| 1 |
################################################################################ |
| 2 |
# Validate that when the sender to which the receiver is connected loses |
| 3 |
# majority, the receiver will switch to another healthy sender. Also validate |
| 4 |
# that the sender which lost majority still has its connection details in |
| 5 |
# replication_asynchronous_connection_failover table. |
| 6 |
# |
| 7 |
# Test: |
| 8 |
# 0. The test requires 4 servers: |
| 9 |
# server1: group primary |
| 10 |
# server2: group secondary |
| 11 |
# server3: single server |
| 12 |
# server4: replica server |
| 13 |
# 1. Deploy a 2 members group in single primary mode. |
| 14 |
# 2. Configure replica to manage the asynchronous connection |
| 15 |
# to group and server3. |
| 16 |
# 3. Create and start the managed channel connected to server1. |
| 17 |
# 4. Kill server1, the group will block due no majority. |
| 18 |
# 5. Despite server2 has the second biggest weight, since |
| 19 |
# the group has no majority, the replica will connect |
| 20 |
# to server3. |
| 21 |
# 6. Wait 30 seconds and validate that replica is still connected |
| 22 |
# to server3. |
| 23 |
# 7. Assert that group information did not change, since it |
| 24 |
# lost the majority. |
| 25 |
# 8. Restore the majority on the group by forcing its membership |
| 26 |
# to only server2. |
| 27 |
# 9. Since server2 has bigger weight than server3, the managed |
| 28 |
# channel must reconnect to server2 now that the group has |
| 29 |
# majority. |
| 30 |
# 10. Clean up. |
| 31 |
################################################################################ |
| 32 |
--source include/big_test.inc |
| 33 |
--source include/not_valgrind.inc |
| 34 |
--source include/have_group_replication_plugin.inc |
| 35 |
--let $rpl_skip_group_replication_start= 1 |
| 36 |
--let $rpl_server_count= 4 |
| 37 |
--let $group_replication_group_name= aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa |
| 38 |
--source include/group_replication.inc |
| 39 |
|
| 40 |
|
| 41 |
--echo |
| 42 |
--echo ############################################################ |
| 43 |
--echo # 1. Deploy a 2 members group in single primary mode. |
| 44 |
--let $rpl_connection_name= server1 |
| 45 |
--source include/rpl_connection.inc |
| 46 |
--let $server1_uuid= `SELECT @@server_uuid` |
| 47 |
--source include/start_and_bootstrap_group_replication.inc |
| 48 |
|
| 49 |
--let $rpl_connection_name= server2 |
| 50 |
--source include/rpl_connection.inc |
| 51 |
--let $server2_uuid= `SELECT @@server_uuid` |
| 52 |
--source include/start_group_replication.inc |
| 53 |
|
| 54 |
|
| 55 |
--echo |
| 56 |
--echo ############################################################ |
| 57 |
--echo # 2. Configure replica to manage the asynchronous connection |
| 58 |
--echo # to group and server3. |
| 59 |
--let $rpl_connection_name= server4 |
| 60 |
--source include/rpl_connection.inc |
| 61 |
|
| 62 |
--replace_result $SERVER_MYPORT_1 SERVER_MYPORT_1 $group_replication_group_name GROUP_NAME |
| 63 |
--eval SELECT asynchronous_connection_failover_add_managed('ch1', 'GroupReplication', '$group_replication_group_name', '127.0.0.1', $SERVER_MYPORT_1, '', 90, 70); |
| 64 |
|
| 65 |
--replace_result $SERVER_MYPORT_3 SERVER_MYPORT_3 |
| 66 |
--eval SELECT asynchronous_connection_failover_add_source('ch1', '127.0.0.1', $SERVER_MYPORT_3, '', 50); |
| 67 |
|
| 68 |
--let $assert_text= 'There is one row in performance_schema.replication_asynchronous_connection_failover for server1' |
| 69 |
--let $assert_cond= [SELECT COUNT(*) count FROM performance_schema.replication_asynchronous_connection_failover WHERE port=$SERVER_MYPORT_1, count, 1] = 1 |
| 70 |
--source include/assert.inc |
| 71 |
|
| 72 |
--let $assert_text= 'There is one row in performance_schema.replication_asynchronous_connection_failover for server3' |
| 73 |
--let $assert_cond= [SELECT COUNT(*) count FROM performance_schema.replication_asynchronous_connection_failover WHERE port=$SERVER_MYPORT_3, count, 1] = 1 |
| 74 |
--source include/assert.inc |
| 75 |
|
| 76 |
--let $assert_text= 'There is one row in performance_schema.replication_asynchronous_connection_failover_managed for server1' |
| 77 |
--let $assert_cond= [SELECT COUNT(*) count FROM performance_schema.replication_asynchronous_connection_failover_managed WHERE managed_name="$group_replication_group_name", count, 1] = 1 |
| 78 |
--source include/assert.inc |
| 79 |
|
| 80 |
|
| 81 |
--echo |
| 82 |
--echo ############################################################ |
| 83 |
--echo # 3. Create and start the managed channel connected to server1. |
| 84 |
--let $rpl_connection_name= server4 |
| 85 |
--source include/rpl_connection.inc |
| 86 |
--replace_result $SERVER_MYPORT_1 SERVER_1_PORT |
| 87 |
--eval CHANGE REPLICATION SOURCE TO SOURCE_HOST='127.0.0.1', SOURCE_USER='root', SOURCE_AUTO_POSITION=1, SOURCE_CONNECTION_AUTO_FAILOVER=1, SOURCE_PORT=$SERVER_MYPORT_1, SOURCE_RETRY_COUNT=0 FOR CHANNEL 'ch1' |
| 88 |
|
| 89 |
--let $rpl_channel_name='ch1' |
| 90 |
--source include/start_slave.inc |
| 91 |
--let $rpl_channel_name= |
| 92 |
|
| 93 |
--let $assert_text= Verify channel ch1 IO_THREAD is ON and connected to server1 |
| 94 |
--let $assert_cond= "[SELECT SERVICE_STATE FROM performance_schema.replication_connection_status WHERE channel_name=\'ch1\' AND source_uuid=\'$server1_uuid\', SERVICE_STATE, 1]" = "ON" |
| 95 |
--source include/assert.inc |
| 96 |
|
| 97 |
--let $wait_condition= SELECT COUNT(*)=3 FROM performance_schema.replication_asynchronous_connection_failover |
| 98 |
--source include/wait_condition.inc |
| 99 |
|
| 100 |
|
| 101 |
--echo |
| 102 |
--echo ############################################################ |
| 103 |
--echo # 4. Kill server1, the group will block due no majority. |
| 104 |
--let $rpl_connection_name= server1 |
| 105 |
--source include/rpl_connection.inc |
| 106 |
--source include/kill_and_restart_mysqld.inc |
| 107 |
|
| 108 |
--let $rpl_server_number= 1 |
| 109 |
--source include/rpl_reconnect.inc |
| 110 |
|
| 111 |
--let $rpl_connection_name= server2 |
| 112 |
--source include/rpl_connection.inc |
| 113 |
--let $group_replication_member_state= UNREACHABLE |
| 114 |
--let $group_replication_member_id= $server1_uuid |
| 115 |
--source include/gr_wait_for_member_state.inc |
| 116 |
|
| 117 |
|
| 118 |
--echo |
| 119 |
--echo ############################################################ |
| 120 |
--echo # 5. Despite server2 has the second biggest weight, since |
| 121 |
--echo # the group has no majority, the replica will connect |
| 122 |
--echo # to server3. |
| 123 |
--let $rpl_connection_name= server3 |
| 124 |
--source include/rpl_connection.inc |
| 125 |
--let $server3_uuid= `SELECT @@server_uuid` |
| 126 |
|
| 127 |
--let $rpl_connection_name= server4 |
| 128 |
--source include/rpl_connection.inc |
| 129 |
--echo Wait until ch1 IO_THREAD is connected to server3 |
| 130 |
--let $wait_condition= SELECT service_state="ON" FROM performance_schema.replication_connection_status WHERE channel_name="ch1" AND source_uuid="$server3_uuid" |
| 131 |
--source include/wait_condition.inc |
| 132 |
|
| 133 |
|
| 134 |
--let $wait_condition= SELECT COUNT(*)>=1 FROM performance_schema.error_log WHERE (error_code='MY-013682' OR error_code='MY-013683') AND data LIKE 'The % thread detected that the source % does not belong to the group majority, thence the channel % will try to connect to another source.' |
| 135 |
--source include/wait_condition.inc |
| 136 |
|
| 137 |
|
| 138 |
--echo |
| 139 |
--echo ############################################################ |
| 140 |
--echo # 6. Wait 30 seconds and validate that replica is still connected |
| 141 |
--echo # to server3. |
| 142 |
--let $rpl_connection_name= server4 |
| 143 |
--source include/rpl_connection.inc |
| 144 |
--sleep 30 |
| 145 |
--let $assert_text= Verify channel ch1 IO_THREAD is ON and connected to server3 |
| 146 |
--let $assert_cond= "[SELECT SERVICE_STATE FROM performance_schema.replication_connection_status WHERE channel_name=\'ch1\' AND source_uuid=\'$server3_uuid\', SERVICE_STATE, 1]" = "ON" |
| 147 |
--source include/assert.inc |
| 148 |
|
| 149 |
|
| 150 |
--echo |
| 151 |
--echo ############################################################ |
| 152 |
--echo # 7. Assert that group information did not change, since it |
| 153 |
--echo # lost the majority. |
| 154 |
--let $rpl_connection_name= server4 |
| 155 |
--source include/rpl_connection.inc |
| 156 |
|
| 157 |
--let $assert_text= 'There is one row in performance_schema.replication_asynchronous_connection_failover for server1' |
| 158 |
--let $assert_cond= [SELECT COUNT(*) count FROM performance_schema.replication_asynchronous_connection_failover WHERE port=$SERVER_MYPORT_1 AND weight=90 AND managed_name="$group_replication_group_name" , count, 1] = 1 |
| 159 |
--source include/assert.inc |
| 160 |
|
| 161 |
--let $assert_text= 'There is one row in performance_schema.replication_asynchronous_connection_failover for server2' |
| 162 |
--let $assert_cond= [SELECT COUNT(*) count FROM performance_schema.replication_asynchronous_connection_failover WHERE port=$SERVER_MYPORT_2 AND weight=70 AND managed_name="$group_replication_group_name" , count, 1] = 1 |
| 163 |
--source include/assert.inc |
| 164 |
|
| 165 |
|
| 166 |
--echo |
| 167 |
--echo ############################################################ |
| 168 |
--echo # 8. Restore the majority on the group by forcing its membership |
| 169 |
--echo # to only server2. |
| 170 |
--let $rpl_connection_name= server2 |
| 171 |
--source include/rpl_connection.inc |
| 172 |
--let $local_address_server2= `SELECT @@GLOBAL.group_replication_local_address` |
| 173 |
--replace_result $local_address_server2 GROUP_REPLICATION_LOCAL_ADDR_SERVER2 |
| 174 |
--eval SET GLOBAL group_replication_force_members= "$local_address_server2" |
| 175 |
|
| 176 |
# Check for the primary member after the group is unblocked |
| 177 |
--let $server1_uuid= `SELECT @@server_uuid` |
| 178 |
--let $group_replication_expected_uuid= $server2_uuid |
| 179 |
--source include/gr_wait_primary_member_uuid.inc |
| 180 |
|
| 181 |
--let $wait_condition=SELECT COUNT(*)=1 FROM performance_schema.replication_group_members where MEMBER_STATE="ONLINE" |
| 182 |
--source include/wait_condition.inc |
| 183 |
|
| 184 |
SET GLOBAL group_replication_force_members= ""; |
| 185 |
|
| 186 |
|
| 187 |
--echo |
| 188 |
--echo ############################################################ |
| 189 |
--echo # 9. Since server2 has bigger weight than server3, the managed |
| 190 |
--echo # channel must reconnect to server2 now that the group has |
| 191 |
--echo # majority. |
| 192 |
--let $rpl_connection_name= server4 |
| 193 |
--source include/rpl_connection.inc |
| 194 |
--echo Wait until ch1 IO_THREAD is connected to server2 |
| 195 |
--let $wait_condition= SELECT service_state="ON" FROM performance_schema.replication_connection_status WHERE channel_name="ch1" AND source_uuid="$server2_uuid" |
| 196 |
--source include/wait_condition.inc |
| 197 |
|
| 198 |
|
| 199 |
--echo |
| 200 |
--echo ############################################################ |
| 201 |
--echo # 10. Clean up. |
| 202 |
--let $rpl_connection_name= server4 |
| 203 |
--source include/rpl_connection.inc |
| 204 |
--let $rpl_channel_name='ch1' |
| 205 |
--source include/stop_slave.inc |
| 206 |
|
| 207 |
--let $rpl_reset_slave_all= 1 |
| 208 |
--let $rpl_multi_source= 1 |
| 209 |
--source include/rpl_reset_slave.inc |
| 210 |
--let $rpl_channel_name= |
| 211 |
--let $rpl_reset_slave_all= |
| 212 |
--let $rpl_multi_source= |
| 213 |
|
| 214 |
--replace_result $group_replication_group_name GROUP_NAME |
| 215 |
--eval SELECT asynchronous_connection_failover_delete_managed('ch1', '$group_replication_group_name') |
| 216 |
|
| 217 |
--let $assert_text= 'There are 0 rows in performance_schema.replication_asynchronous_connection_failover_managed for group' |
| 218 |
--let $assert_cond= [SELECT COUNT(*) count FROM performance_schema.replication_asynchronous_connection_failover_managed WHERE managed_name="$group_replication_group_name", count, 1] = 0 |
| 219 |
--source include/assert.inc |
| 220 |
|
| 221 |
--let $assert_text= 'There are 0 rows in performance_schema.replication_asynchronous_connection_failover for group' |
| 222 |
--let $assert_cond= [SELECT COUNT(*) count FROM performance_schema.replication_asynchronous_connection_failover WHERE managed_name="$group_replication_group_name" , count, 1] = 0 |
| 223 |
|
| 224 |
--replace_result $SERVER_MYPORT_3 SERVER_MYPORT_3 |
| 225 |
--eval SELECT asynchronous_connection_failover_delete_source('ch1', '127.0.0.1', $SERVER_MYPORT_3, ''); |
| 226 |
|
| 227 |
--let $assert_text= 'There are no rows in performance_schema.replication_asynchronous_connection_failover' |
| 228 |
--let $assert_cond= [SELECT COUNT(*) count FROM performance_schema.replication_asynchronous_connection_failover, count, 1] = 0 |
| 229 |
--source include/assert.inc |
| 230 |
|
| 231 |
--let $assert_text= 'There are no rows in performance_schema.replication_asynchronous_connection_failover_managed' |
| 232 |
--let $assert_cond= [SELECT COUNT(*) count FROM performance_schema.replication_asynchronous_connection_failover_managed, count, 1] = 0 |
| 233 |
--source include/assert.inc |
| 234 |
|
| 235 |
SET SESSION sql_log_bin = 0; |
| 236 |
call mtr.add_suppression("The source .* for channel 'ch1' has joined the group .*, and so added its entry into replication_asynchronous_connection_failover table."); |
| 237 |
call mtr.add_suppression("The Monitor IO thread detected that the source .* does not belong to the group majority, thence the channel .* will try to connect to another source.*"); |
| 238 |
call mtr.add_suppression("The IO thread detected that the source .* does not belong to the group majority, thence the channel .* will try to connect to another source."); |
| 239 |
call mtr.add_suppression("The source .* for channel 'ch1' has left the group .*, and so removed its entry from replication_asynchronous_connection_failover table."); |
| 240 |
call mtr.add_suppression("The group .* for the channel .* has been removed, and so removed its entry from replication_asynchronous_connection_failover_managed and all the group members from replication_asynchronous_connection_failover table."); |
| 241 |
call mtr.add_suppression("The group .* for the channel .* has been added, and so added its entry in replication_asynchronous_connection_failover_managed and source to replication_asynchronous_connection_failover table."); |
| 242 |
call mtr.add_suppression("The Monitor IO thread failed to detect if the source belongs to the group majority on the source.*"); |
| 243 |
SET SESSION sql_log_bin = 1; |
| 244 |
|
| 245 |
--let $rpl_connection_name= server2 |
| 246 |
--source include/rpl_connection.inc |
| 247 |
--source include/stop_group_replication.inc |
| 248 |
|
| 249 |
--source include/group_replication_end.inc |