def add(self, deployment_pb):
try:
deployment_spec = deployment_pb.spec
sagemaker_config = deployment_spec.sagemaker_operator_config
sagemaker_config.region = (sagemaker_config.region
or get_default_aws_region())
if sagemaker_config is None:
raise YataiDeploymentException(
"Sagemaker configuration is missing.")
bento_pb = self.yatai_service.GetBento(
GetBentoRequest(
bento_name=deployment_spec.bento_name,
bento_version=deployment_spec.bento_version,
))
if bento_pb.bento.uri.type not in (BentoUri.LOCAL, BentoUri.S3):
raise BentoMLException(
"BentoML currently not support {} repository".format(
BentoUri.StorageType.Name(bento_pb.bento.uri.type)))
return self._add(deployment_pb, bento_pb, bento_pb.bento.uri.uri)
except BentoMLException as error:
deployment_pb.state.state = DeploymentState.ERROR
deployment_pb.state.error_message = (
f"Error creating SageMaker deployment: {str(error)}")
return ApplyDeploymentResponse(status=error.status_proto,
deployment=deployment_pb)
def delete(self, deployment_pb):
try:
logger.debug('Deleting AWS Lambda deployment')
deployment_spec = deployment_pb.spec
lambda_deployment_config = deployment_spec.aws_lambda_operator_config
lambda_deployment_config.region = (lambda_deployment_config.region
or get_default_aws_region())
if not lambda_deployment_config.region:
raise InvalidArgument('AWS region is missing')
cf_client = boto3.client('cloudformation',
lambda_deployment_config.region)
stack_name = generate_aws_compatible_string(
deployment_pb.namespace, deployment_pb.name)
if deployment_pb.state.info_json:
deployment_info_json = json.loads(
deployment_pb.state.info_json)
bucket_name = deployment_info_json.get('s3_bucket')
if bucket_name:
cleanup_s3_bucket_if_exist(bucket_name,
lambda_deployment_config.region)
logger.debug(
'Deleting AWS CloudFormation: %s that includes Lambda function '
'and related resources',
stack_name,
)
cf_client.delete_stack(StackName=stack_name)
return DeleteDeploymentResponse(status=Status.OK())
except BentoMLException as error:
return DeleteDeploymentResponse(status=error.status_proto)
def add(self, deployment_pb):
try:
deployment_spec = deployment_pb.spec
deployment_spec.aws_lambda_operator_config.region = (
deployment_spec.aws_lambda_operator_config.region
or get_default_aws_region())
if not deployment_spec.aws_lambda_operator_config.region:
raise InvalidArgument('AWS region is missing')
bento_pb = self.yatai_service.GetBento(
GetBentoRequest(
bento_name=deployment_spec.bento_name,
bento_version=deployment_spec.bento_version,
))
if bento_pb.bento.uri.type not in (BentoUri.LOCAL, BentoUri.S3):
raise BentoMLException(
'BentoML currently not support {} repository'.format(
BentoUri.StorageType.Name(bento_pb.bento.uri.type)))
return self._add(deployment_pb, bento_pb, bento_pb.bento.uri.uri)
except BentoMLException as error:
deployment_pb.state.state = DeploymentState.ERROR
deployment_pb.state.error_message = f'Error: {str(error)}'
return ApplyDeploymentResponse(status=error.status_proto,
deployment=deployment_pb)
def update(self, deployment_pb, previous_deployment):
try:
ensure_sam_available_or_raise()
ensure_docker_available_or_raise()
deployment_spec = deployment_pb.spec
ec2_deployment_config = deployment_spec.aws_ec2_operator_config
ec2_deployment_config.region = (ec2_deployment_config.region
or get_default_aws_region())
if not ec2_deployment_config.region:
raise InvalidArgument("AWS region is missing")
bento_pb = self.yatai_service.GetBento(
GetBentoRequest(
bento_name=deployment_spec.bento_name,
bento_version=deployment_spec.bento_version,
))
if bento_pb.bento.uri.type not in (BentoUri.LOCAL, BentoUri.S3):
raise BentoMLException(
"BentoML currently not support {} repository".format(
BentoUri.StorageType.Name(bento_pb.bento.uri.type)))
return self._update(
deployment_pb,
previous_deployment,
bento_pb.bento.uri.uri,
ec2_deployment_config.region,
)
except BentoMLException as error:
deployment_pb.state.state = DeploymentState.ERROR
deployment_pb.state.error_message = f"Error: {str(error)}"
return ApplyDeploymentResponse(status=error.status_proto,
deployment=deployment_pb)
def delete(self, deployment_pb):
try:
deployment_spec = deployment_pb.spec
ec2_deployment_config = deployment_spec.aws_ec2_operator_config
ec2_deployment_config.region = (ec2_deployment_config.region
or get_default_aws_region())
if not ec2_deployment_config.region:
raise InvalidArgument("AWS region is missing")
_, deployment_stack_name, repository_name, _ = generate_ec2_resource_names(
deployment_pb.namespace, deployment_pb.name)
# delete stack
delete_cloudformation_stack(deployment_stack_name,
ec2_deployment_config.region)
# delete repo from ecr
delete_ecr_repository(repository_name,
ec2_deployment_config.region)
# remove bucket
if deployment_pb.state.info_json:
deployment_info_json = json.loads(
deployment_pb.state.info_json)
bucket_name = deployment_info_json.get('S3Bucket')
if bucket_name:
cleanup_s3_bucket_if_exist(bucket_name,
ec2_deployment_config.region)
return DeleteDeploymentResponse(status=Status.OK())
except BentoMLException as error:
return DeleteDeploymentResponse(status=error.status_proto)
def delete(self, deployment_pb):
try:
deployment_spec = deployment_pb.spec
sagemaker_config = deployment_spec.sagemaker_operator_config
sagemaker_config.region = (sagemaker_config.region
or get_default_aws_region())
if not sagemaker_config.region:
raise InvalidArgument("AWS region is missing")
delete_sagemaker_deployment_resources_if_exist(deployment_pb)
return DeleteDeploymentResponse(status=Status.OK())
except BentoMLException as error:
return DeleteDeploymentResponse(status=error.status_proto)
def test_aws_ec2_deployment(iris_clf_service):
random_hash = uuid.uuid4().hex[:6]
deployment_name = f'tests-ec2-e2e-{random_hash}'
deployment_namespace = "dev"
deployment_region = get_default_aws_region()
create_deployment_command = [
'bentoml',
'ec2',
'deploy',
deployment_name,
'-b',
iris_clf_service,
'--namespace',
deployment_namespace,
'--region',
deployment_region,
'--verbose',
]
try:
deployment_endpoint = run_aws_ec2_create_command(create_deployment_command)
instance_addresses = wait_for_healthy_targets_from_stack(
name=deployment_name,
namespace=deployment_namespace,
region=deployment_region,
)
assert deployment_endpoint, "AWS EC2 deployment creation should success"
assert instance_addresses, "AWS EC2 deployment should have all targets healthy"
iris = datasets.load_iris()
sample_data = iris.data[0:1]
results = send_test_data_to_multiple_endpoint(
[deployment_endpoint] + instance_addresses, json.dumps(sample_data.tolist())
)
for result in results:
assert result[0] == 200, "prediction request should success"
assert result[1] == '[0]', "prediction result mismatch"
finally:
delete_deployment('ec2', deployment_name, deployment_namespace)
def describe(self, deployment_pb):
try:
deployment_spec = deployment_pb.spec
sagemaker_config = deployment_spec.sagemaker_operator_config
sagemaker_config.region = (sagemaker_config.region
or get_default_aws_region())
if not sagemaker_config.region:
raise InvalidArgument("AWS region is missing")
sagemaker_client = boto3.client("sagemaker",
sagemaker_config.region)
_, _, sagemaker_endpoint_name = _get_sagemaker_resource_names(
deployment_pb)
try:
endpoint_status_response = sagemaker_client.describe_endpoint(
EndpointName=sagemaker_endpoint_name)
except ClientError as e:
raise _aws_client_error_to_bentoml_exception(
e,
f"Failed to fetch current status of sagemaker endpoint "
f"'{sagemaker_endpoint_name}'",
)
logger.debug("AWS describe endpoint response: %s",
endpoint_status_response)
endpoint_status = endpoint_status_response["EndpointStatus"]
service_state = ENDPOINT_STATUS_TO_STATE[endpoint_status]
deployment_state = DeploymentState(
state=service_state,
info_json=json.dumps(endpoint_status_response, default=str),
)
deployment_state.timestamp.GetCurrentTime()
return DescribeDeploymentResponse(state=deployment_state,
status=Status.OK())
except BentoMLException as error:
return DescribeDeploymentResponse(status=error.status_proto)
def _update(self, deployment_pb, current_deployment, bento_pb, bento_path):
if loader._is_remote_path(bento_path):
with loader._resolve_remote_bundle_path(bento_path) as local_path:
return self._update(deployment_pb, current_deployment,
bento_pb, local_path)
updated_deployment_spec = deployment_pb.spec
updated_sagemaker_config = updated_deployment_spec.sagemaker_operator_config
sagemaker_client = boto3.client(
"sagemaker", updated_sagemaker_config.region
or get_default_aws_region())
try:
raise_if_api_names_not_found_in_bento_service_metadata(
bento_pb.bento.bento_service_metadata,
[updated_sagemaker_config.api_name],
)
describe_latest_deployment_state = self.describe(deployment_pb)
current_deployment_spec = current_deployment.spec
current_sagemaker_config = current_deployment_spec.sagemaker_operator_config
latest_deployment_state = json.loads(
describe_latest_deployment_state.state.info_json)
current_ecr_image_tag = latest_deployment_state[
"ProductionVariants"][0]["DeployedImages"][0]["SpecifiedImage"]
if (updated_deployment_spec.bento_name !=
current_deployment_spec.bento_name
or updated_deployment_spec.bento_version !=
current_deployment_spec.bento_version):
logger.debug(
"BentoService tag is different from current deployment, "
"creating new docker image and push to ECR")
with TempDirectory() as temp_dir:
sagemaker_project_dir = os.path.join(
temp_dir, updated_deployment_spec.bento_name)
_init_sagemaker_project(
sagemaker_project_dir,
bento_path,
bento_pb.bento.bento_service_metadata.env.
docker_base_image,
)
ecr_image_path = create_and_push_docker_image_to_ecr(
updated_sagemaker_config.region,
updated_deployment_spec.bento_name,
updated_deployment_spec.bento_version,
sagemaker_project_dir,
)
else:
logger.debug("Using existing ECR image for Sagemaker model")
ecr_image_path = current_ecr_image_tag
(
updated_sagemaker_model_name,
updated_sagemaker_endpoint_config_name,
sagemaker_endpoint_name,
) = _get_sagemaker_resource_names(deployment_pb)
(
current_sagemaker_model_name,
current_sagemaker_endpoint_config_name,
_,
) = _get_sagemaker_resource_names(current_deployment)
if (updated_sagemaker_config.api_name !=
current_sagemaker_config.api_name
or updated_sagemaker_config.
num_of_gunicorn_workers_per_instance !=
current_sagemaker_config.
num_of_gunicorn_workers_per_instance
or ecr_image_path != current_ecr_image_tag):
logger.debug(
"Sagemaker model requires update. Delete current sagemaker model %s"
"and creating new model %s",
current_sagemaker_model_name,
updated_sagemaker_model_name,
)
_delete_sagemaker_model_if_exist(sagemaker_client,
current_sagemaker_model_name)
_create_sagemaker_model(
sagemaker_client,
updated_sagemaker_model_name,
ecr_image_path,
updated_sagemaker_config,
)
# When bento service tag is not changed, we need to delete the current
# endpoint configuration in order to create new one to avoid name collation
if (current_sagemaker_endpoint_config_name ==
updated_sagemaker_endpoint_config_name):
logger.debug(
"Current sagemaker config name %s is same as updated one, "
"delete it before create new endpoint config",
current_sagemaker_endpoint_config_name,
)
_delete_sagemaker_endpoint_config_if_exist(
sagemaker_client, current_sagemaker_endpoint_config_name)
logger.debug(
"Create new endpoint configuration %s",
updated_sagemaker_endpoint_config_name,
)
_create_sagemaker_endpoint_config(
sagemaker_client,
updated_sagemaker_model_name,
updated_sagemaker_endpoint_config_name,
updated_sagemaker_config,
)
logger.debug(
"Updating endpoint to new endpoint configuration %s",
updated_sagemaker_endpoint_config_name,
)
_update_sagemaker_endpoint(
sagemaker_client,
sagemaker_endpoint_name,
updated_sagemaker_endpoint_config_name,
)
if not (current_sagemaker_endpoint_config_name
== updated_sagemaker_endpoint_config_name):
logger.debug(
'Delete old sagemaker endpoint config %s',
current_sagemaker_endpoint_config_name,
)
_delete_sagemaker_endpoint_config_if_exist(
sagemaker_client, current_sagemaker_endpoint_config_name)
except AWSServiceError as e:
delete_sagemaker_deployment_resources_if_exist(deployment_pb)
raise e
return ApplyDeploymentResponse(status=Status.OK(),
deployment=deployment_pb)
def describe(self, deployment_pb):
try:
deployment_spec = deployment_pb.spec
lambda_deployment_config = deployment_spec.aws_lambda_operator_config
lambda_deployment_config.region = (lambda_deployment_config.region
or get_default_aws_region())
if not lambda_deployment_config.region:
raise InvalidArgument('AWS region is missing')
bento_pb = self.yatai_service.GetBento(
GetBentoRequest(
bento_name=deployment_spec.bento_name,
bento_version=deployment_spec.bento_version,
))
bento_service_metadata = bento_pb.bento.bento_service_metadata
api_names = ([lambda_deployment_config.api_name]
if lambda_deployment_config.api_name else
[api.name for api in bento_service_metadata.apis])
try:
cf_client = boto3.client('cloudformation',
lambda_deployment_config.region)
stack_name = generate_aws_compatible_string(
'{ns}-{name}'.format(ns=deployment_pb.namespace,
name=deployment_pb.name))
cloud_formation_stack_result = cf_client.describe_stacks(
StackName=stack_name)
stack_result = cloud_formation_stack_result.get('Stacks')[0]
# https://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/\
# using-cfn-describing-stacks.html
success_status = ['CREATE_COMPLETE', 'UPDATE_COMPLETE']
if stack_result['StackStatus'] in success_status:
if stack_result.get('Outputs'):
outputs = stack_result['Outputs']
else:
return DescribeDeploymentResponse(
status=Status.ABORTED(
'"Outputs" field is not present'),
state=DeploymentState(
state=DeploymentState.ERROR,
error_message='"Outputs" field is not present',
),
)
elif stack_result[
'StackStatus'] in FAILED_CLOUDFORMATION_STACK_STATUS:
state = DeploymentState(state=DeploymentState.FAILED)
state.timestamp.GetCurrentTime()
return DescribeDeploymentResponse(status=Status.OK(),
state=state)
else:
state = DeploymentState(state=DeploymentState.PENDING)
state.timestamp.GetCurrentTime()
return DescribeDeploymentResponse(status=Status.OK(),
state=state)
except Exception as error: # pylint: disable=broad-except
state = DeploymentState(state=DeploymentState.ERROR,
error_message=str(error))
state.timestamp.GetCurrentTime()
return DescribeDeploymentResponse(status=Status.INTERNAL(
str(error)),
state=state)
outputs = {o['OutputKey']: o['OutputValue'] for o in outputs}
info_json = {}
if 'EndpointUrl' in outputs:
info_json['endpoints'] = [
outputs['EndpointUrl'] + '/' + api_name
for api_name in api_names
]
if 'S3Bucket' in outputs:
info_json['s3_bucket'] = outputs['S3Bucket']
state = DeploymentState(state=DeploymentState.RUNNING,
info_json=json.dumps(info_json))
state.timestamp.GetCurrentTime()
return DescribeDeploymentResponse(status=Status.OK(), state=state)
except BentoMLException as error:
return DescribeDeploymentResponse(status=error.status_proto)
def describe(self, deployment_pb):
try:
deployment_spec = deployment_pb.spec
ec2_deployment_config = deployment_spec.aws_ec2_operator_config
ec2_deployment_config.region = (ec2_deployment_config.region
or get_default_aws_region())
if not ec2_deployment_config.region:
raise InvalidArgument("AWS region is missing")
bento_pb = self.yatai_service.GetBento(
GetBentoRequest(
bento_name=deployment_spec.bento_name,
bento_version=deployment_spec.bento_version,
))
bento_service_metadata = bento_pb.bento.bento_service_metadata
api_names = [api.name for api in bento_service_metadata.apis]
deployment_stack_name = generate_aws_compatible_string(
"btml-stack-{namespace}-{name}".format(
namespace=deployment_pb.namespace,
name=deployment_pb.name))
try:
cf_client = boto3.client("cloudformation",
ec2_deployment_config.region)
cloudformation_stack_result = cf_client.describe_stacks(
StackName=deployment_stack_name)
stack_result = cloudformation_stack_result.get("Stacks")[0]
if stack_result.get("Outputs"):
outputs = stack_result.get("Outputs")
else:
return DescribeDeploymentResponse(
status=Status.ABORTED(
'"Outputs" field is not present'),
state=DeploymentState(
state=DeploymentState.ERROR,
error_message='"Outputs" field is not present',
),
)
if stack_result[
"StackStatus"] in FAILED_CLOUDFORMATION_STACK_STATUS:
state = DeploymentState(state=DeploymentState.FAILED)
return DescribeDeploymentResponse(status=Status.OK(),
state=state)
except Exception as error: # pylint: disable=broad-except
state = DeploymentState(state=DeploymentState.ERROR,
error_message=str(error))
return DescribeDeploymentResponse(status=Status.INTERNAL(
str(error)),
state=state)
info_json = {}
outputs = {o["OutputKey"]: o["OutputValue"] for o in outputs}
if "AutoScalingGroup" in outputs:
info_json[
"InstanceDetails"] = get_instance_ip_from_scaling_group(
[outputs["AutoScalingGroup"]],
ec2_deployment_config.region)
info_json["Endpoints"] = get_endpoints_from_instance_address(
info_json["InstanceDetails"], api_names)
if "S3Bucket" in outputs:
info_json["S3Bucket"] = outputs["S3Bucket"]
if "TargetGroup" in outputs:
info_json["TargetGroup"] = outputs["TargetGroup"]
if "Url" in outputs:
info_json["Url"] = outputs["Url"]
healthy_target = get_healthy_target(outputs["TargetGroup"],
ec2_deployment_config.region)
if healthy_target:
deployment_state = DeploymentState.RUNNING
else:
deployment_state = DeploymentState.PENDING
state = DeploymentState(state=deployment_state,
info_json=json.dumps(info_json))
return DescribeDeploymentResponse(status=Status.OK(), state=state)
except BentoMLException as error:
return DescribeDeploymentResponse(status=error.status_proto)
