March 2024 Summaries

Getting started with LaunchDarkly involves setting up projects, choosing an SDK, creating feature flags, expanding flag usage, and setting up access controls. Projects represent applications or components within an application, while environments represent the release stages of that application. The choice of SDK depends on the programming language being used. Feature flags can be boolean, multivariate, or customized according to specific use cases. Access controls involve assigning user roles and creating teams for managing permissions across projects, environments, and flags.

Software deployments are complex due to the intricate application architecture and infrastructure requirements. The process involves pushing new code to an environment, making it available and accessible to end users. Deployment strategies such as blue/green, canary, shadow, and rolling deployments have emerged to streamline this process. However, they all rely on managing infrastructure resources. Feature management solutions like LaunchDarkly help organizations control the release process without disrupting continuous deployment by enabling feature toggling and experimentation. These tools also integrate with APMs for monitoring and incident management, allowing teams to quickly toggle off problematic features without needing a new deployment.

The text discusses six common deployment strategies in software delivery: big-bang, rolling, recreate, canary, blue/green, and shadow. It highlights that the best strategy depends on factors such as application characteristics, team capabilities, and deployment goals. Feature flags are introduced as a better approach to these strategies by separating deployments from releases. The text also explores differences between deployments and releases, types of deployment strategies, choosing the right deployment strategy, comparing deployment strategies, and the role of feature management in deployment.

Images have a unique ability to influence our comprehension and perception due to their processing speed in the brain compared to text. The traditional DevOps infinity loop, attributed to Dustin Whittle, portrays DevOps as a sequence of phases. However, modern DevOps does not occur in a linear structure but rather processes that happen in parallel and often out of order. In the book "Operating Continuously," Edith Harbaugh, Cody De Arkland, and the author focus on continuous operation with steps that happen in parallel, enabling ops teams to operate more efficiently and companies to deliver value to users at a consistent and rapid pace. The deployment process covers the workflow of moving compiled code onto a destination infrastructure or software platform, decoupled via feature flags. Release responsibility has shifted to the ops side of the circle with strategies like canary deployments and kill switches. Monitoring and observability are now blended into the operate step, creating the opportunity to implement practices that quantify the efficacy of the features released to end users more proactively. The dev and ops sides of the infinity loop are interconnected gears, with each side needing the other to turn efficiently to operate at full force.

LaunchDarkly has introduced an improved experience for creating feature flags, as well as the ability to define custom defaults for flag templates. Users can now set custom flag defaults for each of the available flag templates in the flag settings page for a project. This ensures that each time a new flag is created, it is configured according to the user's preferences. The customizable default fields include whether a flag is temporary or permanent, boolean default variations and labels, and tags for categorization, filtering, and permissions management.

In 2024, release management practices are expected to undergo significant shifts due to the emergence of GenAI and its potential risks. Engineering organizations will likely push for standardizing safety nets like feature flags to manage these risks while accelerating innovation. The adoption of progressive delivery is also anticipated to increase as teams aim to control their release process, enhance speed, and mitigate inherent development friction. Additionally, there will be a growing emphasis on defining and tracking key release metrics such as deployment frequency and mean time to recover (MTTR) from an incident.