rpl.rpl_extra_col_master_myisam test unstable on 5.5

[Laurynas Biveinis]

rpl_extra_col_master_myisam test fails on 5.5 with

rpl.rpl_extra_col_master_myisam 'row' w4 [ fail ] Found warnings/errors in server log file!
        Test ended at 2016-07-19 21:13:07
line
160719 21:13:07 [Warning] Slave I/O: SET @master_heartbeat_period to master failed with error: Lost connection to MySQL server during query, Error_code: 2013
^ Found warnings in /mnt/workspace/percona-server-5.5-trunk/BUILD_TYPE/debug/Host/ubuntu-precise-64bit/mysql-test/var/4/log/mysqld.2.err
ok

 - saving '/mnt/workspace/percona-server-5.5-trunk/BUILD_TYPE/debug/Host/ubuntu-precise-64bit/mysql-test/var/4/log/rpl.rpl_extra_col_master_myisam-row/' to '/mnt/workspace/percona-server-5.5-trunk/BUILD_TYPE/debug/Host/ubuntu-precise-64bit/mysql-test/var/log/rpl.rpl_extra_col_master_myisam-row/'

The likely cause is the test doing

START SLAVE;

followed almost immediately by

STOP SLAVE;

as START SLAVE is async command, it might not be fully completed by the time STOP SLAVE comes, getting killed by it, and resulting in the above warning.

This is fixed in 5.6 by [1] (replacing the above with rpl_reset, which synchronously waits for START SLAVE to complete).

[1]:

commit 82320f93675824a165b2bbf2cb2cdb667a9cfa1e
Author: Sven Sandberg <email address hidden>
Date: Mon Dec 19 11:41:54 2011 +0100

    WL#3584 bugfixes
     - bugfix: reset slave should not reset Gtid_state
     - bugfix: always register binlog handler when acquiring ownership of
       GTIDs, so that ownership can be released
     - bugfix: write empty GTID to binary log when releasing ownership of
       GTIDs, if GTID is not already in binary log
     - bugfix: don't acquire ownership for empty statements
     - feature: added @@global.gtid_owned and @@session.gtid_owned
     - cleanup: don't pass gtid_state, global_sid_lock, or global_sid_map as arguments
       in functions in Group_cache, and zgroup_execution.cc
     - cleanup: pass Gtid instead of (rpl_sidno, rpl_gno) to some
       functions
     - cleanup: moved Group_cache::update_gtid_state to Gtid_state::update
     - extra assertions
     - clean up some comments
     - disable use of gtid_next_list until we need it
     - clean up some tests

[Laurynas Biveinis (laurynas-biveinis)]

https://github.com/percona/percona-server/pull/700, https://github.com/percona/percona-server/pull/701, https://github.com/percona/percona-server/pull/702

[Laurynas Biveinis (laurynas-biveinis)]

It's better to fix this in code by not printing the warning in the first place as 5.6+ does, see bug 1606572. But no need to revert the testcase change.

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-3495