August 2020 Summaries

Change Advisory Boards (CABs) have traditionally been used in software delivery to manage risk and ensure safety, but recent studies suggest they may slow down the process without enhancing system stability. Dr. Nicole Fosgren's research indicates that external approvals from CABs negatively impact lead time and deployment frequency without reducing change failure rates. Instead, a lightweight change approval process, such as peer review combined with a deployment pipeline, is recommended for safer and more efficient software delivery. Modern approaches, like those adopted by companies such as Booking.com and Walmart Labs, utilize advanced deployment pipelines to detect and revert bad changes quickly, emphasizing the importance of progressive delivery methods. These methods, which include decoupling deploy from release and selectively exposing new code, are becoming foundational practices in achieving continuous integration and delivery. Feature flags and automated mechanisms for detecting and rejecting problematic changes represent a shift towards impact-driven software delivery, enabling teams to make more informed decisions based on real-time data analytics without the bottleneck of traditional change management processes.

Harness has expanded its capabilities by integrating Drone, a popular open-source continuous integration platform, into its suite as Harness CI, providing comprehensive continuous integration and delivery solutions. Along with this integration, Harness has introduced Cloud Cost Management recommendations to help users reduce cloud expenses efficiently by utilizing real-time usage and billing data. Additionally, updates to the Harness GraphQL API now allow for the creation and setup of deployment Triggers, automating processes based on various conditions such as Git events and new artifacts. Harness also announced its first industry conference, {Unscripted}, scheduled for October, aiming to educate attendees on simplifying software delivery. The company has received media attention for these developments and continues to engage with the community through virtual events and promotional offers like the Harness T-shirt Hall of Fame.

Java and native agents offer distinct capabilities for monitoring and updating JVM applications, with Harness SRM utilizing native agents to perform low-level, real-time error analysis without affecting performance. Java agents, typically .jar files, integrate with the JVM at runtime to modify class methods dynamically through bytecode instrumentation, enabling tasks such as profiling and hot-swapping classes. In contrast, native agents, often written in C++, leverage the JVM Tooling Interface (JVMTI) to access a broad range of JVM internals, allowing for powerful debugging and monitoring capabilities. While native agents provide greater access and control, they are more complex, require platform-specific compilation, and carry risks of JVM instability if improperly managed. Harness SRM primarily operates through native code, minimizing the footprint on server resources and ensuring compatibility with existing tools and agents.

Feature flags are a powerful tool for Node.js development, offering developers the ability to control functionality, test features safely, and iterate quickly without impacting production environments. The article provides a step-by-step guide to creating a simple library API using Node.js and Express, illustrating how feature flags can separate code deployment from feature enablement. By using Split's JavaScript SDK, developers can easily manage which features are available to users, allowing for safe experimentation and deployment of new functionalities. The use of feature flags is demonstrated through enabling and disabling a book deletion feature, showing how developers can control user access based on treatment settings within Split. The guide emphasizes the flexibility and control feature flags offer in software development, encouraging developers to explore advanced options for experimentation and secure deployment.

Feature flags in Spring Boot provide a method for efficient feature management, allowing developers to control the release and testing of new functionalities without redeploying code. This technique is beneficial for A/B testing and canary releases, enabling selective feature exposure to specific user groups to gauge reactions or mitigate risks. The implementation of feature toggles, as demonstrated with Split by Harness in a Spring Boot web API, involves setting up an environment to manage these toggles, configuring the necessary software dependencies, and updating application code to integrate the toggles. This process enhances a software release life cycle by supporting continuous integration and deployment, thereby enabling swift and consistent production changes. The practical application was illustrated through a tutorial on adding a feature toggle to manage reservations in a sample restaurant API, showcasing the real-world utility of feature flags in software development.

Harness integrates Continuous Integration (CI) and Continuous Delivery (CD) to streamline the deployment process from code to production, enabling seamless Docker image builds and Kubernetes cluster deployments with minimal setup. By leveraging a Harness CI/Drone Pipeline, users can build and push a GoLang Docker Image to DockerHub, after which Harness CD deploys the new artifact to a Kubernetes cluster. The process requires setting up a Harness Delegate in the Kubernetes cluster and configuring a Harness Application, including Services and Environments. A rolling deployment workflow is created and triggered by a new artifact in DockerHub, allowing for automated, end-to-end CI/CD pipeline management. The system's ease of use and scalability make it an attractive solution for simplifying and enhancing CI/CD capabilities.

Adopting microservices with CI/CD principles offers increased deployment flexibility and scalability, as highlighted by Tiffany Jachja's insights on foundational strategies. Her session emphasizes the essential principles required to initiate work with microservices, focusing on using CI/CD effectively. Harness is described as a comprehensive AI software delivery platform designed to manage the software development lifecycle through specialized AI agents, aiming to streamline and enhance the process.

Harness CI, enhanced by the acquisition of Drone.io, aims to streamline the continuous integration process by integrating with source code management platforms and operating on a server-runner model, which simplifies setup and build execution. The integration process involves authenticating the Harness CI/Drone server with an SCM provider, setting up a CentOS machine with Docker Engine, and utilizing a Kubernetes cluster for deploying runners. Users must configure OAuth applications in platforms like GitHub and set up a shared secret for secure server-to-runner communication. The system leverages Configuration as Code, allowing seamless pipeline creation and execution through webhook requests from the SCM. Once configured, users can initiate their first build and publish it to Docker Hub, marking a successful deployment. Harness CI and Drone communities are set to benefit from further enhancements in CI and CD capabilities, fostering a more inclusive and democratized environment for developers.

Testing in production, particularly using feature flags, is gaining traction as a method for ensuring safe deployments and validating features in real-world conditions. This approach allows developers to test functionality, design, and performance in the actual production environment by initially exposing features only to internal users, thereby mitigating risks before a broader user rollout. The traditional reliance on staging environments is being questioned, as staging might not accurately replicate production conditions, potentially leading to discrepancies in test results. A structured 90-day implementation plan is proposed to facilitate the transition to testing in production, emphasizing the importance of aligning automation frameworks and feature flag management with existing software development life cycle (SDLC) setups. Real-world testing offers insights into user traffic, load handling, and performance that are challenging to simulate in controlled environments, although staging remains valuable for debugging and issue isolation. The ultimate goal is to leverage feature flags to manage code changes, rollbacks, and continuous delivery efficiently, ensuring that features are thoroughly vetted in production before full release to end users.

Harness's acquisition of Drone.io aims to enhance its CI/CD platform by integrating Drone's open-source, container-native CI solution, which has been supported by a robust community of over 50,000 active users and 250+ contributors. Founded by Jyoti Bansal, who is also the co-founder of Traceable and Unusual Ventures, Harness seeks to simplify software delivery for engineers and DevOps teams. The acquisition is expected to maintain Drone's open-source nature while allowing Harness to invest in the community and platform, promising to deliver a simple yet powerful CI/CD platform focused on eliminating engineering toil and complexity. Harness plans to contribute internal projects to the open-source community and ensure that the existing Drone community and customer base experience no changes. With over 100 million pulls on DockerHub, Drone's integration is seen as a strategic move to empower engineers to focus on innovation rather than the manual building of software delivery pipelines.

Ravi Lachhman, a CI/CD Evangelist, showcases the capabilities of Harness Cloud Cost Management, highlighting how it has enabled customers to achieve significant savings in idle cloud expenditure.