Multiple nodes on a single host/machine with multiple IP addresses
Attempting to install/configure multiple membase nodes on a single physical host/machine with multiple IP addresses.
I appreciate, if anyone could share the installation steps or I could attempt if you could direct me, even if you thought about it but haven't attempted yourself.
background:
Two of our hosts has ample memory; and would like to run multiple nodes and enable the replication and fail over feature.
Does anyone know why 3 nodes are required for the fail over?
With two nodes in place I could live with 1 replica copy in a best scenario and with zero replicas when the second node, whose replica's is available on the first node, goes down (I am okay with warning but still I would like the fail over to the first node be effective)
I could add the second node and re-balance the keys manually later when the seconds node comes back.
Thank you Perry. We are attempting the auto fail over with two machines regardless of loosing all the nodes on the machine, however if at least some node survive the failure (failure not related to network) then the auto fail over would kick in at the cluster level?
Could you pleas post the steps to configure multiple nodes on a single machine?
You'll have to rebuild the software from source, is that something you already have working?
Again, I would say that there is no benefit to running multiple nodes on the same host, and we would not be able to support that configuration in a production environment.
Perry
Nope. This would be brand new cluster installation, so I could give it a try if I could get the steps to build from source with multi IP capability.
Thank you for clarifying the support constraint and I do understand the cons.
This is technically possible, but not very easy at all to do.
Even still, why would you want to run multiple nodes on a single host? That really defeats the replication and failover since, if that host goes down, you would lose all nodes right?
3 nodes are not required for failover...they are required for auto-failover so that we can avoid getting into a split-brain situation and not knowing it. If you only have two servers, and the network connection between them breaks, auto-failover would activate on both of them and you'd have an inconsistent dataset == BAD. With 3 nodes, we can be sure that auto-failover won't do the wrong thing in that situation.
With only 2 nodes, you are still able to failover yourself, or create some intelligence in your monitoring system that can determine if a split-brain has happened and activate the failover programatically...but there's no way for our software to know when that has happened.
Hope that helps, Perry
Forum support is great for free but sometimes you need a guaranteed response time and dedicated resources for your questions or issues.
Consider purchasing enterprise-level support from Couchbase: http://www.couchbase.com/products-and-services/overview
Call or email "sales -at- couchbase-dot- com" today!