Question

I read that once you drain a node you can delete the files and then restart. It works fine,
but I tried it only by draining all nodes, shutting down the the whole cluster, deleting the files and restarting.

What happens if I restart only one node at the time? As far as I understood there is a risk
that the restarted node will accept read requests and perform read repair using data from other replica.

Does anybody know the most failsafe procedure to truncate a keyspace while leaving the whole cluster up and running in order to serve other keyspaces?

Answer 1

$ bin/cassandra-cli -h localhost
[default@unknown] use keyspace1;
Authenticated to keyspace: Keyspace1
[default@Keyspace1] truncate standard1;     
standard1 truncated.

By design, this is not race-proof (that would require heavyweight locking); normally you would only atruncate a CF that isn’t serving live reads anyway. but if for some reason you must, disable read repair first (“update column family standard1 with read_repair_chance=0”).