Details
-
Improvement
-
Status: Resolved
-
Low
-
Resolution: Fixed
-
None
-
None
Description
Allow a Job Configuration to control the consistency level with which reads and writes occur for mapreduce jobs. This would be useful if we were doing statistical rollups and wanted to fail if consistency isn't met (W + R <= N as described by http://wiki.apache.org/cassandra/ArchitectureOverview). An example of a situation where this might happen would be if a set of cassandra nodes became partitioned from the other nodes, but were still accepting some writes with ConsistencyLevel 1/2 (in the case that writes were local to the partition). Then statistical rollup jobs might succeed without those accepted writes being included, and this would not be desirable.