def test_environment():
toolkit = Toolkit(software="spark",
version="2.2.0",
environment="miniconda")
assert toolkit.software == "spark"
assert toolkit.version == "2.2.0"
assert toolkit.environment == "miniconda"
def _merge_dict(self, config):
config = config.get('job')
if config.get('id') is not None:
self.id = config['id']
cluster_configuration = config.get('cluster_configuration')
if cluster_configuration:
self.vm_size = cluster_configuration.get('vm_size')
self.toolkit = Toolkit.from_dict(
cluster_configuration.get('toolkit'))
if cluster_configuration.get('size') is not None:
self.max_dedicated_nodes = cluster_configuration.get('size')
if cluster_configuration.get('size_low_priority') is not None:
self.max_low_pri_nodes = cluster_configuration.get(
'size_low_priority')
self.custom_scripts = cluster_configuration.get('custom_scripts')
self.subnet_id = cluster_configuration.get('subnet_id')
self.worker_on_master = cluster_configuration.get(
"worker_on_master")
scheduling_target = cluster_configuration.get("scheduling_target")
if scheduling_target:
self.scheduling_target = SchedulingTarget(scheduling_target)
applications = config.get('applications')
if applications:
self.applications = []
for application in applications:
self.applications.append(
aztk.spark.models.ApplicationConfiguration(
name=application.get('name'),
application=application.get('application'),
application_args=application.get('application_args'),
main_class=application.get('main_class'),
jars=application.get('jars'),
py_files=application.get('py_files'),
files=application.get('files'),
driver_java_options=application.get(
'driver_java_options'),
driver_library_path=application.get(
'driver_library_path'),
driver_class_path=application.get('driver_class_path'),
driver_memory=application.get('driver_memory'),
executor_memory=application.get('executor_memory'),
driver_cores=application.get('driver_cores'),
executor_cores=application.get('executor_cores')))
spark_configuration = config.get('spark_configuration')
if spark_configuration:
self.spark_defaults_conf = self.__convert_to_path(
spark_configuration.get('spark_defaults_conf'))
self.spark_env_sh = self.__convert_to_path(
spark_configuration.get('spark_env_sh'))
self.core_site_xml = self.__convert_to_path(
spark_configuration.get('core_site_xml'))
self.jars = [
self.__convert_to_path(jar)
for jar in spark_configuration.get('jars') or []
]
def cluster_config_from_dict(config: dict):
output = ClusterConfiguration()
wait = False
if config.get('id') is not None:
output.cluster_id = config['id']
if config.get('vm_size') is not None:
output.vm_size = config['vm_size']
if config.get('size'):
output.vm_count = config['size']
if config.get('size_low_pri'):
output.vm_low_pri_count = config['size_low_pri']
if config.get('subnet_id') is not None:
output.subnet_id = config['subnet_id']
if config.get('username') is not None:
output.user_configuration = UserConfiguration(
username=config['username'])
if config.get('password') is not None:
output.user_configuration.password = config['password']
if config.get('custom_scripts') not in [[None], None]:
output.custom_scripts = []
for custom_script in config['custom_scripts']:
output.custom_scripts.append(
aztk.spark.models.CustomScript(script=custom_script['script'],
run_on=custom_script['runOn']))
if config.get('azure_files') not in [[None], None]:
output.file_shares = []
for file_share in config['azure_files']:
output.file_shares.append(
aztk.spark.models.FileShare(
storage_account_name=file_share['storage_account_name'],
storage_account_key=file_share['storage_account_key'],
file_share_path=file_share['file_share_path'],
mount_path=file_share['mount_path'],
))
if config.get('toolkit') is not None:
output.toolkit = Toolkit.from_dict(config['toolkit'])
if config.get('plugins') not in [[None], None]:
output.plugins = []
for plugin in config['plugins']:
ref = PluginReference.from_dict(plugin)
output.plugins.append(ref.get_plugin())
if config.get('worker_on_master') is not None:
output.worker_on_master = config['worker_on_master']
if config.get('wait') is not None:
wait = config['wait']
return output, wait
def _merge_dict(self, config):
config = config.get("job")
if config.get("id") is not None:
self.id = config["id"]
cluster_configuration = config.get("cluster_configuration")
if cluster_configuration:
self.vm_size = cluster_configuration.get("vm_size")
self.toolkit = Toolkit.from_dict(
cluster_configuration.get("toolkit"))
if cluster_configuration.get("size") is not None:
self.max_dedicated_nodes = cluster_configuration.get("size")
if cluster_configuration.get("size_low_priority") is not None:
self.max_low_pri_nodes = cluster_configuration.get(
"size_low_priority")
self.subnet_id = cluster_configuration.get("subnet_id")
self.worker_on_master = cluster_configuration.get(
"worker_on_master")
scheduling_target = cluster_configuration.get("scheduling_target")
if scheduling_target:
self.scheduling_target = SchedulingTarget(scheduling_target)
applications = config.get("applications")
if applications:
self.applications = []
for application in applications:
self.applications.append(
aztk.spark.models.ApplicationConfiguration(
name=application.get("name"),
application=application.get("application"),
application_args=application.get("application_args"),
main_class=application.get("main_class"),
jars=application.get("jars"),
py_files=application.get("py_files"),
files=application.get("files"),
driver_java_options=application.get(
"driver_java_options"),
driver_library_path=application.get(
"driver_library_path"),
driver_class_path=application.get("driver_class_path"),
driver_memory=application.get("driver_memory"),
executor_memory=application.get("executor_memory"),
driver_cores=application.get("driver_cores"),
executor_cores=application.get("executor_cores"),
))
spark_configuration = config.get("spark_configuration")
if spark_configuration:
self.spark_defaults_conf = _convert_to_path(
spark_configuration.get("spark_defaults_conf"))
self.spark_env_sh = _convert_to_path(
spark_configuration.get("spark_env_sh"))
self.core_site_xml = _convert_to_path(
spark_configuration.get("core_site_xml"))
self.jars = [
_convert_to_path(jar)
for jar in spark_configuration.get("jars") or []
]
def execute(args: typing.NamedTuple):
if not args.toolkit_software:
return print_available_softwares()
if not validate_software(args.toolkit_software):
return None
if not args.version:
return print_available_software_version(args.toolkit_software)
if not args.environment:
print_available_environments(args.toolkit_software)
toolkit = Toolkit(software=args.toolkit_software,
version=args.version,
environment=args.environment)
toolkit.validate()
log.info("Docker image picked for this toolkit: %s",
toolkit.get_docker_repo(args.gpu))
return None
def test_scheduling_target_dedicated_with_no_dedicated_nodes_raise_error():
with pytest.raises(InvalidModelError, match="Scheduling target cannot be Dedicated if dedicated vm size is 0"):
conf = ClusterConfiguration(
cluster_id="abc",
scheduling_target=SchedulingTarget.Dedicated,
vm_size="standard_a2",
size=0,
size_low_priority=2,
toolkit=Toolkit(software="spark", version="1.6.3"),
)
conf.validate()
def test_valid_software_and_version():
Toolkit(software="spark", version="2.2.0").validate()
def test_missing_version_raise_error():
with pytest.raises(InvalidModelError):
Toolkit(software="spark", version=None).validate()
def test_basic_toolkit():
toolkit = Toolkit(software="spark", version="2.2.0")
assert toolkit.software == "spark"
assert toolkit.version == "2.2.0"
def test_get_right_docker_repo_with_env_for_gpu():
repo = Toolkit(software="spark", version="2.2.0",
environment="miniconda").get_docker_repo(True)
assert repo == "aztk/spark:v{0}-spark2.2.0-miniconda-gpu".format(
docker_image_version)
def test_get_right_docker_repo():
repo = Toolkit(software="spark", version="2.2.0").get_docker_repo(False)
assert repo == "aztk/spark:v{0}-spark2.2.0-base".format(
docker_image_version)
def test_get_right_docker_repo_for_gpu():
repo = Toolkit(software="spark", version="2.1.0").get_docker_repo(True)
assert repo == "aztk/spark:v{0}-spark2.1.0-gpu".format(
docker_image_version)
def test_invalid_environment_version_raise_error():
with pytest.raises(InvalidModelError):
Toolkit(software="spark",
version="2.2.0",
environment="miniconda",
environment_version="7.1.9").validate()
def test_invalid_version_raise_error():
with pytest.raises(InvalidModelError):
Toolkit(software="spark", version="780.0").validate()
def test_valid_software_version_and_environment():
Toolkit(software="spark", version="2.2.0",
environment="miniconda").validate()
def test_invalid_software_raise_error():
with pytest.raises(InvalidModelError):
Toolkit(software="non-supported", version="2.2.0").validate()
def test_missing_software_raise_error():
with pytest.raises(InvalidModelError):
Toolkit(software=None, version="2.2.0").validate()
