| 1 |
################################################################################ |
| 2 |
#Bug#24665506 MEMBER RESTART WHILE DML ON GROUP RESULTS IN RESTARTED MEMBER |
| 3 |
# RECOVERING FOREVER |
| 4 |
# Problem: If the group replication is stopped while the node is downloading |
| 5 |
# partial transaction data, starting the group replication is |
| 6 |
# hanging forever. |
| 7 |
# |
| 8 |
# # Test: |
| 9 |
# 0. The test requires three servers: M1, M2 and M3. |
| 10 |
# 1. Create a basic table on all three servers. |
| 11 |
# 2. On M2: To replicate a partial transaction downloaded scenario, |
| 12 |
# setup a debug point to bring down the applier channel |
| 13 |
# if it receives a Write_rows_log_event and restart the |
| 14 |
# group replication to make the debug point effective. |
| 15 |
# 3. On M1: Execute an insert and wait till M2 goes offline |
| 16 |
# (because of the above debug point, applier channel will |
| 17 |
# go down after receiving Write_rows_log_event). |
| 18 |
# 4. On M2: Reset the debug point and restart the group replication in order |
| 19 |
# to download the transaction into the new relay log file fully. |
| 20 |
# 5. On M1: Insert another tuple and do a diff tables with other nodes. |
| 21 |
# (just to check that everything is working fine). |
| 22 |
# 6. Cleanup. |
| 23 |
################################################################################ |
| 24 |
--source include/big_test.inc |
| 25 |
--source include/have_debug.inc |
| 26 |
|
| 27 |
# skip this test if compression is enabled. |
| 28 |
# This test requires that inspection is done on a per event basis |
| 29 |
# on the applier, and act when a Write_rows_log_event is received. |
| 30 |
# With compression, a full transaction is commpressed inside a |
| 31 |
# Transaction_payload event, thence no partial transaction can be |
| 32 |
# received and handled at the expected point in time in the pipeline. |
| 33 |
--source include/not_binlog_transaction_compression_on.inc |
| 34 |
|
| 35 |
--source include/have_group_replication_plugin.inc |
| 36 |
|
| 37 |
--echo ######################################################################## |
| 38 |
--echo # 0. The test requires three servers: M1, M2 and M3. |
| 39 |
--let $rpl_server_count= 3 |
| 40 |
--source include/group_replication.inc |
| 41 |
|
| 42 |
--echo ######################################################################## |
| 43 |
--echo # 1. Create a basic table on all three servers. |
| 44 |
--let $rpl_connection_name= server1 |
| 45 |
--source include/rpl_connection.inc |
| 46 |
CREATE TABLE t1(a INT PRIMARY KEY); |
| 47 |
--source include/rpl_sync.inc |
| 48 |
|
| 49 |
--let $rpl_connection_name= server2 |
| 50 |
--source include/rpl_connection.inc |
| 51 |
SET sql_log_bin=0; |
| 52 |
call mtr.add_suppression("Fatal error during execution on the Applier process of Group Replication."); |
| 53 |
call mtr.add_suppression("The server was automatically set into read only mode after an error was detected."); |
| 54 |
call mtr.add_suppression("Skipping leave operation: concurrent attempt to leave the group is on-going."); |
| 55 |
call mtr.add_suppression("The replica coordinator and worker threads are stopped"); |
| 56 |
SET sql_log_bin=1; |
| 57 |
|
| 58 |
--echo ######################################################################## |
| 59 |
--echo # 2. On M2: To replicate a partial transaction downloaded scenario, |
| 60 |
--echo # setup a debug point to bring down the applier channel |
| 61 |
--echo # if it receives a Write_rows_log_event and restart the |
| 62 |
--echo # group replication to make the debug point effective. |
| 63 |
SET GLOBAL DEBUG="+d,stop_applier_channel_after_reading_write_rows_log_event"; |
| 64 |
--source include/stop_group_replication.inc |
| 65 |
--source include/start_group_replication.inc |
| 66 |
|
| 67 |
--echo ######################################################################## |
| 68 |
--echo # 3. On M1: Execute an insert and wait till M2 goes offline |
| 69 |
--echo # (because of the above debug point, applier channel will |
| 70 |
--echo # go down after receiving Write_rows_log_event). |
| 71 |
--let $rpl_connection_name= server1 |
| 72 |
--source include/rpl_connection.inc |
| 73 |
INSERT INTO t1 values (12); |
| 74 |
|
| 75 |
--let $wait_condition= SELECT COUNT(*) = 2 FROM performance_schema.replication_group_members WHERE member_state='ONLINE'; |
| 76 |
--source include/wait_condition.inc |
| 77 |
|
| 78 |
--echo ######################################################################## |
| 79 |
--echo # 4. On M2: Reset the debug point and restart the group replication in |
| 80 |
--echo # order to download the transaction into the new relay log |
| 81 |
--echo # file fully. |
| 82 |
--let $rpl_connection_name= server2 |
| 83 |
--source include/rpl_connection.inc |
| 84 |
SET GLOBAL DEBUG="-d,stop_applier_channel_after_reading_write_rows_log_event"; |
| 85 |
|
| 86 |
# In case MTS, applier coordinator thread will wait for 60 seconds to get the |
| 87 |
# remaining group of events. Setting the timeout to a bigger number to avoid |
| 88 |
# time out on pb2 machines. |
| 89 |
--let $wait_timeout=120 |
| 90 |
--source include/stop_group_replication.inc |
| 91 |
|
| 92 |
--let $wait_timeout=120 |
| 93 |
--source include/start_group_replication.inc |
| 94 |
|
| 95 |
--let $assert_text= Certifier broadcast thread must be running |
| 96 |
--let $assert_cond= [SELECT COUNT(*) FROM performance_schema.threads WHERE name = "thread/group_rpl/THD_certifier_broadcast"] = 1 |
| 97 |
--source include/assert.inc |
| 98 |
|
| 99 |
--echo ######################################################################## |
| 100 |
--echo # 5. On M1: Insert another tuple and do a diff tables with other nodes. |
| 101 |
--echo # (just to check that everything is working fine). |
| 102 |
--let $rpl_connection_name= server1 |
| 103 |
--source include/rpl_connection.inc |
| 104 |
INSERT INTO t1 values (13); |
| 105 |
--source include/rpl_sync.inc |
| 106 |
|
| 107 |
--let $diff_tables=server1:t1, server2:t1, server3:t1 |
| 108 |
--source include/diff_tables.inc |
| 109 |
|
| 110 |
--echo ################ |
| 111 |
--echo # 6. Cleanup. |
| 112 |
DROP TABLE t1; |
| 113 |
--source include/group_replication_end.inc |