def apply_deployment(deployment_info, yatai_service=None):
if yatai_service is None:
from bentoml.yatai import get_yatai_service
yatai_service = get_yatai_service()
try:
if isinstance(deployment_info, dict):
deployment_pb = deployment_dict_to_pb(deployment_info)
elif isinstance(deployment_info, str):
deployment_pb = deployment_yaml_string_to_pb(deployment_info)
else:
raise YataiDeploymentException(
'Unexpected argument type, expect deployment info to be str in yaml '
'format or a dict, instead got: {}'.format(
str(type(deployment_info))))
validation_errors = validate_deployment_pb_schema(deployment_pb)
if validation_errors:
return ApplyDeploymentResponse(status=Status.INVALID_ARGUMENT(
'Failed to validate deployment: {errors}'.format(
errors=validation_errors)))
return yatai_service.ApplyDeployment(
ApplyDeploymentRequest(deployment=deployment_pb))
except BentoMLException as error:
return ApplyDeploymentResponse(status=Status.INTERNAL(str(error)))
def apply(self, deployment_info, wait):
if isinstance(deployment_info, dict):
deployment_pb = deployment_dict_to_pb(deployment_info)
elif isinstance(deployment_info, str):
deployment_pb = deployment_yaml_string_to_pb(deployment_info)
elif isinstance(deployment_info, Deployment):
deployment_pb = deployment_info
else:
raise YataiDeploymentException(
'Unexpected argument type, expect deployment info to be str in yaml '
'format or a dict or a deployment protobuf obj, instead got: {}'
.format(str(type(deployment_info))))
validation_errors = validate_deployment_pb_schema(deployment_pb)
if validation_errors:
raise YataiDeploymentException(
f'Failed to validate deployment {deployment_pb.name}: '
f'{validation_errors}')
apply_result = self.yatai_service.ApplyDeployment(
ApplyDeploymentRequest(deployment=deployment_pb))
if apply_result.status.status_code != status_pb2.Status.OK:
error_code, error_message = status_pb_to_error_code_and_message(
apply_result.status)
raise YataiDeploymentException(f'{error_code}:{error_message}')
if wait:
self._wait_deployment_action_complete(deployment_pb.name,
deployment_pb.namespace)
return self.get(namespace=deployment_pb.namespace,
name=deployment_pb.name)
def create(self, deployment_info, wait):
if isinstance(deployment_info, dict):
deployment_pb = deployment_dict_to_pb(deployment_info)
elif isinstance(deployment_info, str):
deployment_pb = deployment_yaml_string_to_pb(deployment_info)
elif isinstance(deployment_info, Deployment):
deployment_pb = deployment_info
else:
raise YataiDeploymentException(
'Unexpected argument type, expect deployment info to be str in yaml '
'format or a dict or a deployment protobuf obj, instead got: {}'.format(
str(type(deployment_info))
)
)
validation_errors = validate_deployment_pb_schema(deployment_pb)
if validation_errors:
raise YataiDeploymentException(
f'Failed to validate deployment {deployment_pb.name}: '
f'{validation_errors}'
)
# Make sure there is no active deployment with the same deployment name
get_deployment_pb = self.yatai_service.GetDeployment(
GetDeploymentRequest(
deployment_name=deployment_pb.name, namespace=deployment_pb.namespace
)
)
if get_deployment_pb.status.status_code != status_pb2.Status.NOT_FOUND:
raise BentoMLException(
f'Deployment "{deployment_pb.name}" already existed, use Update or '
f'Apply for updating existing deployment, delete the deployment, '
f'or use a different deployment name'
)
apply_result = self.yatai_service.ApplyDeployment(
ApplyDeploymentRequest(deployment=deployment_pb)
)
if apply_result.status.status_code != status_pb2.Status.OK:
error_code, error_message = status_pb_to_error_code_and_message(
apply_result.status
)
raise YataiDeploymentException(f'{error_code}:{error_message}')
if wait:
self._wait_deployment_action_complete(
deployment_pb.name, deployment_pb.namespace
)
return self.get(namespace=deployment_pb.namespace, name=deployment_pb.name)
def apply(deployment_yaml, output, wait):
track_cli('deploy-apply', deployment_yaml.get('spec', {}).get('operator'))
try:
deployment_pb = deployment_yaml_to_pb(deployment_yaml)
yatai_service = get_yatai_service()
result = yatai_service.ApplyDeployment(
ApplyDeploymentRequest(deployment=deployment_pb)
)
if result.status.status_code != Status.OK:
_echo(
'Failed to apply deployment {name}. code: {error_code}, message: '
'{error_message}'.format(
name=deployment_pb.name,
error_code=Status.Code.Name(result.status.status_code),
error_message=result.status.error_message,
),
CLI_COLOR_ERROR,
)
else:
if wait:
result_state = get_state_after_await_action_complete(
yatai_service=yatai_service,
name=deployment_pb.name,
namespace=deployment_pb.namespace,
message='Applying deployment...',
)
result.deployment.state.CopyFrom(result_state.state)
_echo(
'Successfully applied spec to deployment {}'.format(
deployment_pb.name
),
CLI_COLOR_SUCCESS,
)
display_deployment_info(result.deployment, output)
except BentoMLException as e:
_echo(
'Failed to apply deployment {name}. Error message: {message}'.format(
name=deployment_pb.name, message=e
)
)
def apply_deployment(self, deployment_info):
if isinstance(deployment_info, dict):
deployment_pb = deployment_dict_to_pb(deployment_info)
elif isinstance(deployment_info, str):
deployment_pb = deployment_yaml_string_to_pb(deployment_info)
elif isinstance(deployment_info, Deployment):
deployment_pb = deployment_info
else:
raise YataiDeploymentException(
'Unexpected argument type, expect deployment info to be str in yaml '
'format or a dict or a deployment protobuf obj, instead got: {}'
.format(str(type(deployment_info))))
validation_errors = validate_deployment_pb_schema(deployment_pb)
if validation_errors:
return ApplyDeploymentResponse(status=Status.INVALID_ARGUMENT(
'Failed to validate deployment: {errors}'.format(
errors=validation_errors)))
return self.yatai_service.ApplyDeployment(
ApplyDeploymentRequest(deployment=deployment_pb))
def create(
name,
bento,
platform,
output,
namespace,
labels,
annotations,
region,
instance_type,
instance_count,
api_name,
kube_namespace,
replicas,
service_name,
service_type,
wait,
):
# converting platform parameter to DeploymentOperator name in proto
# e.g. 'aws-lambda' to 'AWS_LAMBDA'
platform = platform.replace('-', '_').upper()
operator = DeploymentSpec.DeploymentOperator.Value(platform)
track_cli('deploy-create', platform)
yatai_service = get_yatai_service()
# Make sure there is no active deployment with the same deployment name
get_deployment = yatai_service.GetDeployment(
GetDeploymentRequest(deployment_name=name, namespace=namespace))
if get_deployment.status.status_code != Status.NOT_FOUND:
raise BentoMLDeploymentException(
'Deployment {name} already existed, please use update or apply command'
' instead'.format(name=name))
if operator == DeploymentSpec.AWS_SAGEMAKER:
if not api_name:
raise click.BadParameter(
'api-name is required for Sagemaker deployment')
sagemaker_operator_config = DeploymentSpec.SageMakerOperatorConfig(
region=region or config().get('aws', 'default_region'),
instance_count=instance_count
or config().getint('sagemaker', 'instance_count'),
instance_type=instance_type
or config().get('sagemaker', 'instance_type'),
api_name=api_name,
)
spec = DeploymentSpec(
sagemaker_operator_config=sagemaker_operator_config)
elif operator == DeploymentSpec.AWS_LAMBDA:
aws_lambda_operator_config = DeploymentSpec.AwsLambdaOperatorConfig(
region=region or config().get('aws', 'default_region'))
if api_name:
aws_lambda_operator_config.api_name = api_name
spec = DeploymentSpec(
aws_lambda_operator_config=aws_lambda_operator_config)
elif operator == DeploymentSpec.GCP_FUNCTION:
gcp_function_operator_config = DeploymentSpec.GcpFunctionOperatorConfig(
region=region
or config().get('google-cloud', 'default_region'))
if api_name:
gcp_function_operator_config.api_name = api_name
spec = DeploymentSpec(
gcp_function_operator_config=gcp_function_operator_config)
elif operator == DeploymentSpec.KUBERNETES:
kubernetes_operator_config = DeploymentSpec.KubernetesOperatorConfig(
kube_namespace=kube_namespace,
replicas=replicas,
service_name=service_name,
service_type=service_type,
)
spec = DeploymentSpec(
kubernetes_operator_config=kubernetes_operator_config)
else:
raise BentoMLDeploymentException(
'Custom deployment is not supported in the current version of BentoML'
)
bento_name, bento_version = bento.split(':')
spec.bento_name = bento_name
spec.bento_version = bento_version
spec.operator = operator
result = yatai_service.ApplyDeployment(
ApplyDeploymentRequest(deployment=Deployment(
namespace=namespace,
name=name,
annotations=parse_key_value_pairs(annotations),
labels=parse_key_value_pairs(labels),
spec=spec,
)))
if result.status.status_code != Status.OK:
_echo(
'Failed to create deployment {name}. {error_code}: '
'{error_message}'.format(
name=name,
error_code=Status.Code.Name(result.status.status_code),
error_message=result.status.error_message,
),
CLI_COLOR_ERROR,
)
else:
if wait:
result_state = get_state_after_await_action_complete(
yatai_service=yatai_service,
name=name,
namespace=namespace,
message='Creating deployment ',
)
result.deployment.state.CopyFrom(result_state.state)
_echo('Successfully created deployment {}'.format(name),
CLI_COLOR_SUCCESS)
display_deployment_info(result.deployment, output)
def apply(
bento,
deployment_name,
platform,
output,
namespace,
labels,
annotations,
region,
stage,
instance_type,
instance_count,
api_name,
kube_namespace,
replicas,
service_name,
service_type,
):
track_cli('deploy-apply', platform)
bento_name, bento_verison = bento.split(':')
spec = DeploymentSpec(
bento_name=bento_name,
bento_verison=bento_verison,
operator=get_deployment_operator_type(platform),
)
if platform == 'aws_sagemaker':
spec.sagemaker_operator_config = DeploymentSpec.SageMakerOperatorConfig(
region=region,
instance_count=instance_count,
instance_type=instance_type,
api_name=api_name,
)
elif platform == 'aws_lambda':
spec.aws_lambda_operator_config = DeploymentSpec.AwsLambdaOperatorConfig(
region=region, stage=stage)
elif platform == 'gcp_function':
spec.gcp_function_operator_config = \
DeploymentSpec.GcpFunctionOperatorConfig(
region=region, stage=stage
)
elif platform == 'kubernetes':
spec.kubernetes_operator_config = DeploymentSpec.KubernetesOperatorConfig(
kube_namespace=kube_namespace,
replicas=replicas,
service_name=service_name,
service_type=service_type,
)
else:
raise BentoMLDeploymentException(
'Custom deployment is not supported in current version of BentoML'
)
result = get_yatai_service().ApplyDeployment(
ApplyDeploymentRequest(deployment=Deployment(
namespace=namespace,
name=deployment_name,
annotations=parse_key_value_pairs(annotations),
labels=parse_key_value_pairs(labels),
spec=spec,
)))
if result.status.status_code != Status.OK:
_echo(
'Failed to apply deployment {name}. code: {error_code}, message: '
'{error_message}'.format(
name=deployment_name,
error_code=Status.Code.Name(result.status.status_code),
error_message=result.status.error_message,
),
CLI_COLOR_ERROR,
)
else:
_echo(
'Successfully apply deployment {}'.format(deployment_name),
CLI_COLOR_SUCCESS,
)
display_deployment_info(result.deployment, output)