def create_store(self, store_type):
if store_type == "redis":
redisStore = pygloo.rendezvous.RedisStore(self._ip_address,
int(self._redis_port))
redis_password = ray_constants.REDIS_DEFAULT_PASSWORD
redisStore.authorize(redis_password)
self._store = redisStore
elif store_type == "file":
store_name = get_store_name(self._group_name)
store_path = gloo_util.get_gloo_store_path(store_name)
if self._context.rank == 0:
if not os.path.exists(store_path):
os.makedirs(store_path)
elif os.listdir(store_path) and os.listdir(store_path):
shutil.rmtree(store_path)
os.makedirs(store_path)
else:
while not os.path.exists(store_path):
time.sleep(0.1)
# Note: multi-machines needs a shared NFS.
fileStore = pygloo.rendezvous.FileStore(store_path)
self._store = pygloo.rendezvous.PrefixStore(
self._group_name, fileStore)
elif store_type == "hash":
raise NotImplementedError("No implementation for hash store.")
else:
raise RuntimeError(
"Unrecognized store type: {}.".format(store_type))
def destroy_group(self):
"""Destroy the group and release GLOO communicators."""
if self._gloo_context is not None:
pygloo.barrier(self._gloo_context)
# destroy the communicator
self._gloo_context = None
if self.rank == 0 and self._rendezvous.store_type == "file":
store_name = get_store_name(self._group_name)
store_path = gloo_util.get_gloo_store_path(store_name)
if os.path.exists(store_path):
shutil.rmtree(store_path)
super(GLOOGroup, self).destroy_group()