| 1 |
################################################################################ |
| 2 |
# This test validates the behavior of the group replication recovery channel |
| 3 |
# against common slave commands like START/STOP/RESET SLAVE. |
| 4 |
# |
| 5 |
# Test: |
| 6 |
# 0. The test requires three servers: M1, M2 and M3. |
| 7 |
# 1. Bootstrap start group replication on M2. Add some data for recovery. |
| 8 |
# Start M1. |
| 9 |
# 2. Stop replication on M1. On M2 lock a table to block recovery. Start M1. |
| 10 |
# 3. Execute a STOP SLAVE command and see the recovery channel still active. |
| 11 |
# 4. Execute a RESET SLAVE ALL command and see the recovery channel is not |
| 12 |
# affected on M1 when GR node is active. |
| 13 |
# 5. Kill the server M1. The recovery channel should still be there but it's |
| 14 |
# not started. Recreate the slave channel to test that it started |
| 15 |
# 6. Clean the created channels and tables. |
| 16 |
################################################################################ |
| 17 |
|
| 18 |
|
| 19 |
# This test does crashes servers, thence we skip it on valgrind. |
| 20 |
--source include/not_valgrind.inc |
| 21 |
--source include/big_test.inc |
| 22 |
--source include/force_restart.inc |
| 23 |
--source include/have_group_replication_plugin.inc |
| 24 |
--let $rpl_skip_group_replication_start= 1 |
| 25 |
--let $rpl_server_count= 3 |
| 26 |
--source include/group_replication.inc |
| 27 |
|
| 28 |
# |
| 29 |
# Phase 1: Start group replication on two servers |
| 30 |
# |
| 31 |
|
| 32 |
--connection server2 |
| 33 |
--echo server2 |
| 34 |
|
| 35 |
--source include/start_and_bootstrap_group_replication.inc |
| 36 |
|
| 37 |
# Add some data for recovery |
| 38 |
CREATE TABLE t1 (c1 INT NOT NULL PRIMARY KEY) ENGINE=InnoDB; |
| 39 |
INSERT INTO t1 VALUES (1); |
| 40 |
|
| 41 |
--connection server1 |
| 42 |
--echo server1 |
| 43 |
--source include/start_group_replication.inc |
| 44 |
--let $member1_uuid= query_get_value(SELECT @@SERVER_UUID, @@SERVER_UUID, 1) |
| 45 |
|
| 46 |
--connection server2 |
| 47 |
--let $group_replication_number_of_members= 2 |
| 48 |
--source include/gr_wait_for_number_of_members.inc |
| 49 |
|
| 50 |
# |
| 51 |
# Phase 2: Stop replication on server 1 |
| 52 |
# On server 2 lock a table to block recovery |
| 53 |
# |
| 54 |
|
| 55 |
--connection server1 |
| 56 |
--source include/stop_group_replication.inc |
| 57 |
|
| 58 |
--connection server2 |
| 59 |
INSERT INTO t1 VALUES (2); |
| 60 |
INSERT INTO t1 VALUES (3); |
| 61 |
|
| 62 |
#lock the table in a different connection |
| 63 |
--connection server_1 |
| 64 |
|
| 65 |
LOCK TABLE t1 READ; |
| 66 |
|
| 67 |
--connection server1 |
| 68 |
|
| 69 |
--let $group_replication_start_member_state= RECOVERING |
| 70 |
--source include/start_group_replication.inc |
| 71 |
|
| 72 |
# |
| 73 |
# Phase 3: Execute a STOP SLAVE command and see the channel still active. |
| 74 |
# |
| 75 |
|
| 76 |
let $wait_condition= |
| 77 |
SELECT COUNT(*)=1 FROM performance_schema.replication_connection_status where CHANNEL_NAME="group_replication_recovery" and SERVICE_STATE="ON"; |
| 78 |
--source include/wait_condition.inc |
| 79 |
|
| 80 |
STOP SLAVE; |
| 81 |
|
| 82 |
--let $assert_text= 'The group replication recovery channel is ON' |
| 83 |
--let $assert_cond= [SELECT COUNT(*) AS count FROM performance_schema.replication_connection_status where CHANNEL_NAME="group_replication_recovery" and SERVICE_STATE="ON", count, 1] = 1 |
| 84 |
--source include/assert.inc |
| 85 |
|
| 86 |
# |
| 87 |
# Phase 4: Execute a RESET SLAVE ALL command and see the recovery channel is |
| 88 |
# not affected as GR node is active. |
| 89 |
# |
| 90 |
|
| 91 |
--connection server1 |
| 92 |
SET GLOBAL SUPER_READ_ONLY=0; # Bug#22097534 |
| 93 |
--replace_result $SERVER_MYPORT_3 SERVER_3_PORT |
| 94 |
--eval CHANGE REPLICATION SOURCE TO SOURCE_HOST="127.0.0.1", SOURCE_USER="root", SOURCE_PASSWORD="", SOURCE_PORT=$SERVER_MYPORT_3, SOURCE_AUTO_POSITION=1 FOR CHANNEL "channel_1" |
| 95 |
|
| 96 |
--let $assert_text= 'The slave channel is present' |
| 97 |
--let $assert_cond= [SELECT COUNT(*) AS count FROM mysql.slave_relay_log_info WHERE channel_name="channel_1", count, 1] = 1 |
| 98 |
--source include/assert.inc |
| 99 |
|
| 100 |
--let $datadir_1= `SELECT @@GLOBAL.datadir` |
| 101 |
|
| 102 |
--let $recovery_relay_log_file= `SELECT CONCAT('$datadir_1', 'mgr-group_replication_recovery.000001')` |
| 103 |
--file_exists $recovery_relay_log_file |
| 104 |
|
| 105 |
--let $recovery_relay_log_index= `SELECT CONCAT('$datadir_1', 'mgr-group_replication_recovery.index')` |
| 106 |
--file_exists $recovery_relay_log_index |
| 107 |
|
| 108 |
--error ER_REPLICA_CHANNEL_OPERATION_NOT_ALLOWED |
| 109 |
RESET SLAVE ALL; |
| 110 |
|
| 111 |
--let $assert_text= 'The group replication recovery channel is present' |
| 112 |
--let $assert_cond= [SELECT COUNT(*) AS count FROM mysql.slave_relay_log_info WHERE channel_name="group_replication_recovery", count, 1] = 1 |
| 113 |
--source include/assert.inc |
| 114 |
|
| 115 |
--let $assert_text= 'The group replication recovery channel is ON' |
| 116 |
--let $assert_cond= [SELECT COUNT(*) AS count FROM performance_schema.replication_connection_status where CHANNEL_NAME="group_replication_recovery" and SERVICE_STATE="ON", count, 1] = 1 |
| 117 |
--source include/assert.inc |
| 118 |
|
| 119 |
--file_exists $recovery_relay_log_file |
| 120 |
|
| 121 |
--file_exists $recovery_relay_log_index |
| 122 |
|
| 123 |
--let $assert_text= 'The slave channel is not present' |
| 124 |
--let $assert_cond= [SELECT COUNT(*) AS count FROM mysql.slave_relay_log_info WHERE channel_name="channel_1", count, 1] = 0 |
| 125 |
--source include/assert.inc |
| 126 |
|
| 127 |
# |
| 128 |
# Phase 5: Kill the server |
| 129 |
# The recovery channel should still be there but it's not started |
| 130 |
# Recreate the slave channel to test that it started |
| 131 |
# |
| 132 |
|
| 133 |
--connection server1 |
| 134 |
--replace_result $SERVER_MYPORT_3 SERVER_3_PORT |
| 135 |
--eval CHANGE REPLICATION SOURCE TO SOURCE_HOST="127.0.0.1", SOURCE_USER="root", SOURCE_PASSWORD="", SOURCE_PORT=$SERVER_MYPORT_3, SOURCE_AUTO_POSITION=1 FOR CHANNEL "channel_1" |
| 136 |
--source include/start_slave.inc |
| 137 |
|
| 138 |
--let $assert_text= 'The created slave channel is ON' |
| 139 |
--let $assert_cond= [SELECT COUNT(*) AS count FROM performance_schema.replication_connection_status where CHANNEL_NAME="channel_1" and SERVICE_STATE="ON", count, 1] = 1 |
| 140 |
--source include/assert.inc |
| 141 |
|
| 142 |
--let $_group_replication_local_address= `SELECT @@GLOBAL.group_replication_local_address` |
| 143 |
--let $_group_replication_group_seeds= `SELECT @@GLOBAL.group_replication_group_seeds` |
| 144 |
|
| 145 |
# kill server 1 |
| 146 |
--let $restart_parameters=restart:--group_replication_local_address=$_group_replication_local_address --group_replication_group_seeds=$_group_replication_group_seeds |
| 147 |
--replace_result $_group_replication_local_address GROUP_REPLICATION_LOCAL_ADDRESS $_group_replication_group_seeds GROUP_REPLICATION_GROUP_SEEDS |
| 148 |
--source include/kill_and_restart_mysqld.inc |
| 149 |
|
| 150 |
#Needed as we are not using rpl_restart_server.inc |
| 151 |
--let $rpl_server_number= 1 |
| 152 |
--source include/rpl_reconnect.inc |
| 153 |
|
| 154 |
# Wait until group has only 1 member, that is, server1 death is detected by |
| 155 |
# the group. |
| 156 |
--let $rpl_connection_name= server2 |
| 157 |
--source include/rpl_connection.inc |
| 158 |
|
| 159 |
--let $group_replication_member_state= UNREACHABLE |
| 160 |
--let $group_replication_member_id= $member1_uuid |
| 161 |
--source include/gr_wait_for_member_state.inc |
| 162 |
|
| 163 |
# unblock the group, since server1 has gone away and |
| 164 |
# there were only two in the group - majority lost |
| 165 |
|
| 166 |
--let $local_address_server2= `SELECT @@GLOBAL.group_replication_local_address` |
| 167 |
--disable_query_log |
| 168 |
# Reseting the group to server2 only |
| 169 |
--eval SET GLOBAL group_replication_force_members= "$local_address_server2" |
| 170 |
--enable_query_log |
| 171 |
|
| 172 |
--let $group_replication_number_of_members= 1 |
| 173 |
--source include/gr_wait_for_number_of_members.inc |
| 174 |
|
| 175 |
--let $rpl_connection_name= server1 |
| 176 |
--source include/rpl_connection.inc |
| 177 |
# Wait for the slave channel start |
| 178 |
--let $rpl_channel_name= 'channel_1' |
| 179 |
--source include/wait_for_slave_to_start.inc |
| 180 |
--let $rpl_channel_name= |
| 181 |
|
| 182 |
--let $assert_text= 'The group replication recovery channel is present' |
| 183 |
--let $assert_cond= [SELECT COUNT(*) AS count FROM mysql.slave_relay_log_info WHERE channel_name="group_replication_recovery", count, 1] = 1 |
| 184 |
--source include/assert.inc |
| 185 |
|
| 186 |
--let $assert_text= 'The group replication recovery channel is OFF' |
| 187 |
--let $assert_cond= [SELECT COUNT(*) AS count FROM performance_schema.replication_connection_status where CHANNEL_NAME="group_replication_recovery" and SERVICE_STATE="OFF", count, 1] = 1 |
| 188 |
--source include/assert.inc |
| 189 |
|
| 190 |
--let $assert_text= 'The created slave channel is ON' |
| 191 |
--let $assert_cond= [SELECT COUNT(*) AS count FROM performance_schema.replication_connection_status where CHANNEL_NAME="channel_1" and SERVICE_STATE="ON", count, 1] = 1 |
| 192 |
--source include/assert.inc |
| 193 |
|
| 194 |
# |
| 195 |
# Phase 6: Clean the created channels and tables |
| 196 |
# |
| 197 |
|
| 198 |
--connection server2 |
| 199 |
|
| 200 |
DROP TABLE t1; |
| 201 |
|
| 202 |
--connection server1 |
| 203 |
|
| 204 |
--source include/stop_slave.inc |
| 205 |
CHANGE REPLICATION SOURCE TO SOURCE_AUTO_POSITION=0 FOR CHANNEL "channel_1"; |
| 206 |
|
| 207 |
--source include/start_group_replication.inc |
| 208 |
|
| 209 |
--let $skip_restore_connection= 0 |
| 210 |
--source include/group_replication_end.inc |