2015-06-24 - production
01:01	<gwicke>	rolling restart of cassandra instances to rule out a single node in funky state causing elevated p99 latency	[production]