Module aws_lambda_powertools.metrics

CloudWatch Embedded Metric Format utility

Expand source code
"""CloudWatch Embedded Metric Format utility
"""
from aws_lambda_powertools.metrics.base import MetricResolution, MetricUnit, single_metric
from aws_lambda_powertools.metrics.exceptions import (
    MetricResolutionError,
    MetricUnitError,
    MetricValueError,
    SchemaValidationError,
)
from aws_lambda_powertools.metrics.metrics import EphemeralMetrics, Metrics

__all__ = [
    "single_metric",
    "MetricUnitError",
    "MetricResolutionError",
    "SchemaValidationError",
    "MetricValueError",
    "Metrics",
    "EphemeralMetrics",
    "MetricResolution",
    "MetricUnit",
]

Sub-modules

aws_lambda_powertools.metrics.base
aws_lambda_powertools.metrics.exceptions
aws_lambda_powertools.metrics.functions
aws_lambda_powertools.metrics.metric
aws_lambda_powertools.metrics.metrics
aws_lambda_powertools.metrics.provider
aws_lambda_powertools.metrics.types

Functions

def single_metric(name: str, unit: MetricUnit, value: float, resolution: MetricResolution | int = 60, namespace: str | None = None, default_dimensions: Dict[str, str] | None = None) ‑> Generator[SingleMetric, None, None]

Context manager to simplify creation of a single metric

Example

Creates cold start metric with function_version as dimension

from aws_lambda_powertools import single_metric
from aws_lambda_powertools.metrics import MetricUnit
from aws_lambda_powertools.metrics import MetricResolution

with single_metric(name="ColdStart", unit=MetricUnit.Count, value=1, resolution=MetricResolution.Standard, namespace="ServerlessAirline") as metric:
    metric.add_dimension(name="function_version", value="47")

Same as above but set namespace using environment variable

$ export POWERTOOLS_METRICS_NAMESPACE="ServerlessAirline"

from aws_lambda_powertools import single_metric
from aws_lambda_powertools.metrics import MetricUnit
from aws_lambda_powertools.metrics import MetricResolution

with single_metric(name="ColdStart", unit=MetricUnit.Count, value=1, resolution=MetricResolution.Standard) as metric:
    metric.add_dimension(name="function_version", value="47")

Parameters

name : str
Metric name
unit : MetricUnit
aws_lambda_powertools.helper.models.MetricUnit
resolution : MetricResolution
aws_lambda_powertools.helper.models.MetricResolution
value : float
Metric value
namespace : str
Namespace for metrics

Yields

SingleMetric
SingleMetric class instance

Raises

MetricUnitError
When metric metric isn't supported by CloudWatch
MetricResolutionError
When metric resolution isn't supported by CloudWatch
MetricValueError
When metric value isn't a number
SchemaValidationError
When metric object fails EMF schema validation
Expand source code
@contextmanager
def single_metric(
    name: str,
    unit: MetricUnit,
    value: float,
    resolution: MetricResolution | int = 60,
    namespace: str | None = None,
    default_dimensions: Dict[str, str] | None = None,
) -> Generator[SingleMetric, None, None]:
    """Context manager to simplify creation of a single metric

    Example
    -------
    **Creates cold start metric with function_version as dimension**

        from aws_lambda_powertools import single_metric
        from aws_lambda_powertools.metrics import MetricUnit
        from aws_lambda_powertools.metrics import MetricResolution

        with single_metric(name="ColdStart", unit=MetricUnit.Count, value=1, resolution=MetricResolution.Standard, namespace="ServerlessAirline") as metric:
            metric.add_dimension(name="function_version", value="47")

    **Same as above but set namespace using environment variable**

        $ export POWERTOOLS_METRICS_NAMESPACE="ServerlessAirline"

        from aws_lambda_powertools import single_metric
        from aws_lambda_powertools.metrics import MetricUnit
        from aws_lambda_powertools.metrics import MetricResolution

        with single_metric(name="ColdStart", unit=MetricUnit.Count, value=1, resolution=MetricResolution.Standard) as metric:
            metric.add_dimension(name="function_version", value="47")

    Parameters
    ----------
    name : str
        Metric name
    unit : MetricUnit
        `aws_lambda_powertools.helper.models.MetricUnit`
    resolution : MetricResolution
        `aws_lambda_powertools.helper.models.MetricResolution`
    value : float
        Metric value
    namespace: str
        Namespace for metrics

    Yields
    -------
    SingleMetric
        SingleMetric class instance

    Raises
    ------
    MetricUnitError
        When metric metric isn't supported by CloudWatch
    MetricResolutionError
        When metric resolution isn't supported by CloudWatch
    MetricValueError
        When metric value isn't a number
    SchemaValidationError
        When metric object fails EMF schema validation
    """  # noqa: E501
    metric_set: Dict | None = None
    try:
        metric: SingleMetric = SingleMetric(namespace=namespace)
        metric.add_metric(name=name, unit=unit, value=value, resolution=resolution)

        if default_dimensions:
            for dim_name, dim_value in default_dimensions.items():
                metric.add_dimension(name=dim_name, value=dim_value)

        yield metric
        metric_set = metric.serialize_metric_set()
    finally:
        print(json.dumps(metric_set, separators=(",", ":")))

Classes

class EphemeralMetrics (metric_set: Dict[str, Any] | None = None, dimension_set: Dict | None = None, namespace: str | None = None, metadata_set: Dict[str, Any] | None = None, service: str | None = None, default_dimensions: Dict[str, Any] | None = None)

AmazonCloudWatchEMFProvider creates metrics asynchronously via CloudWatch Embedded Metric Format (EMF).

CloudWatch EMF can create up to 100 metrics per EMF object and metrics, dimensions, and namespace created via AmazonCloudWatchEMFProvider will adhere to the schema, will be serialized and validated against EMF Schema.

Use Metrics or single_metric() to create EMF metrics.

Environment Variables

POWERTOOLS_METRICS_NAMESPACE : str metric namespace to be set for all metrics POWERTOOLS_SERVICE_NAME : str service name used for default dimension

Raises

MetricUnitError
When metric unit isn't supported by CloudWatch
MetricResolutionError
When metric resolution isn't supported by CloudWatch
MetricValueError
When metric value isn't a number
SchemaValidationError
When metric object fails EMF schema validation
Expand source code
class AmazonCloudWatchEMFProvider(BaseProvider):
    """
    AmazonCloudWatchEMFProvider creates metrics asynchronously via CloudWatch Embedded Metric Format (EMF).

    CloudWatch EMF can create up to 100 metrics per EMF object
    and metrics, dimensions, and namespace created via AmazonCloudWatchEMFProvider
    will adhere to the schema, will be serialized and validated against EMF Schema.

    **Use `aws_lambda_powertools.Metrics` or
    `aws_lambda_powertools.single_metric` to create EMF metrics.**

    Environment variables
    ---------------------
    POWERTOOLS_METRICS_NAMESPACE : str
        metric namespace to be set for all metrics
    POWERTOOLS_SERVICE_NAME : str
        service name used for default dimension

    Raises
    ------
    MetricUnitError
        When metric unit isn't supported by CloudWatch
    MetricResolutionError
        When metric resolution isn't supported by CloudWatch
    MetricValueError
        When metric value isn't a number
    SchemaValidationError
        When metric object fails EMF schema validation
    """

    def __init__(
        self,
        metric_set: Dict[str, Any] | None = None,
        dimension_set: Dict | None = None,
        namespace: str | None = None,
        metadata_set: Dict[str, Any] | None = None,
        service: str | None = None,
        default_dimensions: Dict[str, Any] | None = None,
    ):
        self.metric_set = metric_set if metric_set is not None else {}
        self.dimension_set = dimension_set if dimension_set is not None else {}
        self.default_dimensions = default_dimensions or {}
        self.namespace = resolve_env_var_choice(choice=namespace, env=os.getenv(constants.METRICS_NAMESPACE_ENV))
        self.service = resolve_env_var_choice(choice=service, env=os.getenv(constants.SERVICE_NAME_ENV))
        self.metadata_set = metadata_set if metadata_set is not None else {}

        self._metric_units = [unit.value for unit in MetricUnit]
        self._metric_unit_valid_options = list(MetricUnit.__members__)
        self._metric_resolutions = [resolution.value for resolution in MetricResolution]

        self.dimension_set.update(**self.default_dimensions)

    def add_metric(
        self,
        name: str,
        unit: MetricUnit | str,
        value: float,
        resolution: MetricResolution | int = 60,
    ) -> None:
        """Adds given metric

        Example
        -------
        **Add given metric using MetricUnit enum**

            metric.add_metric(name="BookingConfirmation", unit=MetricUnit.Count, value=1)

        **Add given metric using plain string as value unit**

            metric.add_metric(name="BookingConfirmation", unit="Count", value=1)

        **Add given metric with MetricResolution non default value**

            metric.add_metric(name="BookingConfirmation", unit="Count", value=1, resolution=MetricResolution.High)

        Parameters
        ----------
        name : str
            Metric name
        unit : Union[MetricUnit, str]
            `aws_lambda_powertools.helper.models.MetricUnit`
        value : float
            Metric value
        resolution : Union[MetricResolution, int]
            `aws_lambda_powertools.helper.models.MetricResolution`

        Raises
        ------
        MetricUnitError
            When metric unit is not supported by CloudWatch
        MetricResolutionError
            When metric resolution is not supported by CloudWatch
        """
        if not isinstance(value, numbers.Number):
            raise MetricValueError(f"{value} is not a valid number")

        unit = extract_cloudwatch_metric_unit_value(
            metric_units=self._metric_units,
            metric_valid_options=self._metric_unit_valid_options,
            unit=unit,
        )
        resolution = extract_cloudwatch_metric_resolution_value(
            metric_resolutions=self._metric_resolutions,
            resolution=resolution,
        )
        metric: Dict = self.metric_set.get(name, defaultdict(list))
        metric["Unit"] = unit
        metric["StorageResolution"] = resolution
        metric["Value"].append(float(value))
        logger.debug(f"Adding metric: {name} with {metric}")
        self.metric_set[name] = metric

        if len(self.metric_set) == MAX_METRICS or len(metric["Value"]) == MAX_METRICS:
            logger.debug(f"Exceeded maximum of {MAX_METRICS} metrics - Publishing existing metric set")
            metrics = self.serialize_metric_set()
            print(json.dumps(metrics))

            # clear metric set only as opposed to metrics and dimensions set
            # since we could have more than 100 metrics
            self.metric_set.clear()

    def serialize_metric_set(
        self,
        metrics: Dict | None = None,
        dimensions: Dict | None = None,
        metadata: Dict | None = None,
    ) -> CloudWatchEMFOutput:
        """Serializes metric and dimensions set

        Parameters
        ----------
        metrics : Dict, optional
            Dictionary of metrics to serialize, by default None
        dimensions : Dict, optional
            Dictionary of dimensions to serialize, by default None
        metadata: Dict, optional
            Dictionary of metadata to serialize, by default None

        Example
        -------
        **Serialize metrics into EMF format**

            metrics = MetricManager()
            # ...add metrics, dimensions, namespace
            ret = metrics.serialize_metric_set()

        Returns
        -------
        Dict
            Serialized metrics following EMF specification

        Raises
        ------
        SchemaValidationError
            Raised when serialization fail schema validation
        """
        if metrics is None:  # pragma: no cover
            metrics = self.metric_set

        if dimensions is None:  # pragma: no cover
            dimensions = self.dimension_set

        if metadata is None:  # pragma: no cover
            metadata = self.metadata_set

        if self.service and not self.dimension_set.get("service"):
            # self.service won't be a float
            self.add_dimension(name="service", value=self.service)

        if len(metrics) == 0:
            raise SchemaValidationError("Must contain at least one metric.")

        if self.namespace is None:
            raise SchemaValidationError("Must contain a metric namespace.")

        logger.debug({"details": "Serializing metrics", "metrics": metrics, "dimensions": dimensions})

        # For standard resolution metrics, don't add StorageResolution field to avoid unnecessary ingestion of data into cloudwatch # noqa E501
        # Example: [ { "Name": "metric_name", "Unit": "Count"} ] # noqa ERA001
        #
        # In case using high-resolution metrics, add StorageResolution field
        # Example: [ { "Name": "metric_name", "Unit": "Count", "StorageResolution": 1 } ] # noqa ERA001
        metric_definition: List[MetricNameUnitResolution] = []
        metric_names_and_values: Dict[str, float] = {}  # { "metric_name": 1.0 }

        for metric_name in metrics:
            metric: dict = metrics[metric_name]
            metric_value: int = metric.get("Value", 0)
            metric_unit: str = metric.get("Unit", "")
            metric_resolution: int = metric.get("StorageResolution", 60)

            metric_definition_data: MetricNameUnitResolution = {"Name": metric_name, "Unit": metric_unit}

            # high-resolution metrics
            if metric_resolution == 1:
                metric_definition_data["StorageResolution"] = metric_resolution

            metric_definition.append(metric_definition_data)

            metric_names_and_values.update({metric_name: metric_value})

        return {
            "_aws": {
                "Timestamp": int(datetime.datetime.now().timestamp() * 1000),  # epoch
                "CloudWatchMetrics": [
                    {
                        "Namespace": self.namespace,  # "test_namespace"
                        "Dimensions": [list(dimensions.keys())],  # [ "service" ]
                        "Metrics": metric_definition,
                    },
                ],
            },
            # NOTE: Mypy doesn't recognize splats '** syntax' in TypedDict
            **dimensions,  # type: ignore[misc] # "service": "test_service"
            **metadata,  # "username": "test"
            **metric_names_and_values,  # "single_metric": 1.0
        }

    def add_dimension(self, name: str, value: str) -> None:
        """Adds given dimension to all metrics

        Example
        -------
        **Add a metric dimensions**

            metric.add_dimension(name="operation", value="confirm_booking")

        Parameters
        ----------
        name : str
            Dimension name
        value : str
            Dimension value
        """
        logger.debug(f"Adding dimension: {name}:{value}")
        if len(self.dimension_set) == MAX_DIMENSIONS:
            raise SchemaValidationError(
                f"Maximum number of dimensions exceeded ({MAX_DIMENSIONS}): Unable to add dimension {name}.",
            )
        # Cast value to str according to EMF spec
        # Majority of values are expected to be string already, so
        # checking before casting improves performance in most cases
        self.dimension_set[name] = value if isinstance(value, str) else str(value)

    def add_metadata(self, key: str, value: Any) -> None:
        """Adds high cardinal metadata for metrics object

        This will not be available during metrics visualization.
        Instead, this will be searchable through logs.

        If you're looking to add metadata to filter metrics, then
        use add_dimensions method.

        Example
        -------
        **Add metrics metadata**

            metric.add_metadata(key="booking_id", value="booking_id")

        Parameters
        ----------
        key : str
            Metadata key
        value : any
            Metadata value
        """
        logger.debug(f"Adding metadata: {key}:{value}")

        # Cast key to str according to EMF spec
        # Majority of keys are expected to be string already, so
        # checking before casting improves performance in most cases
        if isinstance(key, str):
            self.metadata_set[key] = value
        else:
            self.metadata_set[str(key)] = value

    def clear_metrics(self) -> None:
        logger.debug("Clearing out existing metric set from memory")
        self.metric_set.clear()
        self.dimension_set.clear()
        self.metadata_set.clear()
        self.set_default_dimensions(**self.default_dimensions)

    def flush_metrics(self, raise_on_empty_metrics: bool = False) -> None:
        """Manually flushes the metrics. This is normally not necessary,
        unless you're running on other runtimes besides Lambda, where the @log_metrics
        decorator already handles things for you.

        Parameters
        ----------
        raise_on_empty_metrics : bool, optional
            raise exception if no metrics are emitted, by default False
        """
        if not raise_on_empty_metrics and not self.metric_set:
            warnings.warn(
                "No application metrics to publish. The cold-start metric may be published if enabled. "
                "If application metrics should never be empty, consider using 'raise_on_empty_metrics'",
                stacklevel=2,
            )
        else:
            logger.debug("Flushing existing metrics")
            metrics = self.serialize_metric_set()
            print(json.dumps(metrics, separators=(",", ":")))
            self.clear_metrics()

    def log_metrics(
        self,
        lambda_handler: Callable[[Dict, Any], Any] | Optional[Callable[[Dict, Any, Optional[Dict]], Any]] = None,
        capture_cold_start_metric: bool = False,
        raise_on_empty_metrics: bool = False,
        **kwargs,
    ):
        """Decorator to serialize and publish metrics at the end of a function execution.

        Be aware that the log_metrics **does call* the decorated function (e.g. lambda_handler).

        Example
        -------
        **Lambda function using tracer and metrics decorators**

            from aws_lambda_powertools import Metrics, Tracer

            metrics = Metrics(service="payment")
            tracer = Tracer(service="payment")

            @tracer.capture_lambda_handler
            @metrics.log_metrics
            def handler(event, context):
                    ...

        Parameters
        ----------
        lambda_handler : Callable[[Any, Any], Any], optional
            lambda function handler, by default None
        capture_cold_start_metric : bool, optional
            captures cold start metric, by default False
        raise_on_empty_metrics : bool, optional
            raise exception if no metrics are emitted, by default False
        **kwargs

        Raises
        ------
        e
            Propagate error received
        """

        default_dimensions = kwargs.get("default_dimensions")

        if default_dimensions:
            self.set_default_dimensions(**default_dimensions)

        return super().log_metrics(
            lambda_handler=lambda_handler,
            capture_cold_start_metric=capture_cold_start_metric,
            raise_on_empty_metrics=raise_on_empty_metrics,
            **kwargs,
        )

    def add_cold_start_metric(self, context: LambdaContext) -> None:
        """Add cold start metric and function_name dimension

        Parameters
        ----------
        context : Any
            Lambda context
        """
        logger.debug("Adding cold start metric and function_name dimension")
        with single_metric(name="ColdStart", unit=MetricUnit.Count, value=1, namespace=self.namespace) as metric:
            metric.add_dimension(name="function_name", value=context.function_name)
            if self.service:
                metric.add_dimension(name="service", value=str(self.service))

    def set_default_dimensions(self, **dimensions) -> None:
        """Persist dimensions across Lambda invocations

        Parameters
        ----------
        dimensions : Dict[str, Any], optional
            metric dimensions as key=value

        Example
        -------
        **Sets some default dimensions that will always be present across metrics and invocations**

            from aws_lambda_powertools import Metrics

            metrics = Metrics(namespace="ServerlessAirline", service="payment")
            metrics.set_default_dimensions(environment="demo", another="one")

            @metrics.log_metrics()
            def lambda_handler():
                return True
        """
        for name, value in dimensions.items():
            self.add_dimension(name, value)

        self.default_dimensions.update(**dimensions)

Ancestors

Methods

def add_cold_start_metric(self, context: LambdaContext) ‑> None

Add cold start metric and function_name dimension

Parameters

context : Any
Lambda context
Expand source code
def add_cold_start_metric(self, context: LambdaContext) -> None:
    """Add cold start metric and function_name dimension

    Parameters
    ----------
    context : Any
        Lambda context
    """
    logger.debug("Adding cold start metric and function_name dimension")
    with single_metric(name="ColdStart", unit=MetricUnit.Count, value=1, namespace=self.namespace) as metric:
        metric.add_dimension(name="function_name", value=context.function_name)
        if self.service:
            metric.add_dimension(name="service", value=str(self.service))
def add_dimension(self, name: str, value: str) ‑> None

Adds given dimension to all metrics

Example

Add a metric dimensions

metric.add_dimension(name="operation", value="confirm_booking")

Parameters

name : str
Dimension name
value : str
Dimension value
Expand source code
def add_dimension(self, name: str, value: str) -> None:
    """Adds given dimension to all metrics

    Example
    -------
    **Add a metric dimensions**

        metric.add_dimension(name="operation", value="confirm_booking")

    Parameters
    ----------
    name : str
        Dimension name
    value : str
        Dimension value
    """
    logger.debug(f"Adding dimension: {name}:{value}")
    if len(self.dimension_set) == MAX_DIMENSIONS:
        raise SchemaValidationError(
            f"Maximum number of dimensions exceeded ({MAX_DIMENSIONS}): Unable to add dimension {name}.",
        )
    # Cast value to str according to EMF spec
    # Majority of values are expected to be string already, so
    # checking before casting improves performance in most cases
    self.dimension_set[name] = value if isinstance(value, str) else str(value)
def add_metadata(self, key: str, value: Any) ‑> None

Adds high cardinal metadata for metrics object

This will not be available during metrics visualization. Instead, this will be searchable through logs.

If you're looking to add metadata to filter metrics, then use add_dimensions method.

Example

Add metrics metadata

metric.add_metadata(key="booking_id", value="booking_id")

Parameters

key : str
Metadata key
value : any
Metadata value
Expand source code
def add_metadata(self, key: str, value: Any) -> None:
    """Adds high cardinal metadata for metrics object

    This will not be available during metrics visualization.
    Instead, this will be searchable through logs.

    If you're looking to add metadata to filter metrics, then
    use add_dimensions method.

    Example
    -------
    **Add metrics metadata**

        metric.add_metadata(key="booking_id", value="booking_id")

    Parameters
    ----------
    key : str
        Metadata key
    value : any
        Metadata value
    """
    logger.debug(f"Adding metadata: {key}:{value}")

    # Cast key to str according to EMF spec
    # Majority of keys are expected to be string already, so
    # checking before casting improves performance in most cases
    if isinstance(key, str):
        self.metadata_set[key] = value
    else:
        self.metadata_set[str(key)] = value
def add_metric(self, name: str, unit: MetricUnit | str, value: float, resolution: MetricResolution | int = 60) ‑> None

Adds given metric

Example

Add given metric using MetricUnit enum

metric.add_metric(name="BookingConfirmation", unit=MetricUnit.Count, value=1)

Add given metric using plain string as value unit

metric.add_metric(name="BookingConfirmation", unit="Count", value=1)

Add given metric with MetricResolution non default value

metric.add_metric(name="BookingConfirmation", unit="Count", value=1, resolution=MetricResolution.High)

Parameters

name : str
Metric name
unit : Union[MetricUnit, str]
aws_lambda_powertools.helper.models.MetricUnit
value : float
Metric value
resolution : Union[MetricResolution, int]
aws_lambda_powertools.helper.models.MetricResolution

Raises

MetricUnitError
When metric unit is not supported by CloudWatch
MetricResolutionError
When metric resolution is not supported by CloudWatch
Expand source code
def add_metric(
    self,
    name: str,
    unit: MetricUnit | str,
    value: float,
    resolution: MetricResolution | int = 60,
) -> None:
    """Adds given metric

    Example
    -------
    **Add given metric using MetricUnit enum**

        metric.add_metric(name="BookingConfirmation", unit=MetricUnit.Count, value=1)

    **Add given metric using plain string as value unit**

        metric.add_metric(name="BookingConfirmation", unit="Count", value=1)

    **Add given metric with MetricResolution non default value**

        metric.add_metric(name="BookingConfirmation", unit="Count", value=1, resolution=MetricResolution.High)

    Parameters
    ----------
    name : str
        Metric name
    unit : Union[MetricUnit, str]
        `aws_lambda_powertools.helper.models.MetricUnit`
    value : float
        Metric value
    resolution : Union[MetricResolution, int]
        `aws_lambda_powertools.helper.models.MetricResolution`

    Raises
    ------
    MetricUnitError
        When metric unit is not supported by CloudWatch
    MetricResolutionError
        When metric resolution is not supported by CloudWatch
    """
    if not isinstance(value, numbers.Number):
        raise MetricValueError(f"{value} is not a valid number")

    unit = extract_cloudwatch_metric_unit_value(
        metric_units=self._metric_units,
        metric_valid_options=self._metric_unit_valid_options,
        unit=unit,
    )
    resolution = extract_cloudwatch_metric_resolution_value(
        metric_resolutions=self._metric_resolutions,
        resolution=resolution,
    )
    metric: Dict = self.metric_set.get(name, defaultdict(list))
    metric["Unit"] = unit
    metric["StorageResolution"] = resolution
    metric["Value"].append(float(value))
    logger.debug(f"Adding metric: {name} with {metric}")
    self.metric_set[name] = metric

    if len(self.metric_set) == MAX_METRICS or len(metric["Value"]) == MAX_METRICS:
        logger.debug(f"Exceeded maximum of {MAX_METRICS} metrics - Publishing existing metric set")
        metrics = self.serialize_metric_set()
        print(json.dumps(metrics))

        # clear metric set only as opposed to metrics and dimensions set
        # since we could have more than 100 metrics
        self.metric_set.clear()
def flush_metrics(self, raise_on_empty_metrics: bool = False) ‑> None

Manually flushes the metrics. This is normally not necessary, unless you're running on other runtimes besides Lambda, where the @log_metrics decorator already handles things for you.

Parameters

raise_on_empty_metrics : bool, optional
raise exception if no metrics are emitted, by default False
Expand source code
def flush_metrics(self, raise_on_empty_metrics: bool = False) -> None:
    """Manually flushes the metrics. This is normally not necessary,
    unless you're running on other runtimes besides Lambda, where the @log_metrics
    decorator already handles things for you.

    Parameters
    ----------
    raise_on_empty_metrics : bool, optional
        raise exception if no metrics are emitted, by default False
    """
    if not raise_on_empty_metrics and not self.metric_set:
        warnings.warn(
            "No application metrics to publish. The cold-start metric may be published if enabled. "
            "If application metrics should never be empty, consider using 'raise_on_empty_metrics'",
            stacklevel=2,
        )
    else:
        logger.debug("Flushing existing metrics")
        metrics = self.serialize_metric_set()
        print(json.dumps(metrics, separators=(",", ":")))
        self.clear_metrics()
def log_metrics(self, lambda_handler: Callable[[Dict, Any], Any] | Optional[Callable[[Dict, Any, Optional[Dict]], Any]] = None, capture_cold_start_metric: bool = False, raise_on_empty_metrics: bool = False, **kwargs)

Decorator to serialize and publish metrics at the end of a function execution.

Be aware that the log_metrics *does call the decorated function (e.g. lambda_handler).

Example

Lambda function using tracer and metrics decorators

from aws_lambda_powertools import Metrics, Tracer

metrics = Metrics(service="payment")
tracer = Tracer(service="payment")

@tracer.capture_lambda_handler
@metrics.log_metrics
def handler(event, context):
        ...

Parameters

lambda_handler : Callable[[Any, Any], Any], optional
lambda function handler, by default None
capture_cold_start_metric : bool, optional
captures cold start metric, by default False
raise_on_empty_metrics : bool, optional
raise exception if no metrics are emitted, by default False
**kwargs
 

Raises

e
Propagate error received
Expand source code
def log_metrics(
    self,
    lambda_handler: Callable[[Dict, Any], Any] | Optional[Callable[[Dict, Any, Optional[Dict]], Any]] = None,
    capture_cold_start_metric: bool = False,
    raise_on_empty_metrics: bool = False,
    **kwargs,
):
    """Decorator to serialize and publish metrics at the end of a function execution.

    Be aware that the log_metrics **does call* the decorated function (e.g. lambda_handler).

    Example
    -------
    **Lambda function using tracer and metrics decorators**

        from aws_lambda_powertools import Metrics, Tracer

        metrics = Metrics(service="payment")
        tracer = Tracer(service="payment")

        @tracer.capture_lambda_handler
        @metrics.log_metrics
        def handler(event, context):
                ...

    Parameters
    ----------
    lambda_handler : Callable[[Any, Any], Any], optional
        lambda function handler, by default None
    capture_cold_start_metric : bool, optional
        captures cold start metric, by default False
    raise_on_empty_metrics : bool, optional
        raise exception if no metrics are emitted, by default False
    **kwargs

    Raises
    ------
    e
        Propagate error received
    """

    default_dimensions = kwargs.get("default_dimensions")

    if default_dimensions:
        self.set_default_dimensions(**default_dimensions)

    return super().log_metrics(
        lambda_handler=lambda_handler,
        capture_cold_start_metric=capture_cold_start_metric,
        raise_on_empty_metrics=raise_on_empty_metrics,
        **kwargs,
    )
def serialize_metric_set(self, metrics: Dict | None = None, dimensions: Dict | None = None, metadata: Dict | None = None) ‑> CloudWatchEMFOutput

Serializes metric and dimensions set

Parameters

metrics : Dict, optional
Dictionary of metrics to serialize, by default None
dimensions : Dict, optional
Dictionary of dimensions to serialize, by default None
metadata : Dict, optional
Dictionary of metadata to serialize, by default None

Example

Serialize metrics into EMF format

metrics = MetricManager()
# ...add metrics, dimensions, namespace
ret = metrics.serialize_metric_set()

Returns

Dict
Serialized metrics following EMF specification

Raises

SchemaValidationError
Raised when serialization fail schema validation
Expand source code
def serialize_metric_set(
    self,
    metrics: Dict | None = None,
    dimensions: Dict | None = None,
    metadata: Dict | None = None,
) -> CloudWatchEMFOutput:
    """Serializes metric and dimensions set

    Parameters
    ----------
    metrics : Dict, optional
        Dictionary of metrics to serialize, by default None
    dimensions : Dict, optional
        Dictionary of dimensions to serialize, by default None
    metadata: Dict, optional
        Dictionary of metadata to serialize, by default None

    Example
    -------
    **Serialize metrics into EMF format**

        metrics = MetricManager()
        # ...add metrics, dimensions, namespace
        ret = metrics.serialize_metric_set()

    Returns
    -------
    Dict
        Serialized metrics following EMF specification

    Raises
    ------
    SchemaValidationError
        Raised when serialization fail schema validation
    """
    if metrics is None:  # pragma: no cover
        metrics = self.metric_set

    if dimensions is None:  # pragma: no cover
        dimensions = self.dimension_set

    if metadata is None:  # pragma: no cover
        metadata = self.metadata_set

    if self.service and not self.dimension_set.get("service"):
        # self.service won't be a float
        self.add_dimension(name="service", value=self.service)

    if len(metrics) == 0:
        raise SchemaValidationError("Must contain at least one metric.")

    if self.namespace is None:
        raise SchemaValidationError("Must contain a metric namespace.")

    logger.debug({"details": "Serializing metrics", "metrics": metrics, "dimensions": dimensions})

    # For standard resolution metrics, don't add StorageResolution field to avoid unnecessary ingestion of data into cloudwatch # noqa E501
    # Example: [ { "Name": "metric_name", "Unit": "Count"} ] # noqa ERA001
    #
    # In case using high-resolution metrics, add StorageResolution field
    # Example: [ { "Name": "metric_name", "Unit": "Count", "StorageResolution": 1 } ] # noqa ERA001
    metric_definition: List[MetricNameUnitResolution] = []
    metric_names_and_values: Dict[str, float] = {}  # { "metric_name": 1.0 }

    for metric_name in metrics:
        metric: dict = metrics[metric_name]
        metric_value: int = metric.get("Value", 0)
        metric_unit: str = metric.get("Unit", "")
        metric_resolution: int = metric.get("StorageResolution", 60)

        metric_definition_data: MetricNameUnitResolution = {"Name": metric_name, "Unit": metric_unit}

        # high-resolution metrics
        if metric_resolution == 1:
            metric_definition_data["StorageResolution"] = metric_resolution

        metric_definition.append(metric_definition_data)

        metric_names_and_values.update({metric_name: metric_value})

    return {
        "_aws": {
            "Timestamp": int(datetime.datetime.now().timestamp() * 1000),  # epoch
            "CloudWatchMetrics": [
                {
                    "Namespace": self.namespace,  # "test_namespace"
                    "Dimensions": [list(dimensions.keys())],  # [ "service" ]
                    "Metrics": metric_definition,
                },
            ],
        },
        # NOTE: Mypy doesn't recognize splats '** syntax' in TypedDict
        **dimensions,  # type: ignore[misc] # "service": "test_service"
        **metadata,  # "username": "test"
        **metric_names_and_values,  # "single_metric": 1.0
    }
def set_default_dimensions(self, **dimensions) ‑> None

Persist dimensions across Lambda invocations

Parameters

dimensions : Dict[str, Any], optional
metric dimensions as key=value

Example

Sets some default dimensions that will always be present across metrics and invocations

from aws_lambda_powertools import Metrics

metrics = Metrics(namespace="ServerlessAirline", service="payment")
metrics.set_default_dimensions(environment="demo", another="one")

@metrics.log_metrics()
def lambda_handler():
    return True
Expand source code
def set_default_dimensions(self, **dimensions) -> None:
    """Persist dimensions across Lambda invocations

    Parameters
    ----------
    dimensions : Dict[str, Any], optional
        metric dimensions as key=value

    Example
    -------
    **Sets some default dimensions that will always be present across metrics and invocations**

        from aws_lambda_powertools import Metrics

        metrics = Metrics(namespace="ServerlessAirline", service="payment")
        metrics.set_default_dimensions(environment="demo", another="one")

        @metrics.log_metrics()
        def lambda_handler():
            return True
    """
    for name, value in dimensions.items():
        self.add_dimension(name, value)

    self.default_dimensions.update(**dimensions)

Inherited members

class MetricResolution (*args, **kwds)

Create a collection of name/value pairs.

Example enumeration:

>>> class Color(Enum):
...     RED = 1
...     BLUE = 2
...     GREEN = 3

Access them by:

  • attribute access::
>>> Color.RED
<Color.RED: 1>
  • value lookup:
>>> Color(1)
<Color.RED: 1>
  • name lookup:
>>> Color['RED']
<Color.RED: 1>

Enumerations can be iterated over, and know how many members they have:

>>> len(Color)
3
>>> list(Color)
[<Color.RED: 1>, <Color.BLUE: 2>, <Color.GREEN: 3>]

Methods can be added to enumerations, and members can have their own attributes – see the documentation for details.

Expand source code
class MetricResolution(Enum):
    Standard = 60
    High = 1

Ancestors

  • enum.Enum

Class variables

var High
var Standard
class MetricResolutionError (*args, **kwargs)

When metric resolution is not supported by CloudWatch

Expand source code
class MetricResolutionError(Exception):
    """When metric resolution is not supported by CloudWatch"""

    pass

Ancestors

  • builtins.Exception
  • builtins.BaseException
class MetricUnit (*args, **kwds)

Create a collection of name/value pairs.

Example enumeration:

>>> class Color(Enum):
...     RED = 1
...     BLUE = 2
...     GREEN = 3

Access them by:

  • attribute access::
>>> Color.RED
<Color.RED: 1>
  • value lookup:
>>> Color(1)
<Color.RED: 1>
  • name lookup:
>>> Color['RED']
<Color.RED: 1>

Enumerations can be iterated over, and know how many members they have:

>>> len(Color)
3
>>> list(Color)
[<Color.RED: 1>, <Color.BLUE: 2>, <Color.GREEN: 3>]

Methods can be added to enumerations, and members can have their own attributes – see the documentation for details.

Expand source code
class MetricUnit(Enum):
    Seconds = "Seconds"
    Microseconds = "Microseconds"
    Milliseconds = "Milliseconds"
    Bytes = "Bytes"
    Kilobytes = "Kilobytes"
    Megabytes = "Megabytes"
    Gigabytes = "Gigabytes"
    Terabytes = "Terabytes"
    Bits = "Bits"
    Kilobits = "Kilobits"
    Megabits = "Megabits"
    Gigabits = "Gigabits"
    Terabits = "Terabits"
    Percent = "Percent"
    Count = "Count"
    BytesPerSecond = "Bytes/Second"
    KilobytesPerSecond = "Kilobytes/Second"
    MegabytesPerSecond = "Megabytes/Second"
    GigabytesPerSecond = "Gigabytes/Second"
    TerabytesPerSecond = "Terabytes/Second"
    BitsPerSecond = "Bits/Second"
    KilobitsPerSecond = "Kilobits/Second"
    MegabitsPerSecond = "Megabits/Second"
    GigabitsPerSecond = "Gigabits/Second"
    TerabitsPerSecond = "Terabits/Second"
    CountPerSecond = "Count/Second"

Ancestors

  • enum.Enum

Class variables

var Bits
var BitsPerSecond
var Bytes
var BytesPerSecond
var Count
var CountPerSecond
var Gigabits
var GigabitsPerSecond
var Gigabytes
var GigabytesPerSecond
var Kilobits
var KilobitsPerSecond
var Kilobytes
var KilobytesPerSecond
var Megabits
var MegabitsPerSecond
var Megabytes
var MegabytesPerSecond
var Microseconds
var Milliseconds
var Percent
var Seconds
var Terabits
var TerabitsPerSecond
var Terabytes
var TerabytesPerSecond
class MetricUnitError (*args, **kwargs)

When metric unit is not supported by CloudWatch

Expand source code
class MetricUnitError(Exception):
    """When metric unit is not supported by CloudWatch"""

    pass

Ancestors

  • builtins.Exception
  • builtins.BaseException
class MetricValueError (*args, **kwargs)

When metric value isn't a valid number

Expand source code
class MetricValueError(Exception):
    """When metric value isn't a valid number"""

    pass

Ancestors

  • builtins.Exception
  • builtins.BaseException
class Metrics (service: str | None = None, namespace: str | None = None, provider: AmazonCloudWatchEMFProvider | None = None)

Metrics create an CloudWatch EMF object with up to 100 metrics

Use Metrics when you need to create multiple metrics that have dimensions in common (e.g. service_name="payment").

Metrics up to 100 metrics in memory and are shared across all its instances. That means it can be safely instantiated outside of a Lambda function, or anywhere else.

A decorator (log_metrics) is provided so metrics are published at the end of its execution. If more than 100 metrics are added at a given function execution, these metrics are serialized and published before adding a given metric to prevent metric truncation.

Example

Creates a few metrics and publish at the end of a function execution

from aws_lambda_powertools import Metrics

metrics = Metrics(namespace="ServerlessAirline", service="payment")

@metrics.log_metrics(capture_cold_start_metric=True)
def lambda_handler():
    metrics.add_metric(name="BookingConfirmation", unit="Count", value=1)
    metrics.add_dimension(name="function_version", value="$LATEST")

    return True

Environment Variables

POWERTOOLS_METRICS_NAMESPACE : str metric namespace POWERTOOLS_SERVICE_NAME : str service name used for default dimension

Parameters

service : str, optional
service name to be used as metric dimension, by default "service_undefined"
namespace : str, optional
Namespace for metrics
provider : AmazonCloudWatchEMFProvider, optional
Pre-configured AmazonCloudWatchEMFProvider provider

Raises

MetricUnitError
When metric unit isn't supported by CloudWatch
MetricResolutionError
When metric resolution isn't supported by CloudWatch
MetricValueError
When metric value isn't a number
SchemaValidationError
When metric object fails EMF schema validation
Expand source code
class Metrics:
    """Metrics create an CloudWatch EMF object with up to 100 metrics

    Use Metrics when you need to create multiple metrics that have
    dimensions in common (e.g. service_name="payment").

    Metrics up to 100 metrics in memory and are shared across
    all its instances. That means it can be safely instantiated outside
    of a Lambda function, or anywhere else.

    A decorator (log_metrics) is provided so metrics are published at the end of its execution.
    If more than 100 metrics are added at a given function execution,
    these metrics are serialized and published before adding a given metric
    to prevent metric truncation.

    Example
    -------
    **Creates a few metrics and publish at the end of a function execution**

        from aws_lambda_powertools import Metrics

        metrics = Metrics(namespace="ServerlessAirline", service="payment")

        @metrics.log_metrics(capture_cold_start_metric=True)
        def lambda_handler():
            metrics.add_metric(name="BookingConfirmation", unit="Count", value=1)
            metrics.add_dimension(name="function_version", value="$LATEST")

            return True

    Environment variables
    ---------------------
    POWERTOOLS_METRICS_NAMESPACE : str
        metric namespace
    POWERTOOLS_SERVICE_NAME : str
        service name used for default dimension

    Parameters
    ----------
    service : str, optional
        service name to be used as metric dimension, by default "service_undefined"
    namespace : str, optional
        Namespace for metrics
    provider: AmazonCloudWatchEMFProvider, optional
        Pre-configured AmazonCloudWatchEMFProvider provider

    Raises
    ------
    MetricUnitError
        When metric unit isn't supported by CloudWatch
    MetricResolutionError
        When metric resolution isn't supported by CloudWatch
    MetricValueError
        When metric value isn't a number
    SchemaValidationError
        When metric object fails EMF schema validation
    """

    # NOTE: We use class attrs to share metrics data across instances
    # this allows customers to initialize Metrics() throughout their code base (and middlewares)
    # and not get caught by accident with metrics data loss, or data deduplication
    # e.g., m1 and m2 add metric ProductCreated, however m1 has 'version' dimension  but m2 doesn't
    # Result: ProductCreated is created twice as we now have 2 different EMF blobs
    _metrics: Dict[str, Any] = {}
    _dimensions: Dict[str, str] = {}
    _metadata: Dict[str, Any] = {}
    _default_dimensions: Dict[str, Any] = {}

    def __init__(
        self,
        service: str | None = None,
        namespace: str | None = None,
        provider: AmazonCloudWatchEMFProvider | None = None,
    ):
        self.metric_set = self._metrics
        self.metadata_set = self._metadata
        self.default_dimensions = self._default_dimensions
        self.dimension_set = self._dimensions

        self.dimension_set.update(**self._default_dimensions)

        if provider is None:
            self.provider = AmazonCloudWatchEMFProvider(
                namespace=namespace,
                service=service,
                metric_set=self.metric_set,
                dimension_set=self.dimension_set,
                metadata_set=self.metadata_set,
                default_dimensions=self._default_dimensions,
            )
        else:
            self.provider = provider

    def add_metric(
        self,
        name: str,
        unit: MetricUnit | str,
        value: float,
        resolution: MetricResolution | int = 60,
    ) -> None:
        self.provider.add_metric(name=name, unit=unit, value=value, resolution=resolution)

    def add_dimension(self, name: str, value: str) -> None:
        self.provider.add_dimension(name=name, value=value)

    def serialize_metric_set(
        self,
        metrics: Dict | None = None,
        dimensions: Dict | None = None,
        metadata: Dict | None = None,
    ) -> CloudWatchEMFOutput:
        return self.provider.serialize_metric_set(metrics=metrics, dimensions=dimensions, metadata=metadata)

    def add_metadata(self, key: str, value: Any) -> None:
        self.provider.add_metadata(key=key, value=value)

    def flush_metrics(self, raise_on_empty_metrics: bool = False) -> None:
        self.provider.flush_metrics(raise_on_empty_metrics=raise_on_empty_metrics)

    def log_metrics(
        self,
        lambda_handler: Callable[[Dict, Any], Any] | Optional[Callable[[Dict, Any, Optional[Dict]], Any]] = None,
        capture_cold_start_metric: bool = False,
        raise_on_empty_metrics: bool = False,
        default_dimensions: Dict[str, str] | None = None,
    ):
        return self.provider.log_metrics(
            lambda_handler=lambda_handler,
            capture_cold_start_metric=capture_cold_start_metric,
            raise_on_empty_metrics=raise_on_empty_metrics,
            default_dimensions=default_dimensions,
        )

    def set_default_dimensions(self, **dimensions) -> None:
        self.provider.set_default_dimensions(**dimensions)
        """Persist dimensions across Lambda invocations

        Parameters
        ----------
        dimensions : Dict[str, Any], optional
            metric dimensions as key=value

        Example
        -------
        **Sets some default dimensions that will always be present across metrics and invocations**

            from aws_lambda_powertools import Metrics

            metrics = Metrics(namespace="ServerlessAirline", service="payment")
            metrics.set_default_dimensions(environment="demo", another="one")

            @metrics.log_metrics()
            def lambda_handler():
                return True
        """
        for name, value in dimensions.items():
            self.add_dimension(name, value)

        self.default_dimensions.update(**dimensions)

    def clear_default_dimensions(self) -> None:
        self.provider.default_dimensions.clear()
        self.default_dimensions.clear()

    def clear_metrics(self) -> None:
        self.provider.clear_metrics()

    # We now allow customers to bring their own instance
    # of the AmazonCloudWatchEMFProvider provider
    # So we need to define getter/setter for namespace and service properties
    # To access these attributes on the provider instance.
    @property
    def namespace(self):
        return self.provider.namespace

    @namespace.setter
    def namespace(self, namespace):
        self.provider.namespace = namespace

    @property
    def service(self):
        return self.provider.service

    @service.setter
    def service(self, service):
        self.provider.service = service

Instance variables

var namespace
Expand source code
@property
def namespace(self):
    return self.provider.namespace
var service
Expand source code
@property
def service(self):
    return self.provider.service

Methods

def add_dimension(self, name: str, value: str) ‑> None
Expand source code
def add_dimension(self, name: str, value: str) -> None:
    self.provider.add_dimension(name=name, value=value)
def add_metadata(self, key: str, value: Any) ‑> None
Expand source code
def add_metadata(self, key: str, value: Any) -> None:
    self.provider.add_metadata(key=key, value=value)
def add_metric(self, name: str, unit: MetricUnit | str, value: float, resolution: MetricResolution | int = 60) ‑> None
Expand source code
def add_metric(
    self,
    name: str,
    unit: MetricUnit | str,
    value: float,
    resolution: MetricResolution | int = 60,
) -> None:
    self.provider.add_metric(name=name, unit=unit, value=value, resolution=resolution)
def clear_default_dimensions(self) ‑> None
Expand source code
def clear_default_dimensions(self) -> None:
    self.provider.default_dimensions.clear()
    self.default_dimensions.clear()
def clear_metrics(self) ‑> None
Expand source code
def clear_metrics(self) -> None:
    self.provider.clear_metrics()
def flush_metrics(self, raise_on_empty_metrics: bool = False) ‑> None
Expand source code
def flush_metrics(self, raise_on_empty_metrics: bool = False) -> None:
    self.provider.flush_metrics(raise_on_empty_metrics=raise_on_empty_metrics)
def log_metrics(self, lambda_handler: Callable[[Dict, Any], Any] | Optional[Callable[[Dict, Any, Optional[Dict]], Any]] = None, capture_cold_start_metric: bool = False, raise_on_empty_metrics: bool = False, default_dimensions: Dict[str, str] | None = None)
Expand source code
def log_metrics(
    self,
    lambda_handler: Callable[[Dict, Any], Any] | Optional[Callable[[Dict, Any, Optional[Dict]], Any]] = None,
    capture_cold_start_metric: bool = False,
    raise_on_empty_metrics: bool = False,
    default_dimensions: Dict[str, str] | None = None,
):
    return self.provider.log_metrics(
        lambda_handler=lambda_handler,
        capture_cold_start_metric=capture_cold_start_metric,
        raise_on_empty_metrics=raise_on_empty_metrics,
        default_dimensions=default_dimensions,
    )
def serialize_metric_set(self, metrics: Dict | None = None, dimensions: Dict | None = None, metadata: Dict | None = None) ‑> CloudWatchEMFOutput
Expand source code
def serialize_metric_set(
    self,
    metrics: Dict | None = None,
    dimensions: Dict | None = None,
    metadata: Dict | None = None,
) -> CloudWatchEMFOutput:
    return self.provider.serialize_metric_set(metrics=metrics, dimensions=dimensions, metadata=metadata)
def set_default_dimensions(self, **dimensions) ‑> None
Expand source code
def set_default_dimensions(self, **dimensions) -> None:
    self.provider.set_default_dimensions(**dimensions)
    """Persist dimensions across Lambda invocations

    Parameters
    ----------
    dimensions : Dict[str, Any], optional
        metric dimensions as key=value

    Example
    -------
    **Sets some default dimensions that will always be present across metrics and invocations**

        from aws_lambda_powertools import Metrics

        metrics = Metrics(namespace="ServerlessAirline", service="payment")
        metrics.set_default_dimensions(environment="demo", another="one")

        @metrics.log_metrics()
        def lambda_handler():
            return True
    """
    for name, value in dimensions.items():
        self.add_dimension(name, value)

    self.default_dimensions.update(**dimensions)
class SchemaValidationError (*args, **kwargs)

When serialization fail schema validation

Expand source code
class SchemaValidationError(Exception):
    """When serialization fail schema validation"""

    pass

Ancestors

  • builtins.Exception
  • builtins.BaseException