I got a cluster of 6 nodes in production which is replicated with XDCR to a double node cluster.
I am running the same view with same input on the two clusters and while the replication cluster returns a correct result, the production cluster with the 6 nodes returns result with duplicate keys, which looks like rereduce didn’t happen at some stage.
Is there any limit to the rereduce level? why does it happen in one cluster and not in the other?
I see this behavior when running the queries from Java client and also from the couchbase’ web console.