| 1 |
############################################################################### |
| 2 |
# |
| 3 |
# When a member has recovery blocked it should not cause problems on rejoin, |
| 4 |
# it shall stop (or try too) recovery module and do another rejoin attempt. |
| 5 |
# |
| 6 |
# Test: |
| 7 |
# 0. The test requires three servers |
| 8 |
# 1. Create table t1 on server1 with binlog disabled and bootstrap the |
| 9 |
# group |
| 10 |
# 2. Create table t1 on server2 with binlog disabled and join to the |
| 11 |
# group |
| 12 |
# 3. Setup Server3. |
| 13 |
# Set the attempts of rejoin to 3 so we validate the process wont |
| 14 |
# block. |
| 15 |
# Reduce stop_component timeout so recovery module stops faster. |
| 16 |
# 4. Server3 disable binlog and supreess warnings, create table t1 and |
| 17 |
# get pid. Pid will be used to expel member from group |
| 18 |
# 5. Enable super read only, to avoid problem set it after FLUSH |
| 19 |
# TABLES WITH READ LOCK. Table locked will prevent server from |
| 20 |
# finishing recovery |
| 21 |
# 6. Insert data on group |
| 22 |
# 7. Join server3 to group, it will stay on recovery due transactions |
| 23 |
# on group that it can't apply as the table is locked |
| 24 |
# 8. Wait until group_replication_recovery channel waits on table t1 |
| 25 |
# commit lock |
| 26 |
# 9. Expel server3 from the group |
| 27 |
# 10. Server3 should change is status to ERROR |
| 28 |
# 11. Waiting for two auto-rejoin attempts to happen, they will fail |
| 29 |
# due to recovery failing to stop |
| 30 |
# 12. After the beginning of the second attempt server3 will stay in |
| 31 |
# ERROR state |
| 32 |
# 13. group_replication_recovery channel still waits on table t1 |
| 33 |
# commit lock |
| 34 |
# 14. UNLOCK tables on server3 and the join shall succeed |
| 35 |
# 15. Server3 shall be ONLINE |
| 36 |
# 16. Cleanup |
| 37 |
# |
| 38 |
############################################################################### |
| 39 |
|
| 40 |
--source include/linux.inc |
| 41 |
--source include/big_test.inc |
| 42 |
--source include/not_have_privilege_checks_user.inc |
| 43 |
--source include/have_group_replication_plugin.inc |
| 44 |
--let $rpl_server_count= 3 |
| 45 |
--let $rpl_skip_group_replication_start= 1 |
| 46 |
--source include/group_replication.inc |
| 47 |
|
| 48 |
--echo |
| 49 |
--echo # 1. Create table t1 on server1 with binlog disabled and bootstrap the |
| 50 |
--echo # group |
| 51 |
|
| 52 |
SET sql_log_bin=0; |
| 53 |
CREATE TABLE t1 (c1 INT NOT NULL AUTO_INCREMENT PRIMARY KEY, c2 INT); |
| 54 |
SET sql_log_bin=1; |
| 55 |
|
| 56 |
--source include/start_and_bootstrap_group_replication.inc |
| 57 |
|
| 58 |
--echo |
| 59 |
--echo # 2. Create table t1 on server2 with binlog disabled and join to the |
| 60 |
--echo # group |
| 61 |
|
| 62 |
--let $rpl_connection_name= server2 |
| 63 |
--source include/rpl_connection.inc |
| 64 |
|
| 65 |
SET sql_log_bin=0; |
| 66 |
CREATE TABLE t1 (c1 INT NOT NULL AUTO_INCREMENT PRIMARY KEY, c2 INT); |
| 67 |
SET sql_log_bin=1; |
| 68 |
|
| 69 |
--source include/start_group_replication.inc |
| 70 |
|
| 71 |
--echo |
| 72 |
--echo # 3. Setup Server3. |
| 73 |
--echo # Set the attempts of rejoin to 3 so we validate the process wont |
| 74 |
--echo # block. |
| 75 |
--echo # Reduce stop_component timeout so recovery module stops faster. |
| 76 |
|
| 77 |
--let $rpl_connection_name= server3 |
| 78 |
--source include/rpl_connection.inc |
| 79 |
|
| 80 |
SET @group_replication_autorejoin_tries_save = @@GLOBAL.group_replication_autorejoin_tries; |
| 81 |
SET @group_replication_components_stop_timeout_save = @@GLOBAL.group_replication_components_stop_timeout; |
| 82 |
|
| 83 |
SET GLOBAL group_replication_autorejoin_tries = 3; |
| 84 |
SET GLOBAL group_replication_components_stop_timeout = 30; |
| 85 |
|
| 86 |
--echo |
| 87 |
--echo # 4. Server3 disable binlog and supreess warnings, create table t1 and |
| 88 |
--echo # get pid. Pid will be used to expel member from group |
| 89 |
|
| 90 |
SET SESSION sql_log_bin= 0; |
| 91 |
|
| 92 |
call mtr.add_suppression('On shutdown there was a timeout on the Group Replication recovery module termination. Check the log for more details'); |
| 93 |
call mtr.add_suppression('Member was expelled from the group due to network failures, changing member status to ERROR.'); |
| 94 |
call mtr.add_suppression('The server was automatically set into read only mode after an error was detected.'); |
| 95 |
call mtr.add_suppression("Error when stopping the group replication incremental recovery's donor connection"); |
| 96 |
call mtr.add_suppression('Unable to confirm whether the server has left the group or not. Check performance_schema.replication_group_members to check group membership information.'); |
| 97 |
|
| 98 |
CREATE TABLE pid_table(pid_no INT PRIMARY KEY); |
| 99 |
--let $pid_file=`SELECT @@pid_file` |
| 100 |
--replace_result $pid_file pid_file |
| 101 |
--eval LOAD DATA LOCAL INFILE '$pid_file' INTO TABLE pid_table |
| 102 |
--let $server_pid=`SELECT pid_no FROM pid_table` |
| 103 |
DROP TABLE pid_table; |
| 104 |
|
| 105 |
CREATE TABLE t1 (c1 INT NOT NULL AUTO_INCREMENT PRIMARY KEY, c2 INT); |
| 106 |
SET SESSION sql_log_bin= 1; |
| 107 |
|
| 108 |
--let $member3_uuid = `SELECT @@GLOBAL.server_uuid` |
| 109 |
|
| 110 |
--echo |
| 111 |
--echo # 5. Enable super read only, to avoid problem set it after FLUSH |
| 112 |
--echo # TABLES WITH READ LOCK. Table locked will prevent server from |
| 113 |
--echo # finishing recovery |
| 114 |
|
| 115 |
--let $rpl_connection_name= server_3 |
| 116 |
--source include/rpl_connection.inc |
| 117 |
SET GLOBAL super_read_only= 1; |
| 118 |
FLUSH TABLES WITH READ LOCK; |
| 119 |
|
| 120 |
--echo |
| 121 |
--echo # 6. Insert data on group |
| 122 |
|
| 123 |
--let $rpl_connection_name= server1 |
| 124 |
--source include/rpl_connection.inc |
| 125 |
|
| 126 |
INSERT INTO test.t1 VALUES (1, 1); |
| 127 |
INSERT INTO test.t1 VALUES (2, 1); |
| 128 |
|
| 129 |
--echo |
| 130 |
--echo # 7. Join server3 to group, it will stay on recovery due transactions |
| 131 |
--echo # on group that it can't apply as the table is locked |
| 132 |
|
| 133 |
--let $rpl_connection_name= server3 |
| 134 |
--source include/rpl_connection.inc |
| 135 |
|
| 136 |
--let $group_replication_start_member_state= RECOVERING |
| 137 |
--source include/start_group_replication.inc |
| 138 |
|
| 139 |
--echo |
| 140 |
--echo # 8. Wait until group_replication_recovery channel waits on table t1 |
| 141 |
--echo # commit lock |
| 142 |
|
| 143 |
--let $wait_condition=SELECT COUNT(*)=1 FROM information_schema.processlist WHERE user='system user' AND state='Waiting for commit lock' |
| 144 |
--source include/wait_condition.inc |
| 145 |
|
| 146 |
--echo |
| 147 |
--echo # 9. Expel server3 from the group |
| 148 |
|
| 149 |
# Send signal SIGSTOP to server 3. |
| 150 |
--exec kill -19 $server_pid |
| 151 |
|
| 152 |
# Wait until the group settles on 2 members |
| 153 |
--let $rpl_connection_name= server1 |
| 154 |
--source include/rpl_connection.inc |
| 155 |
--let $group_replication_number_of_members= 2 |
| 156 |
--source include/gr_wait_for_number_of_members.inc |
| 157 |
|
| 158 |
--let $rpl_connection_name= server2 |
| 159 |
--source include/rpl_connection.inc |
| 160 |
--let $group_replication_number_of_members= 2 |
| 161 |
--source include/gr_wait_for_number_of_members.inc |
| 162 |
|
| 163 |
# Send signal SIGCONT to server 3. |
| 164 |
--let $rpl_connection_name= server3 |
| 165 |
--source include/rpl_connection.inc |
| 166 |
--exec kill -18 $server_pid |
| 167 |
|
| 168 |
--echo |
| 169 |
--echo # 10. Server3 should change is status to ERROR |
| 170 |
|
| 171 |
--let $group_replication_member_state= ERROR |
| 172 |
--let $group_replication_member_id= $member3_uuid |
| 173 |
--source include/gr_wait_for_member_state.inc |
| 174 |
|
| 175 |
--echo |
| 176 |
--echo # 11. Waiting for two auto-rejoin attempts to happen, they will fail |
| 177 |
--echo # due to recovery failing to stop |
| 178 |
|
| 179 |
--let $wait_timeout= 600 |
| 180 |
--let $wait_condition= SELECT WORK_COMPLETED = 2 FROM performance_schema.events_stages_current WHERE EVENT_NAME LIKE '%auto-rejoin%' |
| 181 |
--source include/wait_condition.inc |
| 182 |
|
| 183 |
--echo |
| 184 |
--echo # 12. After the beginning of the second attempt server3 will stay in |
| 185 |
--echo # ERROR state |
| 186 |
|
| 187 |
--let $group_replication_member_state= ERROR |
| 188 |
--let $group_replication_member_id= $member3_uuid |
| 189 |
--source include/gr_wait_for_member_state.inc |
| 190 |
|
| 191 |
--echo |
| 192 |
--echo # 13. group_replication_recovery channel still waits on table t1 |
| 193 |
--echo # commit lock |
| 194 |
|
| 195 |
--let $wait_condition=SELECT COUNT(*)=1 FROM information_schema.processlist WHERE user='system user' AND state='Waiting for commit lock' |
| 196 |
--source include/wait_condition.inc |
| 197 |
|
| 198 |
--echo |
| 199 |
--echo # 14. UNLOCK tables on server3 and the join shall succeed |
| 200 |
|
| 201 |
--let $rpl_connection_name= server_3 |
| 202 |
--source include/rpl_connection.inc |
| 203 |
UNLOCK TABLES; |
| 204 |
|
| 205 |
--echo |
| 206 |
--echo # 15. Server3 shall be ONLINE |
| 207 |
|
| 208 |
--let $rpl_connection_name= server3 |
| 209 |
--source include/rpl_connection.inc |
| 210 |
# Due to concurrency between the completeness of second auto-rejoin attempt |
| 211 |
# and the UNLOCK TABLES, member can successfully rejoin on the second or third |
| 212 |
# attempt, thence the big wait_timeout. |
| 213 |
--let $wait_timeout= 600 |
| 214 |
--let $group_replication_member_state= ONLINE |
| 215 |
--let $group_replication_member_id= $member3_uuid |
| 216 |
--source include/gr_wait_for_member_state.inc |
| 217 |
|
| 218 |
--echo |
| 219 |
--echo # 16. Cleanup |
| 220 |
|
| 221 |
DROP TABLE t1; |
| 222 |
|
| 223 |
SET GLOBAl group_replication_autorejoin_tries = @group_replication_autorejoin_tries_save; |
| 224 |
SET GLOBAl group_replication_components_stop_timeout = @group_replication_components_stop_timeout_save; |
| 225 |
|
| 226 |
--source include/group_replication_end.inc |