March 2017 Summaries

Evan Ralston has joined Sentry as an operations engineer, bringing his experience from various Bay Area startups where he improved uptime of web services. He will focus on proactive project management and system optimization for the Sentry team, rather than just reacting to issues. In his free time, Evan enjoys sharing his love for Steven Universe with others.

To help reduce noisy JavaScript errors when using Sentry for error tracking, consider configuring Sentry to whitelist specific URLs to ignore problematic errors originating from browser extensions, malware, or 3rd-party applications. Using inbound data filters can also discard known low-value errors, such as those triggered by legacy browsers or web crawlers. Keeping the latest version of Raven.js installed and utilizing source maps can improve error grouping and reduce noise in issue streams. Ignoring troublesome errors via Sentry's UI or configuration can prevent them from reaching servers and counting against account quotas. Finally, uploading production JavaScript files and source maps as release artifacts ensures maximum grouping effectiveness by providing direct access to these files. By implementing these strategies, developers can greatly reduce the amount of noisy JavaScript errors they receive.

Kelly Carino is joining the Sentry team as part of a new hire announcement, bringing her experience in customer experience improvement and scaling the support team from Massdrop to the role. She will be focusing on technical support and building out projects with the Sentry team.

The company has introduced a dedicated React Native SDK with some exciting features. Users of raven-js for React error monitoring are encouraged to transition to the new react-native-sentry SDK. This innovative SDK enables mixed stacktraces, allowing users to see the last JavaScript call before a crash in native code. Moreover, it ensures proper handling of native crashes on iOS and provides useful stacktraces for JavaScript errors. The SDK also offers more information about devices and operating systems due to its use of the powerful native Swift SDK. While currently only fully supporting iOS, it will gracefully fall back to raven-js for Android apps. To begin using the new SDK, refer to the react-native-sentry documentation.

Reprocessing is a new feature in Sentry that allows iOS developers to reprocess events from error monitoring that cannot be processed due to missing debug symbols, holding them in the event stream until the symbols are uploaded, reducing noise and preventing email notifications from being sent out prematurely. The feature can be enabled or disabled in project settings, providing an additional layer of control for developers. With reprocessing, events that lack mandatory debug symbols are put on hold and will not show up until their symbols are uploaded, ensuring accurate symbolication and reducing bad grouping. Developers can also trigger reprocessing manually through the API or command line tools.

Dodging S3 Downtime With Nginx and HAProxy We primarily use Amazon S3 as a data store for uploaded artifacts like JavaScript source maps and iOS debug symbols, which are critical in our event processing pipeline. Recently, we experienced an outage that lasted 3 hours, but the impact on our processing pipeline was minimal. To mitigate this issue, we set up an S3 proxy cache using Nginx and HAProxy to minimize risk to production-facing traffic, avoid single points of failure, prove the concept without increasing hardware costs, and avoid committing changes to application code. We used Nginx as a proxy server to leverage its `proxy_cache` feature, which stored all our S3 assets on disk when requested, while using HAProxy to route requests back to S3 if Nginx failed. Our new proposed infrastructure uses HAProxy to direct traffic to the cache server, which then proxies upstream to Amazon, allowing for a failover to occur without interruption. The configuration of HAProxy was quick and only required eight lines, while configuring Nginx involved more steps due to its heavy lifting in caching. The proxy service had been running for a week while we monitored our bandwidth and S3 bill, which dropped significantly. During the S3 outage, the proxy cache served all Sentry's S3 assets, allowing the event processing pipeline to continue flowing smoothly. Overall, implementing this proxy cache took less than a day, reducing our S3 bandwidth cost by 70% and gaining performance and reliability in our event processing pipeline.