Skip to content

Commit

Permalink
MDEV-8518 rpl.sec_behind_master-5114 fails sporadically in buildbot
Browse files Browse the repository at this point in the history
- fix the test to avoid false-negatives before MDEV-5114 patch;
- fix the race condition which made the test fail on slow builders
  • Loading branch information
elenst committed Jan 4, 2017
1 parent bc4cac3 commit 9bf9270
Show file tree
Hide file tree
Showing 2 changed files with 41 additions and 4 deletions.
3 changes: 2 additions & 1 deletion mysql-test/suite/rpl/r/sec_behind_master-5114.result
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,8 @@ include/master-slave.inc
call mtr.add_suppression("Unsafe statement written to the binary log");
CREATE TABLE t1 (a int);
INSERT INTO t1 VALUES(SLEEP(2));
Seconds_Behind_Master: 1
Seconds_Behind_Master_is_less_than_100
1
Warnings:
Note 1592 Unsafe statement written to the binary log using statement format since BINLOG_FORMAT = STATEMENT. Statement is unsafe because it uses a system function that may return a different value on the slave.
drop table t1;
Expand Down
42 changes: 39 additions & 3 deletions mysql-test/suite/rpl/t/sec_behind_master-5114.test
Original file line number Diff line number Diff line change
Expand Up @@ -3,24 +3,60 @@ source include/have_binlog_format_statement.inc;

call mtr.add_suppression("Unsafe statement written to the binary log");


# Make sure that the start time of the first event is certainly different
# from the next event
--let $timestamp= `SELECT @@timestamp`
--disable_query_log
eval SET TIMESTAMP= $timestamp-100;
--enable_query_log
CREATE TABLE t1 (a int);

# Make sure that the slave is done with the first event, and all checks
# that we'll perform later will be really against the second event
sync_slave_with_master;

connection master;

# Restore the timestamp now. It doesn't matter that it's not precise,
# it just needs to be very different from the earlier event
--disable_query_log
eval SET TIMESTAMP= $timestamp;
--enable_query_log

send INSERT INTO t1 VALUES(SLEEP(2));

connection slave;
let $run = 10;

# When the slave starts executing the event, Seconds_Behind_Master
# should start growing steadilly. The bugfix ensures that they are
# calculated based on the start time of the current event, rather
# than the start time of the previous event. To check it, we only need
# the first non-zero value

let $run = 20;
while ($run)
{
dec $run;
let $sbm=query_get_value(SHOW SLAVE STATUS, Seconds_Behind_Master, 1);
# for debugging uncomment echo and remove the if()
#echo Seconds_Behind_Master: $sbm;
# echo Seconds_Behind_Master: $sbm;
if ($sbm)
{
let $run = 0;
}
sleep 0.5;
}
echo Seconds_Behind_Master: $sbm;

# Normally the first non-zero value should be 1. However, due to race
# conditions on slow servers, sometimes the check might miss the value 1,
# and only catch a higher one. It does not matter, we just need to make
# sure it didn't start with 100+, as it would have with bug MDEV-5114

--disable_query_log
eval SELECT $sbm > 0 and $sbm < 99 AS Seconds_Behind_Master_is_less_than_100;
--enable_query_log

connection master;
reap;
drop table t1;
Expand Down

0 comments on commit 9bf9270

Please sign in to comment.