September 2020 Summaries

LaunchDarkly's Data Export feature allows users to analyze raw analytics data, including feature flag requests, analytics events, custom events, and experimentation events. By exporting this data to a location of their choice, users can utilize their own data warehouse and tools for analysis. The blog post demonstrates how to use LaunchDarkly's Data Export with Amazon Kinesis stream as the destination, then sending the data into Elasticsearch/Kibana for analytics purposes. This information can be used to gain insights into user demographics, feature usage, and troubleshooting support issues.

LaunchDarkly introduces Slack notifications for feature flags that are ready to be archived. This helps teams maintain clean codebases and reduce technical debt by prompting them when a flag's status becomes launched or inactive. Users can subscribe to these notifications through the LaunchDarkly Slack app, with options to selectively subscribe to flags for removal or filter by project, environment, flag key, and other details. Additionally, users can choose which actions they want to be notified about, such as create, update, delete, or removal.

LaunchDarkly has introduced new capabilities to its Code References tool to help developers efficiently track and manage feature flag references in their codebase. The features include Reference Aliases, Branch Filtering, and Last Reference Timestamps. These enhancements aim to improve the management of technical debt with feature flags by making it easier to locate all references to a specific flag, narrow down search for flag reference points, and keep track of whether a flag is active or not. LaunchDarkly also offers other tools like flag dashboard, compare and copy settings between environments, and guided Flag Archive workflow to simplify feature management. Developers can start using Code References by installing the latest version of the tool and checking out the updated documentation on GitHub.

Scaling LaunchDarkly's feature management platform across an entire organization involves choosing the right implementation model, with three main options: the single project model, one project per product line, and the shared services model. The single project model is easy to implement and suits smaller teams or monolithic structures but can become complex with numerous flags and permissions. The one project per product line model is ideal for organizations using microservices, as it simplifies permissions and limits scope but lacks cross-project dependencies. The shared services model, popular for managing common service dependencies, allows centralized flag management but requires careful handling to avoid complexity, such as not duplicating flags across projects. Each model has its advantages and challenges, and understanding them is crucial for a smooth scale-out process of LaunchDarkly within an organization.

LaunchDarkly has partnered with Amplitude to introduce Synced Segments, a feature that enables seamless management of user segments across platforms. This integration allows teams to import user cohorts directly from Amplitude and deliver custom experiences using feature flags. Synced Segments automatically synchronize changes in segment membership for feature targeting, eliminating the need for manual updates or coordination between systems. The partnership aims to streamline customer experience delivery by empowering data-driven development teams with precise behavioral cohorts tailored to their customers' needs.

LaunchDarkly has introduced an Experiments overview dashboard designed to help teams manage and review their experiments more efficiently. The new feature enables users to check the status of all experiments in a selected environment, create new experiments from one place, and view detailed results for each experiment's variations. This add-on is available to LaunchDarkly customers on Pro and Enterprise plans, with further updates planned to enhance its capabilities.

Incident resolution is crucial for minimizing the impact of unplanned service interruptions or reductions in quality. Automation plays a significant role in incident management, with many organizations lacking sufficient automation tools and processes. The pandemic has further complicated incident response due to remote work and increased demand for digital services. Feature flags can be used to improve incident management by providing valuable data points during troubleshooting, automating tasks, and reducing mean time to recovery (MTTR). Automation helps eliminate toil and cognitive load during incident resolution, allowing teams to focus on resolving the problem at hand. Regularly updating runbooks and incorporating chaos experiments or game days can help identify areas where operational flags are needed for proactive incident management.

LaunchDarkly introduces flag triggers for Honeycomb and Datadog integrations, allowing users to automate the process of turning feature flags 'on' or 'off'. This integration is designed for specific situations where a performance metric or alert should toggle a flag. Flag triggers work by sending webhooks to a unique URL, enabling automation from various tools such as CI and APM. Currently available on LaunchDarkly's Enterprise plans, future feature workflows will connect health metrics to more flag actions in users' workflows.

Trajectory Conference 2020 was held online due to the pandemic and attracted more attendees than an in-person event could have accommodated. The conference covered topics such as testing in production, scaling systems, industry trends, and maintaining perspective on who we build for and who is part of building. Key discussions included increasing speed of learning through rapid experimentation with safety measures, using feature flags for service protection, quarantining unusual traffic to investigate further, and implementing valid tests that focus on the desired outcome. Other topics covered scaling systems reliably and economically, automating workflows, progressive delivery, and keeping perspective by remembering the wider world. The conference also emphasized the importance of empathy, collaboration, and diversity in experience for building effective teams and products.