Reputation: 984
Given Server 1, 2 and 3, can I have Couchbase instances A1, B1, C1 on Server 1; A2, B2, C2 on Server 2 and A3, B3 and C3 on Server 3, so that A1, A2, A3 forms a cluster, B1, B2 and B3 forms a cluster, etc?
This would be required for production purposes.
Thanks.
Upvotes: 1
Views: 335
Reputation: 420
For your use-case, be certain you need multiple clusters. It's not likely that you'll need separate clusters if you instead structured your data to either span multiple buckets in a single cluster (and hence with one instance per machine). Even combining data that doesn't logically group into single buckets (differentiating different document categories by some sort of key prefix) would be a far better solution. If your concern is that only certain applications need to access certain clusters (well, groups of data), they can still connect to the same cluster, albeit different buckets (or even looking for specific keys in the same bucket), and less considerations would need to make to ensure the virtual machines behaved as expected.
If you absolutely require separate clusters, it is possible, to run multiple Couchbase instances on the same hardware with the use of virtual machines. Your approach to keep nodes within a cluster on separate physical machines is perfect (as it maintains availability with replicas on separate hardware).
For some considerations with regards to running Couchbase Server on virtual machines, there is a good blog post outlining some sizing considerations as to how you would spec such a cluster. There is also a guide on how to deploy Couchbase on Windows Azure virtual machines that may also prove useful.
Upvotes: 2