mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-02-04 21:02:17 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions
mariadb/mysql-test/suite/multi_source/gtid_ignore_duplicates.test
Marko Mäkelä ddd7d5d8e3 MDEV-24035 Failing assertion: UT_LIST_GET_LEN(lock.trx_locks) == 0 causing disruption and replication failure 
Under unknown circumstances, the SQL layer may wrongly disregard an
invocation of thd_mark_transaction_to_rollback() when an InnoDB
transaction had been aborted (rolled back) due to one of the following errors:
* HA_ERR_LOCK_DEADLOCK
* HA_ERR_RECORD_CHANGED (if innodb_snapshot_isolation=ON)
* HA_ERR_LOCK_WAIT_TIMEOUT (if innodb_rollback_on_timeout=ON)

Such an error used to cause a crash of InnoDB during transaction commit.
These changes aim to catch and report the error earlier, so that not only
this crash can be avoided but also the original root cause be found and
fixed more easily later.

The idea of this fix is from Michael 'Monty' Widenius.

HA_ERR_ROLLBACK: A new error code that will be translated into
ER_ROLLBACK_ONLY, signalling that the current transaction
has been aborted and the only allowed action is ROLLBACK.

trx_t::state: Add TRX_STATE_ABORTED that is like
TRX_STATE_NOT_STARTED, but noting that the transaction had been
rolled back and aborted.

trx_t::is_started(): Replaces trx_is_started().

ha_innobase: Check the transaction state in various places.
Simplify the logic around SAVEPOINT.

ha_innobase::is_valid_trx(): Replaces ha_innobase::is_read_only().

The InnoDB logic around transaction savepoints, commit, and rollback
was unnecessarily complex and might have contributed to this
inconsistency. So, we are simplifying that logic as well.

trx_savept_t: Replace with const undo_no_t*. When we rollback to
a savepoint, all we need to know is the number of undo log records
that must survive.

trx_named_savept_t, DB_NO_SAVEPOINT: Remove. We can store undo_no_t
directly in the space allocated at innobase_hton->savepoint_offset.

fts_trx_create(): Do not copy previous savepoints.

fts_savepoint_rollback(): If a savepoint was not found, roll back
everything after the default savepoint of fts_trx_create().
The test innodb_fts.savepoint is extended to cover this code.

Reviewed by: Vladislav Lesin
Tested by: Matthias Leich
2024-12-12 18:02:00 +02:00

518 lines
16 KiB
Text
Raw Permalink Blame History

--source include/not_embedded.inc
--source include/have_innodb.inc
--source include/have_debug.inc
--disable_query_log
call mtr.add_suppression("InnoDB: Transaction was aborted due to ");
--enable_query_log
--echo *** Test all-to-all replication with --gtid-ignore-duplicates ***
--connect (server_1,127.0.0.1,root,,,$SERVER_MYPORT_1)
--connect (server_2,127.0.0.1,root,,,$SERVER_MYPORT_2)
--connect (server_3,127.0.0.1,root,,,$SERVER_MYPORT_3)
--connect (server_4,127.0.0.1,root,,,$SERVER_MYPORT_4)
# Setup A <-> B, B <-> C, C <-> A, and A -> D.
--connection server_1
SET @old_parallel= @@GLOBAL.slave_parallel_threads;
SET GLOBAL slave_parallel_threads=5;
SET @old_ignore_duplicates= @@GLOBAL.gtid_ignore_duplicates;
SET GLOBAL gtid_ignore_duplicates=1;
SET GLOBAL gtid_domain_id= 1;
SET SESSION gtid_domain_id= 1;
--replace_result $SERVER_MYPORT_2 MYPORT_2
eval CHANGE MASTER 'b2a' TO master_port=$SERVER_MYPORT_2, master_host='127.0.0.1', master_user='root', master_use_gtid=slave_pos;
--replace_result $SERVER_MYPORT_3 MYPORT_3
eval CHANGE MASTER 'c2a' TO master_port=$SERVER_MYPORT_3, master_host='127.0.0.1', master_user='root', master_use_gtid=slave_pos;
set default_master_connection = 'b2a';
START SLAVE;
--source include/wait_for_slave_to_start.inc
set default_master_connection = 'c2a';
START SLAVE;
--source include/wait_for_slave_to_start.inc
set default_master_connection = '';
--connection server_2
SET @old_parallel= @@GLOBAL.slave_parallel_threads;
SET GLOBAL slave_parallel_threads=5;
SET @old_ignore_duplicates= @@GLOBAL.gtid_ignore_duplicates;
SET GLOBAL gtid_ignore_duplicates=1;
SET GLOBAL gtid_domain_id= 2;
SET SESSION gtid_domain_id= 2;
--replace_result $SERVER_MYPORT_1 MYPORT_1
eval CHANGE MASTER 'a2b' TO master_port=$SERVER_MYPORT_1, master_host='127.0.0.1', master_user='root', master_use_gtid=slave_pos;
--replace_result $SERVER_MYPORT_3 MYPORT_3
eval CHANGE MASTER 'c2b' TO master_port=$SERVER_MYPORT_3, master_host='127.0.0.1', master_user='root', master_use_gtid=slave_pos;
set default_master_connection = 'a2b';
START SLAVE;
--source include/wait_for_slave_to_start.inc
set default_master_connection = 'c2b';
START SLAVE;
--source include/wait_for_slave_to_start.inc
set default_master_connection = '';
--connection server_3
SET @old_parallel= @@GLOBAL.slave_parallel_threads;
SET GLOBAL slave_parallel_threads=5;
SET @old_ignore_duplicates= @@GLOBAL.gtid_ignore_duplicates;
SET GLOBAL gtid_ignore_duplicates=1;
SET GLOBAL gtid_domain_id= 3;
SET SESSION gtid_domain_id= 3;
--replace_result $SERVER_MYPORT_1 MYPORT_1
eval CHANGE MASTER 'a2c' TO master_port=$SERVER_MYPORT_1, master_host='127.0.0.1', master_user='root', master_use_gtid=slave_pos;
--replace_result $SERVER_MYPORT_2 MYPORT_2
eval CHANGE MASTER 'b2c' TO master_port=$SERVER_MYPORT_2, master_host='127.0.0.1', master_user='root', master_use_gtid=slave_pos;
set default_master_connection = 'a2c';
START SLAVE;
--source include/wait_for_slave_to_start.inc
set default_master_connection = 'b2c';
START SLAVE;
--source include/wait_for_slave_to_start.inc
set default_master_connection = '';
--connection server_4
SET @old_parallel= @@GLOBAL.slave_parallel_threads;
SET GLOBAL slave_parallel_threads=5;
SET @old_ignore_duplicates= @@GLOBAL.gtid_ignore_duplicates;
SET GLOBAL gtid_ignore_duplicates=1;
SET GLOBAL gtid_domain_id= 1;
SET SESSION gtid_domain_id= 1;
--replace_result $SERVER_MYPORT_1 MYPORT_1
eval CHANGE MASTER 'a2d' TO master_port=$SERVER_MYPORT_1, master_host='127.0.0.1', master_user='root', master_use_gtid=slave_pos;
set default_master_connection = 'a2d';
START SLAVE;
--source include/wait_for_slave_to_start.inc
set default_master_connection = '';
--connection server_1
ALTER TABLE mysql.gtid_slave_pos ENGINE=InnoDB;
CALL mtr.add_suppression("This change will not take full effect until all SQL threads have been restarted");
CREATE TABLE t1 (a INT PRIMARY KEY) ENGINE=InnoDB;
INSERT INTO t1 VALUES (1);
BEGIN;
INSERT INTO t1 VALUES (2);
INSERT INTO t1 VALUES (3);
COMMIT;
INSERT INTO t1 VALUES (4), (5);
INSERT INTO t1 VALUES (6);
--source include/save_master_gtid.inc
--connection server_2
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
--connection server_3
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
--connection server_4
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
--connection server_1
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
# Test that we can connect at a GTID position that has not yet reached
# that master server.
# We stop the connections C->B and A->B, create an event on C, Check that
# the event has reached A (but not B). Then let A stop and re-connect to
# B, which will connect at the new event, which is in the future for B.
--connection server_3
INSERT INTO t1 VALUES (10);
--source include/save_master_gtid.inc
--connection server_2
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 10 ORDER BY a;
STOP SLAVE "c2b";
SET default_master_connection = "c2b";
--source include/wait_for_slave_to_stop.inc
STOP SLAVE "a2b";
SET default_master_connection = "a2b";
--source include/wait_for_slave_to_stop.inc
--connection server_3
INSERT INTO t1 VALUES (11);
--source include/save_master_gtid.inc
--connection server_1
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 10 ORDER BY a;
SET default_master_connection = "b2a";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
--connection server_2
INSERT INTO t1 VALUES (12);
SELECT * FROM t1 WHERE a >= 10 ORDER BY a;
--source include/save_master_gtid.inc
--connection server_1
START SLAVE "b2a";
SET default_master_connection = "b2a";
--source include/wait_for_slave_to_start.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 10 ORDER BY a;
--connection server_2
START SLAVE "c2b";
SET default_master_connection = "c2b";
--source include/wait_for_slave_to_start.inc
START SLAVE "a2b";
SET default_master_connection = "a2b";
--source include/wait_for_slave_to_start.inc
--connection server_1
--source include/save_master_gtid.inc
--connection server_2
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 10 ORDER BY a;
--echo *** MDEV-33475: --gtid-ignore-duplicate can double-apply event in case of parallel replication retry
# Create a bunch of transactions that will cause conflicts and retries.
# The bug was that the retry code was not handling the --gtid-ignore-duplicates
# option, so events could be doubly-applied.
--connection server_2
STOP SLAVE "c2b";
SET default_master_connection = "c2b";
--source include/wait_for_slave_to_stop.inc
STOP SLAVE "a2b";
SET default_master_connection = "a2b";
--source include/wait_for_slave_to_stop.inc
--connection server_1
CREATE TABLE t2 (a INT PRIMARY KEY, b INT) ENGINE=InnoDB;
BEGIN;
--let $i= 0
while ($i < 10) {
eval INSERT INTO t2 VALUES ($i, 0);
inc $i;
}
COMMIT;
--let $i= 0
while ($i < 10) {
BEGIN;
eval INSERT INTO t2 VALUES ($i+10, 100);
eval UPDATE t2 SET b=$i WHERE a<10;
eval INSERT INTO t2 VALUES ($i+20, 200);
COMMIT;
inc $i;
}
SELECT COUNT(*), SUM(a), SUM(b) FROM t2;
--source include/save_master_gtid.inc
--connection server_2
SET @old_mode= @@GLOBAL.slave_parallel_mode;
SET GLOBAL slave_parallel_mode=aggressive;
SET default_master_connection = "a2b";
START SLAVE;
--source include/wait_for_slave_to_start.inc
SET default_master_connection = "c2b";
START SLAVE;
--source include/wait_for_slave_to_start.inc
--source include/sync_with_master_gtid.inc
SELECT COUNT(*), SUM(a), SUM(b) FROM t2;
--connection server_3
--source include/sync_with_master_gtid.inc
SELECT COUNT(*), SUM(a), SUM(b) FROM t2;
--connection server_4
--source include/sync_with_master_gtid.inc
SELECT COUNT(*), SUM(a), SUM(b) FROM t2;
--echo *** Test also with not using parallel replication.
--connection server_1
SET default_master_connection = "b2a";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
SET default_master_connection = "c2a";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
SET GLOBAL slave_parallel_threads=0;
SET default_master_connection = "b2a";
START SLAVE;
--source include/wait_for_slave_to_start.inc
SET default_master_connection = "c2a";
START SLAVE;
--source include/wait_for_slave_to_start.inc
--connection server_2
SET default_master_connection = "a2b";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
SET default_master_connection = "c2b";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
SET GLOBAL slave_parallel_threads=0;
SET default_master_connection = "a2b";
START SLAVE;
--source include/wait_for_slave_to_start.inc
SET default_master_connection = "c2b";
START SLAVE;
--source include/wait_for_slave_to_start.inc
--connection server_3
SET default_master_connection = "a2c";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
SET default_master_connection = "b2c";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
SET GLOBAL slave_parallel_threads=0;
SET default_master_connection = "a2c";
START SLAVE;
--source include/wait_for_slave_to_start.inc
SET default_master_connection = "b2c";
START SLAVE;
--source include/wait_for_slave_to_start.inc
--connection server_4
SET default_master_connection = "a2d";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
SET GLOBAL slave_parallel_threads=0;
SET default_master_connection = "a2d";
START SLAVE;
--source include/wait_for_slave_to_start.inc
--connection server_2
INSERT INTO t1 VALUES (21);
BEGIN;
INSERT INTO t1 VALUES (22);
INSERT INTO t1 VALUES (23);
COMMIT;
INSERT INTO t1 VALUES (24), (25);
INSERT INTO t1 VALUES (26);
--source include/save_master_gtid.inc
--connection server_1
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 20 ORDER BY a;
--connection server_3
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 20 ORDER BY a;
--connection server_4
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 20 ORDER BY a;
--connection server_2
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 20 ORDER BY a;
--echo *** MDEV-8354: out-of-order error with --gtid-ignore-duplicates and row-based replication ***
# Have only A->C A->B initially.
--connection server_1
SET default_master_connection = "b2a";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
SET default_master_connection = "c2a";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
--connection server_2
SET default_master_connection = "c2b";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
--connection server_3
SET default_master_connection = "b2c";
STOP SLAVE;
--source include/wait_for_slave_to_stop.inc
SET @old_slave_mode=@@GLOBAL.slave_exec_mode;
SET GLOBAL slave_exec_mode=IDEMPOTENT;
SET @old_strict=@@GLOBAL.gtid_strict_mode;
SET GLOBAL gtid_strict_mode=1;
SET @old_dbug=@@GLOBAL.debug_dbug;
# This will inject a small sleep that helps trigger the race. I did not manage
# to create a non-sleeping version with debug_sync for this; the problem is
# that once the bug is fixed, the race becomes impossible, so even with
# debug_sync at best we can check that the debug_sync times out. Which is
# just another way of adding a sleep.
#
# The bug was a race at this point where another multi-source connection
# could incorrectly re-apply the same GTID, in case of row-based replication.
SET GLOBAL debug_dbug="+d,inject_sleep_gtid_100_x_x";
--connection server_1
SET @old_domain=@@SESSION.gtid_domain_id;
SET @old_format=@@SESSION.binlog_format;
SET SESSION gtid_domain_id=100;
SET SESSION binlog_format='row';
INSERT INTO t1 VALUES (30);
INSERT INTO t1 VALUES (31);
INSERT INTO t1 VALUES (32);
INSERT INTO t1 VALUES (33);
INSERT INTO t1 VALUES (34);
INSERT INTO t1 VALUES (35);
INSERT INTO t1 VALUES (36);
INSERT INTO t1 VALUES (37);
INSERT INTO t1 VALUES (38);
INSERT INTO t1 VALUES (39);
INSERT INTO t1 VALUES (40);
INSERT INTO t1 VALUES (41);
INSERT INTO t1 VALUES (42);
INSERT INTO t1 VALUES (43);
INSERT INTO t1 VALUES (44);
INSERT INTO t1 VALUES (45);
INSERT INTO t1 VALUES (46);
INSERT INTO t1 VALUES (47);
INSERT INTO t1 VALUES (48);
INSERT INTO t1 VALUES (49);
SET SESSION gtid_domain_id=@old_domain;
SET SESSION binlog_format=@old_format;
--source include/save_master_gtid.inc
--connection server_2
--source include/sync_with_master_gtid.inc
INSERT INTO t1 VALUES (50);
--let $gtid=`SELECT @@last_gtid`
--source include/save_master_gtid.inc
--connection server_3
SET default_master_connection = "b2c";
START SLAVE;
--source include/wait_for_slave_to_start.inc
--replace_result $gtid GTID
eval SELECT MASTER_GTID_WAIT("$gtid", 30);
# The bug occurred here, the slave would get an out-of-order binlog error
# due to trying to re-apply the 100-x-x transaction.
# Restart stopped multi-source connections, and sync up.
--connection server_1
SET default_master_connection = "b2a";
START SLAVE;
--source include/wait_for_slave_to_start.inc
SET default_master_connection = "c2a";
START SLAVE;
--source include/wait_for_slave_to_start.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 30 ORDER BY a;
--connection server_2
SET default_master_connection = "c2b";
START SLAVE;
--source include/wait_for_slave_to_start.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 30 ORDER BY a;
--connection server_3
--source include/sync_with_master_gtid.inc
SET GLOBAL debug_dbug=@old_dbug;
SELECT * FROM t1 WHERE a >= 30 ORDER BY a;
SET GLOBAL slave_exec_mode=@old_slave_mode;
SET GLOBAL gtid_strict_mode=@old_strict;
--echo *** MDEV-8496: gtid_ignore_duplicates treats gtid_seq_no as 32-bit ***
--connection server_1
SET @old_domain= @@SESSION.gtid_domain_id;
SET SESSION gtid_domain_id=102;
SET SESSION gtid_seq_no=4294967294;
INSERT INTO t1 VALUES (60);
INSERT INTO t1 VALUES (61);
INSERT INTO t1 VALUES (62);
# The bug was an overflow, the seq_no value 4294967296 (2**32) was treated
# as 0, causing the last transaction to be ignored.
SET SESSION gtid_domain_id= @old_domain;
--source include/save_master_gtid.inc
--connection server_4
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 60 ORDER BY a;
--connection server_2
SET default_master_connection = "c2b";
--source include/sync_with_master_gtid.inc
SET default_master_connection = "a2b";
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 60 ORDER BY a;
--connection server_3
SET default_master_connection = "b2c";
--source include/sync_with_master_gtid.inc
SET default_master_connection = "a2c";
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 60 ORDER BY a;
# Clean up.
--connection server_1
SET GLOBAL gtid_domain_id=0;
--sorted_result
STOP ALL SLAVES;
SET GLOBAL slave_parallel_threads= @old_parallel;
SET GLOBAL gtid_ignore_duplicates= @old_ignore_duplicates;
--connection server_2
SET GLOBAL gtid_domain_id=0;
--sorted_result
STOP ALL SLAVES;
SET GLOBAL slave_parallel_threads= @old_parallel;
SET GLOBAL slave_parallel_mode= @old_mode;
SET GLOBAL gtid_ignore_duplicates= @old_ignore_duplicates;
--connection server_3
SET GLOBAL gtid_domain_id=0;
--sorted_result
STOP ALL SLAVES;
SET GLOBAL slave_parallel_threads= @old_parallel;
SET GLOBAL gtid_ignore_duplicates= @old_ignore_duplicates;
--connection server_4
SET GLOBAL gtid_domain_id=0;
--sorted_result
STOP ALL SLAVES;
SET GLOBAL slave_parallel_threads= @old_parallel;
SET GLOBAL gtid_ignore_duplicates= @old_ignore_duplicates;
--connection server_1
DROP TABLE t1, t2;
ALTER TABLE mysql.gtid_slave_pos ENGINE=Aria;
--source include/reset_master_slave.inc
--disconnect server_1
--connection server_2
DROP TABLE t1, t2;
ALTER TABLE mysql.gtid_slave_pos ENGINE=Aria;
--source include/reset_master_slave.inc
--disconnect server_2
--connection server_3
DROP TABLE t1, t2;
ALTER TABLE mysql.gtid_slave_pos ENGINE=Aria;
--source include/reset_master_slave.inc
--disconnect server_3
--connection server_4
DROP TABLE t1, t2;
ALTER TABLE mysql.gtid_slave_pos ENGINE=Aria;
--source include/reset_master_slave.inc
--disconnect server_4
Reference in a new issue View git blame Copy permalink