I have two Cassandra clusters deployed in k8s (lets say A and B). It is quite dynamic environment and from time to time new Cassandra node from cluster A receives an IP from old dead node which previously was in cluster B.
It happens because system.peer table is not refreshed fast enough.
So apps using python driver and cluster B reports:
From the other hand java-driver based apps reports:
It was already fixed for java driver ( ) I think it would be good to define and describe some common way of handling such cases for all drivers. Especially that there is also some other peeres validation logic which is inconsistent between drivers (RUBY-255, PYTHON-1260, etc)
Java driver notes regarding that issue: