path: root/mysql-test/suite/rpl/t/rpl_parallel_optimistic.test

--source include/have_innodb.inc
--source include/have_debug.inc
--source include/have_debug_sync.inc
--let $rpl_topology=1->2
--source include/rpl_init.inc

call mtr.add_suppression("Deadlock found when trying to get lock; try restarting transaction");
call mtr.add_suppression("Can't find record in 't1'");
call mtr.add_suppression("Can't find record in 't2'");

--connection server_1
ALTER TABLE mysql.gtid_slave_pos ENGINE=InnoDB;
CREATE TABLE t1 (a int PRIMARY KEY, b INT) ENGINE=InnoDB;
--save_master_pos

--connection server_2
--sync_with_master
SET @old_parallel_threads=@@GLOBAL.slave_parallel_threads;
--source include/stop_slave.inc
SET GLOBAL slave_parallel_threads=10;
CHANGE MASTER TO master_use_gtid=slave_pos;
SET @old_parallel_mode=@@GLOBAL.slave_parallel_mode;
SET GLOBAL slave_parallel_mode='optimistic';
# Run the first part of the test with high batch size and see that
# old rows remain in the table.
SET @old_gtid_cleanup_batch_size= @@GLOBAL.gtid_cleanup_batch_size;
SET GLOBAL gtid_cleanup_batch_size= 1000000;


--connection server_1

INSERT INTO t1 VALUES(1,1);
BEGIN;
INSERT INTO t1 VALUES(2,1);
INSERT INTO t1 VALUES(3,1);
COMMIT;

# Do a bunch of INSERT/DELETE on the same rows, bound to conflict.
# We will get a lot of rollbacks, probably, but they should be handled without
# any visible errors.

DELETE FROM t1 WHERE a=2;
INSERT INTO t1 VALUES (2,2);
DELETE FROM t1 WHERE a=2;
INSERT INTO t1 VALUES (2,3);
DELETE FROM t1 WHERE a=2;
INSERT INTO t1 VALUES (2,4);
DELETE FROM t1 WHERE a=2;
INSERT INTO t1 VALUES (2,5);

DELETE FROM t1 WHERE a=3;
INSERT INTO t1 VALUES(3,2);
DELETE FROM t1 WHERE a=1;
INSERT INTO t1 VALUES(1,2);
DELETE FROM t1 WHERE a=3;
INSERT INTO t1 VALUES(3,3);

DELETE FROM t1 WHERE a=2;
INSERT INTO t1 VALUES (2,6);
--source include/save_master_gtid.inc
SELECT * FROM t1 ORDER BY a;

--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
#SHOW STATUS LIKE 'Slave_retried_transactions';


--echo *** Test a bunch of non-transactional/DDL event groups. ***

--connection server_2
--source include/stop_slave.inc

--connection server_1

INSERT INTO t1 VALUES (4,4);
INSERT INTO t1 VALUES (5,5);
CREATE TABLE t2 (a INT PRIMARY KEY) ENGINE=InnoDB;
INSERT INTO t2 VALUES (1);
CREATE TABLE t3 (a INT PRIMARY KEY) ENGINE=MyISAM;
ALTER TABLE t2 ADD b INT;
INSERT INTO t2 VALUES (2,2);
ALTER TABLE t2 DROP b;
INSERT INTO t2 VALUES (3);
ALTER TABLE t2 ADD c INT;
INSERT INTO t2 VALUES (4,5);
INSERT INTO t2 VALUES (5,5);
INSERT INTO t3 VALUES (1);
UPDATE t2 SET c=NULL WHERE a=4;
ALTER TABLE t2 ADD UNIQUE (c);
INSERT INTO t2 VALUES (6,6);
UPDATE t2 SET c=c+100 WHERE a=2;
INSERT INTO t3(a) VALUES (2);
DELETE FROM t3 WHERE a=2;
INSERT INTO t3(a) VALUES (2);
DELETE FROM t3 WHERE a=2;
ALTER TABLE t3 CHANGE a c INT NOT NULL;
INSERT INTO t3(c) VALUES (2);
DELETE FROM t3 WHERE c=2;
INSERT INTO t3 SELECT a+200 FROM t2;
DELETE FROM t3 WHERE c >= 200;
INSERT INTO t3 SELECT a+200 FROM t2;
--source include/save_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
SELECT * FROM t2 ORDER BY a;
SELECT * FROM t3 ORDER BY c;

--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
SELECT * FROM t2 ORDER BY a;
SELECT * FROM t3 ORDER BY c;
# Check that we have a bunch of old rows left-over - they were not deleted
# due to high @@gtid_cleanup_batch_size. Then set a low
# @@gtid_cleanup_batch_size so we can test that rows start being deleted.
SELECT IF(COUNT(*) >= 30, "OK", CONCAT("Error: too few old rows found: ", COUNT(*)))
  FROM mysql.gtid_slave_pos;
SET GLOBAL gtid_cleanup_batch_size=1;


--echo *** Test @@skip_parallel_replication. ***

--connection server_2
--source include/stop_slave.inc
--let $retry1= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', Value, 1)

--connection server_1
# We do a bunch of conflicting transactions on the master with
# skip_parallel_replication set to true, and check that we do not
# get any retries on the slave.

UPDATE t1 SET b=10 WHERE a=3;
SET SESSION skip_parallel_replication=1;
UPDATE t1 SET b=20 WHERE a=3;
UPDATE t1 SET b=30 WHERE a=3;
UPDATE t1 SET b=50 WHERE a=3;
UPDATE t1 SET b=80 WHERE a=3;
UPDATE t1 SET b=130 WHERE a=3;
UPDATE t1 SET b=210 WHERE a=3;
UPDATE t1 SET b=340 WHERE a=3;
UPDATE t1 SET b=550 WHERE a=3;
UPDATE t1 SET b=890 WHERE a=3;
SET SESSION skip_parallel_replication=0;
SELECT * FROM t1 ORDER BY a;
--source include/save_master_gtid.inc

--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
--let $retry2= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', Value, 1)
--disable_query_log
eval SELECT IF($retry1=$retry2, "Ok, no retry",
       CONCAT("ERROR: ", $retry2-$retry1, " retries during replication (was ",
              $retry1, " now ", $retry2, ")")) AS status;
--enable_query_log


--echo *** Test that we do not replicate in parallel transactions that had row lock waits on the master ***

--connection server_2
--source include/stop_slave.inc
--let $retry1= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', Value, 1)

--connection server_1
# Setup a bunch of transactions that all needed to wait.
--connect (m1,127.0.0.1,root,,test,$SERVER_MYPORT_1,)
--connect (m2,127.0.0.1,root,,test,$SERVER_MYPORT_1,)
--connect (m3,127.0.0.1,root,,test,$SERVER_MYPORT_1,)
--connect (m4,127.0.0.1,root,,test,$SERVER_MYPORT_1,)
--connect (m5,127.0.0.1,root,,test,$SERVER_MYPORT_1,)
--connect (m6,127.0.0.1,root,,test,$SERVER_MYPORT_1,)
--connect (m7,127.0.0.1,root,,test,$SERVER_MYPORT_1,)
--connect (m8,127.0.0.1,root,,test,$SERVER_MYPORT_1,)

--connection default
BEGIN; UPDATE t1 SET b=b+1 WHERE a=3;

--connection m1
SET debug_sync='thd_report_wait_for SIGNAL waiting1';
send UPDATE t1 SET b=1001 WHERE a=3;
--connection default
SET debug_sync='now WAIT_FOR waiting1';

--connection m2
BEGIN;
UPDATE t1 SET b=1002 WHERE a=5;
SET debug_sync='thd_report_wait_for SIGNAL waiting2';
send UPDATE t1 SET b=102 WHERE a=3;
--connection default
SET debug_sync='now WAIT_FOR waiting2';

UPDATE t1 SET b=1000 WHERE a=1;
--connection m3
SET debug_sync='thd_report_wait_for SIGNAL waiting3';
send UPDATE t1 SET b=1003 WHERE a=5;
--connection default
SET debug_sync='now WAIT_FOR waiting3';

--connection m4
SET debug_sync='thd_report_wait_for SIGNAL waiting4';
send UPDATE t1 SET b=1004 WHERE a=3;
--connection default
SET debug_sync='now WAIT_FOR waiting4';

--connection m5
SET debug_sync='thd_report_wait_for SIGNAL waiting5';
send UPDATE t1 SET b=1005 WHERE a=5;
--connection default
SET debug_sync='now WAIT_FOR waiting5';

--connection m6
SET debug_sync='thd_report_wait_for SIGNAL waiting6';
send UPDATE t1 SET b=1006 WHERE a=1;
--connection default
SET debug_sync='now WAIT_FOR waiting6';

--connection m7
SET debug_sync='thd_report_wait_for SIGNAL waiting7';
send UPDATE t1 SET b=1007 WHERE a=5;
--connection default
SET debug_sync='now WAIT_FOR waiting7';

--connection m8
SET debug_sync='thd_report_wait_for SIGNAL waiting8';
send UPDATE t1 SET b=1008 WHERE a=3;
--connection default
SET debug_sync='now WAIT_FOR waiting8';

--connection default
COMMIT;
--connection m1
REAP;
--connection m2
REAP;
COMMIT;
--connection m3
REAP;
--connection m4
REAP;
--connection m5
REAP;
--connection m6
REAP;
--connection m7
REAP;
--connection m8
REAP;
--connection default
SET debug_sync='RESET';
SELECT * FROM t1 ORDER BY a;
--source include/save_master_gtid.inc

--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
--let $retry2= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', Value, 1)
--disable_query_log
eval SELECT IF($retry1=$retry2, "Ok, no retry",
       CONCAT("ERROR: ", $retry2-$retry1, " retries during replication (was ",
              $retry1, " now ", $retry2, ")")) AS status;
--enable_query_log


--echo *** Test that we replicate correctly when using READ COMMITTED and binlog_format=MIXED on the slave ***

--connection server_2
--source include/stop_slave.inc
SET @old_format= @@GLOBAL.binlog_format;
# Use MIXED format; we cannot binlog ROW events on slave in STATEMENT format.
SET GLOBAL binlog_format= MIXED;
SET @old_isolation= @@GLOBAL.tx_isolation;
SET GLOBAL TRANSACTION ISOLATION LEVEL READ COMMITTED;
# Reset the worker threads to make the new settings take effect.
SET GLOBAL slave_parallel_threads=0;
SET GLOBAL slave_parallel_threads=10;

--connection server_1
DROP TABLE t1, t2;
CREATE TABLE t1 (a int PRIMARY KEY, b INT) ENGINE=InnoDB;
CREATE TABLE t2 (a int PRIMARY KEY, b INT) ENGINE=InnoDB;
INSERT INTO t1 VALUES (1,0), (2,0), (3,0);
INSERT INTO t2 VALUES (1,0), (2,0);
INSERT INTO t1 SELECT 4, COUNT(*) FROM t2;
INSERT INTO t2 SELECT 4, COUNT(*) FROM t1;

INSERT INTO t1 SELECT 5, COUNT(*) FROM t2;
INSERT INTO t2 SELECT 5, COUNT(*) FROM t1;

INSERT INTO t2 SELECT 6, COUNT(*) FROM t1;
INSERT INTO t1 SELECT 6, COUNT(*) FROM t2;

INSERT INTO t1 SELECT 7, COUNT(*) FROM t2;
INSERT INTO t2 SELECT 7, COUNT(*) FROM t1;

INSERT INTO t2 SELECT 8, COUNT(*) FROM t1;
INSERT INTO t1 SELECT 8, COUNT(*) FROM t2;

INSERT INTO t2 SELECT 9, COUNT(*) FROM t1;
INSERT INTO t1 SELECT 9, COUNT(*) FROM t2;

INSERT INTO t1 SELECT 10, COUNT(*) FROM t2;
INSERT INTO t2 SELECT 10, COUNT(*) FROM t1;

SELECT * FROM t1 ORDER BY a;
SELECT * FROM t2 ORDER BY a;
--source include/save_master_gtid.inc

--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 ORDER BY a;
SELECT * FROM t2 ORDER BY a;

--source include/stop_slave.inc
SET GLOBAL binlog_format= @old_format;
SET GLOBAL tx_isolation= @old_isolation;
--source include/start_slave.inc


--echo *** MDEV-7888: ANALYZE TABLE does wakeup_subsequent_commits(), causing wrong binlog order and parallel replication hang ***

--connection server_1
DROP TABLE t1, t2, t3;
CREATE TABLE t1 (a INT PRIMARY KEY, b INT) ENGINE=InnoDB;
CREATE TABLE t2 (a INT PRIMARY KEY, b INT) ENGINE=InnoDB;
CREATE TABLE t3 (a INT PRIMARY KEY, b INT) ENGINE=MyISAM;
INSERT INTO t2 VALUES (1,1), (2,1), (3,1), (4,1), (5,1);
--source include/save_master_gtid.inc

--connection server_2
--source include/sync_with_master_gtid.inc
--source include/stop_slave.inc
SET @old_debug= @@GLOBAL.debug_dbug;
SET GLOBAL debug_dbug= '+d,inject_analyze_table_sleep';

--connection server_1
# The bug was that ANALYZE TABLE would call
# wakeup_subsequent_commits() too early, allowing the following
# transaction in the same group to run ahead and binlog and free the
# GCO. Then we get wrong binlog order and later access freed GCO,
# which causes lost wakeup of following GCO and thus replication hang.
# We injected a small sleep in ANALYZE to make the race easier to hit (this
# can only cause false negatives in versions with the bug, not false positives,
# so sleep is ok here. And it's in general not possible to trigger reliably
# the race with debug_sync, since the bugfix makes the race impossible).

ALTER TABLE t2 COMMENT "123abc";
ANALYZE TABLE t2;
INSERT INTO t1 VALUES (1,2);
INSERT INTO t1 VALUES (2,2);
INSERT INTO t1 VALUES (3,2);
INSERT INTO t1 VALUES (4,2);
INSERT INTO t3 VALUES (1,3);
ALTER TABLE t2 COMMENT "hello, world";
BEGIN;
INSERT INTO t1 VALUES (5,4);
INSERT INTO t1 VALUES (6,4);
INSERT INTO t1 VALUES (7,4);
INSERT INTO t1 VALUES (8,4);
INSERT INTO t1 VALUES (9,4);
INSERT INTO t1 VALUES (10,4);
INSERT INTO t1 VALUES (11,4);
INSERT INTO t1 VALUES (12,4);
INSERT INTO t1 VALUES (13,4);
INSERT INTO t1 VALUES (14,4);
INSERT INTO t1 VALUES (15,4);
INSERT INTO t1 VALUES (16,4);
INSERT INTO t1 VALUES (17,4);
INSERT INTO t1 VALUES (18,4);
INSERT INTO t1 VALUES (19,4);
INSERT INTO t1 VALUES (20,4);
COMMIT;
INSERT INTO t1 VALUES (21,5);
INSERT INTO t1 VALUES (22,5);

SELECT * FROM t1 ORDER BY a;
SELECT * FROM t2 ORDER BY a;
SELECT * FROM t3 ORDER BY a;
--source include/save_master_gtid.inc

--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc

SELECT * FROM t1 ORDER BY a;
SELECT * FROM t2 ORDER BY a;
SELECT * FROM t3 ORDER BY a;

--source include/stop_slave.inc
SET GLOBAL debug_dbug= @old_debug;
--source include/start_slave.inc

--echo *** MDEV-7929: record_gtid() for non-transactional event group calls wakeup_subsequent_commits() too early, causing slave hang. ***

--connection server_2
--source include/stop_slave.inc
SET @old_dbug= @@GLOBAL.debug_dbug;
# The bug was that record_gtid(), when there is no existing transaction from
# a DML event being replicated, would commit its own transaction. This wrongly
# caused wakeup_subsequent_commits(), with similar consequences as MDEV-7888
# above. We simulate this condition with a small sleep in record_gtid() for
# a specific ANALYZE that we binlog with server id 100.
SET GLOBAL debug_dbug= '+d,inject_record_gtid_serverid_100_sleep';

--connection server_1

ALTER TABLE t3 COMMENT "DDL statement 1";
INSERT INTO t1 VALUES (30,0);
INSERT INTO t1 VALUES (31,0);
INSERT INTO t1 VALUES (32,0);
INSERT INTO t1 VALUES (33,0);
INSERT INTO t1 VALUES (34,0);
INSERT INTO t1 VALUES (35,0);
INSERT INTO t1 VALUES (36,0);
SET @old_server_id= @@SESSION.server_id;
SET SESSION server_id= 100;
ANALYZE TABLE t2;
SET SESSION server_id= @old_server_id;
INSERT INTO t1 VALUES (37,0);
ALTER TABLE t3 COMMENT "DDL statement 2";
INSERT INTO t1 VALUES (38,0);
INSERT INTO t1 VALUES (39,0);
ALTER TABLE t3 COMMENT "DDL statement 3";

SELECT * FROM t1 WHERE a >= 30 ORDER BY a;

--source include/save_master_gtid.inc


--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 30 ORDER BY a;


--source include/stop_slave.inc
SET GLOBAL debug_dbug= @old_debug;
--source include/start_slave.inc


--echo *** MDEV-8113: ALTER TABLE causes slave hang in optimistic parallel replication ***

--connection server_2
--source include/stop_slave.inc

--connection server_1
ALTER TABLE t2 ADD c INT;
INSERT INTO t2 (a,b) VALUES (50, 0);
INSERT INTO t2 (a,b) VALUES (51, 1);
INSERT INTO t2 (a,b) VALUES (52, 2);
INSERT INTO t2 (a,b) VALUES (53, 3);
INSERT INTO t2 (a,b) VALUES (54, 4);
INSERT INTO t2 (a,b) VALUES (55, 5);
INSERT INTO t2 (a,b) VALUES (56, 6);
INSERT INTO t2 (a,b) VALUES (57, 7);
INSERT INTO t2 (a,b) VALUES (58, 8);
INSERT INTO t2 (a,b) VALUES (59, 9);
ALTER TABLE t2 DROP COLUMN c;
SELECT * FROM t2 WHERE a >= 50 ORDER BY a;
--source include/save_master_gtid.inc

--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t2 WHERE a >= 50 ORDER BY a;


--echo *** MDEV-8075: DROP TEMPORARY TABLE not marked as ddl, causing optimistic parallel replication to fail ***

--connection server_2
--source include/stop_slave.inc

--connection server_1
INSERT INTO t1 VALUES (40, 10);
CREATE TEMPORARY TABLE t1 (a INT PRIMARY KEY) ENGINE=InnoDB;
INSERT INTO t1 VALUES (41);
BEGIN;
INSERT INTO t2 SELECT a, 20 FROM t1;
DROP TEMPORARY TABLE t1;
COMMIT;
INSERT INTO t1 VALUES (42, 10);
--source include/save_master_gtid.inc
SELECT * FROM t1 WHERE a >= 40 ORDER BY a;
SELECT * FROM t2 WHERE a >= 40 ORDER BY a;

--connection server_2
--source include/start_slave.inc
--source include/sync_with_master_gtid.inc
SELECT * FROM t1 WHERE a >= 40 ORDER BY a;
SELECT * FROM t2 WHERE a >= 40 ORDER BY a;

# partial cleanup to reuse the tables by following tests
--connection server_1
DELETE FROM t1;
DELETE FROM t2;
--source include/save_master_gtid.inc

--connection server_2
--source include/sync_with_master_gtid.inc

#
# MDEV-13577 optimistic parallel slave errors out to error log unnecessary
#

# The 1st of the following two trx:s a blocker on slave
--connection server_2
set global log_warnings=2;
BEGIN;
INSERT INTO t1 SET a=1;

--connection server_1
SET @save.binlog_format=@@session.binlog_format;
SET @@SESSION.binlog_format=row;

BEGIN;
  INSERT INTO t1 SET a=1;
  INSERT INTO t2 SET a=1;
COMMIT;

# This transaction is going to win optimistical race with above INSERT
# on slave while being depend on it. That means it will face a kind of temporary error
# and then will retry to succeed.
BEGIN;
  DELETE FROM t2;
COMMIT;

# First make sure DELETE raced indeed to get stuck at retrying stage
# where it runs "realistically" now. There is nomore optimistic error
# in the errorlog, which is downgraded to the warning level (when
# --log-warnings > 1), see above suppression.
--connection server_2
--let $wait_condition= SELECT COUNT(*) = 1 FROM information_schema.processlist WHERE state = "Waiting for prior transaction to commit"
--source include/wait_condition.inc

# Next release the 1st trx to commit.
--connection server_2
ROLLBACK;

# MDEV-13577 local cleanup:
--connection server_1
SET @@SESSION.binlog_format= @save.binlog_format;
DELETE FROM t1;
DELETE FROM t2;
--source include/save_master_gtid.inc

--connection server_2
--source include/sync_with_master_gtid.inc

#
# Clean up.
#
--connection server_2
--source include/stop_slave.inc
set global log_warnings=default;
SET GLOBAL slave_parallel_mode=@old_parallel_mode;
SET GLOBAL slave_parallel_threads=@old_parallel_threads;
--source include/start_slave.inc

--connection server_1
DROP TABLE t1, t2, t3;
--source include/save_master_gtid.inc

--connection server_2
--source include/sync_with_master_gtid.inc
# Check that old rows are deleted from mysql.gtid_slave_pos.
# Deletion is asynchronous, so use wait_condition.inc.
# Also, there is a small amount of non-determinism in the deletion of old
# rows, so it is not guaranteed that there can never be more than
# @@gtid_cleanup_batch_size rows in the table; so allow a bit of slack
# here.
let $wait_condition=
  SELECT COUNT(*) <= 5*@@GLOBAL.gtid_cleanup_batch_size
    FROM mysql.gtid_slave_pos;
--source include/wait_condition.inc
eval $wait_condition;
SET GLOBAL gtid_cleanup_batch_size= @old_gtid_cleanup_batch_size;

--connection server_1
--source include/rpl_end.inc