Architecture Requirements
Summary
The OpenTelemetry Community Demo application is intended to be a ‘showcase’ for OpenTelemetry API, SDK, and tools in a production-lite cloud native application. The overall goal of this application is not only to provide a canonical ‘demo’ of OpenTelemetry components, but also to act as a framework for further customization by end-users, vendors, and other stakeholders.
Requirements
Application Goals
- Provide developers with a robust sample application they can use in learning OpenTelemetry instrumentation.
- Provide observability vendors with a single, well-supported, demo platform that they can further customize (or simply use OOB).
- Provide the OpenTelemetry community with a living artifact that demonstrates the features and capabilities of OTel APIs, SDKs, and tools.
- Provide OpenTelemetry maintainers and WGs a platform to demonstrate new features/concepts ‘in the wild’.
The following is a general description of the logical components of the demo application.
Main Application
The bulk of the demo app is a self-contained microservice-based application that does some useful ‘real-world’ work, such as an eCommerce site. This application is composed of multiple services that communicate with each other over gRPC and HTTP and runs on Kubernetes (or Docker, locally).
Each service shall be instrumented with OpenTelemetry for traces, metrics, and logs (as applicable/available).
Each service should be interchangeable with a service that performs the same business logic, implementing the same gRPC endpoints, but written in a different language/implementation. For the initial implementation of the demo, we should focus on adding as many missing languages as possible by swapping out existing services with implementations in un-represented languages. For future versions we will look to add more distinct language options per service.
Each service should communicate with a feature flag service in order to enable/disable ‘faults’ that can be used to illustrate how telemetry helps solve problems in distributed applications.
A PHP service should be added to the main application as an ‘admin service’. A Database should be added to enable CRUD functionality on the Product Catalog.
The ‘shippingservice’ should be reimplemented in Rust.
The ‘currencyservice’ should be reimplemented in C++.
The ’emailservice’ should be reimplemented in Ruby.
For future iterations, the ‘frontend’ service can be extended with a mobile application written in Swift.
Feature Flag Component
Feature flagging is a crucial part of cloud native application development. The demo uses OpenFeature, a CNCF incubating project, to manage feature flags.
Orchestration and Deployment
All services should run on Kubernetes. The OpenTelemetry Collector should be deployed via the OpenTelemetry Operator, and run in a sidecar + gateway mode. Telemetry from each pod should be routed from agents to a gateway, and the gateway should export telemetry by default to an open source trace + metrics visualizer.
For local/non-Kubernetes deployment, the Collector should be deployed via
compose file and monitor not only traces/metrics from applications, but also the
docker containers via dockerstatsreceiver.
A design goal of this project is to include a CI/CD pipeline for self-deployment into cloud environments. This could be skipped for local development.