DCAwareRoundRobinPolicy and RackAwareRoundRobinPolicy populate does no register all the nodes in some cases

[dkropachev]

populate is using groupby in both policies:

python-driver/cassandra/policies.py

Lines 255 to 257 in 7e47772

	def populate(self, cluster, hosts):
	for dc, dc_hosts in groupby(hosts, lambda h: self._dc(h)):
	self._dc_live_hosts[dc] = tuple(set(dc_hosts))

python-driver/cassandra/policies.py

Lines 375 to 379 in 7e47772

	def populate(self, cluster, hosts):
	for (dc, rack), rack_hosts in groupby(hosts, lambda host: (self._dc(host), self._rack(host))):
	self._live_hosts[(dc, rack)] = tuple(set(rack_hosts))
	for dc, dc_hosts in groupby(hosts, lambda host: self._dc(host)):
	self._dc_live_hosts[dc] = tuple(set(dc_hosts))

groupby will produce another key every time it encounter next group, example:

for dc, dc_hosts in groupby([{'k': 1}, {'k': 1}, {'k': 2}, {'k': 2}], lambda v: v['k'] ):
    print(dc, list(dc_hosts))
# 1 [{'k': 1}, {'k': 1}]
# 2 [{'k': 2}, {'k': 2}] 

for dc, dc_hosts in groupby([{'k': 1}, {'k': 2}, {'k': 1}, {'k': 2}], lambda v: v['k'] ):
    print(dc, list(dc_hosts))
# 1 [{'k': 1}]
# 2 [{'k': 2}]
# 1 [{'k': 1}]
# 2 [{'k': 2}]

But code in both policies does not consider that and overwrite original value, as result, there is a case when policy will loos some of the hosts, which results in weird routing behavior.