December 2020 Summaries

Feature flags, also known as feature toggles, are a software development technique that allows teams to enable or disable parts of a codebase without altering the code itself. This approach supports DevOps practices by increasing deployment speeds, system stability, and cross-team collaboration. By using feature flags, developers can perform continuous integration/continuous delivery (CI/CD), A/B testing, and partial rollouts or canary releases more effectively. Feature flag management software like LaunchDarkly helps teams manage multiple flags from a centralized place, enabling faster code deployments and better user experiences. Overall, feature flags are an essential tool for organizations looking to adopt DevOps practices and improve their software development life cycle (SDLC) pipeline.

Canary testing is a method used in software development to test new features or updates with minimal impact on users. It involves deploying the updated version of an application to a small subset of users, allowing developers to monitor and fix any issues before rolling it out to all users. The term "canary testing" originates from coal mining, where canaries were used as early indicators of toxic gases. Feature flags are often utilized in canary testing to expose new features to specific groups of users within a single production environment. This technique reduces the risk associated with software releases and allows for faster deployment of updates while minimizing potential negative impacts on end-users.

In her exploration of the scientific method's application to software development, Dawn Parzych emphasizes the importance of critical thinking over mere knowledge in problem-solving. She illustrates how breaking down problems into smaller parts, forming hypotheses, and conducting controlled experiments can effectively address software issues, as demonstrated in a case at LaunchDarkly involving performance improvements with pagination. Parzych highlights the necessity of creating specific, measurable hypotheses and conducting experiments that yield repeatable results, while also sharing insights gained from these experiments. Additionally, she warns of biases such as anchoring and framing, which can distort decision-making, urging practitioners to remain aware of these biases when crafting hypotheses and interpreting outcomes.

Clubhouse, a project management software for development teams, has integrated LaunchDarkly into its operations to enhance product collaboration and accelerate experimentation. Since adopting LaunchDarkly, Clubhouse has experienced faster validation of ideas, rapid iteration with engineering, and shorter release cycles. The integration also enabled the company to separate deployment from release, allowing product managers to control targeting rules without relying on engineering resources. This approach led to a 30% increase in trial retention for Clubhouse. As the product team continues to mature, they plan to add more experiments and fine-tune customer experiences while exploring additional integrations with tools like Segment, Amplitude, and Slack.

LaunchDarkly has introduced Flag Approvals to streamline the process of reviewing and approving feature flag changes before they are applied in production. This feature allows users to request a review from teammates for proposed flag updates directly within the platform, replicating common code-review workflows such as PR reviews in GitHub. The approval process helps teams collaborate more effectively on updating flags, ensuring that all changes are thoroughly reviewed before being merged into production. Flag Approvals also provide users with granular control over how approvals work in different environments and support continuous delivery by allowing multiple approval requests for the same flag. Future enhancements include mandatory approvals for flag changes in specific environments, approval requests for scheduling flag changes, and integration with ServiceNow for centralized change management.

LaunchDarkly introduces saved dashboards to simplify flag management and improve the overall user experience. This new feature allows users to save frequent filter views, making it easier for teams within an organization to navigate between flags relevant to them. Users can set up personalized views tailored to their workflows or team practices, and share dashboard links with teammates. LaunchDarkly aims to continuously enhance the dashboard experience by adding features like saved dashboards, enhanced filters, and cross-environment flag status to save time and effort for users.

Jared Jones discusses the integration of Rollbar's Automated Feature Flag Triggers with LaunchDarkly, which enhances risk management in software deployment by automatically deactivating feature flags in response to new or reactivated errors. This integration aims to improve the confidence and speed of feature deployment by allowing developers to address errors in real time, thus reducing downtime and enhancing user experience. The feature supports progressive delivery, test-in-production, and A/B testing by using error data to inform deployment decisions, ensuring that any problematic features are quickly rolled back. Developers are encouraged to participate in the beta program for these Automated Feature Flag Triggers or develop their own integrations through LaunchDarkly’s partner program.

Last week, a session was held with customers from Xero and National Australia Bank (NAB) discussing Feature Management in financial services and other regulated industries. Key points included working closely with change management teams to understand regulatory guardrails and best practices, leveraging feature management solutions for collaboration across teams and departments, and maximizing value by starting small and expanding use cases over time. Additionally, the upcoming Feature Workflows functionality was highlighted as a tool to further ensure compliance with change management processes.

AWS re:Invent 2020 has commenced virtually with three weeks of programming focused on various aspects of software development and deployment. The virtual format allows participants to choose from multiple time slots for each session. Key sessions related to releasing and deploying software include "Choosing the Right Modern Deployment Strategy," "Mistakes were made, lessons were learned," "De-risking migration and modernization initiatives," "Learn how teams at Amazon rapidly release features at scale," "Canaries in the code mines: Monitoring deployment pipelines," and "To monitor or to observe: The million-dollar question." These sessions offer insights into various aspects of software delivery, such as choosing the right deployment strategy, learning from mistakes, safely migrating and modernizing tech stacks, monitoring deployment pipelines, and understanding observability.