Details
-
New Feature
-
Status: Open
-
Low
-
Resolution: Unresolved
-
None
Description
A simple query like `select * from table` may time out if one of the nodes fail.
We had a 4-node cassandra cluster with RF=3 and CL=LOCAL_QUORUM.
The range query is issued to only two as per ConsistencyLevel.java: liveEndpoints.subList(0, Math.min(liveEndpoints.size(), blockFor(keyspace)));
If a node fails amongst this sublist, the query will time out. Why don't you issue range queries to all replicas?