August 2021 Summaries

Enrique Fueyo Ramírez, Co-founder and CTO of Lang.ai, shares how his team instrumented performance monitoring for GraphQL resolvers using Sentry. They used Apollo to connect the frontend and backend, and created a custom plugin to intercept resolver lifecycles, starting and finishing spans for each request. The plugin allows them to monitor performance without adding explicit start/stop code to their functions, providing fine-grained information on individual transactions and distributed tracing capabilities.

Sentry's mobile error monitoring provides real-time visibility into how fast an app starts, the duration of HTTP requests, and other performance metrics such as slow and frozen frames. The four key metrics every mobile team should track are cold starts, warm starts, slow frames, and frozen frames, which can help identify bottlenecks and speed up apps. Sentry's auto-instrumentation allows developers to easily monitor their app's performance without additional setup or accumulating testing devices, providing detailed insights into transactions, spans, and Mobile Vitals. By tracking these metrics, developers can prioritize critical performance issues, trace the issue down to the root cause, and solve them faster, ultimately providing a delightful user experience.

Distributed tracing is a method of recording connected operations across multiple services, allowing developers to pinpoint errors or performance bottlenecks in individual services that affect the overall system. Distributed tracing involves two fundamental components: spans, which describe an operation or work taking place on a service, and traces, which describe the end-to-end journey of one or more connected spans. Traces are linked together using a unique trace identifier and span identifiers, with the trace context being propagated between services to connect them. By attaching metadata describing operations, parent operations, and tracing identifiers, developers can augment logging and telemetry data to better understand the sequence of events in their distributed services, and tools like OpenTelemetry and Sentry provide libraries and APIs for instrumenting and exporting this data.

Komodor is a Kubernetes-native platform designed to streamline troubleshooting by consolidating changes and impacts from various tools into a single timeline view, allowing developers to focus on innovation rather than hours of troubleshooting. The platform integrates with existing tools such as CI/CD, repo, monitoring, alerting, and communication, providing a comprehensive view of the service's context. By integrating Komodor with Sentry, teams can triage incidents triggered by feature flags, including exception data, and jump to the full timeline of changes for specific services, enabling faster debugging and deployment with confidence.

The Sentry team developed a mobile app for managing releases, focusing on Release Health, using Flutter as the primary framework to build both iOS and Android versions. The app fetches project data from linked organizations and displays session health information, including healthy, errored, abnormal, and crashed sessions, with customizable settings and widget integration. The team used Redux for state management, leveraging Flutter's widgets and ViewModels to update the UI automatically when state changes. They also utilized Sentry's SDK for issue detection and custom drawing using the canvas API to improve performance and pixel-perfect rendering. The app was developed during Hackweek and is now available on the Apple App Store and Google Play Store, with plans for future releases based on user feedback.