[sheepdog] [PATCH 1/4] sheep: fix typo in help information

[Robin Dong]

2013/11/27 MORITA Kazutaka <morita.kazutaka at gmail.com>

> At Tue, 26 Nov 2013 16:38:01 +0800,
> Robin Dong wrote:
> >
> > > Getting a distributed lock is an expensive operation and it can causes
> > > a severe performance problem if we do it for each object creation.
> > > Can we find another way?  Sheepdog is not designed to allow concurrent
> > > write access.
> > >
> >
> > It will hurt performance if the object is very small, but for big object
> > (1GB,10GB,100GB), we only need to lock at "create object inode" moment,
> > after that, the object-uploading operation do not need the lock.
> >
> > I have tested this zookeeper lock, it could lock/unlock 200 times per
> > second, which I think is not too slow even for small objects.
>
> Can it scale when there are many containers who need distributed lock?
>

Hi Kazutaka,

Zookeeper is scalable enough as many hadoop clusters use zookeeper as its
lock server.
Even if there is too many containers (distributed locks) for one zookeeper
cluster,
we can create multi zookeeperk clusters and dispatch these locks to them.


>
> Is there any chance the distributed lock support corosync users?
>

Of course we can implement a distributed lock on corosync（the present
sample is dlm
https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/High_Availability_Add-On_Overview/ch-dlm.html
).

Maybe we can add "->lock" for cluster framework in sheepdog and just
implement "zookeeper ->lock" but leave "corosync ->lock" empty.
Will this be better?


> Thanks,
>
> Kazutaka
>



-- 
--
Best Regard
Robin Dong
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.wpkg.org/pipermail/sheepdog/attachments/20131127/2b95dce1/attachment-0004.html>