Since XDCR is an eventually consistent architecture, synchronous checks of documents is not a possibility.
Currently, you can use the following ways to monitor the XDCR replication and check documents in the replicated buckets.

• Prometheus metrics – there are over 60 XDCR and related metrics, including xdcr_docs_written_total, xdcr_pipeline_errors, xdcr_pipeline_status, xdcr_wtavg_docs_latency_seconds, xdcr_docs_failed_cr_target_total, xdcr_docs_failed_cr_source_total, and more to monitor the XDCR replications.

• You can use the xdcrDiffer utility (https://github.com/couchbase/xdcrDiffer) to run a report. (xdcrDiffer is planned to be included in the Morpheus Server package as a diagnostic utility for Support.)

• You can programmatically check the CAS of documents on source and target clusters. 

• You can use the Kafka connector to retrieve changes from a source bucket and confirm the changes on the target.

Just commenting here since this one comes up frequently…

At the high level, the user may think what they're asking for is fairly simple, but the semantics of replication are very different between in-cluster and across cluster by design. Also, the failure modes of WAN are are rather different.

If a user asks for this feature, it may be a good idea to approach this with an answer, e.g. "we don't have that today", and then try to dive into what they're trying to achieve. Chances are they don't want writes to stop when the remote cluster is no longer available. And there would be no compensating logic anyway for this 'observe'. It might be that they want some way to have confidence around a recovery point object (a.k.a. RPO) which may be able to be satisfied with stats. Or it might be that they want to consider how to be cache coherent across clusters… which has other techniques.

If you arrive here asking "when will this be available", please consider trying to gather more info about what the user is trying to do with this feature and open a CBSE.

Currently in feature-backlog – requirements are not clear.

Based on the discussions and reviews with field and customers, this is not a top priority anymore. Will share the details of the learning in an email. Hence, moving this to feature backlog.

can you provide a summary on this requirement from your discussions with customers?