodeport=:worker_2_port;
SELECT run_command_on_workers('DELETE FROM pg_dist_node WHERE nodeport=' || :'worker_2_port');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,"DELETE 1")
(1 row)
SELECT * FROM cluster_management_test;
ERROR: there is a shard placement in node group 6 but there are no nodes in that group
-- clean-up
SELECT * INTO old_placements FROM pg_dist_placement WHERE groupid = :worker_2_group;
DELETE FROM pg_dist_placement WHERE groupid = :worker_2_group;
SELECT master_add_node('localhost', :worker_2_port) AS new_node \gset
WARNING: could not find any shard placements for shardId 1220001
WARNING: could not find any shard placements for shardId 1220001
WARNING: could not find any shard placements for shardId 1220003
WARNING: could not find any shard placements for shardId 1220005
WARNING: could not find any shard placements for shardId 1220007
WARNING: could not find any shard placements for shardId 1220009
WARNING: could not find any shard placements for shardId 1220011
WARNING: could not find any shard placements for shardId 1220013
WARNING: could not find any shard placements for shardId 1220015
INSERT INTO pg_dist_placement SELECT * FROM old_placements;
SELECT groupid AS new_group FROM pg_dist_node WHERE nodeid = :new_node \gset
UPDATE pg_dist_placement SET groupid = :new_group WHERE groupid = :worker_2_group;
SELECT run_command_on_workers('UPDATE pg_dist_placement SET groupid = ' || :'new_group' || ' WHERE groupid = ' || :'worker_2_group');
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,"UPDATE 8")
(localhost,57638,t,"UPDATE 0")
(2 rows)
SELECT start_metadata_sync_to_node('localhost', :worker_2_port);
start_metadata_sync_to_node
---------------------------------------------------------------------
(1 row)
-- test that you are allowed to remove secondary nodes even if there are placements
SELECT 1 FROM master_add_node('localhost', 9990, groupid => :new_group, noderole => 'secondary');
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT master_remove_node('localhost', :worker_2_port);
ERROR: cannot remove or disable the node localhost:xxxxx because because it contains the only shard placement for shard xxxxx
DETAIL: One of the table(s) that prevents the operation complete successfully is public.cluster_management_test
HINT: To proceed, either drop the tables or use undistribute_table() function to convert them to local tables
SELECT master_remove_node('localhost', 9990);
master_remove_node
---------------------------------------------------------------------
(1 row)
-- clean-up
DROP TABLE cluster_management_test;
-- check that adding/removing nodes are propagated to nodes with metadata
SELECT master_remove_node('localhost', :worker_2_port);
master_remove_node
---------------------------------------------------------------------
(1 row)
SELECT start_metadata_sync_to_node('localhost', :worker_1_port);
start_metadata_sync_to_node
---------------------------------------------------------------------
(1 row)
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
\c - - - :worker_1_port
SELECT nodename, nodeport FROM pg_dist_node WHERE nodename='localhost' AND nodeport=:worker_2_port;
nodename | nodeport
---------------------------------------------------------------------
localhost | 57638
(1 row)
\c - - - :master_port
SELECT master_remove_node('localhost', :worker_2_port);
master_remove_node
---------------------------------------------------------------------
(1 row)
\c - - - :worker_1_port
SELECT nodename, nodeport FROM pg_dist_node WHERE nodename='localhost' AND nodeport=:worker_2_port;
nodename | nodeport
---------------------------------------------------------------------
(0 rows)
\c - - - :master_port
-- check that added nodes are not propagated to nodes without metadata
SELECT stop_metadata_sync_to_node('localhost', :worker_1_port);
NOTICE: dropping metadata on the node (localhost,57637)
stop_metadata_sync_to_node
---------------------------------------------------------------------
(1 row)
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
\c - - - :worker_1_port
SELECT nodename, nodeport FROM pg_dist_node WHERE nodename='localhost' AND nodeport=:worker_2_port;
nodename | nodeport
---------------------------------------------------------------------
(0 rows)
\c - - - :master_port
-- check that removing two nodes in the same transaction works
SELECT
master_remove_node('localhost', :worker_1_port),
master_remove_node('localhost', :worker_2_port);
master_remove_node | master_remove_node
---------------------------------------------------------------------
|
(1 row)
SELECT count(1) FROM pg_dist_node;
count
---------------------------------------------------------------------
0
(1 row)
-- check that adding two nodes in the same transaction works
SELECT
master_add_node('localhost', :worker_1_port),
master_add_node('localhost', :worker_2_port);
master_add_node | master_add_node
---------------------------------------------------------------------
11 | 12
(1 row)
SELECT * FROM pg_dist_node ORDER BY nodeid;
nodeid | groupid | nodename | nodeport | noderack | hasmetadata | isactive | noderole | nodecluster | metadatasynced | shouldhaveshards
---------------------------------------------------------------------
11 | 9 | localhost | 57637 | default | t | t | primary | default | t | t
12 | 10 | localhost | 57638 | default | t | t | primary | default | t | t
(2 rows)
-- check that mixed add/remove node commands work fine inside transaction
BEGIN;
SELECT master_remove_node('localhost', :worker_2_port);
master_remove_node
---------------------------------------------------------------------
(1 row)
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT master_remove_node('localhost', :worker_2_port);
master_remove_node
---------------------------------------------------------------------
(1 row)
COMMIT;
SELECT nodename, nodeport FROM pg_dist_node WHERE nodename='localhost' AND nodeport=:worker_2_port;
nodename | nodeport
---------------------------------------------------------------------
(0 rows)
SELECT start_metadata_sync_to_node('localhost', :worker_1_port);
start_metadata_sync_to_node
---------------------------------------------------------------------
(1 row)
BEGIN;
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT master_remove_node('localhost', :worker_2_port);
master_remove_node
---------------------------------------------------------------------
(1 row)
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
COMMIT;
SELECT nodename, nodeport FROM pg_dist_node WHERE nodename='localhost' AND nodeport=:worker_2_port;
nodename | nodeport
---------------------------------------------------------------------
localhost | 57638
(1 row)
\c - - - :worker_1_port
SELECT nodename, nodeport FROM pg_dist_node WHERE nodename='localhost' AND nodeport=:worker_2_port;
nodename | nodeport
---------------------------------------------------------------------
localhost | 57638
(1 row)
\c - - - :master_port
SELECT master_remove_node(nodename, nodeport) FROM pg_dist_node;
master_remove_node
---------------------------------------------------------------------
(2 rows)
SELECT 1 FROM master_add_node('localhost', :worker_1_port);
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
-- check that a distributed table can be created after adding a node in a transaction
SET citus.shard_count TO 4;
SELECT master_remove_node('localhost', :worker_2_port);
master_remove_node
---------------------------------------------------------------------
(1 row)
BEGIN;
SELECT 1 FROM master_add_node('localhost', :worker_2_port);
?column?
---------------------------------------------------------------------
1
(1 row)
CREATE TABLE temp(col1 text, col2 int);
SELECT create_distributed_table('temp', 'col1');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO temp VALUES ('row1', 1);
INSERT INTO temp VALUES ('row2', 2);
COMMIT;
SELECT col1, col2 FROM temp ORDER BY col1;
col1 | col2
---------------------------------------------------------------------
row1 | 1
row2 | 2
(2 rows)
SELECT
count(*)
FROM
pg_dist_shard_placement, pg_dist_shard
WHERE
pg_dist_shard_placement.shardid = pg_dist_shard.shardid
AND pg_dist_shard.logicalrelid = 'temp'::regclass
AND pg_dist_shard_placement.nodeport = :worker_2_port;
count
---------------------------------------------------------------------
4
(1 row)
DROP TABLE temp;
\c - - - :worker_1_port
DELETE FROM pg_dist_partition;
DELETE FROM pg_dist_shard;
DELETE FROM pg_dist_placement;
DELETE FROM pg_dist_node;
\c - - - :master_port
SELECT stop_metadata_sync_to_node('localhost', :worker_1_port);
NOTICE: dropping metadata on the node (localhost,57637)
stop_metadata_sync_to_node
---------------------------------------------------------------------
(1 row)
SELECT stop_metadata_sync_to_node('localhost', :worker_2_port);
NOTICE: dropping metadata on the node (localhost,57638)
stop_metadata_sync_to_node
---------------------------------------------------------------------
(1 row)
-- check that you can't add a primary to a non-default cluster
SELECT master_add_node('localhost', 9999, nodecluster => 'olap');
ERROR: primaries must be added to the default cluster
-- check that you can't add more than one primary to a group
SELECT groupid AS worker_1_group FROM pg_dist_node WHERE nodeport = :worker_1_port \gset
SELECT master_add_node('localhost', 9999, groupid => :worker_1_group, noderole => 'primary');
ERROR: group 14 already has a primary node
-- check that you can add secondaries and unavailable nodes to a group
SELECT groupid AS worker_2_group FROM pg_dist_node WHERE nodeport = :worker_2_port \gset
SELECT 1 FROM master_add_node('localhost', 9998, groupid => :worker_1_group, noderole => 'secondary');
?column?
---------------------------------------------------------------------
1
(1 row)
SELECT 1 FROM master_add_node('localhost', 9997, groupid => :worker_1_group, noderole => 'unavailable');
?column?
---------------------------------------------------------------------
1
(1 row)
-- add_inactive_node also works with secondaries
SELECT 1 FROM master_add_inactive_node('localhost', 9996, groupid => :worker_2_group, noderole => 'secondary');
?column?
---------------------------------------------------------------------
1
(1 row)
-- check that you can add a seconary to a non-default cluster, and activate it, and remove it
SELECT master_add_inactive_node('localhost', 9999, groupid => :worker_2_group, nodecluster => 'olap', noderole => 'secondary');
master_add_inactive_node
---------------------------------------------------------------------
22
(1 row)
SELECT master_activate_node('localhost', 9999);
master_activate_node
---------------------------------------------------------------------
22
(1 row)
SELECT citus_disable_node('localhost', 9999);
citus_disable_node
---------------------------------------------------------------------
(1 row)
SELECT public.wait_until_metadata_sync();
wait_until_metadata_sync
---------------------------------------------------------------------
(1 row)
SELECT master_remove_node('localhost', 9999);
master_remove_node
---------------------------------------------------------------------
(1 row)
-- check that you can't manually add two primaries to a group
INSERT INTO pg_dist_node (nodename, nodeport, groupid, noderole)
VALUES ('localhost', 5000, :worker_1_group, 'primary');
ERROR: there cannot be two primary nodes in a group
CONTEXT: PL/pgSQL function citus_internal.pg_dist_node_trigger_func() line XX at RAISE
UPDATE pg_dist_node SET noderole = 'primary'
WHERE groupid = :worker_1_group AND nodeport = 9998;
ERROR: there cannot be two primary nodes in a group
CONTEXT: PL/pgSQL function citus_internal.pg_dist_node_trigger_func() line XX at RAISE
-- check that you can't manually add a primary to a non-default cluster
INSERT INTO pg_dist_node (nodename, nodeport, groupid, noderole, nodecluster)
VALUES ('localhost', 5000, 1000, 'primary', 'olap');
ERROR: new row for relation "pg_dist_node" violates check constraint "primaries_are_only_allowed_in_the_default_cluster"
DETAIL: Failing row contains (24, 1000, localhost, 5000, default, f, t, primary, olap, f, t).
UPDATE pg_dist_node SET nodecluster = 'olap'
WHERE nodeport = :worker_1_port;
ERROR: new row for relation "pg_dist_node" violates check constraint "primaries_are_only_allowed_in_the_default_cluster"
DETAIL: Failing row contains (16, 14, localhost, 57637, default, f, t, primary, olap, f, t).
-- check that you /can/ add a secondary node to a non-default cluster
SELECT groupid AS worker_2_group FROM pg_dist_node WHERE nodeport = :worker_2_port \gset
SELECT master_add_node('localhost', 8888, groupid => :worker_1_group, noderole => 'secondary', nodecluster=> 'olap');
master_add_node
---------------------------------------------------------------------
25
(1 row)
-- check that super-long cluster names are truncated
SELECT master_add_node('localhost', 8887, groupid => :worker_1_group, noderole => 'secondary', nodecluster=>
'thisisasixtyfourcharacterstringrepeatedfourtimestomake256chars.'
'thisisasixtyfourcharacterstringrepeatedfourtimestomake256chars.'
'thisisasixtyfourcharacterstringrepeatedfourtimestomake256chars.'
'thisisasixtyfourcharacterstringrepeatedfourtimestomake256chars.'
'overflow'
);
master_add_node
---------------------------------------------------------------------
26
(1 row)
SELECT * FROM pg_dist_node WHERE nodeport=8887;
nodeid | groupid | nodename | nodeport | noderack | hasmetadata | isactive | noderole | nodecluster | metadatasynced | shouldhaveshards
---------------------------------------------------------------------
26 | 14 | localhost | 8887 | default | f | t | secondary | thisisasixtyfourcharacterstringrepeatedfourtimestomake256chars. | f | t
(1 row)
-- don't remove the secondary and unavailable nodes, check that no commands are sent to
-- them in any of the remaining tests
-- master_add_secondary_node lets you skip looking up the groupid
SELECT master_add_secondary_node('localhost', 9995, 'localhost', :worker_1_port);
master_add_secondary_node
---------------------------------------------------------------------
27
(1 row)
SELECT master_add_secondary_node('localhost', 9994, primaryname => 'localhost', primaryport => :worker_2_port);
master_add_secondary_node
---------------------------------------------------------------------
28
(1 row)
SELECT master_add_secondary_node('localhost', 9993, 'localhost', 2000);
ERROR: node at "localhost:xxxxx" does not exist
SELECT master_add_secondary_node('localhost', 9992, 'localhost', :worker_1_port, nodecluster => 'second-cluster');
master_add_secondary_node
---------------------------------------------------------------------
29
(1 row)
SELECT nodeid AS worker_1_node FROM pg_dist_node WHERE nodeport=:worker_1_port \gset
-- master_update_node checks node exists
SELECT master_update_node(100, 'localhost', 8000);
ERROR: node 100 not found
-- master_update_node disallows aliasing existing node
SELECT master_update_node(:worker_1_node, 'localhost', :worker_2_port);
ERROR: there is already another node with the specified hostname and port
-- master_update_node moves a node
SELECT master_update_node(:worker_1_node, 'somehost', 9000);
master_update_node
---------------------------------------------------------------------
(1 row)
SELECT * FROM pg_dist_node WHERE nodeid = :worker_1_node;
nodeid | groupid | nodename | nodeport | noderack | hasmetadata | isactive | noderole | nodecluster | metadatasynced | shouldhaveshards
---------------------------------------------------------------------
16 | 14 | somehost | 9000 | default | f | t | primary | default | f | t
(1 row)
-- cleanup
SELECT master_update_node(:worker_1_node, 'localhost', :worker_1_port);
master_update_node
---------------------------------------------------------------------
(1 row)
SELECT * FROM pg_dist_node WHERE nodeid = :worker_1_node;
nodeid | groupid | nodename | nodeport | noderack | hasmetadata | isactive | noderole | nodecluster | metadatasynced | shouldhaveshards
---------------------------------------------------------------------
16 | 14 | localhost | 57637 | default | f | t | primary | default | f | t
(1 row)
SET client_min_messages TO ERROR;
SELECT start_metadata_sync_to_node(nodename, nodeport) FROM pg_dist_node WHERE isactive = 't' and noderole = 'primary';
start_metadata_sync_to_node
---------------------------------------------------------------------
(2 rows)
RESET client_min_messages;
SET citus.shard_replication_factor TO 1;
CREATE TABLE test_dist (x int, y int);
SELECT create_distributed_table('test_dist', 'x');
create_distributed_table
---------------------------------------------------------------------
(1 row)
-- testing behaviour when setting shouldhaveshards to false on partially empty node
SELECT * from master_set_node_property('localhost', :worker_2_port, 'shouldhaveshards', false);
master_set_node_property
---------------------------------------------------------------------
(1 row)
CREATE TABLE test_dist_colocated (x int, y int);
CREATE TABLE test_dist_non_colocated (x int, y int);
CREATE TABLE test_dist_colocated_with_non_colocated (x int, y int);
CREATE TABLE test_ref (a int, b int);
SELECT create_distributed_table('test_dist_colocated', 'x');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('test_dist_non_colocated', 'x', colocate_with => 'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('test_dist_colocated_with_non_colocated', 'x', colocate_with => 'test_dist_non_colocated');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_reference_table('test_ref');
create_reference_table
---------------------------------------------------------------------
(1 row)
-- colocated tables should still be placed on shouldhaveshards false nodes for safety
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_dist_colocated'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 2
57638 | 2
(2 rows)
-- non colocated tables should not be placed on shouldhaveshards false nodes anymore
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_dist_non_colocated'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 4
(1 row)
-- this table should be colocated with the test_dist_non_colocated table
-- correctly only on nodes with shouldhaveshards true
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_dist_colocated_with_non_colocated'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 4
(1 row)
-- reference tables should be placed on with shouldhaveshards false
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_ref'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 1
57638 | 1
(2 rows)
-- cleanup for next test
DROP TABLE test_dist, test_ref, test_dist_colocated, test_dist_non_colocated, test_dist_colocated_with_non_colocated;
-- testing behaviour when setting shouldhaveshards to false on fully empty node
SELECT * from master_set_node_property('localhost', :worker_2_port, 'shouldhaveshards', false);
master_set_node_property
---------------------------------------------------------------------
(1 row)
CREATE TABLE test_dist (x int, y int);
CREATE TABLE test_dist_colocated (x int, y int);
CREATE TABLE test_dist_non_colocated (x int, y int);
CREATE TABLE test_ref (a int, b int);
SELECT create_distributed_table('test_dist', 'x');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_reference_table('test_ref');
create_reference_table
---------------------------------------------------------------------
(1 row)
-- distributed tables should not be placed on nodes with shouldhaveshards false
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_dist'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 4
(1 row)
-- reference tables should be placed on nodes with shouldhaveshards false
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_ref'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 1
57638 | 1
(2 rows)
SELECT * from master_set_node_property('localhost', :worker_2_port, 'shouldhaveshards', true);
master_set_node_property
---------------------------------------------------------------------
(1 row)
-- distributed tables should still not be placed on nodes that were switched to
-- shouldhaveshards true
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_dist'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 4
(1 row)
-- reference tables should still be placed on all nodes with isdatanode 'true'
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_ref'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 1
57638 | 1
(2 rows)
SELECT create_distributed_table('test_dist_colocated', 'x');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('test_dist_non_colocated', 'x', colocate_with => 'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
-- colocated tables should not be placed on nodedes that were switched to
-- shouldhaveshards true
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_dist_colocated'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 4
(1 row)
-- non colocated tables should be placed on nodedes that were switched to
-- shouldhaveshards true
SELECT nodeport, count(*)
FROM pg_dist_shard JOIN pg_dist_shard_placement USING (shardid)
WHERE logicalrelid = 'test_dist_non_colocated'::regclass GROUP BY nodeport ORDER BY nodeport;
nodeport | count
---------------------------------------------------------------------
57637 | 2
57638 | 2
(2 rows)
SELECT * from master_set_node_property('localhost', :worker_2_port, 'bogusproperty', false);
ERROR: only the 'shouldhaveshards' property can be set using this function
DROP TABLE test_dist, test_ref, test_dist_colocated, test_dist_non_colocated;
-- verify that at the end of this file, all primary nodes have metadata synced
SELECT bool_and(hasmetadata) AND bool_and(metadatasynced) FROM pg_dist_node WHERE isactive = 't' and noderole = 'primary';
?column?
---------------------------------------------------------------------
t
(1 row)