Once you’ve integrated with AWS CloudWatch, you have access to metrics from AWS Lambda, which is a compute service that enables code execution without the need to provision or manage servers. Lambda executes your code on a high-availability compute infrastructure while handling all compute resource management tasks, such as server and operating system upkeep, capacity provisioning and automatic scaling, and logging.
See all AWS integrations.
To verify metrics are reporting, search for the metrics on the Metric details page in Settings.
The following table shows the Lambda metrics ingested by Cloud Observability.
| Metric Name | Unit | Description |
|---|---|---|
| aws.lambda.invocations | count | The total number of times your function code is called, both successful and unsuccessful calls with regard to function errors. If an invocation request is throttled or encounters another invocation error, no invocations are recorded. This is the same as how many requests were billed. |
| aws.lambda.errors | count | The number of function calls that end in an error. Exceptions, including timeouts, thrown by your code or the Lambda runtime are both considered function errors. |
| aws.lambda.dead_letter_errors | count | The number of times Lambda tries to send an event to a dead-letter queue during an asynchronous invocation but is unsuccessful. Dead-letter errors can be the result of resource configuration issues, permissions issues, or size's limitations. |
| aws.lambda.destination_delivery_failures | count | The number of times Lambda tries to send an event to a destination during an asynchronous invocation but is unsuccessful. Size restrictions, incorrectly configured resources, and permissions issues can all result in delivery errors. |
| aws.lambda.throttles | count | The volume of throttled invocation requests. Lambda rejects extra requests with a TooManyRequestsException exception when all function' instances are handling requests and there is no concurrency to scale up. Requests that have been throttled and other invocation errors are not included in this calculation. |
| aws.lambda.provisioned_concurrency_invocations | count | The quantity of times provisioned concurrency calls your function code. |
| aws.lambda.provisioned_concurrency_spillover_invocations | count | The number of times the code for your function is run when all provisioned concurrency is active when using standard concurrency. |
| aws.lambda.duration | gauge | How long it takes your function code to process an event. The value of Duration, rounded up to the nearest millisecond, represents the invocation's billed duration. |
| aws.lambda.post_runtime_extensions_duration | gauge | The total time spent by the runtime after the function code has finished running code for "extensions." |
| aws.lambda.iterator_age | gauge | Age of the most recent record in the event for event source mappings that read from streams. The age is the interval of time between the record's receipt by a stream and the event's transmission to the function by the event source mapping. |
| aws.lambda.offset_lag | gauge | The difference in offset between the most recent record written to a topic and the most recent record your Lambda function's consumer group processed for self-managed Apache Kafka and Amazon Managed Streaming for Apache Kafka (Amazon MSK) event sources. Despite the possibility of numerous partitions for a Kafka topic, this statistic assesses the offset latency at the topic level. |
| aws.lambda.concurrent_executions | gauge | The number of function instances handling events. Invocation requests are throttled when this number exceeds the function's reserved concurrency limit or your concurrent executions quota for the region. |
| aws.lambda.provisioned_concurrent_executions | gauge | The number of provisioned concurrency function instances processing events Lambda emits the current count for each invocation of an alias or version with provisioned concurrency. |
| aws.lambda.provisioned_concurrency_utilization | gauge | A version's or alias's value of ProvisionedConcurrentExecutions divided by the overall amount of provisioned concurrency allocated. A value of.5, for example, indicates that 50% of the provisioned concurrency allotted is "used." |
| aws.lambda.unreserved_concurrent_executions | gauge | The quantity of events that functions with no reserved concurrency are processing for a Region. |
See also
Updated Dec 2, 2022