The digest is based on the results of the same query as applied on different replicas. See the following for more details: http://wiki.apache.org/cassandra/ReadRepair http://www.datastax.com/docs/1.0/dml/data_consistency
On Wed, Nov 30, 2011 at 11:38 PM, Thorsten von Eicken <t...@rightscale.com> wrote: > Looking at the docs, I can't conclusively answer this question: > > Suppose I make this CQL query with consistency factor 1 and read-repair > 100%: > select 'a'..'z' from cf where key = 'xyz' limit 5; > Suppose the node I connect to has the key and responds with (improvised > syntax): > ['a'->0, 'c'->2, 'e'->4, 'g'->6, 'i'->8] > Suppose another node has a column 'b'->1, would this be caught by the > read repair? > > The question really boils down to whether the digest query being sent is > the same as the one above, or whether it's more of the form "select a, > c, e, g, i from cf where key = xyz" and thus only checks whether the > column values are in agreement. > > Thanks! > Thorsten
