[sheepdog] [PATCH] sheep: stop special casing local objects on gateways

[Liu Yuan]

On 06/06/2012 07:50 PM, Bastian Scholz wrote:

> Hi all,
> 
> firstly, its cool stuff you made :-)
> 
> Thanks to all participant.
> 
> Maybe it helps, if we can collect some use-cases here?
> 


Hi Bastian,

   Thanks for your feedbacks. Yes, we are highly appreciated at
constructive feedback from end users and it will definitely make a
difference on Sheepdog, towards a better product.

> Am 2012-06-06 12:59, schrieb Liu Yuan:
>> On 06/06/2012 06:54 PM, Christoph Hellwig wrote:
>>> I'd say performance numbers only start to really matter for 20,30+
>>> nodes, or does anyone disagree?
>>
>> Some users in the list claims that they use 2 nodes sheepdog cluster
>> similar to a raid storage. But I think inherently sheepdog should run
>> much more nodes. let's keep 'small' as is 5~15, which corosync can
>> manage well.
> 
> I am a user of a small setup. It includes six sheeps in two herds
> (zones, meaning two server) in a Test-Setup at the moment...
> 
> For production we use actually "cluster" some hosts for virtual
> machines and replicating their data with drbd. We are
> experimenting with sheepdog, because we need to extend these
> solution in the near future.
> 
> Sheepdog seems to be a very nice solution for this, get some
> servers, put a few disks into and start a sheep for each disk
> and let sheepdog take care of the redundancy...
> 
> If we are running out of diskspace, add some new discs or change
> to bigger ones and let sheepdog handle that, too. Similar, if we
> need more cpu-power, with sheepdog, we can easily add a new
> server which integrates seamlessly.
> 
> Such _easy_ and powerfull implementation I didnt see with other
> solutions at the moment. If we starting sheepdog in a production
> environment I think we will have three herds with overall 9 to
> 15 sheeps, which _will_ extend in the future...
> 


Well, the membership management backend such as corosync can only
reliably support less than 20 nodes. This means you can't add more nodes
into a running cluster with Corosync when number exceeds 15~20, See more
info about it, https://github.com/collie/sheepdog/wiki

> As Yuan mentioned, it is maybe not the designed purpose or
> the average use-case, but in my eyes, sheepdog could be a
> reasonable solution even for small numbers of nodes.
> 

Yes, Christoph's patch will hit the performance a lot if the cluster is
less than 10 I guess, because with smaller cluster, more data of the
virtual disk will be located in local node but we won't get a shortcut
to the local storage with the patch.

Thanks,
Yuan