January 2024 Summaries

In January 2024, Cron Monitoring became generally available and Metrics Alpha was released in alpha mode. The new feature allows users to monitor their scheduled jobs and set up alerting parameters such as margins and thresholds. Sentry also improved its Slack notification integration by adding additional details like impact, category, and the developer who owns it. Additionally, the SDK updates include a new Hydration Diff tool for Next.js and backpressure management features that reduce the number of transactions sent to Sentry during traffic spikes. The latest Sentry Java, Ruby, and Python SDKs automatically lower the sample rate to decrease Sentry SDK load on systems.

When dealing with API rate limits, it's essential to understand why they exist and the inconsistencies in handling them. Rate limiting is a traffic management strategy that maintains API stability and prevents users from exhausting system resources. However, there are no published standards for building and consuming APIs that implement rate limiting, resulting in varied HTTP response status codes and headers. To handle rate-limited responses, developers should use a combination of HTTP status codes and available response headers to determine how to proceed, considering factors such as request urgency, application type, and user feedback. In some cases, retrying the request with exponential backoff can be an effective strategy, but it's not always necessary or appropriate. Ultimately, being rate-limited is a "nice to have" problem that indicates the presence of users, making it essential for developers to upgrade their SaaS tool plans accordingly.

The company is updating its Terms of Service to address concerns about the use of customer data in research and development (R&D) projects, particularly those involving generative AI. They have identified two critical missteps: lack of clear communication on their intentions and not considering variable use cases requiring documented consent mechanisms. To re-align these goals with customer trust, they are scaling noise to improve error fingerprinting and deduplication, while ensuring data security and privacy remain paramount. The company will rely on explicit consent for using identifiable or potentially identifiable customer data in shared functionality. For non-identifiable, aggregatable data, no consent is required. They aim to provide granular control over the ways Sentry utilizes customers' data.

Debugging weird stack traces with Session Replay` is a topic that highlights the limitations of traditional debugging methods when faced with complex and nuanced issues. The example from Syntax.fm illustrates how a non-illustrative stack trace can lead to incorrect assumptions about the root cause of an error, and how Session Replay can provide valuable insights into user behavior, helping developers to identify and resolve issues more efficiently. By analyzing the replay of a specific issue, Wes Bos was able to see that the audio player was not loading due to the `preload="none"` attribute, which was causing the browser to prevent auto-play altogether. The fix involved loading the audio file, waiting for it to resolve, and then playing it, adjusting the timestamp accordingly. This example demonstrates the importance of using Session Replay as a tool in the debugging process, especially when dealing with complex issues that may not be immediately apparent from traditional stack traces.

Sentry has updated its Performance Score calculation methodology for its Web Vitals module to improve the accuracy of scores that reflect the user experience of web applications more precisely. Previously, Performance Scores were determined by aggregating Web Vitals data and applying static weights, which sometimes led to skewed results due to outlier pageloads and uneven representation of Web Vitals. The new approach calculates scores based on individual pageloads, allowing for dynamic weight adjustments when certain Web Vitals are missing and reducing the influence of outliers by averaging pageload scores. This change is intended to provide a more accurate reflection of an app's performance, aligning scores more closely with actual user experience, and it is expected to result in higher Performance Scores for many users by mitigating the negative impact of outliers. As a result, Opportunity Scores, which are also derived from Performance Scores, will see improved accuracy, and users may notice changes in their data upon visiting the Web Vitals module.

Sentry has released Cron Monitoring, which alerts developers when scheduled jobs fail and provides context to fix them. This feature focuses on serving code-level insights behind missed and failed jobs, relating them to errors already in the issue stream. The tool helps organizations like fintech startup Wrapbook detect and solve minor issues before they grow into large problems, saving time and resources. Cron Monitoring is integrated into Sentry SDKs for easy setup, with features such as alerting rules, error tracing, and recovery thresholds that help avoid human errors. The feature aims to provide code-level insights that contribute to the overall uptime and reliability of an application, and is available on all Sentry account plans, including a free one monitor with unlimited check-ins.

Ahead-of-Time (AOT) compilation in .NET 8.0 allows ASP.NET Core applications to be compiled into native machine code, eliminating the need for the .NET runtime to be installed on the executing machine, similar to how applications in languages like C or Rust operate. AOT compilation offers faster startup times and reduced memory and disk usage, which is particularly beneficial for short-lived services or applications with scaling policies. However, there are several limitations and compatibility issues, such as incompatibility with certain ASP.NET Core features and challenges with trimming, reflection, and dependency management, that developers must navigate to benefit from AOT. The Sentry SDK case study reveals these challenges, particularly the need to replace Newtonsoft.Json with System.Text.Json for serialization, adjust configuration bindings, and handle dependencies that aren't AOT-compatible, illustrating a practical approach to achieving AOT compatibility. Despite the complexities, the advantages of AOT, such as improved performance and lower operational costs, make it an attractive option for many developers looking to optimize their ASP.NET Core applications.

Sentry, a company born and bred in the Open Source community, has been exploring ways to balance user freedom with developer sustainability, leading to an "Open Source sustainability crisis." The company has found success with time-delayed licensing, which allows it to preserve commercial advantage while sharing its software's source code under an Open Source license. A research paper by the Open Source Initiative and Open Tech Strategies titled "Delayed Open Source Publication" reveals that this approach has been used since the early days of Open Source, with companies wanting to keep as many advantages of Open Source as possible while preserving commercial advantage. Sentry's experience with time-delayed licensing has allowed it to prevent larger competitors from eating its lunch and preserve its ability to produce products while sharing with the community. The paper concludes that there is a convergence towards a recognizable set of delayed Open Source publication licenses, including the newly launched Functional Source License (FSL).

Sentry removed advertising cookies from their website in July 2023, impacting them both positively and negatively. The removal of cookies has made performance marketing more challenging, with issues including reduced targeting options, increased CPCs, and changes to reporting and attribution models. To mitigate these effects, Sentry shifted towards engagement retargeting on ad channels, relied on UTM parameters and manual bidding for display ads, and prioritized content marketing. They also encountered difficulties in procurement, as many tools struggled to navigate user tracking and cookieless environments. Despite the challenges, Sentry found that going completely cookieless allowed them to gain new insights and perspectives, and they would do it again if given the choice. The future of performance marketing is likely to be cookieless, with companies needing to adapt to new models and tools that respect user privacy.

While companies often tout the importance of user privacy, few actually live and breathe it as a company like Sentry does. The company's take on privacy is rooted in three key fundamentals: Don't make me choose, think like your customer, and build for tomorrow today. This approach extends beyond just privacy to how the company thinks about its product, marketing, and business. Recent regulations around PII and GDPR have led Sentry to believe that tracking users in a "personalized" way will cease to exist, and instead, they're focusing on creating a better user experience that respects users' privacy. The company has removed all advertising cookies from their website, which is a bold change, and believes this move will force marketing teams to rethink how they scale customer growth without relying on data or insights. By taking action now, Sentry is getting ahead of the paradigm shift towards protecting users' rights and creating a better user experience.

Sentry has released several product updates in December, including the introduction of Spotlight, which brings telemetry to local development environments, allowing developers to see errors and other signals in real-time. The company also released new performance issues for regression and slow database queries, as well as an improved feedback widget with user installation capabilities. Additionally, Sentry upgraded its Android SDK, added support for Terraform, and launched a beta version of the improved feedback widget. These updates aim to improve developer productivity and debugging experiences.

Sentry.io has been using browser profiling to collect over 2M profiles and gain insights into how its users experience its dashboard, revealing issues such as UI jank that were causing slow interactions for users. Profiling helps developers proactively identify performance issues without relying on customers to report slow behavior, enabling them to get ahead of performance problems. By analyzing the most commonly collected spans during page-load events, Sentry identified a UI long task event span with a high frequency and p75 time, which was causing jank due to inefficient code in React tree diffing. The team used profiling data to pinpoint the issue, analyze the source code, and implement a solution that mitigated the problem, resulting in improved runtime performance for users.

This comprehensive guide discusses how to effectively distribute platform-specific binaries using npm, focusing on the challenges and solutions encountered by Sentry with their @sentry/cli package. The guide highlights the difficulty of including large binaries for different operating systems and architectures within a single npm package due to size constraints. It explores two primary strategies: using `optionalDependencies`, which allows package managers to install only the relevant binary for the user's system based on `os` and `cpu` fields, and employing a `postinstall` script to download the necessary binary if the corresponding optional dependency is not installed. The post also addresses potential pitfalls, such as supply chain attacks and the disabling of these features in some setups, recommending the implementation of both strategies to ensure reliability. Additionally, the guide provides detailed instructions and code snippets for implementing these strategies, considerations for optimizing the process, and advice on handling errors and inefficiencies that may arise during binary distribution.

We are taking steps to update our AI and machine learning practices to better respect the security and privacy of our users' data. We will delay the effective date of our Terms of Service change indefinitely until we've satisfied shared concerns, including providing a better consent mechanism for customers. We will not push customers to agree to use cases that are not well-defined or understood and will explore an opt-out mechanism for other uses of data. We need time to explore what these changes would look like and how we would implement them in a way that stays true to protecting our customers. We are updating our Terms of Service to allow us to use service data for AI, with controls in place to protect customer data, including scrubbing data for PII and ensuring the only service data presented in the output belongs to the customer using the feature.

Slow-loading assets on web pages can negatively impact user experience and conversion rates. To address this issue, Sentry has introduced Resource Monitoring, which helps identify slow-loading JavaScript, CSS, and font assets in production workloads. This tool provides information about asset load duration and size, allowing developers to understand how their resources are used in the real world and make meaningful performance optimizations. By monitoring render-blocking resources and filtering for problematic assets, developers can focus on fixing issues that have the most significant impact on page loading speed. Additionally, Resource Monitoring helps identify third-party scripts added outside of the development process and their effect on site performance. In the future, Sentry plans to connect resource monitoring with its web vitals module and add more network-related information for resources.