Hi Zhenxing,
Actually, START SLAVE UNTIL ... never stops the IO thread, only the SQL
thread. So I think that it should be ok.
/Sven
He Zhenxing wrote:
> Hi Sven
>
> I think your fix also has race condition.
>
> 1) start slave until ...;
> 2) source include/wait_for_slave_io_to_start.inc;
> 3) source include/wait_for_slave_sql_to_stop.inc;
>
> if before running 2), the slave I/O thread has already started and
> finished replication to given position and stopped, then 2) will fail to
> wait the slave I/O start and thus the test will fail.
>
> Sven Sandberg wrote:
>> #At file:///home/sven/bzr/b37717-rpl_stm_until/5.1-rpl/
>>
>> 2633 Sven Sandberg 2008-07-23
>> BUG#37717: rpl.rpl_stm_until 'stmt' fails sporadically on pushbuild
>> Problem: After START SLAVE, the Slave_IO_Status column of
>> SHOW SLAVE STATUS goes from No to Yes asynchronously. That
>> caused sporadic failures on pushbuild in rpl_stm_until since
>> the test contains SHOW SLAVE STATUS right after START SLAVE.
>> Fix: Wait until Slave_IO_Status becomes Yes after each
>> START SLAVE.
>> added:
>> mysql-test/include/wait_for_slave_io_to_start.inc
>> modified:
>> mysql-test/suite/rpl/r/rpl_stm_until.result
>> mysql-test/suite/rpl/t/rpl_stm_until.test
>>
>> per-file messages:
>> mysql-test/include/wait_for_slave_io_to_start.inc
>> Macro that waits until the Slave_IO_Running field of
>> SHOW SLAVE STATUS becomes Yes.
>> mysql-test/suite/rpl/r/rpl_stm_until.result
>> updated result file
>> mysql-test/suite/rpl/t/rpl_stm_until.test
>> - Added wait_for_slave_io_to_start after each start slave.
>> - Removed unused initialization of test variable $VERSION
>> - Added comments.
>> === added file 'mysql-test/include/wait_for_slave_io_to_start.inc'
>> --- a/mysql-test/include/wait_for_slave_io_to_start.inc 1970-01-01 00:00:00
> +0000
>> +++ b/mysql-test/include/wait_for_slave_io_to_start.inc 2008-07-23 11:23:52
> +0000
>> @@ -0,0 +1,19 @@
>> +# ==== Purpose ====
>> +#
>> +# Waits until the IO thread of the current connection has started and
>> +# connected to the master (i.e., until SHOW SLAVE STATUS returns Yes
>> +# in the Slave_IO_Running field), or until a timeout is reached.
>> +#
>> +# ==== Usage ====
>> +#
>> +# source include/wait_for_slave_io_to_start.inc;
>> +#
>> +# Parameters to this macro are $slave_timeout and
>> +# $slave_keep_connection. See wait_for_slave_param.inc for
>> +# descriptions.
>> +
>> +let $slave_param= Slave_IO_Running;
>> +let $slave_param_value= Yes;
>> +let $slave_error_message= Failed while waiting for slave IO thread to start;
>> +source include/wait_for_slave_param.inc;
>> +let $slave_error_message= ;
>>
>> === modified file 'mysql-test/suite/rpl/r/rpl_stm_until.result'
>> --- a/mysql-test/suite/rpl/r/rpl_stm_until.result 2008-07-10 16:09:39 +0000
>> +++ b/mysql-test/suite/rpl/r/rpl_stm_until.result 2008-07-23 11:23:52 +0000
>> @@ -4,7 +4,10 @@ reset master;
>> reset slave;
>> drop table if exists t1,t2,t3,t4,t5,t6,t7,t8,t9;
>> start slave;
>> +[on slave]
>> include/stop_slave.inc
>> +==== Create some events on master ====
>> +[on master]
>> create table t1(n int not null auto_increment primary key);
>> insert into t1 values (1),(2),(3),(4);
>> drop table t1;
>> @@ -12,6 +15,8 @@ create table t2(n int not null auto_incr
>> insert into t2 values (1),(2);
>> insert into t2 values (3),(4);
>> drop table t2;
>> +==== Replicate one event at a time on slave ====
>> +[on slave]
>> start slave until master_log_file='master-bin.000001', master_log_pos=323;
>> select * from t1;
>> n
>> @@ -149,6 +154,8 @@ Last_IO_Error #
>> Last_SQL_Errno 0
>> Last_SQL_Error
>> start slave;
>> +[on master]
>> +[on slave]
>> include/stop_slave.inc
>> start slave until master_log_file='master-bin.000001', master_log_pos=776;
>> SHOW SLAVE STATUS;
>> @@ -190,6 +197,7 @@ Last_IO_Errno #
>> Last_IO_Error #
>> Last_SQL_Errno 0
>> Last_SQL_Error
>> +==== Test various error conditions ====
>> start slave until master_log_file='master-bin', master_log_pos=561;
>> ERROR HY000: Incorrect parameter or combination of parameters for START SLAVE
> UNTIL
>> start slave until master_log_file='master-bin.000001', master_log_pos=561,
> relay_log_pos=12;
>>
>> === modified file 'mysql-test/suite/rpl/t/rpl_stm_until.test'
>> --- a/mysql-test/suite/rpl/t/rpl_stm_until.test 2008-07-10 16:09:39 +0000
>> +++ b/mysql-test/suite/rpl/t/rpl_stm_until.test 2008-07-23 11:23:52 +0000
>> @@ -1,18 +1,37 @@
>> +# ==== Purpose ====
>> +#
>> +# Verify that START SLAVE UNTIL replicates until the given binlog
>> +# position but not longer. Verify that START SLAVE UNTIL with various
>> +# incorrect arguments gives an error.
>> +#
>> +# ==== Method ====
>> +#
>> +# On master, create a table and insert some rows. On slave, START
>> +# SLAVE UNTIL so that it reads one event at a time, and check the
>> +# table and the slave status each time.
>> +#
>> +# Then, on slave, run START SLAVE UNTIL with incorrect arguments and
>> +# verify that it gives an error.
>> +#
>> +# ==== Related bugs ====
>> +#
>> +# Bug in this test: BUG#37717: rpl.rpl_stm_until 'stmt' fails sporadically on
> pushbuild
>> +
>> -- source include/have_binlog_format_mixed_or_statement.inc
>> -- source include/master-slave.inc
>>
>> # Test is dependent on binlog positions
>>
>> -# prepare version for substitutions
>> -let $VERSION=`select version()`;
>> -
>> -# stop slave before he will start replication also sync with master
>> -# for avoiding undetermenistic behaviour
>> +# Stop slave before it starts replication. Also sync with master
>> +# to avoid nondeterministic behaviour.
>> +--echo [on slave]
>> sync_slave_with_master;
>> --source include/stop_slave.inc
>>
>> +--echo ==== Create some events on master ====
>> +
>> +--echo [on master]
>> connection master;
>> -# create some events on master
>> create table t1(n int not null auto_increment primary key);
>> insert into t1 values (1),(2),(3),(4);
>> drop table t1;
>> @@ -21,9 +40,13 @@ insert into t2 values (1),(2);
>> insert into t2 values (3),(4);
>> drop table t2;
>>
>> +--echo ==== Replicate one event at a time on slave ====
>> +
>> # try to replicate all queries until drop of t1
>> +--echo [on slave]
>> connection slave;
>> start slave until master_log_file='master-bin.000001', master_log_pos=323;
>> +--source include/wait_for_slave_io_to_start.inc
>> --source include/wait_for_slave_sql_to_stop.inc
>> # here table should be still not deleted
>> select * from t1;
>> @@ -31,6 +54,7 @@ source include/show_slave_status2.inc;
>>
>> # this should fail right after start
>> start slave until master_log_file='master-no-such-bin.000001',
> master_log_pos=291;
>> +--source include/wait_for_slave_io_to_start.inc
>> --source include/wait_for_slave_sql_to_stop.inc
>> # again this table should be still not deleted
>> select * from t1;
>> @@ -38,24 +62,28 @@ source include/show_slave_status2.inc;
>>
>> # try replicate all up to and not including the second insert to t2;
>> start slave until relay_log_file='slave-relay-bin.000004', relay_log_pos=746;
>> +--source include/wait_for_slave_io_to_start.inc
>> --source include/wait_for_slave_sql_to_stop.inc
>> select * from t2;
>> source include/show_slave_status2.inc;
>>
>> # clean up
>> start slave;
>> +--echo [on master]
>> connection master;
>> +--echo [on slave]
>> sync_slave_with_master;
>> --source include/stop_slave.inc
>>
>> # this should stop immediately as we are already there
>> start slave until master_log_file='master-bin.000001', master_log_pos=776;
>> +--source include/wait_for_slave_io_to_start.inc
>> --source include/wait_for_slave_sql_to_stop.inc
>> -# here the sql slave thread should be stopped
>> --replace_result bin.000005 bin.000004 bin.000006 bin.000004 bin.000007
> bin.000004
>> source include/show_slave_status2.inc;
>>
>> -#testing various error conditions
>> +--echo ==== Test various error conditions ====
>> +
>> --error 1277
>> start slave until master_log_file='master-bin', master_log_pos=561;
>> --error 1277
>> @@ -70,4 +98,3 @@ start slave until relay_log_file='slave-
>> start slave sql_thread;
>> start slave until master_log_file='master-bin.000001', master_log_pos=776;
>>
>> -# End of 4.1 tests
>>
>
>
--
Sven Sandberg, Software Engineer
MySQL AB, www.mysql.com
