Debian Codegraph @

mysql-test/suite/rpl/t/rpl_parallel_autoinc.test

raw
1 --source include/have_binlog_format_statement.inc
2 --source include/have_innodb.inc
3 --source include/master-slave.inc
4
5 --echo MDEV-31482: Lock wait timeout with INSERT-SELECT, autoinc, and statement-based replication
6
7 # The scenario is transactions T1, T2, T3:
8 #
9 # T1 is waiting for T3 on an autoinc lock
10 # T2 is waiting for T1 to commit
11 # T3 is waiting on a normal row lock held by T2
12 #
13 # This caused a hang until innodb_lock_wait_timeout, because autoinc
14 # locks were not reported to the in-order parallel replication, so T3
15 # was not deadlock killed.
16
17 --let $lock_wait_timeout=20
18
19 --let $rpl_connection_name= slave2
20 --let $rpl_server_number= 2
21 --source include/rpl_connect.inc
22
23 --let $rpl_connection_name= slave3
24 --let $rpl_server_number= 2
25 --source include/rpl_connect.inc
26
27 --connection master
28 ALTER TABLE mysql.gtid_slave_pos ENGINE=InnoDB;
29
30 # A table as destination for INSERT-SELECT
31 CREATE TABLE t1 (a INT PRIMARY KEY AUTO_INCREMENT, b INT, c INT, INDEX (c)) ENGINE=InnoDB;
32 INSERT INTO t1 (b,c) VALUES (0, 1), (0, 1), (0, 2), (0,3), (0, 5), (0, 7), (0, 8);
33
34 # A table as source for INSERT-SELECT.
35 CREATE TABLE t2 (a INT PRIMARY KEY, b INT) ENGINE=InnoDB;
36 INSERT INTO t2 VALUES (10,1), (20,2), (30,3), (40,4), (50,5);
37
38 # A table to help order slave worker threads to setup the desired scenario.
39 CREATE TABLE t3 (a VARCHAR(20) PRIMARY KEY, b INT) ENGINE=InnoDB;
40 INSERT INTO t3 VALUES ('row for T1', 0), ('row for T2', 0), ('row for T3', 0);
41 --source include/save_master_gtid.inc
42
43 --connection slave
44 --source include/sync_with_master_gtid.inc
45 --source include/stop_slave.inc
46 --let $save_innodb_lock_wait_timeout= `SELECT @@global.innodb_lock_wait_timeout`
47 --let $save_slave_parallel_threads= `SELECT @@global.slave_parallel_threads`
48 --let $save_slave_parallel_mode= `SELECT @@global.slave_parallel_mode`
49 set @@global.slave_parallel_threads= 3;
50 set @@global.slave_parallel_mode= OPTIMISTIC;
51 eval set @@global.innodb_lock_wait_timeout= $lock_wait_timeout;
52
53 --connection master
54 # Transaction T1.
55 BEGIN;
56 UPDATE t3 SET b=b+1 where a="row for T1";
57 INSERT INTO t1(b, c) SELECT 1, t2.b FROM t2 WHERE a=10;
58 COMMIT;
59
60 # Transaction T2.
61 DELETE FROM t1 WHERE c >= 4 and c < 6;
62
63 # Transaction T3.
64 BEGIN;
65 UPDATE t3 SET b=b+1 where a="row for T3";
66 INSERT INTO t1(b, c) SELECT 3, t2.b FROM t2 WHERE a >= 20 AND a <= 40;
67 COMMIT;
68
69 --source include/save_master_gtid.inc
70
71 --connection slave1
72 # Temporarily block T1 to create the scheduling that triggers the bug.
73 BEGIN;
74 SELECT * FROM t3 WHERE a="row for T1" FOR UPDATE;
75
76 --connection slave2
77 # Temporarily block T3 from starting (so T2 can reach commit).
78 BEGIN;
79 SELECT * FROM t3 WHERE a="row for T3" FOR UPDATE;
80
81 --connection slave3
82 # This critical step blocks T3 after it has inserted its first row,
83 # and thus taken the auto-increment lock, but before it has reached
84 # the point where it gets a row lock wait on T2. Even though
85 # auto-increment lock waits were not reported due to the bug,
86 # transitive lock waits (T1 waits on autoinc of T3 which waits on row
87 # on T2) _were_ reported as T1 waiting on T2, and thus a deadlock kill
88 # happened and the bug was not triggered.
89 BEGIN;
90 DELETE FROM t2 WHERE a=30;
91
92 --connection slave
93 --source include/start_slave.inc
94
95 # First let T2 complete until it is waiting for T1 to commit.
96 --let $wait_condition= SELECT count(*)=1 FROM information_schema.processlist WHERE state='Waiting for prior transaction to commit' and command LIKE 'Slave_worker';
97 --source include/wait_condition.inc
98
99 # Then let T3 reach the point where it has obtained the autoinc lock,
100 # but it is not yet waiting for a row lock held by T2.
101 --connection slave2
102 ROLLBACK;
103 --let $wait_condition= SELECT count(*)=1 FROM information_schema.processlist WHERE state='Sending data' and info LIKE 'INSERT INTO t1(b, c) SELECT 3, t2.b%' and time_ms > 500 and command LIKE 'Slave_worker';
104 --source include/wait_condition.inc
105
106 # Now let T1 continue, while T3 is holding the autoinc lock but before
107 # it is waiting for T2. Wait a short while to give the hang a chance to
108 # happen; T1 needs to get to request the autoinc lock before we let T3
109 # continue. (There's a small chance the sleep will be too small, which will
110 # let the test occasionally pass on non-fixed server).
111 --connection slave1
112 ROLLBACK;
113 --sleep 0.5
114
115 # Now let T3 continue; the bug was that this lead to an undetected
116 # deadlock that remained until innodb lock wait timeout.
117 --connection slave3
118 ROLLBACK;
119
120 --connection slave
121 --let $slave_timeout= `SELECT $lock_wait_timeout/2`
122 --source include/sync_with_master_gtid.inc
123 --let $slave_timeout=
124 SELECT * FROM t1 ORDER BY a;
125 SELECT * FROM t2 ORDER BY a;
126 SELECT * FROM t3 ORDER BY a;
127
128 # Cleanup.
129 --connection master
130 CALL mtr.add_suppression("Unsafe statement written to the binary log using statement format");
131 DROP TABLE t1, t2, t3;
132
133 --connection slave
134 --source include/stop_slave.inc
135 eval SET @@global.slave_parallel_threads= $save_slave_parallel_threads;
136 eval SET @@global.slave_parallel_mode= $save_slave_parallel_mode;
137 eval SET @@global.innodb_lock_wait_timeout= $save_innodb_lock_wait_timeout;
138 --source include/start_slave.inc
139 SELECT @@GLOBAL.innodb_autoinc_lock_mode;
140 --source include/rpl_end.inc