See that last part on this page: http://wiki.apache.org/cassandra/ReadRepair
This doc is dated, but I'm pretty sure it still works this way. On Sun, Aug 25, 2013 at 3:38 PM, John Sanda <john.sa...@gmail.com> wrote: > How is conflict resolution done with range scans? I understand when > reading a single column, the latest timestamp wins. Is the timestamp for > each column compared with a range scan such that some columns in the result > could come from one replica while other columns come from another? > > > - John >
