maria-developers team mailing list archive

[陈福荣 <vinchen13@xxxxxxxxx>]

Hi knielsen,

The new code is here:
https://github.com/vinchen/server/commits/GCSAdmin-10.2-binlog-compressed2-2

based on
https://github.com/knielsen/server/commits/GCSAdmin-10.2-binlog-compressed2-2

And added two fixed:
1.Avoid overflowing buffers in case of corrupt events
2.Check the compressed algorithm.

2016-10-28 21:06 GMT+08:00 Simon Mudd <simon.mudd@xxxxxxxxxxx>:

> Hi,
>
> On 28 Oct 2016, at 06:18, 陈福荣 <vinchen13@xxxxxxxxx> wrote:
>
> Thanks for knielsen's nice replay. I would like to add some comments
>
>
> …
>
> >> (2) Senders would try to compress the event as requested. If the
>> >> compressed event is not any smaller then do not bother compressing it,
>> >> just send the original event.
>> >
>> > This is not in the current patch, but it could be added easily enough.
>> It
>> > might not be needed though, the log_bin_compress_min_len kind of serves
>> a
>> > similar purpose.
>>
>> It just saves the receiver doing extra work and I’d expect that allowing
>> that
>> would be trivial even if not implemented now. Just a thought anyway.
>>
>>
> As knielsen said, the main purpose of log_bin_compress_min_len is avoiding
> the too small binlog events.
>
>
> Can’t you just find out how big the compressed event is and if it’s not
> smaller than the original event send the original?
> That way you know you get some extra compression which if you’re enabling
> this feature it’s what you want anyway.
> Having to configure this minimum size means you need to set a specific
> value. Maybe you can use this
> new setting with it’s current usage and use -1 to mean only write a
> compressed event if it’s actually smaller ?
>
> …
>
> SBR is also work in this feature.
>
>
> That’s good to hear.
>
> Simon
>