mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-09-13 21:02:21 +02:00
Code Issues Projects Releases Packages Wiki Activity
mariadb/mysql-test/suite/rpl/t/rpl_gtid_startpos.test
Kristian Nielsen 0a68328673 MDEV-34705: Binlog-in-engine: Protect against concurrent RESET MASTER and dump threads 
This is actually an existing problem in the old binlog implementation, and
this patch is applicable to old binlog also. The problem is that RESET
MASTER can run concurrently with binlog dump threads / connected slaves.
This will remove the binlog from under the feet of the reader, which can
cause all sorts of strange behaviour.

This patch fixes the problem by disallowing to run RESET MASTER when dump
threads (or other RESET MASTER or SHOW BINARY LOGS) are running. An error is
thrown in this case, user must stop slaves and/or kill dump threads to make
the RESET MASTER go through. A slave that connects in the middle of RESET
MASTER will wait for it to complete.

Fix a lot of test cases to kill any lingering dump threads before doing
RESET MASTER, mostly just by sourcing include/kill_binlog_dump_threads.inc.

Signed-off-by: Kristian Nielsen <knielsen@knielsen-hq.org>
2025-06-11 11:32:10 +02:00

364 lines
11 KiB
Text
Raw Permalink Blame History

--source include/have_innodb.inc
--let $rpl_topology=1->2
--source include/rpl_init.inc
--echo *** Test connecting with empty GTID state to start from very beginning of binlog ***
--connection server_2
--source include/stop_slave.inc
RESET MASTER;
RESET SLAVE;
--connection server_1
--source include/kill_binlog_dump_threads.inc
RESET MASTER;
# Create an empty binlog file, to check that empty binlog state is handled correctly.
FLUSH LOGS;
CREATE TABLE t1 (a INT PRIMARY KEY) ENGINE=InnoDB;
--save_master_pos
--connection server_2
SET GLOBAL gtid_slave_pos="";
--replace_result $MASTER_MYPORT MASTER_PORT
eval CHANGE MASTER TO master_host = '127.0.0.1', master_port = $MASTER_MYPORT,
MASTER_USE_GTID=SLAVE_POS;
--source include/start_slave.inc
--sync_with_master
SELECT * FROM t1;
--connection server_1
INSERT INTO t1 VALUES (1);
--save_master_pos
--connection server_2
--sync_with_master
SELECT * FROM t1;
--source include/stop_slave.inc
--echo *** Test that master gives error when slave asks for empty gtid pos and binlog files have been purged. ***
--connection server_1
FLUSH LOGS;
INSERT INTO t1 VALUES (2);
--save_master_pos
--let $purge_binlogs_to=master-bin.000003
--source include/wait_for_purge.inc
--source include/show_binary_logs.inc
--connection server_2
SET GLOBAL gtid_slave_pos="";
--replace_result $MASTER_MYPORT MASTER_PORT
eval CHANGE MASTER TO master_host = '127.0.0.1', master_port = $MASTER_MYPORT,
MASTER_USE_GTID=SLAVE_POS;
START SLAVE;
--let $slave_io_errno= 1236
--source include/wait_for_slave_io_error.inc
--source include/stop_slave_sql.inc
--replace_result $MASTER_MYPORT MASTER_PORT
eval CHANGE MASTER TO master_host = '127.0.0.1', master_port = $MASTER_MYPORT,
MASTER_LOG_FILE="master-bin.000003", MASTER_LOG_POS=4, MASTER_USE_GTID=NO;
--source include/start_slave.inc
--sync_with_master
SELECT * FROM t1 ORDER BY a;
SET sql_log_bin=0;
call mtr.add_suppression('Could not find GTID state requested by slave in any binlog files');
SET sql_log_bin=1;
--echo *** Test that we give warning when explict @@gtid_slave_pos=xxx that conflicts with what is in our binary log ***
--source include/stop_slave.inc
--connection server_1
INSERT INTO t1 VALUES(3);
--source include/save_master_gtid.inc
--connection server_2
SET GLOBAL gtid_slave_pos='0-1-3';
--replace_result $MASTER_MYPORT MASTER_PORT
eval CHANGE MASTER TO master_host = '127.0.0.1', master_port = $MASTER_MYPORT,
MASTER_USE_GTID=SLAVE_POS;
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER by a;
--source include/stop_slave.inc
--connection server_1
INSERT INTO t1 VALUES (4);
--source include/save_master_gtid.inc
--connection server_2
# Now add some local transactions that conflict with the GTID position
# being set for @@gtid_slave_pos.
INSERT INTO t1 VALUES (10);
DELETE FROM t1 WHERE a=10;
SET GLOBAL gtid_slave_pos='0-1-4';
# Try again after RESET MASTER to remove the conflicting binlog.
RESET MASTER;
SET GLOBAL gtid_slave_pos='0-1-4';
START SLAVE;
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER by a;
--connection server_1
DROP TABLE t1;
--save_master_pos
--connection server_2
--sync_with_master
--echo *** MDEV-4275: I/O thread restart duplicates events in relay log ***
--connection server_2
--source include/stop_slave.inc
RESET SLAVE ALL;
RESET MASTER;
--connection server_1
--source include/kill_binlog_dump_threads.inc
RESET MASTER;
--connection server_2
SET GLOBAL gtid_slave_pos='';
--replace_result $MASTER_MYPORT MASTER_PORT
eval CHANGE MASTER TO master_host='127.0.0.1', master_port=$MASTER_MYPORT, master_user='root', master_use_gtid=slave_pos;
--source include/start_slave.inc
--connection server_1
CREATE TABLE t1 (a INT PRIMARY KEY);
INSERT INTO t1 VALUES (1);
--save_master_pos
--connection server_2
--sync_with_master
SELECT * FROM t1;
--source include/stop_slave_io.inc
START SLAVE IO_THREAD;
--source include/wait_for_slave_io_to_start.inc
--connection server_1
INSERT INTO t1 VALUES (2);
--save_master_pos
--connection server_2
--sync_with_master
SELECT * FROM t1 ORDER BY a;
--echo *** MDEV-4329: GTID_SLAVE_POS='' is not checked for conflicts with binlog ***
# Test starting the slave completely from scratch, deleting all tables and
# replicating from the start of the master's binlog. This requires RESET
# MASTER is run on the slave to avoid old junk in the binlog. The bug was
# that the code did not catch the error of missing RESET MASTER when an
# empty GTID_SLAVE_POS='' was specified.
--connection server_2
--source include/stop_slave.inc
DROP TABLE t1;
RESET SLAVE;
SET GLOBAL gtid_slave_pos="";
RESET MASTER;
SET GLOBAL gtid_slave_pos="";
--source include/start_slave.inc
--sync_with_master
SELECT * FROM t1 ORDER BY a;
# Same thing, but this time using SQL_LOG_BIN=0 to avoid polluting the
# slave binlog.
--connection server_2
--source include/stop_slave.inc
SET SQL_LOG_BIN=0;
DROP TABLE t1;
SET SQL_LOG_BIN=1;
RESET SLAVE;
SET GLOBAL gtid_slave_pos="";
# Ensure that the slave fails because of missing table to be dropped
SET @save_slave_ddl_exec_mode=@@global.slave_ddl_exec_mode;
SET GLOBAL slave_ddl_exec_mode=STRICT;
--source include/start_slave.inc
--sync_with_master
SELECT * FROM t1 ORDER BY a;
--echo *** Test that RESET SLAVE clears the Using_Gtid flag. ***
--source include/stop_slave.inc
--let $master_use_gtid_option= No
--source include/reset_slave.inc
--let $status_items= Using_Gtid
--source include/show_slave_status.inc
# Starting the slave now reverts to old-style position which defaults to
# the first non-purged binlog file on the master.
# This should give error due to table already existing.
START SLAVE;
--let $slave_sql_errno= 1050
--source include/wait_for_slave_sql_error.inc
# Going back to using GTID should fix things.
STOP SLAVE IO_THREAD;
CHANGE MASTER TO MASTER_USE_GTID=SLAVE_POS;
--source include/start_slave.inc
--connection server_1
INSERT INTO t1 VALUES(3);
--source include/save_master_gtid.inc
--connection server_2
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
SET SQL_LOG_BIN=0;
call mtr.add_suppression("Slave: Table 't1' already exists error.* 1050");
SET SQL_LOG_BIN=1;
--echo *** Test reconnecting slave with GTID after purge logs on master. ***
--connection server_1
FLUSH LOGS;
INSERT INTO t1 VALUES (4);
--source include/save_master_gtid.inc
--connection server_2
--source include/sync_with_master_gtid.inc
--source include/stop_slave.inc
--connection server_1
FLUSH LOGS;
FLUSH LOGS;
--let $purge_binlogs_to=master-bin.000004
--source include/wait_for_purge.inc
--source include/show_binary_logs.inc
INSERT INTO t1 VALUES (5);
--source include/save_master_gtid.inc
--connection server_2
--replace_result $MASTER_MYPORT MASTER_PORT
eval CHANGE MASTER TO master_host = '127.0.0.1', master_port = $MASTER_MYPORT;
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
--echo *** Test modifying binlog on slave and the effect on GTID state. ***
--connection server_2
--source include/stop_slave.inc
RESET MASTER;
SET GLOBAL gtid_slave_pos="";
--connection server_1
--source include/kill_binlog_dump_threads.inc
RESET MASTER;
TRUNCATE TABLE t1;
INSERT INTO t1 VALUES (10); # Will be GTID 0-1-2
--source include/save_master_gtid.inc
--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1;
--let $value= query_get_value(SHOW SLAVE STATUS, "Using_Gtid", 1)
eval SELECT '$value' AS Using_Gtid;
--let $value= query_get_value(SHOW ALL SLAVES STATUS, "Gtid_Slave_Pos", 1)
eval SELECT '$value' AS Gtid_Slave_Pos;
UPDATE t1 SET a=9 WHERE a=10;
UPDATE t1 SET a=10 WHERE a=9;
--let $value= query_get_value(SHOW ALL SLAVES STATUS, "Gtid_Slave_Pos", 1)
eval SELECT '$value' AS Gtid_Slave_Pos;
--source include/stop_slave.inc
SET GLOBAL gtid_slave_pos='0-1-2';
RESET MASTER;
--let $value= query_get_value(SHOW ALL SLAVES STATUS, "Gtid_Slave_Pos", 1)
eval SELECT '$value' AS Gtid_Slave_Pos;
SET GLOBAL gtid_slave_pos='0-1-2';
--source include/start_slave.inc
--let $value= query_get_value(SHOW ALL SLAVES STATUS, "Gtid_Slave_Pos", 1)
eval SELECT '$value' AS Gtid_Slave_Pos;
--echo *** MDEV-4483: Slave loses traditional master coordinates immediately on CHANGE MASTER TO MASTER_USE_GTID = 1 ***
--connection server_2
--source include/stop_slave.inc
DROP TABLE t1;
RESET SLAVE ALL;
CHANGE MASTER TO MASTER_USE_GTID=NO;
RESET MASTER;
SET GLOBAL gtid_slave_pos= "";
# Set up old-style replication. The bug was that CHANGE MASTER would clear
# out the old-style binlog/relaylog coordinates when it should not.
--replace_result $MASTER_MYPORT MASTER_PORT
eval CHANGE MASTER TO master_host='127.0.0.1', master_port=$MASTER_MYPORT, master_user='root', master_use_gtid=no, master_log_file="", master_log_pos= 4;
--connection server_1
DROP TABLE t1;
--source include/kill_binlog_dump_threads.inc
RESET MASTER;
CREATE TABLE t1 (a INT PRIMARY KEY);
--save_master_pos
--connection server_2
--source include/start_slave.inc
--sync_with_master
--source include/stop_slave.inc
--connection server_1
INSERT INTO t1 VALUES (1);
--let $log_file1= query_get_value(SHOW MASTER STATUS, "File", 1)
--let $log_pos1= query_get_value(SHOW MASTER STATUS, "Position", 1)
INSERT INTO t1 VALUES (2);
--let $log_file2= query_get_value(SHOW MASTER STATUS, "File", 1)
--let $log_pos2= query_get_value(SHOW MASTER STATUS, "Position", 1)
--save_master_pos
--connection server_2
# Get the slave to a point where the IO thread has fetched one event ahead
# of the SQL thread (we want to test that CHANGE MASTER does not mess with
# existing relay logs).
--replace_result $log_file1 LOG_FILE1 $log_pos1 LOG_POS1
eval START SLAVE UNTIL master_log_file='$log_file1', master_log_pos=$log_pos1;
--source include/wait_for_slave_sql_to_stop.inc
SELECT * FROM t1;
--let $slave_param= Read_Master_Log_Pos
--let $slave_param_value= $log_pos2
--source include/wait_for_slave_param.inc
--source include/stop_slave_io.inc
# Test that we can change to GTID and back without loosing our
# old-style slave position.
CHANGE MASTER TO master_use_gtid=slave_pos;
# Unknown GTID, so slave will fail to connect.
SET GLOBAL gtid_slave_pos="0-42-42";
SET sql_log_bin=0;
call mtr.add_suppression("Error: connecting slave requested to start from GTID");
SET sql_log_bin=1;
START SLAVE;
--let $slave_io_errno= 1236
--source include/wait_for_slave_io_error.inc
STOP SLAVE SQL_THREAD;
--source include/wait_for_slave_sql_to_stop.inc
CHANGE MASTER TO master_use_gtid=no;
--source include/start_slave.inc
--sync_with_master
SELECT * FROM t1 ORDER BY a;
# Clean up.
--connection server_1
DROP TABLE t1;
--connection server_2
set @@global.slave_ddl_exec_mode=@save_slave_ddl_exec_mode;
--source include/rpl_end.inc
Reference in a new issue View git blame Copy permalink