I’d like you all to help me on XDCR replication across 3 DataCenters.
Here’s our use-case and present architecture of Couchbase Cluster
We are running on Couchbase Community Edition - 4.5.1 across all 18 nodes distributed along 3 DC’s and enabled replication as follows:-
Option - 1:- A bi-cential cyclic approach
1 XDCR -> 2 XDCR & 3 XDCR
2 XDCR -> 1 XDCR & 3 XDCR and
3 XDCR -> 1 XDCR & 2 XDCR
Option - 2:- A cyclic approach
1 XDCR -> 2 XDCR -> 3 XDCR -> 1 XDCR
So, could anyone from here guide me, which of the above is the right approach of the replication process for better efficiency and performance of small metadata documents with heavy read/writes and iops storing in our buckets.