Buggy --slave-info code in innobackupex

[Alexey Kopytov]

innobackupex --slave-info executes the following code to get the master binlog coordinates for CHANGE MASTER TO written xtrabackup_slave_info:

    mysql_send 'SHOW SLAVE STATUS\G';

    # get output of the "show slave status" command from mysql output
    # and extract binlog position of the master server
    file_to_array($mysql_stdout, \@lines);
    for (@lines) {
        $master = $1 if /Master_Host:\s*(\S*)\s*$/;
        $filename = $1 if /Master_Log_File:\s*(\S*)\s*$/;
        $position = $1 if /Master_Log_Pos:\s*(\S*)\s*$/;
    }

That code works, but there's a couple of problems with it, making it dependent on specific SHOW SLAVE STATUS output format:

- we actually want to read Relay_Master_Log_File rather than Master_Log_File. And we do read it, but only because Relay_Master_Log_File happens to be after Master_Log_File in the SHOW SLAVE STATUS output, so the previously read Master_Log_File value (incorrect) is later overwritten by the Relay_Master_Log_File value (i.e. the correct one)

- there's no column in the SHOW SLAVE STATUS output named "Master_Log_Pos", but there are "Read_Master_Log_Pos" and "Exec_Master_Log_Pos". Again, we need the latter value for --slave-info, and we are just lucky it goes after Read_Master_Log_Pos.

The fix is to change the column names in the above code to not depend on specific order and other column naming in SHOW SLAVE STATUS.

[Alexey Kopytov (akopytov)]

This is also fixed in lp:~akopytov/percona-xtrabackup/use-dbd-mysql-in-innobackupex

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PXB-45