diff options
Diffstat (limited to 'mysql-test/suite/rpl/t/rpl_parallel.test')
| -rw-r--r-- | mysql-test/suite/rpl/t/rpl_parallel.test | 2414 |
1 files changed, 1 insertions, 2413 deletions
diff --git a/mysql-test/suite/rpl/t/rpl_parallel.test b/mysql-test/suite/rpl/t/rpl_parallel.test index ace9e8e37f4..b7c4bb429a4 100644 --- a/mysql-test/suite/rpl/t/rpl_parallel.test +++ b/mysql-test/suite/rpl/t/rpl_parallel.test @@ -1,2413 +1 @@ ---source include/have_innodb.inc ---source include/have_debug.inc ---source include/have_debug_sync.inc ---let $rpl_topology=1->2 ---source include/rpl_init.inc - -# Test various aspects of parallel replication. - ---connection server_2 -SET @old_parallel_threads=@@GLOBAL.slave_parallel_threads; ---error ER_SLAVE_MUST_STOP -SET GLOBAL slave_parallel_threads=10; ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads=10; - -# Check that we do not spawn any worker threads when no slave is running. -SELECT IF(COUNT(*) < 10, "OK", CONCAT("Found too many system user processes: ", COUNT(*))) FROM information_schema.processlist WHERE user = "system user"; - -CHANGE MASTER TO master_use_gtid=slave_pos; ---source include/start_slave.inc - -# Check that worker threads get spawned when slave starts. -SELECT IF(COUNT(*) >= 10, "OK", CONCAT("Found too few system user processes: ", COUNT(*))) FROM information_schema.processlist WHERE user = "system user"; -# ... and that worker threads get removed when slave stops. ---source include/stop_slave.inc -SELECT IF(COUNT(*) < 10, "OK", CONCAT("Found too many system user processes: ", COUNT(*))) FROM information_schema.processlist WHERE user = "system user"; ---source include/start_slave.inc -SELECT IF(COUNT(*) >= 10, "OK", CONCAT("Found too few system user processes: ", COUNT(*))) FROM information_schema.processlist WHERE user = "system user"; - ---echo *** Test long-running query in domain 1 can run in parallel with short queries in domain 0 *** - ---connection server_1 -ALTER TABLE mysql.gtid_slave_pos ENGINE=InnoDB; -CREATE TABLE t1 (a int PRIMARY KEY) ENGINE=MyISAM; -CREATE TABLE t2 (a int PRIMARY KEY) ENGINE=InnoDB; -INSERT INTO t1 VALUES (1); -INSERT INTO t2 VALUES (1); ---save_master_pos - ---connection server_2 ---sync_with_master - -# Block the table t1 to simulate a replicated query taking a long time. ---connect (con_temp1,127.0.0.1,root,,test,$SERVER_MYPORT_2,) -LOCK TABLE t1 WRITE; - ---connection server_1 -SET gtid_domain_id=1; -# This query will be blocked on the slave until UNLOCK TABLES. -INSERT INTO t1 VALUES (2); -SET gtid_domain_id=0; -# These t2 queries can be replicated in parallel with the prior t1 query, as -# they are in a separate replication domain. -INSERT INTO t2 VALUES (2); -INSERT INTO t2 VALUES (3); -BEGIN; -INSERT INTO t2 VALUES (4); -INSERT INTO t2 VALUES (5); -COMMIT; -INSERT INTO t2 VALUES (6); - ---connection server_2 ---let $wait_condition= SELECT COUNT(*) = 6 FROM t2 ---source include/wait_condition.inc - -SELECT * FROM t2 ORDER by a; - ---connection con_temp1 -SELECT * FROM t1; -UNLOCK TABLES; - ---connection server_2 ---let $wait_condition= SELECT COUNT(*) = 2 FROM t1 ---source include/wait_condition.inc - -SELECT * FROM t1 ORDER BY a; - - ---echo *** Test two transactions in different domains committed in opposite order on slave but in a single group commit. *** ---connection server_2 ---source include/stop_slave.inc - ---connection server_1 -# Use a stored function to inject a debug_sync into the appropriate THD. -# The function does nothing on the master, and on the slave it injects the -# desired debug_sync action(s). -SET sql_log_bin=0; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; - -SET @old_format= @@SESSION.binlog_format; -SET binlog_format='statement'; -SET gtid_domain_id=1; -INSERT INTO t2 VALUES (foo(10, - 'commit_before_enqueue SIGNAL ready1 WAIT_FOR cont1', - 'commit_after_release_LOCK_prepare_ordered SIGNAL ready2')); - ---connection server_2 -FLUSH LOGS; ---source include/wait_for_binlog_checkpoint.inc -SET sql_log_bin=0; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - IF d1 != '' THEN - SET debug_sync = d1; - END IF; - IF d2 != '' THEN - SET debug_sync = d2; - END IF; - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; -SET @old_format=@@GLOBAL.binlog_format; -SET GLOBAL binlog_format=statement; -# We need to restart all parallel threads for the new global setting to -# be copied to the session-level values. -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - -# First make sure the first insert is ready to commit, but not queued yet. -SET debug_sync='now WAIT_FOR ready1'; - ---connection server_1 -SET gtid_domain_id=2; -INSERT INTO t2 VALUES (foo(11, - 'commit_before_enqueue SIGNAL ready3 WAIT_FOR cont3', - 'commit_after_release_LOCK_prepare_ordered SIGNAL ready4 WAIT_FOR cont4')); -SET gtid_domain_id=0; -SELECT * FROM t2 WHERE a >= 10 ORDER BY a; - ---connection server_2 -# Now wait for the second insert to queue itself as the leader, and then -# wait for more commits to queue up. -SET debug_sync='now WAIT_FOR ready3'; -SET debug_sync='now SIGNAL cont3'; -SET debug_sync='now WAIT_FOR ready4'; -# Now allow the first insert to queue up to participate in group commit. -SET debug_sync='now SIGNAL cont1'; -SET debug_sync='now WAIT_FOR ready2'; -# Finally allow the second insert to proceed and do the group commit. -SET debug_sync='now SIGNAL cont4'; - ---let $wait_condition= SELECT COUNT(*) = 2 FROM t2 WHERE a >= 10 ---source include/wait_condition.inc -SELECT * FROM t2 WHERE a >= 10 ORDER BY a; -# The two INSERT transactions should have been committed in opposite order, -# but in the same group commit (seen by precense of cid=# in the SHOW -# BINLOG output). ---let $binlog_file= slave-bin.000002 ---source include/show_binlog_events.inc -FLUSH LOGS; ---source include/wait_for_binlog_checkpoint.inc - -# Restart all the slave parallel worker threads, to clear all debug_sync actions. ---connection server_2 ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; -SET debug_sync='RESET'; ---source include/start_slave.inc - - ---echo *** Test that group-committed transactions on the master can replicate in parallel on the slave. *** ---connection server_1 -SET debug_sync='RESET'; -FLUSH LOGS; ---source include/wait_for_binlog_checkpoint.inc -CREATE TABLE t3 (a INT PRIMARY KEY, b INT) ENGINE=InnoDB; -# Create some sentinel rows so that the rows inserted in parallel fall into -# separate gaps and do not cause gap lock conflicts. -INSERT INTO t3 VALUES (1,1), (3,3), (5,5), (7,7); ---save_master_pos ---connection server_2 ---sync_with_master - -# We want to test that the transactions can execute out-of-order on -# the slave, but still end up committing in-order, and in a single -# group commit. -# -# The idea is to group-commit three transactions together on the master: -# A, B, and C. On the slave, C will execute the insert first, then A, -# and then B. But B manages to complete before A has time to commit, so -# all three end up committing together. -# -# So we start by setting up some row locks that will block transactions -# A and B from executing, allowing C to run first. - ---connection con_temp1 -BEGIN; -INSERT INTO t3 VALUES (2,102); ---connect (con_temp2,127.0.0.1,root,,test,$SERVER_MYPORT_2,) -BEGIN; -INSERT INTO t3 VALUES (4,104); - -# On the master, queue three INSERT transactions as a single group commit. ---connect (con_temp3,127.0.0.1,root,,test,$SERVER_MYPORT_1,) -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -SET binlog_format=statement; -send INSERT INTO t3 VALUES (2, foo(12, - 'commit_after_release_LOCK_prepare_ordered SIGNAL slave_queued1 WAIT_FOR slave_cont1', - '')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connect (con_temp4,127.0.0.1,root,,test,$SERVER_MYPORT_1,) -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -SET binlog_format=statement; -send INSERT INTO t3 VALUES (4, foo(14, - 'commit_after_release_LOCK_prepare_ordered SIGNAL slave_queued2', - '')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; - ---connect (con_temp5,127.0.0.1,root,,test,$SERVER_MYPORT_1,) -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued3'; -SET binlog_format=statement; -send INSERT INTO t3 VALUES (6, foo(16, - 'group_commit_waiting_for_prior SIGNAL slave_queued3', - '')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued3'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con_temp3 -REAP; ---connection con_temp4 -REAP; ---connection con_temp5 -REAP; -SET debug_sync='RESET'; - ---connection server_1 -SELECT * FROM t3 ORDER BY a; ---let $binlog_file= master-bin.000002 ---source include/show_binlog_events.inc - -# First, wait until insert 3 is ready to queue up for group commit, but is -# waiting for insert 2 to commit before it can do so itself. ---connection server_2 -SET debug_sync='now WAIT_FOR slave_queued3'; - -# Next, let insert 1 proceed, and allow it to queue up as the group commit -# leader, but let it wait for insert 2 to also queue up before proceeding. ---connection con_temp1 -ROLLBACK; ---connection server_2 -SET debug_sync='now WAIT_FOR slave_queued1'; - -# Now let insert 2 proceed and queue up. ---connection con_temp2 -ROLLBACK; ---connection server_2 -SET debug_sync='now WAIT_FOR slave_queued2'; -# And finally, we can let insert 1 proceed and do the group commit with all -# three insert transactions together. -SET debug_sync='now SIGNAL slave_cont1'; - -# Wait for the commit to complete and check that all three transactions -# group-committed together (will be seen in the binlog as all three having -# cid=# on their GTID event). ---let $wait_condition= SELECT COUNT(*) = 3 FROM t3 WHERE a IN (2,4,6) ---source include/wait_condition.inc -SELECT * FROM t3 ORDER BY a; ---let $binlog_file= slave-bin.000003 ---source include/show_binlog_events.inc - - ---echo *** Test STOP SLAVE in parallel mode *** ---connection server_2 ---source include/stop_slave.inc -# Respawn all worker threads to clear any left-over debug_sync or other stuff. -SET debug_sync='RESET'; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; - ---connection server_1 -# Set up a couple of transactions. The first will be blocked halfway -# through on a lock, and while it is blocked we initiate STOP SLAVE. -# We then test that the halfway-initiated transaction is allowed to -# complete, but no subsequent ones. -# We have to use statement-based mode and set -# binlog_direct_non_transactional_updates=0; otherwise the binlog will -# be split into two event groups, one for the MyISAM part and one for the -# InnoDB part. -SET binlog_direct_non_transactional_updates=0; -SET sql_log_bin=0; -CALL mtr.add_suppression("Statement is unsafe because it accesses a non-transactional table after accessing a transactional table within the same transaction"); -SET sql_log_bin=1; -BEGIN; -INSERT INTO t2 VALUES (20); ---disable_warnings -INSERT INTO t1 VALUES (20); ---enable_warnings -INSERT INTO t2 VALUES (21); -INSERT INTO t3 VALUES (20, 20); -COMMIT; -INSERT INTO t3 VALUES(21, 21); -INSERT INTO t3 VALUES(22, 22); -SET binlog_format=@old_format; ---save_master_pos - -# Start a connection that will block the replicated transaction halfway. ---connection con_temp1 -BEGIN; -INSERT INTO t2 VALUES (21); - ---connection server_2 -START SLAVE; -# Wait for the MyISAM change to be visible, after which replication will wait -# for con_temp1 to roll back. ---let $wait_condition= SELECT COUNT(*) = 1 FROM t1 WHERE a=20 ---source include/wait_condition.inc - ---connection con_temp2 -# Initiate slave stop. It will have to wait for the current event group -# to complete. -# The dbug injection causes debug_sync to signal 'wait_for_done_waiting' -# when the SQL driver thread is ready. -SET @old_dbug= @@GLOBAL.debug_dbug; -SET GLOBAL debug_dbug="+d,rpl_parallel_wait_for_done_trigger"; -send STOP SLAVE; - ---connection con_temp1 -SET debug_sync='now WAIT_FOR wait_for_done_waiting'; -ROLLBACK; - ---connection con_temp2 -reap; -SET GLOBAL debug_dbug=@old_dbug; -SET debug_sync='RESET'; - ---connection server_2 ---source include/wait_for_slave_to_stop.inc -# We should see the first transaction applied, but not the two others. -SELECT * FROM t1 WHERE a >= 20 ORDER BY a; -SELECT * FROM t2 WHERE a >= 20 ORDER BY a; -SELECT * FROM t3 WHERE a >= 20 ORDER BY a; - ---source include/start_slave.inc ---sync_with_master -SELECT * FROM t1 WHERE a >= 20 ORDER BY a; -SELECT * FROM t2 WHERE a >= 20 ORDER BY a; -SELECT * FROM t3 WHERE a >= 20 ORDER BY a; - - ---connection server_2 -# Respawn all worker threads to clear any left-over debug_sync or other stuff. ---source include/stop_slave.inc -SET GLOBAL binlog_format=@old_format; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - - ---echo *** Test killing slave threads at various wait points *** ---echo *** 1. Test killing transaction waiting in commit for previous transaction to commit *** - -# Set up three transactions on the master that will be group-committed -# together so they can be replicated in parallel on the slave. ---connection con_temp3 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -SET binlog_format=statement; -send INSERT INTO t3 VALUES (31, foo(31, - 'commit_before_prepare_ordered WAIT_FOR t2_waiting', - 'commit_after_prepare_ordered SIGNAL t1_ready WAIT_FOR t1_cont')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connection con_temp4 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -SET binlog_format=statement; -BEGIN; -# This insert is just so we can get T2 to wait while a query is running that we -# can see in SHOW PROCESSLIST so we can get its thread_id to kill later. -INSERT INTO t3 VALUES (32, foo(32, - 'ha_write_row_end SIGNAL t2_query WAIT_FOR t2_cont', - '')); -# This insert sets up debug_sync points so that T2 will tell when it is at its -# wait point where we want to kill it - and when it has been killed. -INSERT INTO t3 VALUES (33, foo(33, - 'group_commit_waiting_for_prior SIGNAL t2_waiting', - 'group_commit_waiting_for_prior_killed SIGNAL t2_killed')); -send COMMIT; - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; - ---connection con_temp5 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued3'; -SET binlog_format=statement; -send INSERT INTO t3 VALUES (34, foo(34, - '', - '')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued3'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con_temp3 -REAP; ---connection con_temp4 -REAP; ---connection con_temp5 -REAP; - ---connection server_1 -SELECT * FROM t3 WHERE a >= 30 ORDER BY a; -SET debug_sync='RESET'; - ---connection server_2 -SET sql_log_bin=0; -CALL mtr.add_suppression("Query execution was interrupted"); -CALL mtr.add_suppression("Commit failed due to failure of an earlier commit on which this one depends"); -CALL mtr.add_suppression("Slave: Connection was killed"); -SET sql_log_bin=1; -# Wait until T2 is inside executing its insert of 32, then find it in SHOW -# PROCESSLIST to know its thread id for KILL later. -SET debug_sync='now WAIT_FOR t2_query'; ---let $thd_id= `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE INFO LIKE '%foo(32%' AND INFO NOT LIKE '%LIKE%'` -SET debug_sync='now SIGNAL t2_cont'; - -# Wait until T2 has entered its wait for T1 to commit, and T1 has -# progressed into its commit phase. -SET debug_sync='now WAIT_FOR t1_ready'; - -# Now kill the transaction T2. ---replace_result $thd_id THD_ID -eval KILL $thd_id; - -# Wait until T2 has reacted on the kill. -SET debug_sync='now WAIT_FOR t2_killed'; - -# Now we can allow T1 to proceed. -SET debug_sync='now SIGNAL t1_cont'; - ---let $slave_sql_errno= 1317,1927,1964 ---source include/wait_for_slave_sql_error.inc -STOP SLAVE IO_THREAD; -SELECT * FROM t3 WHERE a >= 30 ORDER BY a; - -# Now we have to disable the debug_sync statements, so they do not trigger -# when the events are retried. -SET debug_sync='RESET'; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; -SET sql_log_bin=0; -DROP FUNCTION foo; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; - ---connection server_1 -INSERT INTO t3 VALUES (39,0); ---save_master_pos - ---connection server_2 ---source include/start_slave.inc ---sync_with_master -SELECT * FROM t3 WHERE a >= 30 ORDER BY a; -# Restore the foo() function. -SET sql_log_bin=0; -DROP FUNCTION foo; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - IF d1 != '' THEN - SET debug_sync = d1; - END IF; - IF d2 != '' THEN - SET debug_sync = d2; - END IF; - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; - - ---connection server_2 -# Respawn all worker threads to clear any left-over debug_sync or other stuff. ---source include/stop_slave.inc -SET GLOBAL binlog_format=@old_format; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - - ---echo *** 2. Same as (1), but without restarting IO thread after kill of SQL threads *** - -# Set up three transactions on the master that will be group-committed -# together so they can be replicated in parallel on the slave. ---connection con_temp3 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -SET binlog_format=statement; -send INSERT INTO t3 VALUES (41, foo(41, - 'commit_before_prepare_ordered WAIT_FOR t2_waiting', - 'commit_after_prepare_ordered SIGNAL t1_ready WAIT_FOR t1_cont')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connection con_temp4 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -SET binlog_format=statement; -BEGIN; -# This insert is just so we can get T2 to wait while a query is running that we -# can see in SHOW PROCESSLIST so we can get its thread_id to kill later. -INSERT INTO t3 VALUES (42, foo(42, - 'ha_write_row_end SIGNAL t2_query WAIT_FOR t2_cont', - '')); -# This insert sets up debug_sync points so that T2 will tell when it is at its -# wait point where we want to kill it - and when it has been killed. -INSERT INTO t3 VALUES (43, foo(43, - 'group_commit_waiting_for_prior SIGNAL t2_waiting', - 'group_commit_waiting_for_prior_killed SIGNAL t2_killed')); -send COMMIT; - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; - ---connection con_temp5 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued3'; -SET binlog_format=statement; -send INSERT INTO t3 VALUES (44, foo(44, - '', - '')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued3'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con_temp3 -REAP; ---connection con_temp4 -REAP; ---connection con_temp5 -REAP; - ---connection server_1 -SELECT * FROM t3 WHERE a >= 40 ORDER BY a; -SET debug_sync='RESET'; - ---connection server_2 -# Wait until T2 is inside executing its insert of 42, then find it in SHOW -# PROCESSLIST to know its thread id for KILL later. -SET debug_sync='now WAIT_FOR t2_query'; ---let $thd_id= `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE INFO LIKE '%foo(42%' AND INFO NOT LIKE '%LIKE%'` -SET debug_sync='now SIGNAL t2_cont'; - -# Wait until T2 has entered its wait for T1 to commit, and T1 has -# progressed into its commit phase. -SET debug_sync='now WAIT_FOR t1_ready'; - -# Now kill the transaction T2. ---replace_result $thd_id THD_ID -eval KILL $thd_id; - -# Wait until T2 has reacted on the kill. -SET debug_sync='now WAIT_FOR t2_killed'; - -# Now we can allow T1 to proceed. -SET debug_sync='now SIGNAL t1_cont'; - ---let $slave_sql_errno= 1317,1927,1964 ---source include/wait_for_slave_sql_error.inc - -# Now we have to disable the debug_sync statements, so they do not trigger -# when the events are retried. -SET debug_sync='RESET'; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; -SET sql_log_bin=0; -DROP FUNCTION foo; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; - ---connection server_1 -INSERT INTO t3 VALUES (49,0); ---save_master_pos - ---connection server_2 -START SLAVE SQL_THREAD; ---sync_with_master -SELECT * FROM t3 WHERE a >= 40 ORDER BY a; -# Restore the foo() function. -SET sql_log_bin=0; -DROP FUNCTION foo; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - IF d1 != '' THEN - SET debug_sync = d1; - END IF; - IF d2 != '' THEN - SET debug_sync = d2; - END IF; - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; - - ---connection server_2 -# Respawn all worker threads to clear any left-over debug_sync or other stuff. ---source include/stop_slave.inc -SET GLOBAL binlog_format=@old_format; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - - ---echo *** 3. Same as (2), but not using gtid mode *** - ---connection server_2 ---source include/stop_slave.inc -CHANGE MASTER TO master_use_gtid=no; ---source include/start_slave.inc - ---connection server_1 -# Set up three transactions on the master that will be group-committed -# together so they can be replicated in parallel on the slave. ---connection con_temp3 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -SET binlog_format=statement; -send INSERT INTO t3 VALUES (51, foo(51, - 'commit_before_prepare_ordered WAIT_FOR t2_waiting', - 'commit_after_prepare_ordered SIGNAL t1_ready WAIT_FOR t1_cont')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connection con_temp4 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -SET binlog_format=statement; -BEGIN; -# This insert is just so we can get T2 to wait while a query is running that we -# can see in SHOW PROCESSLIST so we can get its thread_id to kill later. -INSERT INTO t3 VALUES (52, foo(52, - 'ha_write_row_end SIGNAL t2_query WAIT_FOR t2_cont', - '')); -# This insert sets up debug_sync points so that T2 will tell when it is at its -# wait point where we want to kill it - and when it has been killed. -INSERT INTO t3 VALUES (53, foo(53, - 'group_commit_waiting_for_prior SIGNAL t2_waiting', - 'group_commit_waiting_for_prior_killed SIGNAL t2_killed')); -send COMMIT; - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; - ---connection con_temp5 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued3'; -SET binlog_format=statement; -send INSERT INTO t3 VALUES (54, foo(54, - '', - '')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued3'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con_temp3 -REAP; ---connection con_temp4 -REAP; ---connection con_temp5 -REAP; - ---connection server_1 -SELECT * FROM t3 WHERE a >= 50 ORDER BY a; -SET debug_sync='RESET'; - ---connection server_2 -# Wait until T2 is inside executing its insert of 52, then find it in SHOW -# PROCESSLIST to know its thread id for KILL later. -SET debug_sync='now WAIT_FOR t2_query'; ---let $thd_id= `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE INFO LIKE '%foo(52%' AND INFO NOT LIKE '%LIKE%'` -SET debug_sync='now SIGNAL t2_cont'; - -# Wait until T2 has entered its wait for T1 to commit, and T1 has -# progressed into its commit phase. -SET debug_sync='now WAIT_FOR t1_ready'; - -# Now kill the transaction T2. ---replace_result $thd_id THD_ID -eval KILL $thd_id; - -# Wait until T2 has reacted on the kill. -SET debug_sync='now WAIT_FOR t2_killed'; - -# Now we can allow T1 to proceed. -SET debug_sync='now SIGNAL t1_cont'; - ---let $slave_sql_errno= 1317,1927,1964 ---source include/wait_for_slave_sql_error.inc -SELECT * FROM t3 WHERE a >= 50 ORDER BY a; - -# Now we have to disable the debug_sync statements, so they do not trigger -# when the events are retried. -SET debug_sync='RESET'; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; -SET sql_log_bin=0; -DROP FUNCTION foo; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; - ---connection server_1 -INSERT INTO t3 VALUES (59,0); ---save_master_pos - ---connection server_2 -START SLAVE SQL_THREAD; ---sync_with_master -SELECT * FROM t3 WHERE a >= 50 ORDER BY a; -# Restore the foo() function. -SET sql_log_bin=0; -DROP FUNCTION foo; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - IF d1 != '' THEN - SET debug_sync = d1; - END IF; - IF d2 != '' THEN - SET debug_sync = d2; - END IF; - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; - - ---source include/stop_slave.inc -CHANGE MASTER TO master_use_gtid=slave_pos; ---source include/start_slave.inc - ---connection server_2 -# Respawn all worker threads to clear any left-over debug_sync or other stuff. ---source include/stop_slave.inc -SET GLOBAL binlog_format=@old_format; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=4; ---source include/start_slave.inc - - ---echo *** 4. Test killing thread that is waiting to start transaction until previous transaction commits *** - -# We set up four transactions T1, T2, T3, and T4 on the master. T2, T3, and T4 -# can run in parallel with each other (same group commit and commit id), -# but not in parallel with T1. -# -# We use four worker threads, each Ti will be queued on each their own -# worker thread. We will delay T1 commit, T3 will wait for T1 to begin -# commit before it can start. We will kill T3 during this wait, and -# check that everything works correctly. -# -# It is rather tricky to get the correct thread id of the worker to kill. -# We start by injecting four dummy transactions in a debug_sync-controlled -# manner to be able to get known thread ids for the workers in a pool with -# just 4 worker threads. Then we let in each of the real test transactions -# T1-T4 one at a time in a way which allows us to know which transaction -# ends up with which thread id. - ---connection server_1 -SET binlog_format=statement; -SET gtid_domain_id=2; -BEGIN; -# This debug_sync will linger on and be used to control T4 later. -INSERT INTO t3 VALUES (70, foo(70, - 'rpl_parallel_start_waiting_for_prior SIGNAL t4_waiting', '')); -INSERT INTO t3 VALUES (60, foo(60, - 'ha_write_row_end SIGNAL d2_query WAIT_FOR d2_cont2', - 'rpl_parallel_end_of_group SIGNAL d2_done WAIT_FOR d2_cont')); -COMMIT; -SET gtid_domain_id=0; - ---connection server_2 -SET debug_sync='now WAIT_FOR d2_query'; ---let $d2_thd_id= `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE INFO LIKE '%foo(60%' AND INFO NOT LIKE '%LIKE%'` - ---connection server_1 -SET gtid_domain_id=1; -BEGIN; -# These debug_sync's will linger on and be used to control T3 later. -INSERT INTO t3 VALUES (61, foo(61, - 'rpl_parallel_start_waiting_for_prior SIGNAL t3_waiting', - 'rpl_parallel_start_waiting_for_prior_killed SIGNAL t3_killed')); -INSERT INTO t3 VALUES (62, foo(62, - 'ha_write_row_end SIGNAL d1_query WAIT_FOR d1_cont2', - 'rpl_parallel_end_of_group SIGNAL d1_done WAIT_FOR d1_cont')); -COMMIT; -SET gtid_domain_id=0; - ---connection server_2 -SET debug_sync='now WAIT_FOR d1_query'; ---let $d1_thd_id= `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE INFO LIKE '%foo(62%' AND INFO NOT LIKE '%LIKE%'` - ---connection server_1 -SET gtid_domain_id=0; -INSERT INTO t3 VALUES (63, foo(63, - 'ha_write_row_end SIGNAL d0_query WAIT_FOR d0_cont2', - 'rpl_parallel_end_of_group SIGNAL d0_done WAIT_FOR d0_cont')); - ---connection server_2 -SET debug_sync='now WAIT_FOR d0_query'; ---let $d0_thd_id= `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE INFO LIKE '%foo(63%' AND INFO NOT LIKE '%LIKE%'` - ---connection server_1 -SET gtid_domain_id=3; -BEGIN; -# These debug_sync's will linger on and be used to control T2 later. -INSERT INTO t3 VALUES (68, foo(68, - 'rpl_parallel_start_waiting_for_prior SIGNAL t2_waiting', '')); -INSERT INTO t3 VALUES (69, foo(69, - 'ha_write_row_end SIGNAL d3_query WAIT_FOR d3_cont2', - 'rpl_parallel_end_of_group SIGNAL d3_done WAIT_FOR d3_cont')); -COMMIT; -SET gtid_domain_id=0; - ---connection server_2 -SET debug_sync='now WAIT_FOR d3_query'; ---let $d3_thd_id= `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE INFO LIKE '%foo(69%' AND INFO NOT LIKE '%LIKE%'` - -SET debug_sync='now SIGNAL d2_cont2'; -SET debug_sync='now WAIT_FOR d2_done'; -SET debug_sync='now SIGNAL d1_cont2'; -SET debug_sync='now WAIT_FOR d1_done'; -SET debug_sync='now SIGNAL d0_cont2'; -SET debug_sync='now WAIT_FOR d0_done'; -SET debug_sync='now SIGNAL d3_cont2'; -SET debug_sync='now WAIT_FOR d3_done'; - -# Now prepare the real transactions T1, T2, T3, T4 on the master. - ---connection con_temp3 -# Create transaction T1. -SET binlog_format=statement; -INSERT INTO t3 VALUES (64, foo(64, - 'rpl_parallel_before_mark_start_commit SIGNAL t1_waiting WAIT_FOR t1_cont', '')); - -# Create transaction T2, as a group commit leader on the master. -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2 WAIT_FOR master_cont2'; -send INSERT INTO t3 VALUES (65, foo(65, '', '')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; - ---connection con_temp4 -# Create transaction T3, participating in T2's group commit. -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued3'; -send INSERT INTO t3 VALUES (66, foo(66, '', '')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued3'; - ---connection con_temp5 -# Create transaction T4, participating in group commit with T2 and T3. -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued4'; -send INSERT INTO t3 VALUES (67, foo(67, '', '')); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued4'; -SET debug_sync='now SIGNAL master_cont2'; - ---connection con_temp3 -REAP; ---connection con_temp4 -REAP; ---connection con_temp5 -REAP; - ---connection server_1 -SELECT * FROM t3 WHERE a >= 60 ORDER BY a; -SET debug_sync='RESET'; - ---connection server_2 -# Now we have the four transactions pending for replication on the slave. -# Let them be queued for our three worker threads in a controlled fashion. -# We put them at a stage where T1 is delayed and T3 is waiting for T1 to -# commit before T3 can start. Then we kill T3. - -# Make the worker D0 free, and wait for T1 to be queued in it. -SET debug_sync='now SIGNAL d0_cont'; -SET debug_sync='now WAIT_FOR t1_waiting'; - -# Make the worker D3 free, and wait for T2 to be queued in it. -SET debug_sync='now SIGNAL d3_cont'; -SET debug_sync='now WAIT_FOR t2_waiting'; - -# Now release worker D1, and wait for T3 to be queued in it. -# T3 will wait for T1 to commit before it can start. -SET debug_sync='now SIGNAL d1_cont'; -SET debug_sync='now WAIT_FOR t3_waiting'; - -# Release worker D2. Wait for T4 to be queued, so we are sure it has -# received the debug_sync signal (else we might overwrite it with the -# next debug_sync). -SET debug_sync='now SIGNAL d2_cont'; -SET debug_sync='now WAIT_FOR t4_waiting'; - -# Now we kill the waiting transaction T3 in worker D1. ---replace_result $d1_thd_id THD_ID -eval KILL $d1_thd_id; - -# Wait until T3 has reacted on the kill. -SET debug_sync='now WAIT_FOR t3_killed'; - -# Now we can allow T1 to proceed. -SET debug_sync='now SIGNAL t1_cont'; - ---let $slave_sql_errno= 1317,1927,1964 ---source include/wait_for_slave_sql_error.inc -STOP SLAVE IO_THREAD; -# Since T2, T3, and T4 run in parallel, we can not be sure if T2 will have time -# to commit or not before the stop. However, T1 should commit, and T3/T4 may -# not have committed. (After slave restart we check that all become committed -# eventually). -SELECT * FROM t3 WHERE a >= 60 AND a != 65 ORDER BY a; - -# Now we have to disable the debug_sync statements, so they do not trigger -# when the events are retried. -SET debug_sync='RESET'; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; -SET sql_log_bin=0; -DROP FUNCTION foo; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; - ---connection server_1 -UPDATE t3 SET b=b+1 WHERE a=60; ---save_master_pos - ---connection server_2 ---source include/start_slave.inc ---sync_with_master -SELECT * FROM t3 WHERE a >= 60 ORDER BY a; -# Restore the foo() function. -SET sql_log_bin=0; -DROP FUNCTION foo; ---delimiter || -CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500)) - RETURNS INT DETERMINISTIC - BEGIN - IF d1 != '' THEN - SET debug_sync = d1; - END IF; - IF d2 != '' THEN - SET debug_sync = d2; - END IF; - RETURN x; - END -|| ---delimiter ; -SET sql_log_bin=1; - ---connection server_2 -# Respawn all worker threads to clear any left-over debug_sync or other stuff. ---source include/stop_slave.inc -SET GLOBAL binlog_format=@old_format; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - - ---echo *** 5. Test killing thread that is waiting for queue of max length to shorten *** - -# Find the thread id of the driver SQL thread that we want to kill. ---let $wait_condition= SELECT COUNT(*) = 1 FROM INFORMATION_SCHEMA.PROCESSLIST WHERE STATE LIKE '%Slave has read all relay log%' ---source include/wait_condition.inc ---let $thd_id= `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE STATE LIKE '%Slave has read all relay log%'` -SET @old_max_queued= @@GLOBAL.slave_parallel_max_queued; -SET GLOBAL slave_parallel_max_queued=9000; - ---connection server_1 ---let bigstring= `SELECT REPEAT('x', 10000)` -SET binlog_format=statement; -# Create an event that will wait to be signalled. -INSERT INTO t3 VALUES (80, foo(0, - 'ha_write_row_end SIGNAL query_waiting WAIT_FOR query_cont', '')); - ---connection server_2 -SET debug_sync='now WAIT_FOR query_waiting'; -# Inject that the SQL driver thread will signal `wait_queue_ready' to debug_sync -# as it goes to wait for the event queue to become smaller than the value of -# @@slave_parallel_max_queued. -SET @old_dbug= @@GLOBAL.debug_dbug; -SET GLOBAL debug_dbug="+d,rpl_parallel_wait_queue_max"; - ---connection server_1 ---disable_query_log -# Create an event that will fill up the queue. -# The Xid event at the end of the event group will have to wait for the Query -# event with the INSERT to drain so the queue becomes shorter. However that in -# turn waits for the prior event group to continue. -eval INSERT INTO t3 VALUES (81, LENGTH('$bigstring')); ---enable_query_log -SELECT * FROM t3 WHERE a >= 80 ORDER BY a; - ---connection server_2 -SET debug_sync='now WAIT_FOR wait_queue_ready'; - ---replace_result $thd_id THD_ID -eval KILL $thd_id; - -SET debug_sync='now WAIT_FOR wait_queue_killed'; -SET debug_sync='now SIGNAL query_cont'; - ---let $slave_sql_errno= 1317,1927,1964 ---source include/wait_for_slave_sql_error.inc -STOP SLAVE IO_THREAD; - -SET GLOBAL debug_dbug=@old_dbug; -SET GLOBAL slave_parallel_max_queued= @old_max_queued; - ---connection server_1 -INSERT INTO t3 VALUES (82,0); -SET binlog_format=@old_format; ---save_master_pos - ---connection server_2 -SET debug_sync='RESET'; ---source include/start_slave.inc ---sync_with_master -SELECT * FROM t3 WHERE a >= 80 ORDER BY a; - - ---connection server_2 ---source include/stop_slave.inc -SET GLOBAL binlog_format=@old_format; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - ---echo *** MDEV-5788 Incorrect free of rgi->deferred_events in parallel replication *** - ---connection server_2 -# Use just two worker threads, so we are sure to get the rpl_group_info added -# to the free list, which is what triggered the bug. ---source include/stop_slave.inc -SET GLOBAL replicate_ignore_table="test.t3"; -SET GLOBAL slave_parallel_threads=2; ---source include/start_slave.inc - ---connection server_1 -INSERT INTO t3 VALUES (100, rand()); -INSERT INTO t3 VALUES (101, rand()); - ---save_master_pos - ---connection server_2 ---sync_with_master - ---connection server_1 -INSERT INTO t3 VALUES (102, rand()); -INSERT INTO t3 VALUES (103, rand()); -INSERT INTO t3 VALUES (104, rand()); -INSERT INTO t3 VALUES (105, rand()); - ---save_master_pos - ---connection server_2 ---sync_with_master ---source include/stop_slave.inc -SET GLOBAL replicate_ignore_table=""; ---source include/start_slave.inc - ---connection server_1 -INSERT INTO t3 VALUES (106, rand()); -INSERT INTO t3 VALUES (107, rand()); ---save_master_pos - ---connection server_2 ---sync_with_master ---replace_column 2 # -SELECT * FROM t3 WHERE a >= 100 ORDER BY a; - - ---echo *** MDEV-5921: In parallel replication, an error is not correctly signalled to the next transaction *** - ---connection server_2 ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - ---connection server_1 -INSERT INTO t3 VALUES (110, 1); ---save_master_pos - ---connection server_2 ---sync_with_master -SELECT * FROM t3 WHERE a >= 110 ORDER BY a; -# Inject a duplicate key error. -SET sql_log_bin=0; -INSERT INTO t3 VALUES (111, 666); -SET sql_log_bin=1; - ---connection server_1 - -# Create a group commit with two inserts, the first one conflicts with a row on the slave ---connect (con1,127.0.0.1,root,,test,$SERVER_MYPORT_1,) -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -send INSERT INTO t3 VALUES (111, 2); ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connect (con2,127.0.0.1,root,,test,$SERVER_MYPORT_1,) -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -send INSERT INTO t3 VALUES (112, 3); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con1 -REAP; ---connection con2 -REAP; -SET debug_sync='RESET'; ---save_master_pos - ---connection server_2 ---let $slave_sql_errno= 1062 ---source include/wait_for_slave_sql_error.inc ---source include/wait_for_slave_sql_to_stop.inc -# We should not see the row (112,3) here, it should be rolled back due to -# error signal from the prior transaction. -SELECT * FROM t3 WHERE a >= 110 ORDER BY a; -SET sql_log_bin=0; -DELETE FROM t3 WHERE a=111 AND b=666; -SET sql_log_bin=1; -START SLAVE SQL_THREAD; ---sync_with_master -SELECT * FROM t3 WHERE a >= 110 ORDER BY a; - - ---echo ***MDEV-5914: Parallel replication deadlock due to InnoDB lock conflicts *** ---connection server_2 ---source include/stop_slave.inc - ---connection server_1 -CREATE TABLE t4 (a INT PRIMARY KEY, b INT, KEY b_idx(b)) ENGINE=InnoDB; -INSERT INTO t4 VALUES (1,NULL), (2,2), (3,NULL), (4,4), (5, NULL), (6, 6); - -# Create a group commit with UPDATE and DELETE, in that order. -# The bug was that while the UPDATE's row lock does not block the DELETE, the -# DELETE's gap lock _does_ block the UPDATE. This could cause a deadlock -# on the slave. ---connection con1 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -send UPDATE t4 SET b=NULL WHERE a=6; ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connection con2 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -send DELETE FROM t4 WHERE b <= 3; - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con1 -REAP; ---connection con2 -REAP; -SET debug_sync='RESET'; ---save_master_pos - ---connection server_2 ---source include/start_slave.inc ---sync_with_master ---source include/stop_slave.inc - -SELECT * FROM t4 ORDER BY a; - - -# Another example, this one with INSERT vs. DELETE ---connection server_1 -DELETE FROM t4; -INSERT INTO t4 VALUES (1,NULL), (2,2), (3,NULL), (4,4), (5, NULL), (6, 6); - -# Create a group commit with INSERT and DELETE, in that order. -# The bug was that while the INSERT's insert intention lock does not block -# the DELETE, the DELETE's gap lock _does_ block the INSERT. This could cause -# a deadlock on the slave. ---connection con1 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -send INSERT INTO t4 VALUES (7, NULL); ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connection con2 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -send DELETE FROM t4 WHERE b <= 3; - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con1 -REAP; ---connection con2 -REAP; -SET debug_sync='RESET'; ---save_master_pos - ---connection server_2 ---source include/start_slave.inc ---sync_with_master ---source include/stop_slave.inc - -SELECT * FROM t4 ORDER BY a; - - -# MDEV-6549, failing to update gtid_slave_pos for a transaction that was retried. -# The problem was that when a transaction updates the mysql.gtid_slave_pos -# table, it clears the flag that marks that there is a GTID position that -# needs to be updated. Then, if the transaction got killed after that due -# to a deadlock, the subsequent retry would fail to notice that the GTID needs -# to be recorded in gtid_slave_pos. -# -# (In the original bug report, the symptom was an assertion; this was however -# just a side effect of the missing update of gtid_slave_pos, which also -# happened to cause a missing clear of OPTION_GTID_BEGIN). ---connection server_1 -DELETE FROM t4; -INSERT INTO t4 VALUES (1,NULL), (2,2), (3,NULL), (4,4), (5, NULL), (6, 6); - -# Create two transactions that can run in parallel on the slave but cause -# a deadlock if the second runs before the first. ---connection con1 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -send UPDATE t4 SET b=NULL WHERE a=6; ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connection con2 -# Must use statement-based binlogging. Otherwise the transaction will not be -# binlogged at all, as it modifies no rows. -SET @old_format= @@SESSION.binlog_format; -SET binlog_format='statement'; -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -send DELETE FROM t4 WHERE b <= 1; - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con1 -REAP; ---connection con2 -REAP; -SET @old_format=@@GLOBAL.binlog_format; -SET debug_sync='RESET'; ---save_master_pos ---let $last_gtid= `SELECT @@last_gtid` - ---connection server_2 -# Disable the usual skip of gap locks for transactions that are run in -# parallel, using DBUG. This allows the deadlock to occur, and this in turn -# triggers a retry of the second transaction, and the code that was buggy and -# caused the gtid_slave_pos update to be skipped in the retry. -SET @old_dbug= @@GLOBAL.debug_dbug; -SET GLOBAL debug_dbug="+d,disable_thd_need_ordering_with"; ---source include/start_slave.inc ---sync_with_master -SET GLOBAL debug_dbug=@old_dbug; - -SELECT * FROM t4 ORDER BY a; -# Check that the GTID of the second transaction was correctly recorded in -# gtid_slave_pos, in the variable as well as in the table. ---replace_result $last_gtid GTID -eval SET @last_gtid= '$last_gtid'; -SELECT IF(@@gtid_slave_pos LIKE CONCAT('%',@last_gtid,'%'), "GTID found ok", - CONCAT("GTID ", @last_gtid, " not found in gtid_slave_pos=", @@gtid_slave_pos)) - AS result; -SELECT "ROW FOUND" AS `Is the row found?` - FROM mysql.gtid_slave_pos - WHERE CONCAT(domain_id, "-", server_id, "-", seq_no) = @last_gtid; - - ---echo *** MDEV-5938: Exec_master_log_pos not updated at log rotate in parallel replication *** ---connection server_2 ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads=1; -SET DEBUG_SYNC= 'RESET'; ---source include/start_slave.inc - ---connection server_1 -CREATE TABLE t5 (a INT PRIMARY KEY, b INT); -INSERT INTO t5 VALUES (1,1); -INSERT INTO t5 VALUES (2,2), (3,8); -INSERT INTO t5 VALUES (4,16); ---save_master_pos - ---connection server_2 ---sync_with_master -let $io_file= query_get_value(SHOW SLAVE STATUS, Master_Log_File, 1); -let $io_pos= query_get_value(SHOW SLAVE STATUS, Read_Master_Log_Pos, 1); -let $sql_file= query_get_value(SHOW SLAVE STATUS, Relay_Master_Log_File, 1); -let $sql_pos= query_get_value(SHOW SLAVE STATUS, Exec_Master_Log_Pos, 1); ---disable_query_log -eval SELECT IF('$io_file' = '$sql_file', "OK", "Not ok, $io_file <> $sql_file") AS test_check; -eval SELECT IF('$io_pos' = '$sql_pos', "OK", "Not ok, $io_pos <> $sql_pos") AS test_check; ---enable_query_log - ---connection server_1 -FLUSH LOGS; ---source include/wait_for_binlog_checkpoint.inc ---save_master_pos - ---connection server_2 ---sync_with_master -let $io_file= query_get_value(SHOW SLAVE STATUS, Master_Log_File, 1); -let $io_pos= query_get_value(SHOW SLAVE STATUS, Read_Master_Log_Pos, 1); -let $sql_file= query_get_value(SHOW SLAVE STATUS, Relay_Master_Log_File, 1); -let $sql_pos= query_get_value(SHOW SLAVE STATUS, Exec_Master_Log_Pos, 1); ---disable_query_log -eval SELECT IF('$io_file' = '$sql_file', "OK", "Not ok, $io_file <> $sql_file") AS test_check; -eval SELECT IF('$io_pos' = '$sql_pos', "OK", "Not ok, $io_pos <> $sql_pos") AS test_check; ---enable_query_log - - ---echo *** MDEV_6435: Incorrect error handling when query binlogged partially on master with "killed" error *** - ---connection server_1 -CREATE TABLE t6 (a INT) ENGINE=MyISAM; -CREATE TRIGGER tr AFTER INSERT ON t6 FOR EACH ROW SET @a = 1; - ---connection con1 -SET @old_format= @@binlog_format; -SET binlog_format= statement; ---let $conid = `SELECT CONNECTION_ID()` -SET debug_sync='sp_head_execute_before_loop SIGNAL ready WAIT_FOR cont'; -send INSERT INTO t6 VALUES (1), (2), (3); - ---connection server_1 -SET debug_sync='now WAIT_FOR ready'; ---replace_result $conid CONID -eval KILL QUERY $conid; -SET debug_sync='now SIGNAL cont'; - ---connection con1 ---error ER_QUERY_INTERRUPTED ---reap -SET binlog_format= @old_format; -SET debug_sync='RESET'; ---let $after_error_gtid_pos= `SELECT @@gtid_binlog_pos` - ---connection server_1 -SET debug_sync='RESET'; - - ---connection server_2 ---let $slave_sql_errno= 1317 ---source include/wait_for_slave_sql_error.inc -STOP SLAVE IO_THREAD; ---replace_result $after_error_gtid_pos AFTER_ERROR_GTID_POS -eval SET GLOBAL gtid_slave_pos= '$after_error_gtid_pos'; ---source include/start_slave.inc - ---connection server_1 -INSERT INTO t6 VALUES (4); -SELECT * FROM t6 ORDER BY a; ---save_master_pos - ---connection server_2 ---sync_with_master -SELECT * FROM t6 ORDER BY a; - - ---echo *** MDEV-6551: Some replication errors are ignored if slave_parallel_threads > 0 *** - ---connection server_1 -INSERT INTO t2 VALUES (31); ---let $gtid1= `SELECT @@LAST_GTID` ---source include/save_master_gtid.inc - ---connection server_2 ---source include/sync_with_master_gtid.inc ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads= 0; ---source include/start_slave.inc - -# Force a duplicate key error on the slave. -SET sql_log_bin= 0; -INSERT INTO t2 VALUES (32); -SET sql_log_bin= 1; - ---connection server_1 -INSERT INTO t2 VALUES (32); ---let $gtid2= `SELECT @@LAST_GTID` -# Rotate the binlog; the bug is triggered when the master binlog file changes -# after the event group that causes the duplicate key error. -FLUSH LOGS; -INSERT INTO t2 VALUES (33); -INSERT INTO t2 VALUES (34); -SELECT * FROM t2 WHERE a >= 30 ORDER BY a; ---source include/save_master_gtid.inc - ---connection server_2 ---let $slave_sql_errno= 1062 ---source include/wait_for_slave_sql_error.inc - ---connection server_2 ---source include/stop_slave_io.inc -SET GLOBAL slave_parallel_threads=10; -START SLAVE; - ---let $slave_sql_errno= 1062 ---source include/wait_for_slave_sql_error.inc - -# Note: IO thread is still running at this point. -# The bug seems to have been that restarting the SQL thread after an error with -# the IO thread still running, somehow picks up a later relay log position and -# thus ends up skipping the failing event, rather than re-executing. - -START SLAVE SQL_THREAD; ---let $slave_sql_errno= 1062 ---source include/wait_for_slave_sql_error.inc - -SELECT * FROM t2 WHERE a >= 30 ORDER BY a; - -# Skip the duplicate error, so we can proceed. ---error ER_SLAVE_SKIP_NOT_IN_GTID -SET sql_slave_skip_counter= 1; ---source include/stop_slave_io.inc ---disable_query_log -eval SET GLOBAL gtid_slave_pos = REPLACE(@@gtid_slave_pos, "$gtid1", "$gtid2"); ---enable_query_log ---source include/start_slave.inc ---source include/sync_with_master_gtid.inc - -SELECT * FROM t2 WHERE a >= 30 ORDER BY a; - - ---echo *** MDEV-6775: Wrong binlog order in parallel replication *** ---connection server_1 -# A bit tricky bug to reproduce. On the master, we binlog in statement-mode -# two transactions, an UPDATE followed by a DELETE. On the slave, we replicate -# with binlog-mode set to ROW, which means the DELETE, which modifies no rows, -# is not binlogged. Then we inject a wait in the group commit code on the -# slave, shortly before the actual commit of the UPDATE. The bug was that the -# DELETE could wake up from wait_for_prior_commit() before the commit of the -# UPDATE. So the test could see the slave position updated to after DELETE, -# while the UPDATE was still not visible. -DELETE FROM t4; -INSERT INTO t4 VALUES (1,NULL), (3,NULL), (4,4), (5, NULL), (6, 6); ---source include/save_master_gtid.inc - ---connection server_2 ---source include/sync_with_master_gtid.inc ---source include/stop_slave.inc -SET @old_dbug= @@GLOBAL.debug_dbug; -SET GLOBAL debug_dbug="+d,inject_binlog_commit_before_get_LOCK_log"; -SET @old_format=@@GLOBAL.binlog_format; -SET GLOBAL binlog_format=ROW; -# Re-spawn the worker threads to be sure they pick up the new binlog format -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; - ---connection con1 -SET @old_format= @@binlog_format; -SET binlog_format= statement; -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -send UPDATE t4 SET b=NULL WHERE a=6; ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connection con2 -SET @old_format= @@binlog_format; -SET binlog_format= statement; -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -send DELETE FROM t4 WHERE b <= 3; - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con1 -REAP; -SET binlog_format= @old_format; ---connection con2 -REAP; -SET binlog_format= @old_format; -SET debug_sync='RESET'; ---save_master_pos -SELECT * FROM t4 ORDER BY a; - ---connection server_2 ---source include/start_slave.inc -SET debug_sync= 'now WAIT_FOR waiting'; ---sync_with_master -SELECT * FROM t4 ORDER BY a; -SET debug_sync= 'now SIGNAL cont'; - -# Re-spawn the worker threads to remove any DBUG injections or DEBUG_SYNC. ---source include/stop_slave.inc -SET GLOBAL debug_dbug=@old_dbug; -SET GLOBAL binlog_format= @old_format; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - - ---echo *** MDEV-7237: Parallel replication: incorrect relaylog position after stop/start the slave *** ---connection server_1 -INSERT INTO t2 VALUES (40); ---save_master_pos - ---connection server_2 ---sync_with_master ---source include/stop_slave.inc -CHANGE MASTER TO master_use_gtid=no; -SET @old_dbug= @@GLOBAL.debug_dbug; -# This DBUG injection causes a DEBUG_SYNC signal "scheduled_gtid_0_x_100" when -# GTID 0-1-100 has been scheduled for and fetched by a worker thread. -SET GLOBAL debug_dbug="+d,rpl_parallel_scheduled_gtid_0_x_100"; -# This DBUG injection causes a DEBUG_SYNC signal "wait_for_done_waiting" when -# STOP SLAVE has signalled all worker threads to stop. -SET GLOBAL debug_dbug="+d,rpl_parallel_wait_for_done_trigger"; -# Reset worker threads to make DBUG setting catch on. -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; - - ---connection server_1 -# Setup some transaction for the slave to replicate. -INSERT INTO t2 VALUES (41); -INSERT INTO t2 VALUES (42); -# Need to log the DELETE in statement format, so we can see it in processlist. -SET @old_format= @@binlog_format; -SET binlog_format= statement; -DELETE FROM t2 WHERE a=40; -SET binlog_format= @old_format; -INSERT INTO t2 VALUES (43); -INSERT INTO t2 VALUES (44); -# Force the slave to switch to a new relay log file. -FLUSH LOGS; -INSERT INTO t2 VALUES (45); -# Inject a GTID 0-1-100, which will trigger a DEBUG_SYNC signal when this -# transaction has been fetched by a worker thread. -SET gtid_seq_no=100; -INSERT INTO t2 VALUES (46); ---save_master_pos - ---connection con_temp2 -# Temporarily block the DELETE on a=40 from completing. -BEGIN; -SELECT * FROM t2 WHERE a=40 FOR UPDATE; - - ---connection server_2 ---source include/start_slave.inc - -# Wait for a worker thread to start on the DELETE that will be blocked -# temporarily by the SELECT FOR UPDATE. ---let $wait_condition= SELECT count(*) > 0 FROM information_schema.processlist WHERE state='updating' and info LIKE '%DELETE FROM t2 WHERE a=40%' ---source include/wait_condition.inc - -# The DBUG injection set above will make the worker thread signal the following -# debug_sync when the GTID 0-1-100 has been reached by a worker thread. -# Thus, at this point, the SQL driver thread has reached the next -# relay log file name, while a worker thread is still processing a -# transaction in the previous relay log file, blocked on the SELECT FOR -# UPDATE. -SET debug_sync= 'now WAIT_FOR scheduled_gtid_0_x_100'; -# At this point, the SQL driver thread is in the new relay log file, while -# the DELETE from the old relay log file is not yet complete. We will stop -# the slave at this point. The bug was that the DELETE statement would -# update the slave position to the _new_ relay log file name instead of -# its own old file name. Thus, by stoping and restarting the slave at this -# point, we would get an error at restart due to incorrect position. (If -# we would let the slave catch up before stopping, the incorrect position -# would be corrected by a later transaction). - -send STOP SLAVE; - ---connection con_temp2 -# Wait for STOP SLAVE to have proceeded sufficiently that it has signalled -# all worker threads to stop; this ensures that we will stop after the DELETE -# transaction (and not after a later transaction that might have been able -# to set a fixed position). -SET debug_sync= 'now WAIT_FOR wait_for_done_waiting'; -# Now release the row lock that was blocking the replication of DELETE. -ROLLBACK; - ---connection server_2 -reap; ---source include/wait_for_slave_sql_to_stop.inc -SELECT * FROM t2 WHERE a >= 40 ORDER BY a; -# Now restart the slave. With the bug present, this would start at an -# incorrect relay log position, causing relay log read error (or if unlucky, -# silently skip a number of events). ---source include/start_slave.inc ---sync_with_master -SELECT * FROM t2 WHERE a >= 40 ORDER BY a; ---source include/stop_slave.inc -SET GLOBAL debug_dbug=@old_dbug; -SET DEBUG_SYNC= 'RESET'; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; -CHANGE MASTER TO master_use_gtid=slave_pos; ---source include/start_slave.inc - - ---echo *** MDEV-7326 Server deadlock in connection with parallel replication *** -# We use three transactions, each in a separate group commit. -# T1 does mark_start_commit(), then gets a deadlock error. -# T2 wakes up and starts running -# T1 does unmark_start_commit() -# T3 goes to wait for T2 to start its commit -# T2 does mark_start_commit() -# The bug was that at this point, T3 got deadlocked. Because T1 has unmarked(), -# T3 did not yet see the count_committing_event_groups reach its target value -# yet. But when T1 later re-did mark_start_commit(), it failed to send a wakeup -# to T3. - ---connection server_2 ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=3; -SET GLOBAL debug_dbug="+d,rpl_parallel_simulate_temp_err_xid"; ---source include/start_slave.inc - ---connection server_1 -SET @old_format= @@SESSION.binlog_format; -SET binlog_format= STATEMENT; -# This debug_sync will linger on and be used to control T3 later. -INSERT INTO t1 VALUES (foo(50, - "rpl_parallel_start_waiting_for_prior SIGNAL t3_ready", - "rpl_parallel_end_of_group SIGNAL prep_ready WAIT_FOR prep_cont")); ---save_master_pos ---connection server_2 -# Wait for the debug_sync point for T3 to be set. But let the preparation -# transaction remain hanging, so that T1 and T2 will be scheduled for the -# remaining two worker threads. -SET DEBUG_SYNC= "now WAIT_FOR prep_ready"; - ---connection server_1 -INSERT INTO t2 VALUES (foo(50, - "rpl_parallel_simulate_temp_err_xid SIGNAL t1_ready1 WAIT_FOR t1_cont1", - "rpl_parallel_retry_after_unmark SIGNAL t1_ready2 WAIT_FOR t1_cont2")); ---save_master_pos - ---connection server_2 -SET DEBUG_SYNC= "now WAIT_FOR t1_ready1"; -# T1 has now done mark_start_commit(). It will later do a rollback and retry. - ---connection server_1 -# Use a MyISAM table for T2 and T3, so they do not trigger the -# rpl_parallel_simulate_temp_err_xid DBUG insertion on XID event. -INSERT INTO t1 VALUES (foo(51, - "rpl_parallel_before_mark_start_commit SIGNAL t2_ready1 WAIT_FOR t2_cont1", - "rpl_parallel_after_mark_start_commit SIGNAL t2_ready2")); - ---connection server_2 -SET DEBUG_SYNC= "now WAIT_FOR t2_ready1"; -# T2 has now started running, but has not yet done mark_start_commit() -SET DEBUG_SYNC= "now SIGNAL t1_cont1"; -SET DEBUG_SYNC= "now WAIT_FOR t1_ready2"; -# T1 has now done unmark_start_commit() in preparation for its retry. - ---connection server_1 -INSERT INTO t1 VALUES (52); -SET BINLOG_FORMAT= @old_format; -SELECT * FROM t2 WHERE a>=50 ORDER BY a; -SELECT * FROM t1 WHERE a>=50 ORDER BY a; - ---connection server_2 -# Let the preparation transaction complete, so that the same worker thread -# can continue with the transaction T3. -SET DEBUG_SYNC= "now SIGNAL prep_cont"; -SET DEBUG_SYNC= "now WAIT_FOR t3_ready"; -# T3 has now gone to wait for T2 to start committing -SET DEBUG_SYNC= "now SIGNAL t2_cont1"; -SET DEBUG_SYNC= "now WAIT_FOR t2_ready2"; -# T2 has now done mark_start_commit(). -# Let things run, and check that T3 does not get deadlocked. -SET DEBUG_SYNC= "now SIGNAL t1_cont2"; ---sync_with_master - ---connection server_1 ---save_master_pos ---connection server_2 ---sync_with_master -SELECT * FROM t2 WHERE a>=50 ORDER BY a; -SELECT * FROM t1 WHERE a>=50 ORDER BY a; -SET DEBUG_SYNC="reset"; - -# Re-spawn the worker threads to remove any DBUG injections or DEBUG_SYNC. ---source include/stop_slave.inc -SET GLOBAL debug_dbug=@old_dbug; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - - ---echo *** MDEV-7326 Server deadlock in connection with parallel replication *** -# Similar to the previous test, but with T2 and T3 in the same GCO. -# We use three transactions, T1 in one group commit and T2/T3 in another. -# T1 does mark_start_commit(), then gets a deadlock error. -# T2 wakes up and starts running -# T1 does unmark_start_commit() -# T3 goes to wait for T1 to start its commit -# T2 does mark_start_commit() -# The bug was that at this point, T3 got deadlocked. T2 increments the -# count_committing_event_groups but does not signal T3, as they are in -# the same GCO. Then later when T1 increments, it would also not signal -# T3, because now the count_committing_event_groups is not equal to the -# wait_count of T3 (it is one larger). - ---connection server_2 ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=3; -SET GLOBAL debug_dbug="+d,rpl_parallel_simulate_temp_err_xid"; ---source include/start_slave.inc - ---connection server_1 -SET @old_format= @@SESSION.binlog_format; -SET binlog_format= STATEMENT; -# This debug_sync will linger on and be used to control T3 later. -INSERT INTO t1 VALUES (foo(60, - "rpl_parallel_start_waiting_for_prior SIGNAL t3_ready", - "rpl_parallel_end_of_group SIGNAL prep_ready WAIT_FOR prep_cont")); ---save_master_pos ---connection server_2 -# Wait for the debug_sync point for T3 to be set. But let the preparation -# transaction remain hanging, so that T1 and T2 will be scheduled for the -# remaining two worker threads. -SET DEBUG_SYNC= "now WAIT_FOR prep_ready"; - ---connection server_1 -INSERT INTO t2 VALUES (foo(60, - "rpl_parallel_simulate_temp_err_xid SIGNAL t1_ready1 WAIT_FOR t1_cont1", - "rpl_parallel_retry_after_unmark SIGNAL t1_ready2 WAIT_FOR t1_cont2")); ---save_master_pos - ---connection server_2 -SET DEBUG_SYNC= "now WAIT_FOR t1_ready1"; -# T1 has now done mark_start_commit(). It will later do a rollback and retry. - -# Do T2 and T3 in a single group commit. -# Use a MyISAM table for T2 and T3, so they do not trigger the -# rpl_parallel_simulate_temp_err_xid DBUG insertion on XID event. ---connection con_temp3 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1'; -SET binlog_format=statement; -send INSERT INTO t1 VALUES (foo(61, - "rpl_parallel_before_mark_start_commit SIGNAL t2_ready1 WAIT_FOR t2_cont1", - "rpl_parallel_after_mark_start_commit SIGNAL t2_ready2")); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued1'; - ---connection con_temp4 -SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2'; -send INSERT INTO t6 VALUES (62); - ---connection server_1 -SET debug_sync='now WAIT_FOR master_queued2'; -SET debug_sync='now SIGNAL master_cont1'; - ---connection con_temp3 -REAP; ---connection con_temp4 -REAP; - ---connection server_1 -SET debug_sync='RESET'; -SET BINLOG_FORMAT= @old_format; -SELECT * FROM t2 WHERE a>=60 ORDER BY a; -SELECT * FROM t1 WHERE a>=60 ORDER BY a; -SELECT * FROM t6 WHERE a>=60 ORDER BY a; - ---connection server_2 -SET DEBUG_SYNC= "now WAIT_FOR t2_ready1"; -# T2 has now started running, but has not yet done mark_start_commit() -SET DEBUG_SYNC= "now SIGNAL t1_cont1"; -SET DEBUG_SYNC= "now WAIT_FOR t1_ready2"; -# T1 has now done unmark_start_commit() in preparation for its retry. - ---connection server_2 -# Let the preparation transaction complete, so that the same worker thread -# can continue with the transaction T3. -SET DEBUG_SYNC= "now SIGNAL prep_cont"; -SET DEBUG_SYNC= "now WAIT_FOR t3_ready"; -# T3 has now gone to wait for T2 to start committing -SET DEBUG_SYNC= "now SIGNAL t2_cont1"; -SET DEBUG_SYNC= "now WAIT_FOR t2_ready2"; -# T2 has now done mark_start_commit(). -# Let things run, and check that T3 does not get deadlocked. -SET DEBUG_SYNC= "now SIGNAL t1_cont2"; ---sync_with_master - ---connection server_1 ---save_master_pos ---connection server_2 ---sync_with_master -SELECT * FROM t2 WHERE a>=60 ORDER BY a; -SELECT * FROM t1 WHERE a>=60 ORDER BY a; -SELECT * FROM t6 WHERE a>=60 ORDER BY a; -SET DEBUG_SYNC="reset"; - -# Re-spawn the worker threads to remove any DBUG injections or DEBUG_SYNC. ---source include/stop_slave.inc -SET GLOBAL debug_dbug=@old_dbug; -SET GLOBAL slave_parallel_threads=0; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - - ---echo *** MDEV-7335: Potential parallel slave deadlock with specific binlog corruption *** - ---connection server_2 ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads=1; -SET @old_dbug= @@GLOBAL.debug_dbug; -SET GLOBAL debug_dbug="+d,slave_discard_xid_for_gtid_0_x_1000"; - ---connection server_1 -INSERT INTO t2 VALUES (101); -INSERT INTO t2 VALUES (102); -INSERT INTO t2 VALUES (103); -INSERT INTO t2 VALUES (104); -INSERT INTO t2 VALUES (105); -# Inject a partial event group (missing XID at the end). The bug was that such -# partial group was not handled appropriately, leading to server deadlock. -SET gtid_seq_no=1000; -INSERT INTO t2 VALUES (106); -INSERT INTO t2 VALUES (107); -INSERT INTO t2 VALUES (108); -INSERT INTO t2 VALUES (109); -INSERT INTO t2 VALUES (110); -INSERT INTO t2 VALUES (111); -INSERT INTO t2 VALUES (112); -INSERT INTO t2 VALUES (113); -INSERT INTO t2 VALUES (114); -INSERT INTO t2 VALUES (115); -INSERT INTO t2 VALUES (116); -INSERT INTO t2 VALUES (117); -INSERT INTO t2 VALUES (118); -INSERT INTO t2 VALUES (119); -INSERT INTO t2 VALUES (120); -INSERT INTO t2 VALUES (121); -INSERT INTO t2 VALUES (122); -INSERT INTO t2 VALUES (123); -INSERT INTO t2 VALUES (124); -INSERT INTO t2 VALUES (125); -INSERT INTO t2 VALUES (126); -INSERT INTO t2 VALUES (127); -INSERT INTO t2 VALUES (128); -INSERT INTO t2 VALUES (129); -INSERT INTO t2 VALUES (130); ---source include/save_master_gtid.inc - ---connection server_2 ---source include/start_slave.inc ---source include/sync_with_master_gtid.inc -# The partial event group (a=106) should be rolled back and thus missing. -SELECT * FROM t2 WHERE a >= 100 ORDER BY a; - ---source include/stop_slave.inc -SET GLOBAL debug_dbug=@old_dbug; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - - ---echo *** MDEV-7847: "Slave worker thread retried transaction 10 time(s) in vain, giving up", followed by replication hanging *** ---echo *** MDEV-7882: Excessive transaction retry in parallel replication *** - ---connection server_1 -CREATE TABLE t7 (a int PRIMARY KEY, b INT) ENGINE=InnoDB; -CREATE TABLE t8 (a int PRIMARY KEY, b INT) ENGINE=InnoDB; ---save_master_pos - ---connection server_2 ---sync_with_master ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads=40; -SELECT @old_retries:=@@GLOBAL.slave_transaction_retries; -SET GLOBAL slave_transaction_retries= 5; - - -# Using dbug error injection, we artificially create event groups with a lot of -# conflicting transactions in each event group. The bugs were originally seen -# "in the wild" with transactions that did not conflict on the master, and only -# conflicted very rarely on the slave (maybe some edge case with InnoDB btree -# page splits or something like that). The event groups here loosely reflect -# the structure of the original failure's group commits. - - ---connection server_1 -INSERT INTO t7 VALUES (1,1), (2,2), (3,3), (4,4), (5,5); -SET @old_dbug= @@SESSION.debug_dbug; -SET @commit_id= 42; -SET SESSION debug_dbug="+d,binlog_force_commit_id"; -INSERT INTO t8 VALUES (1,1); -INSERT INTO t8 VALUES (2,2); -INSERT INTO t8 VALUES (3,3); -INSERT INTO t8 VALUES (4,4); -INSERT INTO t8 VALUES (5,5); -INSERT INTO t8 VALUES (6,6); -INSERT INTO t8 VALUES (7,7); -INSERT INTO t8 VALUES (8,8); - -UPDATE t7 SET b=9 WHERE a=3; -UPDATE t7 SET b=10 WHERE a=3; -UPDATE t7 SET b=11 WHERE a=3; - -INSERT INTO t8 VALUES (12,12); -INSERT INTO t8 VALUES (13,13); - -UPDATE t7 SET b=14 WHERE a=3; -UPDATE t7 SET b=15 WHERE a=3; - -INSERT INTO t8 VALUES (16,16); - -UPDATE t7 SET b=17 WHERE a=3; - -INSERT INTO t8 VALUES (18,18); -INSERT INTO t8 VALUES (19,19); - -UPDATE t7 SET b=20 WHERE a=3; - -INSERT INTO t8 VALUES (21,21); - -UPDATE t7 SET b=22 WHERE a=3; - -INSERT INTO t8 VALUES (23,24); -INSERT INTO t8 VALUES (24,24); - -UPDATE t7 SET b=25 WHERE a=3; - -INSERT INTO t8 VALUES (26,26); - -UPDATE t7 SET b=27 WHERE a=3; - -BEGIN; -INSERT INTO t8 VALUES (28,28); -INSERT INTO t8 VALUES (29,28), (30,28); -INSERT INTO t8 VALUES (31,28); -INSERT INTO t8 VALUES (32,28); -INSERT INTO t8 VALUES (33,28); -INSERT INTO t8 VALUES (34,28); -INSERT INTO t8 VALUES (35,28); -INSERT INTO t8 VALUES (36,28); -INSERT INTO t8 VALUES (37,28); -INSERT INTO t8 VALUES (38,28); -INSERT INTO t8 VALUES (39,28); -INSERT INTO t8 VALUES (40,28); -INSERT INTO t8 VALUES (41,28); -INSERT INTO t8 VALUES (42,28); -COMMIT; - - -SET @commit_id=43; -INSERT INTO t8 VALUES (43,43); -INSERT INTO t8 VALUES (44,44); - -UPDATE t7 SET b=45 WHERE a=3; - -INSERT INTO t8 VALUES (46,46); -INSERT INTO t8 VALUES (47,47); - -UPDATE t7 SET b=48 WHERE a=3; - -INSERT INTO t8 VALUES (49,49); -INSERT INTO t8 VALUES (50,50); - - -SET @commit_id=44; -INSERT INTO t8 VALUES (51,51); -INSERT INTO t8 VALUES (52,52); - -UPDATE t7 SET b=53 WHERE a=3; - -INSERT INTO t8 VALUES (54,54); -INSERT INTO t8 VALUES (55,55); - -UPDATE t7 SET b=56 WHERE a=3; - -INSERT INTO t8 VALUES (57,57); - -UPDATE t7 SET b=58 WHERE a=3; - -INSERT INTO t8 VALUES (58,58); -INSERT INTO t8 VALUES (59,59); -INSERT INTO t8 VALUES (60,60); -INSERT INTO t8 VALUES (61,61); - -UPDATE t7 SET b=62 WHERE a=3; - -INSERT INTO t8 VALUES (63,63); -INSERT INTO t8 VALUES (64,64); -INSERT INTO t8 VALUES (65,65); -INSERT INTO t8 VALUES (66,66); - -UPDATE t7 SET b=67 WHERE a=3; - -INSERT INTO t8 VALUES (68,68); - -UPDATE t7 SET b=69 WHERE a=3; -UPDATE t7 SET b=70 WHERE a=3; -UPDATE t7 SET b=71 WHERE a=3; - -INSERT INTO t8 VALUES (72,72); - -UPDATE t7 SET b=73 WHERE a=3; -UPDATE t7 SET b=74 WHERE a=3; -UPDATE t7 SET b=75 WHERE a=3; -UPDATE t7 SET b=76 WHERE a=3; - -INSERT INTO t8 VALUES (77,77); - -UPDATE t7 SET b=78 WHERE a=3; - -INSERT INTO t8 VALUES (79,79); - -UPDATE t7 SET b=80 WHERE a=3; - -INSERT INTO t8 VALUES (81,81); - -UPDATE t7 SET b=82 WHERE a=3; - -INSERT INTO t8 VALUES (83,83); - -UPDATE t7 SET b=84 WHERE a=3; - - -SET @commit_id=45; -INSERT INTO t8 VALUES (85,85); -UPDATE t7 SET b=86 WHERE a=3; -INSERT INTO t8 VALUES (87,87); - - -SET @commit_id=46; -INSERT INTO t8 VALUES (88,88); -INSERT INTO t8 VALUES (89,89); -INSERT INTO t8 VALUES (90,90); - -SET SESSION debug_dbug=@old_dbug; - -INSERT INTO t8 VALUES (91,91); -INSERT INTO t8 VALUES (92,92); -INSERT INTO t8 VALUES (93,93); -INSERT INTO t8 VALUES (94,94); -INSERT INTO t8 VALUES (95,95); -INSERT INTO t8 VALUES (96,96); -INSERT INTO t8 VALUES (97,97); -INSERT INTO t8 VALUES (98,98); -INSERT INTO t8 VALUES (99,99); - - -SELECT * FROM t7 ORDER BY a; -SELECT * FROM t8 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 t7 ORDER BY a; -SELECT * FROM t8 ORDER BY a; - ---source include/stop_slave.inc -SET GLOBAL slave_transaction_retries= @old_retries; -SET GLOBAL slave_parallel_threads=10; ---source include/start_slave.inc - - ---echo *** MDEV-7888: ANALYZE TABLE does wakeup_subsequent_commits(), causing wrong binlog order and parallel replication hang *** - ---connection server_2 ---source include/stop_slave.inc -SET @old_dbug= @@GLOBAL.debug_dbug; -SET GLOBAL debug_dbug= '+d,inject_analyze_table_sleep'; - ---connection server_1 -# Inject two group commits. 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). - -SET @old_dbug= @@SESSION.debug_dbug; -SET SESSION debug_dbug="+d,binlog_force_commit_id"; - -# Group commit with cid=10000, two event groups. -SET @commit_id= 10000; -ANALYZE TABLE t2; -INSERT INTO t3 VALUES (120, 0); - -# Group commit with cid=10001, one event group. -SET @commit_id= 10001; -INSERT INTO t3 VALUES (121, 0); - -SET SESSION debug_dbug=@old_dbug; - -SELECT * FROM t3 WHERE a >= 120 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 t3 WHERE a >= 120 ORDER BY a; - ---source include/stop_slave.inc -SET GLOBAL debug_dbug= @old_dbug; ---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; -SET GLOBAL debug_dbug= '+d,inject_record_gtid_serverid_100_sleep'; - ---connection server_1 -# Inject two group commits. The bug was that record_gtid for a -# non-transactional event group would commit its own transaction, which would -# cause ha_commit_trans() to call wakeup_subsequent_commits() too early. This -# in turn lead to access to freed group_commit_orderer object, losing a wakeup -# and causing slave threads to hang. -# We inject a small sleep in the corresponding record_gtid() to make the race -# easier to hit. - -SET @old_dbug= @@SESSION.debug_dbug; -SET SESSION debug_dbug="+d,binlog_force_commit_id"; - -# Group commit with cid=10010, two event groups. -SET @old_server_id= @@SESSION.server_id; -SET SESSION server_id= 100; -SET @commit_id= 10010; -ALTER TABLE t1 COMMENT "Hulubulu!"; -SET SESSION server_id= @old_server_id; -INSERT INTO t3 VALUES (130, 0); - -# Group commit with cid=10011, one event group. -SET @commit_id= 10011; -INSERT INTO t3 VALUES (131, 0); - -SET SESSION debug_dbug=@old_dbug; - -SELECT * FROM t3 WHERE a >= 130 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 t3 WHERE a >= 130 ORDER BY a; - ---source include/stop_slave.inc -SET GLOBAL debug_dbug= @old_dbug; ---source include/start_slave.inc - - ---echo *** MDEV-8031: Parallel replication stops on "connection killed" error (probably incorrectly handled deadlock kill) *** - ---connection server_1 -INSERT INTO t3 VALUES (201,0), (202,0); ---source include/save_master_gtid.inc - ---connection server_2 ---source include/sync_with_master_gtid.inc ---source include/stop_slave.inc -SET @old_dbug= @@GLOBAL.debug_dbug; -SET GLOBAL debug_dbug= '+d,inject_mdev8031'; - ---connection server_1 -# We artificially create a situation that hopefully resembles the original -# bug which was only seen "in the wild", and only once. -# Setup a fake group commit with lots of conflicts that will lead to deadloc -# kill. The slave DBUG injection causes the slave to be deadlock killed at -# a particular point during the retry, and then later do a small sleep at -# another critical point where the prior transaction then has a chance to -# complete. Finally an extra KILL check catches an unhandled, lingering -# deadlock kill. So rather artificial, but at least it exercises the -# relevant code paths. -SET @old_dbug= @@SESSION.debug_dbug; -SET SESSION debug_dbug="+d,binlog_force_commit_id"; - -SET @commit_id= 10200; -INSERT INTO t3 VALUES (203, 1); -INSERT INTO t3 VALUES (204, 1); -INSERT INTO t3 VALUES (205, 1); -UPDATE t3 SET b=b+1 WHERE a=201; -UPDATE t3 SET b=b+1 WHERE a=201; -UPDATE t3 SET b=b+1 WHERE a=201; -UPDATE t3 SET b=b+1 WHERE a=202; -UPDATE t3 SET b=b+1 WHERE a=202; -UPDATE t3 SET b=b+1 WHERE a=202; -UPDATE t3 SET b=b+1 WHERE a=202; -UPDATE t3 SET b=b+1 WHERE a=203; -UPDATE t3 SET b=b+1 WHERE a=203; -UPDATE t3 SET b=b+1 WHERE a=204; -UPDATE t3 SET b=b+1 WHERE a=204; -UPDATE t3 SET b=b+1 WHERE a=204; -UPDATE t3 SET b=b+1 WHERE a=203; -UPDATE t3 SET b=b+1 WHERE a=205; -UPDATE t3 SET b=b+1 WHERE a=205; -SET SESSION debug_dbug=@old_dbug; - -SELECT * FROM t3 WHERE a>=200 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 t3 WHERE a>=200 ORDER BY a; ---source include/stop_slave.inc -SET GLOBAL debug_dbug= @old_dbug; ---source include/start_slave.inc - - ---echo *** Check getting deadlock killed inside open_binlog() during retry. *** - ---connection server_2 ---source include/stop_slave.inc -SET @old_dbug= @@GLOBAL.debug_dbug; -SET GLOBAL debug_dbug= '+d,inject_retry_event_group_open_binlog_kill'; -SET @old_max= @@GLOBAL.max_relay_log_size; -SET GLOBAL max_relay_log_size= 4096; - ---connection server_1 -SET @old_dbug= @@SESSION.debug_dbug; -SET SESSION debug_dbug="+d,binlog_force_commit_id"; - ---let $large= `SELECT REPEAT("*", 8192)` -SET @commit_id= 10210; ---echo Omit long queries that cause relaylog rotations and transaction retries... ---disable_query_log -eval UPDATE t3 SET b=b+1 WHERE a=201 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=201 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=201 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=202 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=202 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=202 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=202 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=203 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=203 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=204 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=204 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=204 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=203 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=205 /* $large */; -eval UPDATE t3 SET b=b+1 WHERE a=205 /* $large */; ---enable_query_log -SET SESSION debug_dbug=@old_dbug; - -SELECT * FROM t3 WHERE a>=200 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 t3 WHERE a>=200 ORDER BY a; ---source include/stop_slave.inc -SET GLOBAL debug_dbug= @old_debg; -SET GLOBAL max_relay_log_size= @old_max; ---source include/start_slave.inc - - ---echo *** MDEV-8302: Duplicate key with parallel replication *** - ---connection server_2 ---source include/stop_slave.inc -/* Inject a small sleep which makes the race easier to hit. */ -SET @old_dbug=@@GLOBAL.debug_dbug; -SET GLOBAL debug_dbug="+d,inject_mdev8302"; - - ---connection server_1 -INSERT INTO t7 VALUES (100,1), (101,2), (102,3), (103,4), (104,5); - -# Artificially create a bunch of group commits with conflicting transactions. -# The bug happened when T1 and T2 was in one group commit, and T3 was in the -# following group commit. T2 is a DELETE of a row with same primary key as a -# row that T3 inserts. T1 and T2 can conflict, causing T2 to be deadlock -# killed after starting to commit. The bug was that T2 could roll back before -# doing unmark_start_commit(); this could allow T3 to run before the retry -# of T2, causing duplicate key violation. - -SET @old_dbug= @@SESSION.debug_dbug; -SET @commit_id= 20000; -SET SESSION debug_dbug="+d,binlog_force_commit_id"; - ---let $n = 100 ---disable_query_log -while ($n) -{ - eval UPDATE t7 SET b=b+1 WHERE a=100+($n MOD 5); - eval DELETE FROM t7 WHERE a=100+($n MOD 5); - - SET @commit_id = @commit_id + 1; - eval INSERT INTO t7 VALUES (100+($n MOD 5), $n); - SET @commit_id = @commit_id + 1; - dec $n; -} ---enable_query_log -SET SESSION debug_dbug=@old_dbug; - - -SELECT * FROM t7 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 t7 ORDER BY a; - ---source include/stop_slave.inc -SET GLOBAL debug_dbug=@old_dbug; ---source include/start_slave.inc - - - ---echo *** MDEV-8725: Assertion on ROLLBACK statement in the binary log *** ---connection server_1 -# Inject an event group terminated by ROLLBACK, by mixing MyISAM and InnoDB -# in a transaction. The bug was an assertion on the ROLLBACK due to -# mark_start_commit() being already called. ---disable_warnings -BEGIN; -INSERT INTO t2 VALUES (200); -INSERT INTO t1 VALUES (200); -INSERT INTO t2 VALUES (201); -ROLLBACK; ---enable_warnings -SELECT * FROM t1 WHERE a>=200 ORDER BY a; -SELECT * FROM t2 WHERE a>=200 ORDER BY a; ---source include/save_master_gtid.inc - ---connection server_2 ---source include/sync_with_master_gtid.inc -SELECT * FROM t1 WHERE a>=200 ORDER BY a; -SELECT * FROM t2 WHERE a>=200 ORDER BY a; - - -# Clean up. ---connection server_2 ---source include/stop_slave.inc -SET GLOBAL slave_parallel_threads=@old_parallel_threads; ---source include/start_slave.inc -SET DEBUG_SYNC= 'RESET'; - ---connection server_1 -DROP function foo; -DROP TABLE t1,t2,t3,t4,t5,t6,t7,t8; -SET DEBUG_SYNC= 'RESET'; - ---source include/rpl_end.inc +--source extra/rpl_tests/rpl_parallel.inc |