def __init__(self, app: App, id: str, **kwargs) -> None:
super().__init__(app, id, **kwargs)
self.template_options.description = "(SO0123) Improving Forecast Accuracy with Machine Learning %%VERSION%% - This solution provides a mechanism to automate Amazon Forecast predictor and forecast generation and visualize it via an Amazon SageMaker Jupyter Notebook"
# set up the template parameters
email = CfnParameter(
self,
id="Email",
type="String",
description="Email to notify with forecast results",
default="",
max_length=50,
allowed_pattern=
r"(^[a-zA-Z0-9_.+-]+@[a-zA-Z0-9-]+\.[a-zA-Z0-9-.]+$|^$)",
constraint_description="Must be a valid email address or blank",
)
lambda_log_level = CfnParameter(
self,
id="LambdaLogLevel",
type="String",
description="Change the verbosity of the logs output to CloudWatch",
default="WARNING",
allowed_values=["DEBUG", "INFO", "WARNING", "ERROR", "CRITICAL"],
)
notebook_deploy = CfnParameter(
self,
id="NotebookDeploy",
type="String",
description="Deploy an Amazon SageMaker Jupyter Notebook instance",
default="No",
allowed_values=["Yes", "No"],
)
notebook_volume_size = CfnParameter(
self,
id="NotebookVolumeSize",
type="Number",
description=
"Enter the size of the notebook instance EBS volume in GB",
default=10,
min_value=5,
max_value=16384,
constraint_description=
"Must be an integer between 5 (GB) and 16384 (16 TB)",
)
notebook_instance_type = CfnParameter(
self,
id="NotebookInstanceType",
type="String",
description="Enter the type of the notebook instance",
default="ml.t2.medium",
allowed_values=[
"ml.t2.medium",
"ml.t3.medium",
"ml.r5.large",
"ml.c5.large",
],
)
quicksight_analysis_owner = CfnParameter(
self,
id="QuickSightAnalysisOwner",
description=
"With QuickSight Enterprise enabled, provide a QuickSight ADMIN user ARN to automatically create QuickSight analyses",
default="",
allowed_pattern="(^arn:.*:quicksight:.*:.*:user.*$|^$)",
)
# set up the metadata/ cloudformation interface
template_options = TemplateOptions()
template_options.add_parameter_group(
label=
"Improving Forecast Accuracy with Machine Learning Configuration",
parameters=[email],
)
template_options.add_parameter_group(
label="Visualization Options",
parameters=[
quicksight_analysis_owner,
notebook_deploy,
notebook_instance_type,
notebook_volume_size,
],
)
template_options.add_parameter_group(label="Deployment Configuration",
parameters=[lambda_log_level])
template_options.add_parameter_label(email, "Email")
template_options.add_parameter_label(lambda_log_level,
"CloudWatch Log Level")
template_options.add_parameter_label(notebook_deploy,
"Deploy Jupyter Notebook")
template_options.add_parameter_label(notebook_volume_size,
"Jupyter Notebook volume size")
template_options.add_parameter_label(notebook_instance_type,
"Jupyter Notebook instance type")
template_options.add_parameter_label(quicksight_analysis_owner,
"Deploy QuickSight Dashboards")
self.template_options.metadata = template_options.metadata
solution_mapping = CfnMapping(
self,
"Solution",
mapping={
"Data": {
"ID": "SO0123",
"Version": "%%VERSION%%",
"SendAnonymousUsageData": "Yes",
}
},
)
source_mapping = CfnMapping(
self,
"SourceCode",
mapping={
"General": {
"S3Bucket": "%%BUCKET_NAME%%",
"KeyPrefix": "%%SOLUTION_NAME%%/%%VERSION%%",
"QuickSightSourceTemplateArn": "%%QUICKSIGHT_SOURCE%%",
}
},
)
# conditions
create_notebook = CfnCondition(
self,
"CreateNotebook",
expression=Fn.condition_equals(notebook_deploy, "Yes"),
)
email_provided = CfnCondition(
self,
"EmailProvided",
expression=Fn.condition_not(Fn.condition_equals(email, "")),
)
send_anonymous_usage_data = CfnCondition(
self,
"SendAnonymousUsageData",
expression=Fn.condition_equals(
Fn.find_in_map("Solution", "Data", "SendAnonymousUsageData"),
"Yes"),
)
create_analysis = CfnCondition(
self,
"CreateAnalysis",
expression=Fn.condition_not(
Fn.condition_equals(quicksight_analysis_owner, ""), ),
)
# Step function and state machine
fns = LambdaFunctions(self, "Functions", log_level=lambda_log_level)
# SNS
notifications = Notifications(
self,
"NotificationConfiguration",
lambda_function=fns.functions["SNS"],
email=email,
email_provided=email_provided,
)
# Custom Resources
unique_name = CfnResource(
self,
"UniqueName",
type="Custom::UniqueName",
properties={
"ServiceToken":
fns.functions["CfnResourceUniqueName"].function_arn
},
)
unique_name.override_logical_id("UniqueName")
data_bucket_name_resource = CfnResource(
self,
"DataBucketName",
type="Custom::BucketName",
properties={
"ServiceToken":
fns.functions["CfnResourceBucketName"].function_arn,
"BucketPurpose": "data-bucket",
"StackName": Aws.STACK_NAME,
"Id": unique_name.get_att("Id"),
},
)
data_bucket_name_resource.override_logical_id("DataBucketName")
# Buckets
access_logs_bucket = self.secure_bucket(
"AccessLogsBucket",
suppressions=[
CfnNagSuppression(
"W35",
"This bucket is used as the logging destination for forecast datasets and exports",
)
],
access_control=BucketAccessControl.LOG_DELIVERY_WRITE,
)
athena_bucket = self.secure_bucket(
"AthenaBucket",
server_access_logs_bucket=access_logs_bucket,
server_access_logs_prefix="athena-bucket-access-logs/",
)
data_bucket = self.secure_bucket(
"ForecastBucket",
lifecycle_rules=[
LifecycleRule(
abort_incomplete_multipart_upload_after=Duration.days(3),
enabled=True,
),
LifecycleRule(expiration=Duration.days(1),
prefix="raw/",
enabled=True),
],
bucket_name=data_bucket_name_resource.get_att("Name").to_string(),
server_access_logs_bucket=access_logs_bucket,
server_access_logs_prefix="forecast-bucket-access-logs/",
)
data_bucket.node.default_child.add_property_override(
"NotificationConfiguration",
{
"LambdaConfigurations": [{
"Function":
fns.functions["S3NotificationLambda"].function_arn,
"Event":
"s3:ObjectCreated:*",
"Filter": {
"S3Key": {
"Rules": [
{
"Name": "prefix",
"Value": "train/"
},
{
"Name": "suffix",
"Value": ".csv"
},
]
}
},
}]
},
)
# Glue and Athena
glue = Glue(self, "GlueResources", unique_name)
athena = Athena(self, "AthenaResources", athena_bucket=athena_bucket)
# Configure permissions for functions
fns.set_s3_notification_permissions(data_bucket_name_resource)
fns.set_forecast_s3_access_permissions(
name="DatasetImport",
function=fns.functions["CreateDatasetImportJob"],
data_bucket_name_resource=data_bucket_name_resource,
)
fns.set_forecast_s3_access_permissions(
name="ForecastExport",
function=fns.functions["CreateForecast"],
data_bucket_name_resource=data_bucket_name_resource,
)
fns.set_forecast_etl_permissions(
function=fns.functions["PrepareForecastExport"],
database=glue.database,
workgroup=athena.workgroup,
quicksight_principal=quicksight_analysis_owner,
quicksight_source=source_mapping,
athena_bucket=athena_bucket,
data_bucket_name_resource=data_bucket_name_resource,
)
fns.set_forecast_permissions(
"CreateDatasetGroup",
data_bucket_name_resource=data_bucket_name_resource)
fns.set_forecast_permissions(
"CreateDatasetImportJob",
data_bucket_name_resource=data_bucket_name_resource,
)
fns.set_forecast_permissions(
"CreateForecast",
data_bucket_name_resource=data_bucket_name_resource)
fns.set_forecast_permissions(
"CreatePredictor",
data_bucket_name_resource=data_bucket_name_resource)
fns.set_forecast_permissions(
"PrepareForecastExport",
data_bucket_name_resource=data_bucket_name_resource)
# notebook (conditional on 'create_notebook')
notebook = Notebook(
self,
"Notebook",
buckets=[data_bucket],
instance_type=notebook_instance_type.value_as_string,
instance_volume_size=notebook_volume_size.value_as_number,
notebook_path=Path(__file__).parent.parent.parent.joinpath(
"notebook", "samples", "notebooks"),
notebook_destination_bucket=data_bucket,
notebook_destination_prefix="notebooks",
)
Aspects.of(notebook).add(ConditionalResources(create_notebook))
# solutions metrics (conditional on 'send_anonymous_usage_data')
metrics = Metrics(
self,
"SolutionMetrics",
metrics_function=fns.functions["CfnResourceSolutionMetrics"],
metrics={
"Solution":
solution_mapping.find_in_map("Data", "ID"),
"Version":
solution_mapping.find_in_map("Data", "Version"),
"Region":
Aws.REGION,
"NotebookDeployed":
Fn.condition_if(create_notebook.node.id, "Yes", "No"),
"NotebookType":
Fn.condition_if(
create_notebook.node.id,
notebook_instance_type.value_as_string,
Aws.NO_VALUE,
),
"QuickSightDeployed":
Fn.condition_if(create_analysis.node.id, "Yes", "No"),
},
)
Aspects.of(metrics).add(
ConditionalResources(send_anonymous_usage_data))
# outputs
CfnOutput(self, "ForecastBucketName", value=data_bucket.bucket_name)
CfnOutput(self, "AthenaBucketName", value=athena_bucket.bucket_name)
CfnOutput(self,
"StepFunctionsName",
value=fns.state_machine.state_machine_name)
def __init__(self, scope: Construct, id: str, vpc: _ec2.IVpc,
codebucket: IBucket, s3_deploy, metrics) -> None:
super().__init__(scope, id)
self._metrics_mapping = CfnMapping(
self,
'AnonymousData',
mapping={'SendAnonymousData': {
'Data': 'Yes'
}})
self._metrics_condition = CfnCondition(
self,
'AnonymousDatatoAWS',
expression=Fn.condition_equals(
self._metrics_mapping.find_in_map('SendAnonymousData', 'Data'),
'Yes'))
self._helper_func = _lambda.SingletonFunction(
self,
'SolutionHelper',
uuid='75248a81-9138-468c-9ba1-bca6c7137599',
runtime=_lambda.Runtime.PYTHON_3_8,
handler='lambda_function.handler',
description=
'This function generates UUID for each deployment and sends anonymous data to the AWS Solutions team',
code=_lambda.Code.from_bucket(bucket=codebucket,
key='app_code/solution_helper.zip'),
vpc=vpc,
timeout=Duration.seconds(30))
self._helper_func.add_dependency(s3_deploy)
self._lambda_provider = _custom_resources.Provider(
self,
'LambdaProvider',
on_event_handler=self._helper_func,
vpc=vpc)
self._uuid = CustomResource(
self,
'UUIDCustomResource',
service_token=self._lambda_provider.service_token,
properties={"Resource": "UUID"},
resource_type="Custom::CreateUUID",
removal_policy=RemovalPolicy.DESTROY)
self._send_data = CustomResource(
self,
'SendDataCustomResource',
service_token=self._lambda_provider.service_token,
properties={
"Resource": "AnonymousMetric",
"UUID": self._uuid.get_att_string("UUID"),
"Solution": metrics["Solution"],
"Data": metrics
},
resource_type='Custom::AnonymousData',
removal_policy=RemovalPolicy.DESTROY)
self._send_data.node.add_dependency(self._uuid)
Aspects.of(self._helper_func).add(Condition(self._metrics_condition))
Aspects.of(self._uuid).add(Condition(self._metrics_condition))
Aspects.of(self._send_data).add(Condition(self._metrics_condition))
