Clustering Across Multiple Data Centers

Hi,

thanks again for helping us on Warm nodes on spinning disks not stable, falling out of cluster during recovery. Cluster is working fine since than (8 months). For last 5 months there are three more servers. Total: 3 in one DC, and 3 in another DC, and 1 master voting-only node in third DC (stretched cluster). We read it first (over year ago) Clustering Across Multiple Data Centers | Elastic Blog and noticed that you support stretched cluster on amazon and google within same region. Our DC-s are like AZ-s in amazon region, even better, like amazon local zones (very close, single digit latency, dedicated fc links (prim and backup. Balancing), very high bandwidth, QoS ...). That was our reason for stretched cluster. We are planning to add more servers to cluster so please give us your opinion on our conclusion about choosing stretched cluster.

Even though your linked blog post is rather old with an age of six years, its main points are still valid.

It's less about "don't run across data centers", but more about "make sure the connection is absolutely low latency", so that sounds fine in your case.

In order to support higher latency use-cases, you can go with CCR (Cross Cluster Replication) and also CCS (Cross cluster search).

Hope that helps!

1 Like

Thanks Alexander. Unfortunately, I found nothing newer on the subject from elastic.co

Some thoughts from "our elastic.co guys" DavidTurner and Warkolm ?