July 2024 Summaries

Monitoring tools are great at identifying issues but don't provide the context needed to solve them. Observability tools offer deeper insights into an application's internal state but don't help developers fix issues directly. Debuggability, on the other hand, is the practice of writing and shipping code that delivers exceptional user experiences by providing actionable data and context to developers when they need it most. Debuggability tools like Sentry enhance developer workflows with connected, contextualized, and intuitive information to quickly identify and resolve root causes of issues across distributed systems, making it easier for every team member to solve problems without switching tools or wasting time guessing. A renewed focus on debuggability benefits all members of a development team by providing the necessary context to positively impact an application's reliability, robustness, and quality.

Tracing is a debugging technique used to capture debugging data that visualizes operation flows, such as a page load. A trace consists of spans with start and end times and arbitrary data attached to them. To get started with capturing traces, one needs to install the Sentry React SDK, initialize it at the top of the app, and deploy changes. The SDK automatically instruments the majority of the codebase, allowing for immediate inspection of tracing data. Tracing can help identify performance issues like fetch waterfalls, where multiple fetch requests are invoked in a sequential manner, leading to significant performance degradation. Identifying these waterfalls involves inspecting slower page loads, which may reveal a fetch waterfall that can be optimized to improve page load times.

Here's a neutral, objective, and interesting summary of the text in one paragraph: Sentry has launched its Session Replay feature for mobile devices, allowing developers to capture video-like reproductions of user sessions to debug issues such as crashes, jankiness, and unresponsiveness. This feature expands on Sentry's existing Session Replay capabilities for web applications, providing a visual representation of user interactions and debugging context to help identify the root cause of problems. With Session Replay, developers can prioritize sampling sessions when users encounter errors, quickly pinpointing how issues occurred and reducing time spent waiting for data. The feature is now available in open beta and free for Sentry customers, with support for Android, iOS, React Native, and other frameworks planned in the future.

This article discusses the importance of identifying and fixing slow page loads caused by slow backends, and how using Sentry can help developers debug performance issues. The author shares their experience with a dynamic website that experiences slow load times due to navigating through multiple routes, each dynamically rendering a custom Pokemon-like card for a user based on their in-game scores. They use Sentry's User Feedback & Session Replay features to identify the issue, and then utilize Tracing to track the performance of their code and pinpoint the problem. The author finds that making unnecessary requests over and over again is causing the slow load times, and by running requests in parallel instead of series, they are able to significantly improve the page load speed. They also discuss how promise streaming can further improve performance by allowing pages to render before all requests have finished.

On May 13, 2024, Sentry's JavaScript SDK team released v8.0.0 of their JavaScript SDKs, marking a significant update that involved extensive changes, primarily the integration of OpenTelemetry. This shift necessitated numerous adjustments to performance APIs, leading to a complex migration path for users transitioning from v7 to v8. The team identified several takeaways, emphasizing the need for smaller, more frequent major releases to simplify the process for users and the importance of versioned documentation to prevent lag behind implementation. While they succeeded in testing and uncovering bugs during pre-release phases, the team acknowledged the need to better align deprecations with updated documentation and be more cautious with unnecessary deprecations. Additionally, they plan to engage more actively with the OpenTelemetry community to enhance collaboration and integration. Despite challenges, the team celebrated the success of the release and continued to refine the SDK, already progressing to version 8.18.0.

Sentry has added new Insights to provide visibility into common backend building blocks such as Caches, Queues, Queries, and Outbound Requests. This helps developers troubleshoot and debug issues when they occur. The Insights feature provides actionable workflows for improving application performance and understanding interactions with infrastructure like databases, queues, upstream APIs, and caches. Sentry's backend Insights are available to all Business and Enterprise plans and can be set up by following the instructions in their docs.

Requests is a feature in the Insights tool that helps developers debug outgoing HTTP requests, providing insights into call frequency, response times, and error rates to troubleshoot issues with third-party APIs. It offers features such as checking if a service is down for everyone or just the user, identifying rate limiting issues, and tracking down bugs in integrations with third-party APIs. Requests is included in Sentry's Business and Enterprise plans and can be accessed through the docs or by trying it out in a Sentry account.

Sentry has introduced a new feature called LLM Monitoring to help developers monitor their AI-powered applications' performance in production. This tool is designed specifically for large language model (LLM) applications and helps debug issues, control token costs, and improve overall application efficiency. With dashboards and alerts, developers can keep track of token usage and cost across all models or through an AI pipeline. Sentry LLM Monitoring also provides performance insights by tracing slowdowns back to the sequence of events leading to the LLM call. Additionally, it aggregates error events related to LLM projects into a single issue for efficient debugging. The feature is currently in beta and available to all Business and Enterprise plans users.

Introducing Insights: Tailored debugging workflows for your application` Insights is a new set of debugging workflows designed to help developers quickly identify and troubleshoot performance issues in their applications. Available on Sentry's Business and Enterprise plans, Insights provides a tailored approach to debugging by tracking key components of the application, infrastructure, and APIs, allowing users to drill down to the specific function or line of code causing issues. With Insights, developers can trace slow Web Vitals directly to their root cause, debug slow mobile screen loads, and connect slow database queries to code, providing a comprehensive view of performance issues in their applications. The feature is designed to help developers save time and reduce frustration when debugging performance issues, allowing them to quickly identify and fix problems before they affect users.

The Sentry Python SDK underwent a significant refactoring to simplify its API, improve compatibility with OpenTelemetry, and maintain backward compatibility, culminating in the release of version 2.0. The refactoring process involved transitioning from a Hub-based API to a Scope-based system that aligns with OpenTelemetry's Contexts, allowing for better data encapsulation and future-proofing the SDK. The project was executed in stages, starting with moving functionalities from the Hub to the Scope, refactoring the Scope, updating integrations to use the new API, and rigorous load testing, including dogfooding at Sentry.io. Despite some minor issues, such as a bug in the Celery integration and a bug report involving the Starlette framework and Uvicorn, these were promptly addressed, leading to a stable release. The changes have been well-received, with significant adoption by users, ensuring that the SDK is prepared for future developments while maintaining performance efficiency.

"LCP score issues often stem from backend problems that aren't immediately apparent. To identify the root cause of slow LCP scores, developers can use tracing to track the complete end-to-end user journey across all downstream systems and services. This involves installing an Application Performance Monitoring (APM) tool like Sentry, collecting data from real users, exploring Core Web Vitals scores, finding sampled page load events with poor LCP scores, and then exploring the full trace of those events to discover performance bottlenecks."

"Session Replay" is a feature of Sentry that records and replays all user interactions on web apps and websites, capturing user actions such as page visits, mouse movements, clicks, and scrolls, and network logs during those times. This allows developers to identify the exact point where an error occurs in their application, eliminating the need for manual steps to reproduce the issue. By using Session Replay, developers can easily solve issues without having to go through a plethora of manual steps, providing benefits such as easy issue reproduction, understanding the origin of reports, enhanced proactive support, fantastic user experience insights with Replay Details, and efficient debugging with network traces. It also eliminates the need for reporters to create screen recordings and attach them to project management software, reducing back-and-forth with reporters and improving overall user experience. By integrating Session Replay with other Sentry features, developers can gain a better understanding of their users' interactions and identify areas for improvement, ultimately resulting in a better UX for everyone.