Hi Christian,

One of the reasons we always recommend to have a minimum of 5 nodes in the 
cluster is that the responsibility for fall-back partitions can be distributed 
fairly evenly among the remaining nodes if one goes down. With only 4 nodes in 
the cluster it is much less likely that the responsibility for fall-back 
partitions will be evenly distributed, and you may experience some nodes 
getting much more than others.

If you were to add a node to your cluster and perform the test again you should 
see a much more even distribution.

Best regards,

Christian


On 12 May 2013, at 12:48, Christian Steinmann <christian.steinman...@gmail.com> 
wrote:

> Hello,
> 
> i made a test with a 4 node cluster, where 1 node (node 4) is stopped after 
> half an hour. Half an hour later i restarted this node. I analyzed the logs 
> of all nodes after 2 hours. Node 1 and 2 had handoff data for node 4, node 3 
> not. The Bench Tool was running all the time and sending inserts to node 1,2 
> and 3. 
> 
> My questions are: Are only some nodes responsible for hinted handoff, when a 
> specific node fails? Can i find some documentation about this behavior 
> somewhere?
> 
> Greetings Christian
> _______________________________________________
> riak-users mailing list
> riak-users@lists.basho.com
> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com

_______________________________________________
riak-users mailing list
riak-users@lists.basho.com
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com

Reply via email to