OpenTelemetry Semantic Conventions

OpenTelemetry Semantic Conventions provide a standardized set of names and structures for attributes, events, logs, metrics, and spans. Using these conventions ensures that telemetry data produced by different instruments and libraries is interoperable and can be consistently queried and analyzed by backend systems.

This repository defines the following core areas of general semantic conventions:

• General attributes: Standardized attribute keys used across different telemetry types.
• Events: Conventions for structured events within telemetry.
• Logs: Conventions for log data.
• Metrics: Conventions for metric data.
• Spans: Conventions for traces and spans.

This document provides the standardized semantic conventions for observability signals in Function-as-a-Service (FaaS) environments. These conventions ensure that spans, metrics, exceptions, and service-specific details (like AWS Lambda) are named and structured consistently across different providers and instrumentation libraries.

The conventions are categorized into the following signal types:

• FaaS Spans: Standardized attributes and names for spans representing FaaS executions.
• FaaS Metrics: Standardized metric names and attributes for monitoring FaaS performance and health.
• FaaS Exceptions: Standardized conventions for capturing exceptions within FaaS functions.

Additionally, technology-specific conventions are provided for services such as AWS Lambda.

The AWS semantic conventions define standardized resource attributes specifically for Amazon Web Services (AWS) resources. These conventions ensure consistent naming and structure for telemetry data originating from AWS.

Note that attributes applicable to multiple cloud providers (such as account.id or host.operating_system) are not defined here; they are located in the parent model directory. This directory focuses exclusively on AWS-specific terminology and service-specific attributes.

OpenTelemetry provides a standardized set of instruments and attributes for monitoring hardware-level metrics. These conventions ensure that metrics from different hardware components (like CPUs, GPUs, or Disks) are reported using consistent names and attribute keys, enabling interoperability across different observability tools.

When implementing new hardware instruments that are not explicitly covered by these specifications, you should follow the general metric semantic conventions.

OpenTelemetry Semantic Conventions define a standardized set of names and structures for telemetry data emitted by OpenTelemetry SDK components. These conventions ensure that telemetry (such as metrics, logs, and traces) is consistent across different implementations, making it easier to query, visualize, and analyze data in observability backends.

Currently, semantic conventions are defined for the following signal:

• Metrics: Standardized names and attributes for metric data.

This project defines standardized semantic conventions for observability signals within Continuous Integration and Continuous Deployment (CI/CD) pipelines. These conventions ensure that spans, metrics, and logs generated by CI/CD processes are interoperable and can be analyzed consistently across different tools and platforms.

The conventions are categorized into three primary signal types:

• CI/CD Spans: Standardized attributes for trace spans representing CI/CD activities.
• CI/CD Metrics: Standardized names and attributes for metrics related to CI/CD processes.
• CI/CD Logs: Standardized attributes for log records produced during CI/CD execution.

This document defines the semantic conventions for describing Function as a Service (FaaS) or serverless function instances using spans.

Key Requirements:

• The span name should be set to the function name being executed.
• Depending on the faas.trigger value, additional attributes must be set (e.g., an http trigger should follow HTTP Server semantic conventions).
• If these conventions are followed, a Resource of type faas MUST exist following the FaaS resource semantic convention.

The gRPC semantic conventions extend and override the standard RPC Semantic Conventions. These conventions define how to instrument gRPC services using OpenTelemetry to ensure consistent observability across different implementations.

There are two primary ways to observe gRPC:

1. Spans: Used for tracing individual RPC calls (Client and Server side).
2. Metrics: Used for aggregating statistical data about gRPC performance.

Note on Compatibility: The gRPC project maintains its own native conventions for metrics and spans. If you are reconciling OpenTelemetry data with native gRPC telemetry, refer to the gRPC compatibility documentation to understand how these definitions relate.

OpenTelemetry defines semantic conventions specifically for CLI (Command Line Interface) programs to ensure consistent observability. Currently, the defined conventions focus on CLI Spans, which describe the execution and lifecycle of commands run in a terminal environment.

For detailed attribute names and span structures, refer to the CLI Spans documentation.