July 2019 Summaries

In June, GitHub hosted a Test in Production Meetup at the Intercom offices in London. Claire Knight, Senior Engineer at GitHub, discussed how her team uses feature flags for experimentation and testing in production. They involve users in experiments, particularly around rate limiting strategies. The team also conducts science experiments to test real-world scenarios and collect data to inform decisions. Feature flagging is a crucial tool for continuous testing and deployment, allowing the team to control access and make progress while involving customers and building rapport with their community.

Michael Gillett from The Betway Group discussed how his team embraces hypothesis-driven development and testing in production during a Test in Production Meetup at Intercom's London offices. He explained that this approach involves setting out requirements scientifically, analyzing data even when tests fail, and understanding what to measure and the expected outcomes. Gillett shared an example of how his team used A/B testing for their homepage, resulting in a 25% better successful registration rate. They also utilized feature toggling for technical improvements and load testing without impacting other applications. He emphasized the importance of setting goals with all stakeholders involved, understanding what success and failure look like, and having an agreement on the sample size for experiments.

LaunchDarkly has introduced support for the OAuth 2.0 standard as a service provider, making it easier for customers to build their own applications on top of the platform. This feature allows second-party built applications and services to authorize with LaunchDarkly without exposing user tokens or requiring passwords. Identity Providers can also authorize with LaunchDarkly and enable SCIM support, including custom role data to align with user groups. Third-party applications and integrations will be able to securely authorize using OAuth. For more information, read the documentation and contact [email protected] for obtaining OAuth credentials for your applications.

Kailin Ke, a recent graduate from Balboa High School, interned at LaunchDarkly for five weeks through Code Nation's program. During the internship, she worked with the Engineering team and learned React, Redux, and TypeScript while building landing pages and improving customer account information management systems. Kailin also developed communication and presentation skills during meetings and appreciated LaunchDarkly's company values emphasizing work-life balance.

Titan Jamison, a recent high school graduate, spent his summer developing technical and professional skills with industry professionals at LaunchDarkly through Code Nation, an organization that teaches coding fundamentals in public schools. He worked on projects for internal applications used by LaunchDarkly employees, learning frameworks like React, Redux, and Typescript along the way. The experience solidified his interest in computer science and highlighted the company's values of teamwork, respect, and growth.

LaunchDarkly introduces project-level view access control for organizations, allowing administrators to manage team members' access to features and information within projects. This feature enables the creation of private projects by restricting visibility on a per-user basis. View Restrictions apply to all resources in a project, including flags, dashboard views, audit logs, and APIs. The ability to manage flag dashboard content, create private projects, and implement security restrictions are some benefits of this feature. Custom roles can be configured using the Policy editor in Account Settings to specify access permissions for standard and custom resources. This feature is currently available for a portion of LaunchDarkly's user base, with plans to expand its availability.

LaunchDarkly has introduced two new SDKs - a client-side SDK for Node.js and a server-side SDK for C, both now in general availability (GA). The client-side SDK is designed for single-user contexts like desktop, mobile, and embedded applications, while the server-side SDK is intended for multi-user systems such as web application servers. These new SDKs expand LaunchDarkly's support to new sets of software, including client-side applications built on Node.js and server-side applications using C or C++. Users can refer to the documentation and reference guides for more information, and reach out to [email protected] with any questions.

Operational feature flags are used to manage and control the behavior of a system, often referred to as kill switches or circuit breakers. They can be short-term or long-term flags, with common use cases including operational interaction testing, API rate limiting, modifying logging levels, release management, and load shedding. Best practices for using operational feature flags include documenting the flag, indicating whether it is permanent or temporary, controlling who can change a flag, and using a relay proxy to reduce stress on infrastructure.

Feature flags, also known as feature toggles, are a common practice in software development. The use of feature flags can vary depending on whether they are short-term or permanent and the specific business needs. This blog series provides best practices for managing feature flags effectively. Key recommendations include determining whether a flag will be short-term or permanent, standardizing naming conventions, minimizing the reach of a flag, reviewing flag use at regular intervals, establishing access control and safety checks for permanent flags, creating a process for removing short-term flags, and conducting regular clean-up and review cycles. These practices can help improve feature flag management workflow and avoid technical debt.

At a recent Test in Production Meetup, Christine Yen, CEO and Co-Founder of Honeycomb, discussed how observability can help with "owning code in production." She emphasized the importance of developers being comfortable in production and considering not just whether their code works, but also if it works in production. Observability is an ability that a team has to understand what their systems are doing in production, supported by tools but ultimately tied to processes and people on the team. By looking at production during various stages of software development, developers can debug issues, build the right software, write better pre-production tests, test in production, and improve their understanding of how code impacts performance. This approach fosters a collaborative culture between dev and ops teams, allowing them to work together to deliver better software to customers.

LaunchDarkly has made enhancements to its platform for users with multiple profiles. Previously, the navigation panel did not clearly indicate which profile was active, and the logout button's location led to accidental logouts. The new updates include a menu at the bottom of the main navigation panel displaying the account name or email, a link to the profile within the menu, and relocation of the logout button. Additionally, improvements have been made to the Profile and Team tabs in account settings, showing assigned roles on the Profile tab and listing team members on the Team tab.