[Gluster-devel] Removal of use-compound-fops option in afr

Amar Tumballi atumball at redhat.com
Mon Mar 5 04:22:03 UTC 2018


On Mon, Mar 5, 2018 at 9:48 AM, Pranith Kumar Karampuri <pkarampu at redhat.com
> wrote:

>
>
> On Mon, Mar 5, 2018 at 9:19 AM, Amar Tumballi <atumball at redhat.com> wrote:
>
>> Pranith,
>>
>>
>>
>>>       We found that compound fops is not giving better performance in
>>> replicate and I am thinking of removing that code. Sent the patch at
>>> https://review.gluster.org/19655
>>>
>>>
>> If I understand it right, as of now AFR is the only component which uses
>> Compound FOP. If it stops using that code, should we maintain the compound
>> fop codebase at all in other places, like protocol, volgen (decompounder
>> etc?)
>>
>
> Infra was also supposed to be used by gfapi when compound fops is
> introduced. So I think it is not a bad idea to keep it until at least there
> is a decision about it. It will be similar to loading feature modules like
> quota even when quota is not used on the volume.
>

Got it, makes sense!


>
>
>>
>> Because, if AFR as a module decides compound fop is not useful, and other
>> method is better, it is completely a decision of AFR maintainers. I don't
>> see a concern there.
>>
>> Only if it had an option and if people are using it, then warning them
>> upfront about this change is better.
>>
>
> By default it is off, so I am not expecting it to be in wide use. Even if
> they are using it, I don't see any dramatic improvement in their workload
> performance based on the numbers we got. This is going to be affective in
> 4.1 release.
>
> I also added gluster-users, so that it is communicated to wider audience.
>

That's nice!


>> Regards,
>> Amar
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-devel/attachments/20180305/987ef843/attachment-0001.html>


More information about the Gluster-devel mailing list