nodetool flush usage

["Christopher J. Bottaro" <cj...@academicworks.com>]

Am I correct in understanding that it needs to be run on each node in the
cluster?  For example, if I have a three node cluster, I'd have to run:

nodetool -h node-1 flush
nodetool -h node-2 flush
nodetool -h node-3 flush

?

Also, does block until it's done?

My use case is recreating a keyspace.  According to this:

https://issues.apache.org/jira/browse/CASSANDRA-4857

The only safe way to recreate a keyspace is to drop -> flush -> create.
 This will be done programmatically and I need to know when the flush
operation has completed.

Thanks!