[sheepdog] [PATCH] sheep: remove master node

[Liu Yuan]

On Thu, Jul 18, 2013 at 05:56:05PM +0900, MORITA Kazutaka wrote:
> At Thu, 18 Jul 2013 16:45:40 +0800,
> Liu Yuan wrote:
> > 
> > > > I noticed zookeeper just send one event to watcher on my test box even if there
> > > > are multiple updater to one member of the queue. But I think there is problem
> > > > like above example. I think we need to check if there someone updates the join
> > > > event already in the queue inside push_join_response(), to allow only one
> > > > updater thus one update event to watcher of all nodes.
> > > 
> > > It is not easy to make sure that there is only one updater in the
> > > cluster.  I think of keeping the master of the zookeeper driver in
> > > this patch.  This patch cannot remove the master in either way, and
> > > introduces another complexity rather than simplifying the code.
> > > 
> > 
> > Why hard? We can read the zk node and check ev->type == EVENT_ACCEPT or not. no?
> 
> While pushing a join response, how to prevent another node from
> updating the same zk node?  For example,
> 
>  1. The node A reads the znode and ev->type is not EVENT_ACCEPT.
>  2. The node B reads the znode before the node A updates it with
>     EVENT_ACCEPT.
>  3. The node A calls zk_set_data() with EVENT_ACCEPT.
>  4. The node B calls zk_set_data() with EVENT_ACCEPT.

I can't testify it, but I notice following scenario:

1 A and B try to update one zk node in a very short time window
2 the sequence might be (A, B) or (B, A), but it doesn't matter.
3 zookeeper will only send one update event!

I guess if the time windoe is less than a specific time in zk, it just
notify one update event for that node.

My zk is 3.3.5.

Kai, can you verify this?

Thanks
Yuan