November 2022 Summaries

The article discusses how to integrate LaunchDarkly, a feature management platform, into AWS Lambda functions. It provides step-by-step instructions on setting up the LaunchDarkly SDK in Node.js and using it within Lambda functions. Additionally, it covers creating a Lambda layer for easier access to the SDK across multiple functions, configuring DynamoDB as a data store for faster flag retrieval, and utilizing the Relay Proxy for increased performance and security. The article also addresses handling LaunchDarkly analytics events in serverless environments.

Team collaboration is crucial in ensuring developers work together to develop and deliver products on time. Two main approaches for version control are feature branches and trunk-based development. Trunk-based development involves dividing tasks into smaller batches, collaborating from separate branches, and frequently merging them into a single branch called the trunk. This approach accelerates software release and allows for microservice architectures. Trunk-based development with Kubernetes can be challenging due to isolated services developed by different team members. However, it is possible to power this strategy using Kubernetes, which increases infrastructure use by efficiently sharing computing resources. To implement trunk-based development with Kubernetes, developers should create a feature branch from the source code main repository, perform adequate testing before merging changes to the trunk, and merge the feature branch into the trunk when work is complete. Trunk-based development significantly accelerates CI/CD processes in DevOps teams by facilitating and streamlining the merging and integration phases. It allows for building and deploying services independently, implementing new features faster, improving software quality, and encouraging cross-collaboration among development teams. Best practices for using trunk-based development include developing in small batches, using feature flags, executing comprehensive automated testing, conducting code reviews, and frequent merging with the trunk. Kubernetes microservice architecture offers an excellent way to achieve complete CI/CD through trunk-based development by splitting monolithic applications into microservice clusters, speeding up the development process, and ensuring efficient resource allocation.

In the blog post, Peter McCarron explores the integration of feature flags within Kubernetes and microservice environments, emphasizing their utility beyond frontend development. He delves into the use of GraphQL, a query language that enhances data retrieval processes, to demonstrate how feature flags can dynamically alter database tables and manage relational data in applications. McCarron provides an example using "Star Wars" and "Lord of the Rings" datasets to illustrate the implementation of feature flags in server-side and frontend components of GraphQL applications, utilizing tools like Join Monster for SQL translation and the Apollo React Client for state management. He highlights the potential of feature flags to facilitate seamless transitions between data sources, such as migrating from Postgres to MongoDB, by employing strategies like progressive rollouts in Kubernetes clusters. Ultimately, the post underscores the broader applicability of feature flags across various layers of application architecture to enable dynamic and flexible code variations.

Deploying new features in high-volume production environments can be risky due to potential bugs or issues affecting users. A percentage rollout strategy gradually introduces a feature to an increasing percentage of users over time, allowing for better control and observation during the release process. This approach is often used with feature flags, which are toggles or switches that allow developers to easily toggle code execution for specific user subsets without needing to redeploy. There are various rollout strategies, including full rollouts (big-bang adoption), canary releases, ring deployment, entitlement releases, and targeted releases. Percentage rollouts minimize risks by being safe and flexible, limiting the scope of errors, providing a kill switch for easy rollback, enabling feature refinement through iterative feedback, meeting customer expectations with quicker feedback cycles, saving development costs, and facilitating application in a progressive delivery pipeline. Major companies like Facebook, Microsoft, and Google use percentage rollouts as part of their progressive delivery strategies when launching new features. Feature flags enable developers to safely implement this approach by allowing for controlled, gradual introduction of new code and features with reduced risks.

In the 2022 State of Feature Management report, it was found that companies are increasingly investing in feature management despite uncertain economic conditions. The study surveyed 1,000 software and IT professionals, revealing that 70% view feature management as a mission-critical or high-priority investment. Virtually all respondents using feature flags (98%) agreed they save their company money and deliver a demonstrable ROI. Additionally, the report found that investments in feature management are rising, with 74% of surveyed companies planning to increase their IT budget for feature management in the upcoming year. The majority of respondents have standardized feature management on a single team, multiple teams, or across the entire organization. Feature flags play a critical role in enabling organizations to deliver and maintain software reliability.

LaunchDarkly has introduced updates to its core flagging, experimentation, and automation capabilities aimed at minimizing risk and maximizing value in software delivery. These updates include Automatically disabling flags during maintenance windows and re-enabling afterward, reviewing and approving feature changes with ease through Slack notifications and reviews, visualizing experiment results for easier interpretation, avoiding sample ratio mismatch errors, reducing friction when importing metrics, setting flag defaults to create optimal configurations, using Helm Chart for Kubernetes-based deployment, integrating Cloudflare for greater flag evaluation speeds, and scaling LaunchDarkly to teams of teams. Additionally, custom contexts are being developed to deliver software based on users or other chosen contexts, and Accelerate is a new product that measures software delivery performance and reliability across teams.

This article discusses how to eliminate risk and deliver value in Angular projects using feature flags. It builds on the Tour of Heroes (TOH) demonstrative app, which showcases basic concepts of Angular. The author explains how to create two feature flags, toh-search and toh-modify, and integrate them into the TOH project. They demonstrate how to use LaunchDarkly's JavaScript SDK to handle feature flag updates within the app whenever the user is changed or the feature flag configurations are modified. The author also provides code snippets for integrating the service with the hero-search component and modifying the hero component to introduce the toh-modify flag. The complete project can be found on GitHub.

This guest post from Hightouch discusses their use case of Reverse ETL to automate feedback loops between customers, sales, and engineering teams. By leveraging their data warehouse as a central source of truth, they streamline the process of collecting customer feature requests, prioritizing development work, and scaling LaunchDarkly feature toggles. The solution involves using Fivetran to ingest data into Snowflake, Hightouch for Reverse ETL, and integrating with Salesforce, Jira, Slack, Google Sheets, and LaunchDarkly. This approach allows teams to operate within their preferred tools while maintaining a bi-directional integration between customer requests and engineering progress.

The DevOps workforce is experiencing high levels of stress, burnout, and cognitive load due to additional roles and responsibilities. A paper published in IT Revolution's DevOps Enterprise Fall 2022 Journal discusses the adoption of a standardized platform to support a "You Build It, You Operate It" (YBYO) model and its impact on developers' cognitive load. LaunchDarkly, a company founded by two developers focused on making developers' lives easier, has implemented changes in their on-call coverage process to reduce burnout. The paper also highlights the importance of psychological safety and experimentation for increased innovation. Monitoring developers' happiness and job satisfaction can help ease the burden of the YBYO model and improve overall performance. Both publications provide valuable insights and tips for DevOps professionals dealing with cognitive load and job satisfaction issues.