java version "1.6.0_20"
OpenJDK Runtime Environment (IcedTea6 1.9.9) (6b20-1.9.9-0ubuntu1~10.10.2)
OpenJDK 64-Bit Server VM (build 19.0-b09, mixed mode)
Always the same node. No other nodes in this cluster, which all have the same hardware and OS, have this issue.
I don't see any resource contention. Cassandra is the only application running on this machine.
What OS? JVM version? is it always on the same node or all nodes? i had a similar problem in the past in that the OS killed Cassandra because it felt threatened and needed more resources.