Site-site cluster node replication

[vAJ]

I've been searching through a couple of docs listed here and elsewhere about Elastisearch clusters, but haven't really found anything that talk about nodes that might be separated by single point of failure links.

We have two primary datacenters with some additional sites coming online. We will be expanding the single node NLS we have right now with a new cluster node in our other DC. We maintain a 10GB private links between both sites, but that circuit has an infrequent/occasional drop (SLAs are odd for such a high bandwidth circuit).

Anyhow, should connectivity break between the cluster nodes, what are the mechanics for replication picking back up? Does it try to dump all of the replica shards back & forth or does it more gradually update?

Thanks,
AJ

[Box293]

I don't have an exact answer for you but this forum thread covers this topic:

http://support.nagios.com/forum/viewtop ... hilit=site

Does this help?

[vAJ]

Thanks Troy. I think this gets me most of the way there.

The burst of data replication is what I was expecting. I think I'll just have to bring up the second node and take measurements on the replication.

I assume the worst case example will be on the initial push of replica shards once the second node is added to the cluster.

Thanks
-Andrew

[jolson]

Sounds good - let us know if you have any trouble along the way.