-- -- failure_mx_metadata_sync.sql -- CREATE SCHEMA IF NOT EXISTS mx_metadata_sync; SET SEARCH_PATH = mx_metadata_sync; SET citus.shard_count TO 2; SET citus.next_shard_id TO 16000000; SET citus.shard_replication_factor TO 1; SELECT pg_backend_pid() as pid \gset SELECT citus.mitmproxy('conn.allow()'); mitmproxy --------------------------------------------------------------------- (1 row) \set VERBOSITY terse SET client_min_messages TO ERROR; CREATE TABLE t1 (id int PRIMARY KEY); SELECT create_distributed_table('t1', 'id'); create_distributed_table --------------------------------------------------------------------- (1 row) INSERT INTO t1 SELECT x FROM generate_series(1,100) AS f(x); -- Initially turn metadata sync off because we'll ingest errors to start/stop metadata sync operations SELECT stop_metadata_sync_to_node('localhost', :worker_2_proxy_port); stop_metadata_sync_to_node --------------------------------------------------------------------- (1 row) SELECT hasmetadata FROM pg_dist_node WHERE nodeport=:worker_2_proxy_port; hasmetadata --------------------------------------------------------------------- f (1 row) -- Failure to set groupid in the worker SELECT citus.mitmproxy('conn.onQuery(query="^UPDATE pg_dist_local_group SET groupid").cancel(' || :pid || ')'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT citus_activate_node('localhost', :worker_2_proxy_port); ERROR: canceling statement due to user request SELECT citus.mitmproxy('conn.onQuery(query="^UPDATE pg_dist_local_group SET groupid").kill()'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT citus_activate_node('localhost', :worker_2_proxy_port); ERROR: failure on connection marked as essential: localhost:xxxxx -- Failure to drop all tables in pg_dist_partition SELECT citus.mitmproxy('conn.onQuery(query="DELETE FROM pg_dist_partition").cancel(' || :pid || ')'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT citus_activate_node('localhost', :worker_2_proxy_port); ERROR: canceling statement due to user request SELECT citus.mitmproxy('conn.onQuery(query="DELETE FROM pg_dist_partition").kill()'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT citus_activate_node('localhost', :worker_2_proxy_port); ERROR: failure on connection marked as essential: localhost:xxxxx -- Failure to delete pg_dist_node entries from the worker SELECT citus.mitmproxy('conn.onQuery(query="DELETE FROM pg_dist_node").cancel(' || :pid || ')'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT start_metadata_sync_to_node('localhost', :worker_2_proxy_port); ERROR: canceling statement due to user request SELECT citus.mitmproxy('conn.onQuery(query="DELETE FROM pg_dist_node").kill()'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT start_metadata_sync_to_node('localhost', :worker_2_proxy_port); ERROR: failure on connection marked as essential: localhost:xxxxx -- Failure to populate pg_dist_node in the worker SELECT citus.mitmproxy('conn.onQuery(query="INSERT INTO pg_dist_node").cancel(' || :pid || ')'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT start_metadata_sync_to_node('localhost', :worker_2_proxy_port); ERROR: canceling statement due to user request SELECT citus.mitmproxy('conn.onQuery(query="INSERT INTO pg_dist_node").kill()'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT start_metadata_sync_to_node('localhost', :worker_2_proxy_port); ERROR: failure on connection marked as essential: localhost:xxxxx -- Verify that coordinator knows worker does not have valid metadata SELECT hasmetadata FROM pg_dist_node WHERE nodeport=:worker_2_proxy_port; hasmetadata --------------------------------------------------------------------- f (1 row) -- Verify we can activate node after unsuccessful attempts SELECT citus.mitmproxy('conn.allow()'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT 1 FROM citus_activate_node('localhost', :worker_2_proxy_port); ?column? --------------------------------------------------------------------- 1 (1 row) SELECT hasmetadata FROM pg_dist_node WHERE nodeport=:worker_2_proxy_port; hasmetadata --------------------------------------------------------------------- t (1 row) -- Check failures on DDL command propagation CREATE TABLE t2 (id int PRIMARY KEY); SELECT citus.mitmproxy('conn.onParse(query="citus_internal_add_placement_metadata").kill()'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT create_distributed_table('t2', 'id'); ERROR: server closed the connection unexpectedly This probably means the server terminated abnormally before or while processing the request. SELECT citus.mitmproxy('conn.onParse(query="citus_internal_add_shard_metadata").cancel(' || :pid || ')'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT create_distributed_table('t2', 'id'); ERROR: canceling statement due to user request -- Verify that the table was not distributed SELECT count(*) > 0 AS is_table_distributed FROM pg_dist_partition WHERE logicalrelid='t2'::regclass; is_table_distributed --------------------------------------------------------------------- f (1 row) -- Failure to set groupid in the worker SELECT citus.mitmproxy('conn.onQuery(query="^UPDATE pg_dist_local_group SET groupid").cancel(' || :pid || ')'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT stop_metadata_sync_to_node('localhost', :worker_2_proxy_port); ERROR: canceling statement due to user request SELECT citus.mitmproxy('conn.onQuery(query="^UPDATE pg_dist_local_group SET groupid").kill()'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT stop_metadata_sync_to_node('localhost', :worker_2_proxy_port); stop_metadata_sync_to_node --------------------------------------------------------------------- (1 row) -- Failure to delete pg_dist_node entries from the worker SELECT citus.mitmproxy('conn.onQuery(query="DELETE FROM pg_dist_node").cancel(' || :pid || ')'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT stop_metadata_sync_to_node('localhost', :worker_2_proxy_port); ERROR: canceling statement due to user request SELECT citus.mitmproxy('conn.onQuery(query="DELETE FROM pg_dist_node").kill()'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT stop_metadata_sync_to_node('localhost', :worker_2_proxy_port); stop_metadata_sync_to_node --------------------------------------------------------------------- (1 row) \c - - - :worker_2_port SELECT count(*) FROM pg_dist_node; count --------------------------------------------------------------------- 2 (1 row) \c - - - :master_port SELECT hasmetadata FROM pg_dist_node WHERE nodeport=:worker_2_proxy_port; hasmetadata --------------------------------------------------------------------- f (1 row) -- Verify we can stop metadata sync after unsuccessful attempts SELECT citus.mitmproxy('conn.allow()'); mitmproxy --------------------------------------------------------------------- (1 row) SELECT stop_metadata_sync_to_node('localhost', :worker_2_proxy_port); NOTICE: dropping metadata on the node (localhost,9060) stop_metadata_sync_to_node --------------------------------------------------------------------- (1 row) \c - - - :worker_2_port SELECT count(*) FROM pg_dist_node; count --------------------------------------------------------------------- 0 (1 row) \c - - - :master_port SELECT hasmetadata FROM pg_dist_node WHERE nodeport=:worker_2_proxy_port; hasmetadata --------------------------------------------------------------------- f (1 row) -- turn metadata sync back on SELECT start_metadata_sync_to_node('localhost', :worker_2_proxy_port); start_metadata_sync_to_node --------------------------------------------------------------------- (1 row) SET SEARCH_PATH = mx_metadata_sync; DROP TABLE t1; DROP TABLE t2; DROP SCHEMA mx_metadata_sync CASCADE;