I know there are many questions about this already, but I have another angle on this.
First of all what I've confirmed is that in Cassandra < 2.2 the COUNT(*) LIMIT N works as one would expect - it returns N if there are more results.
In the later versions, however, this behavior was changed - now the LIMIT has practically no effect as it refers to the number of rows which is always 1 when using COUNT.
Furthermore, here are the Datastax docs on the SELECT for 2.1.
Here they correctly describe the COUNT behavior.
And here are the docs for 3.0 version:
where they simply omit the use-with COUNT-LIMIT.
Here is the post about counting, where they never mention the version:
What I am concerned is the performance of this aggregation: Does it mean that in > 2.2 it gets much slower than before as it now has to go through all the rows matching the criteria?
