def test_host_hash(self):
hash = host_hash()
salted = host_hash('salt')
empty_salted = host_hash('')
self.assertNotEqual(salted, hash)
self.assertEqual(empty_salted, hash)
def _task_fn(index, num_hosts, driver_addresses, settings):
print("Instantiating HorovodRunTaskService: " + str(settings.nics))
task = task_service.HorovodRunTaskService(index, settings.key,
settings.nics)
try:
driver = driver_service.HorovodRunDriverClient(driver_addresses,
settings.key,
settings.verbose)
driver.register_task(index, task.addresses(), host_hash.host_hash())
task.wait_for_initial_registration(settings.start_timeout)
# Tasks ping each other in a circular fashion to determine interfaces
# reachable within the cluster.
next_task_index = (index + 1) % num_hosts
next_task_addresses = driver.all_task_addresses(next_task_index)
# We request interface matching to weed out all the NAT'ed interfaces.
next_task = task_service.HorovodRunTaskClient(next_task_index,
next_task_addresses,
settings.key,
settings.verbose,
match_intf=True,
attempts=10)
driver.register_task_to_task_addresses(next_task_index,
next_task.addresses())
# Notify the next task that the address checks are completed.
next_task.task_to_task_address_check_completed()
# Wait to get a notification from previous task that its address checks
# are completed as well.
task.wait_for_task_to_task_address_check_finish_signal(
settings.start_timeout)
finally:
task.shutdown()
def host_hash(salt=None):
"""
Computes this host's host hash by invoking horovod.runner.common.util.host_hash.host_hash.
Consider environment variable CONTAINER_ID which is present when running Spark via YARN.
A YARN container does not share memory with other containers on the same host,
so it must be considered a `host` in the sense of the `host_hash`.
:param salt: extra information to include in the hash, ignores Falsy values
:return: host hash
"""
# turn salt into an array of a single string if given
salt = [str(salt)] if salt else []
# We would violate resource allocation if we run all tasks of a host in one container.
# See [issues 1497](https://github.com/horovod/horovod/issues/1497) for details.
container = os.environ.get("CONTAINER_ID")
if container is not None:
salt.append(container)
return hh.host_hash(salt='-'.join(salt))
