Couchbase 3.0 XDCR not replication with some keys (after network failure)

I have XDCR with 2 clusters with bi-directional replication(one cluster running Couchbase 3.0, another running Couchbase 2.5). After a network failure (40% timeout connections between 2 cluters), some keys are not able to replication.
I tried to change, delete documents at both clusters but nothing happened.
I found out one key has different CAS value in each cluster.

Does anyone know any workaround? Thanks in advance.

Hi @cuongdd2, this is not supposed to happen. do you see percent complete under the stats for outbound XDCR operations on both sides?

Also you can enable XDCR Trace to log all the activities. Here is how you can do that -

either edit corresponding line in static_config to β€œ{loglevel_xdcr_trace, debug}.” and restart the server or post β€œale:set_loglevel(xdcr_trace, debug).” to /diag/eval.