[Sheepdog] zookeeper vs. corosync

[Liu Yuan]

On 05/07/2012 04:56 PM, David Douard wrote:

> Hi,
> 
> I'm beginning to look at the debian package for sheepdog (which ships
> 0.2.4 only). For now, the package only support corosync. I am wondering
> if the zookeeper backend should be supported in debian.
> 
> What are the differences between zookeper and corosync for managing a
> sheepdog cluster?
> 


Both corosync and zookeeper (and Accord too) are used to manage
memebership and broadcast messages to the cluster nodes.

Corosync can only works reliably with less than 15 nodes in our testbed,
due to its implementation and design goal for small sized cluster.

If we want to add more nodes into sheepdog cluster, we should run
sheepdog against zookeeper. We (from Taobao.com) have been working with
the scalability of the sheepdog and currently running with around 1000
nodes in our test environment. For several month expediencies with
zookeeper, we have found that it works well for node number below 1000,
with object cache enabled.

For a cluster more than 1000 nodes, I think Accord would come up to our
rescue, but it is currently in a unstable development state. When the
sheepdog scales up to 1000 nodes reliably, we might go to look at Accord
and refine it to be a working state.

Thanks,
Yuan