Galera MTR Tests: a test for async slave + non-prim

mysql-test/suite/galera/galera_3nodes_as_slave.cnf

@@ -0,0 +1,91 @@
+#
+# This .cnf file creates a setup with 1 standard MySQL server, followed by a 2-node Galera cluster
+#
+
+# Use default setting for mysqld processes
+!include include/default_mysqld.cnf
+
+[mysqld]
+log-slave-updates
+log-bin=mysqld-bin
+binlog-format=row
+
+[mysqld.1]
+server-id=1
+
+[mysqld.2]
+server-id=2
+
+wsrep_provider=@ENV.WSREP_PROVIDER
+wsrep_cluster_address='gcomm://'
+wsrep_provider_options='base_port=@mysqld.2.#galera_port;evs.install_timeout = PT15S; evs.max_install_timeouts=1;'
+
+# enforce read-committed characteristics across the cluster
+wsrep_causal_reads=ON
+wsrep_sync_wait = 7
+
+wsrep_node_address=127.0.0.1
+wsrep_sst_receive_address=127.0.0.2:@mysqld.2.#sst_port
+wsrep_node_incoming_address=127.0.0.1:@mysqld.2.port
+
+# Required for Galera
+innodb_autoinc_lock_mode=2
+
+innodb_flush_log_at_trx_commit=2
+
+[mysqld.3]
+wsrep_provider=@ENV.WSREP_PROVIDER
+wsrep_cluster_address='gcomm://127.0.0.1:@mysqld.2.#galera_port'
+wsrep_provider_options='base_port=@mysqld.3.#galera_port;evs.install_timeout = PT15S; evs.max_install_timeouts = 1;'
+
+# enforce read-committed characteristics across the cluster
+wsrep_causal_reads=ON
+wsrep_sync_wait = 7
+
+wsrep_node_address=127.0.0.1
+wsrep_sst_receive_address=127.0.0.2:@mysqld.3.#sst_port
+wsrep_node_incoming_address=127.0.0.1:@mysqld.3.port
+
+# Required for Galera
+innodb_autoinc_lock_mode=2
+
+innodb_flush_log_at_trx_commit=2
+
+[mysqld.4]
+wsrep_provider=@ENV.WSREP_PROVIDER
+wsrep_cluster_address='gcomm://127.0.0.1:@mysqld.2.#galera_port'
+wsrep_provider_options='base_port=@mysqld.4.#galera_port;evs.install_timeout = PT15S; evs.max_install_timeouts = 1;'
+
+# enforce read-committed characteristics across the cluster
+wsrep_causal_reads=ON
+wsrep_sync_wait = 7
+
+wsrep_node_address=192.168.95.128
+wsrep_sst_receive_address=192.168.95.128:@mysqld.4.#sst_port
+wsrep_node_incoming_address=192.168.95.128:@mysqld.4.port
+
+# Required for Galera
+innodb_autoinc_lock_mode=2
+
+innodb_flush_log_at_trx_commit=2
+
+[ENV]
+NODE_MYPORT_1= @mysqld.1.port
+NODE_MYSOCK_1= @mysqld.1.socket
+
+NODE_MYPORT_2= @mysqld.2.port
+NODE_MYSOCK_2= @mysqld.2.socket
+
+NODE_MYPORT_3= @mysqld.3.port
+NODE_MYSOCK_3= @mysqld.3.socket
+
+NODE_MYPORT_3= @mysqld.4.port
+NODE_MYSOCK_3= @mysqld.4.socket
+
+NODE_GALERAPORT_2= @mysqld.2.#galera_port
+NODE_GALERAPORT_3= @mysqld.3.#galera_port
+NODE_GALERAPORT_4= @mysqld.4.#galera_port
+
+NODE_SSTPORT_2= @mysqld.2.#sst_port
+NODE_SSTPORT_3= @mysqld.3.#sst_port
+NODE_SSTPORT_4= @mysqld.4.#sst_port

mysql-test/suite/galera/r/galera_as_slave_nonprim.result

@@ -0,0 +1,20 @@
+START SLAVE USER='root';
+Warnings:
+Note	1759	Sending passwords in plain text without SSL/TLS is extremely insecure.
+SET SESSION wsrep_sync_wait = 0;
+CREATE TABLE t1 (f1 INTEGER PRIMARY KEY) ENGINE=InnoDB;
+SET GLOBAL wsrep_provider_options = 'gmcast.isolate=1';
+INSERT INTO t1 VALUES (1),(2),(3),(4),(5);
+expected_error
+1
+SET GLOBAL wsrep_provider_options = 'gmcast.isolate=0';
+START SLAVE;
+DROP TABLE t1;
+STOP SLAVE;
+RESET SLAVE ALL;
+CALL mtr.add_suppression("Slave SQL: Error 'Unknown command' on query");
+CALL mtr.add_suppression("Slave: Unknown command Error_code: 1047");
+CALL mtr.add_suppression("Send action {\\(nil\\), 328, TORDERED} returned -107 \\(Transport endpoint is not connected\\)");
+CALL mtr.add_suppression("Slave SQL: Error in Xid_log_event: Commit could not be completed, 'Deadlock found when trying to get lock; try restarting transaction', Error_code: 1213");
+CALL mtr.add_suppression("Slave SQL: Node has dropped from cluster, Error_code: 1047");
+RESET MASTER;

mysql-test/suite/galera/t/galera_as_slave_nonprim.cnf

@@ -0,0 +1 @@
+!include ../galera_3nodes_as_slave.cnf

mysql-test/suite/galera/t/galera_as_slave_nonprim.test

@@ -0,0 +1,96 @@
+#
+# Test the behavior of a Galera async slave if it goes non-prim. Async replication
+# should abort with an error but it should be possible to restart it.
+#
+# The galera/galera_2node_slave.cnf describes the setup of the nodes
+#
+
+--source include/have_innodb.inc
+--source include/have_log_bin.inc
+--source include/big_test.inc
+
+# Step #1. Establish replication
+#
+# As node #1 is not a Galera node, we connect to node #2 in order to run include/galera_cluster.inc
+#
+--connect node_2a, 127.0.0.1, root, , test, $NODE_MYPORT_2
+--source include/galera_cluster.inc
+
+--connection node_2
+--disable_query_log
+--eval CHANGE MASTER TO MASTER_HOST='127.0.0.1', MASTER_PORT=$NODE_MYPORT_1;
+--enable_query_log
+START SLAVE USER='root';
+SET SESSION wsrep_sync_wait = 0;
+
+--connection node_1
+CREATE TABLE t1 (f1 INTEGER PRIMARY KEY) ENGINE=InnoDB;
+
+--connection node_2
+--let $wait_condition = SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.TABLES WHERE TABLE_NAME = 't1';
+--source include/wait_condition.inc
+
+# Step #2. Force async slave to go non-primary
+
+SET GLOBAL wsrep_provider_options = 'gmcast.isolate=1';
+
+--connect node_3, 127.0.0.1, root, , test, $NODE_MYPORT_3
+--connection node_3
+--source include/wait_until_connected_again.inc
+--let $wait_condition = SELECT VARIABLE_VALUE = 2 FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'wsrep_cluster_size'
+--source include/wait_condition.inc
+
+# Step #3. Force async replication to fail by creating a replication event while the slave is non-prim
+
+--connection node_1
+INSERT INTO t1 VALUES (1),(2),(3),(4),(5);
+
+--connection node_2
+--sleep 5
+
+--let $value = query_get_value(SHOW SLAVE STATUS, Last_SQL_Error, 1)
+--connection node_3
+--disable_query_log
+--eval SELECT "$value" IN ("Error 'Unknown command' on query. Default database: 'test'. Query: 'BEGIN'", "Node has dropped from cluster") AS expected_error
+--enable_query_log
+
+# Step #4. Bring back the async slave and restart replication
+--connection node_2
+SET GLOBAL wsrep_provider_options = 'gmcast.isolate=0';
+
+--connection node_3
+--source include/wait_until_connected_again.inc
+--let $wait_condition = SELECT VARIABLE_VALUE = 3 FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'wsrep_cluster_size'
+--source include/wait_condition.inc
+
+--connection node_2
+--source include/galera_wait_ready.inc
+--source include/wait_until_connected_again.inc
+
+START SLAVE;
+
+# Confirm that the replication events have arrived
+
+--let $wait_condition = SELECT COUNT(*) = 5 FROM t1;
+--source include/wait_condition.inc
+
+--connection node_1
+DROP TABLE t1;
+
+--sleep 2
+--connection node_2
+--let $wait_condition = SELECT COUNT(*) = 0 FROM INFORMATION_SCHEMA.TABLES WHERE TABLE_NAME = 't1';
+--source include/wait_condition.inc
+
+STOP SLAVE;
+RESET SLAVE ALL;
+
+CALL mtr.add_suppression("Slave SQL: Error 'Unknown command' on query");
+CALL mtr.add_suppression("Slave: Unknown command Error_code: 1047");
+CALL mtr.add_suppression("Send action {\\(nil\\), 328, TORDERED} returned -107 \\(Transport endpoint is not connected\\)");
+CALL mtr.add_suppression("Slave SQL: Error in Xid_log_event: Commit could not be completed, 'Deadlock found when trying to get lock; try restarting transaction', Error_code: 1213");
+CALL mtr.add_suppression("Slave SQL: Node has dropped from cluster, Error_code: 1047");
+
+
+--connection node_1
+RESET MASTER;