KB Article #177286

Data stored in Cassandra "vanishes" when multiple instances are run on the same physical host

Problem

* Data in Cassandra "vanishes" when multiple instances are run on the same physical host and cassandra.yaml has never been edited or configured.

Resolution

-- The data is not gone in this case.

The default cassandra.yaml is set such that each host uses the same ports. If multiple instances are supposed to have multiple copies of Cassandra, these ports need to be moved so as not to conflict. If no configuration is done at all, the first instance to start up will hold all of the data and the other Cassandras will never start and thus remain empty unless someone adds data to one after the data "vanishes."

This can lead to a situation where the instances are bounced and then restarted in a different order, leading to a situation where an instance with no data is started first, preventing the instance which has all of the data from starting at all. At this point, it will appear as if all of your Cassandra data has "vanished."

The solution is to either configure each instance on the host to run separately, or to disable Cassandra for instances where it is not wanted. Cassandra can be disabled on an instance by renaming the cassandra.yaml file in that instance's configuration and restarting the instance. Refer to the KPS guide for further information on configuring Cassandra.