The InnoDB log buffer size recommendation may be tested too strictly

[64kramsystem]

Currently, if there is any InnoDB log wait, the recommendation to increase the log buffer size is given.

The problem is that this is given even for cases where the the number of waits is extremely low:

[!!] InnoDB log waits: 0.00% (30 waits / 712207732 writes)

An amount like this is essentially noise under regular circumstances, so I think that the threshold should be increased, even if by a tiny amount.

The source code is:

    if ( defined $mystat{'Innodb_log_waits'}
        && $mystat{'Innodb_log_waits'} > 0 )
    {
        badprint "InnoDB log waits: "
          . percentage( $mystat{'Innodb_log_waits'},
            $mystat{'Innodb_log_writes'} )
          . "% ("
          . $mystat{'Innodb_log_waits'}
          . " waits / "
          . $mystat{'Innodb_log_writes'}
          . " writes)";
        push( @adjvars,
                "innodb_log_buffer_size (>= "
              . hr_bytes_rnd( $myvar{'innodb_log_buffer_size'} )
              . ")" );
    }

It's hard to give a threshold, but I think that even something very little as 0.000001 (1.0 / 10^6) is sensible:

    if ( defined $mystat{'Innodb_log_waits'}
        && (computed_log_waits_occurrence > 0.000001 )

[jmrenouard]

Fixed ;)