April 2018 Summaries

InVision uses LaunchDarkly as its feature flag service provider, which has significantly changed how teams think about feature deployment. One notable feature is the ability to mark certain user data as "private," allowing Personally Identifiable Information (PII) to be used in rule targeting without leaking beyond the application's boundaries. The most challenging aspect of using LaunchDarkly is understanding its user targeting mechanism, which involves syncing targeting rules with an application's local cache through asynchronous streaming. This approach enables designing targeting rules that depend on information only known by the client, such as email addresses marked as private. By denoting specific fields as "private," LaunchDarkly allows for safe targeting of PII while ensuring data security and compliance with privacy regulations.

Konmari is an old concept that emphasizes living with purposeful items and discarding unnecessary ones. Marie Kondo popularized this idea in her book "The Life-Changing Magic of Tidying Up." The author suggests applying these principles to digital lives, particularly codebases. To maintain a clean and organized codebase, follow best practices such as adding comments, using meaningful tags and variable names, and mentioning dependencies. Automate deletion of outdated elements to prevent clutter accumulation. Establish regular cleanup times during the development cycle and reward good behavior to encourage tidiness.

Isaac Mosquera, CTO at Armory.io, discussed canary deployments and their role in increasing development productivity, reducing risk, and keeping developers happy. He explained the concept of Canaries as a method to test new software versions in production by deploying one node into production, assessing its performance, and then scaling out if it performs well or rolling back if not. Automated canary analysis helps eliminate human bias from the decision-making process during deployment. Mosquera emphasized that while automated canaries are helpful, they should be used in conjunction with other testing methods such as unit tests and integration tests to ensure a safe and efficient software delivery system. He also shared lessons learned from implementing Canaries, including using simple statistics instead of machine learning for better transparency, creating separate baseline clusters for accurate comparisons, and choosing the right metrics based on each application's profile. The talk concluded with an overview of Armory's Spinnaker platform, which automates canary analysis and provides a visual representation of pipeline stages, enabling developers to monitor and control their deployments effectively.

In this episode of Greymatter, Armon Dadgar and Glenn Solomon join the show to discuss HashiCorp's journey from an open source project to a successful commercial company. Armon Dadgar, co-founder and CTO of HashiCorp, shares how they built their community around their open source projects and how they managed to grow their user base over time. Glenn Solomon, Managing Partner at GGV Capital, discusses the challenges of investing in open source companies and how he assessed HashiCorp's progress during its early days. The conversation also touches on the paradox of automation and how it affects continuous delivery.

In this episode of Greymatter, Armon Dadgar and Glenn Solomon join the show to discuss HashiCorp's journey from an open-source project to a successful commercial company. Armon shares how he and his co-founder Mitchell Hashimoto started HashiCorp with the goal of building tools that would help developers manage their infrastructure more efficiently. They initially built eight open-source projects, but only one gained significant traction - Vagrant. Armon explains how they used local meetups and conferences to build a community around their products. He also talks about the importance of having a strong commercial product alongside an open-source project. Glenn Solomon, Managing Partner at GGV Capital, shares his perspective on investing in open-source companies and how HashiCorp's approach has evolved over time. The discussion then shifts to continuous development and the challenges that come with automating infrastructure management. Armon talks about the paradox of automation and how it can lead to both increased efficiency and increased risk. He also discusses HashiCorp's approach to mitigating these risks through meta-automation and policy as code. Finally, the conversation wraps up with some final thoughts from Armon and Glenn on the importance of agility, adaptability, and perseverance in building a successful open-source company.

In this March Meetup, Aaron Gentleman from Atlassian shared how his team tested in production during the rollout of ADG3 to Confluence Cloud. The team addressed three main concerns: whether the change was ready for customers, its impact on both customers and the company, and defining when they were done with testing. They used feature flags in LaunchDarkly to control risky changes and test new features in production. Key points include considering customer feedback, ensuring support readiness, and managing multiple code paths during testing. The team also emphasized the importance of keeping customers in mind while rolling out changes.

In March, a Test in Production Meetup was held at Atlassian's Mountain View offices where software engineer Traci Lopez discussed the importance of helpful comments for gaining visibility into events and troubleshooting issues. She emphasized how comments can provide context about why certain actions were taken within an application, which is crucial for collaboration among team members. By using comments effectively, teams can better understand each other's decisions and work more efficiently towards their goals.

The text discusses the process of driving cultural change in an organization, emphasizing that it should start with people rather than technology or processes. It highlights a case study where deployment velocity was used as a guiding metric to improve the company's culture. The author suggests surveying employees to identify problems and selecting a single metric for the team to rally around. He also recommends focusing on trusting developers, buying technology when possible, and aligning people, processes, and tools towards a common goal.