I have a question around XDCR. When a document is updated and written to memory on the node that stores the active copy of the document, CB then also writes the document to replication and disk writes queues. Once flushed from the disk queue and written to disk it will then be written to the xdcr queue and eventually be written to disk in the other data centre.
Is my understanding correct that in a data centre failure scenario:
- any document updates waiting in the failed data centre disk queues will be lost
- there would be no way to identify what document updates have been lost