May 2023 Summaries

Organizations often face the challenge of deciding whether to build their own software solutions or purchase commercially available ones, with factors such as cost, maintenance, support, and security influencing the decision. Building a custom solution can offer benefits like tailored features and no recurring costs, but it demands significant time, maintenance, and potential hidden expenses related to staffing and resources. Conversely, purchasing software often involves subscription costs and limitations in customization, but it includes updates, security patches, and support services that are often more robust. The decision hinges on evaluating long-term value, scalability, and organizational needs, with examples showing both successful in-house solutions and advantages of commercial products like Octopus Deploy, illustrating the importance of aligning the choice with strategic goals.

Octopus is revamping its user interface to address customer feedback about navigation difficulties, focusing on simplifying complex information structures and interactions. The redesign process involved reimagining the UI for better harmony among elements, consulting with customers and DevOps professionals, and implementing changes such as using squares instead of circles for spaces and projects, enhancing visual hierarchy with contrasting colors, and adopting familiar navigation patterns like a cog icon for configuration. These updates, part of a broader modernization effort, also include refining design components through design tokens to ensure adherence to accessibility guidelines and provide more theming options. Octopus is committed to further improvements, with plans to enhance the page header section for easier navigation, aiming for a consistent and accessible user experience.

The redesigned Tenants overview dashboard in Octopus enhances the management of customer-specific deployment pipelines by providing improved visibility and functionality for handling multiple instances of applications. The dashboard now features a data table format that addresses previous navigation challenges, allowing users to view tenant-related information, such as tenant tags, projects, and environments, more efficiently. It includes expandable rows for detailed views and offers pagination options to suit different user preferences. The dashboard maintains the existing filtering and search functions while ensuring fast performance even under large data loads, making it a reliable tool for administering hundreds or thousands of tenants. Currently available as an early access preview for Octopus Cloud customers, the update aims to provide a more responsive and organized experience when managing multi-tenancy deployments.

In a conversation with Joanna Wyganowska, Dale Francis, Director of Product Development at Climavision, shares insights from his DevOps journey, emphasizing the significance of collaboration, automation, and continuous delivery in enhancing software quality and delivery speed. He recounts his initial engagement with DevOps, driven by a need to solve problems and enhance capabilities, which led to his affinity for Octopus, a tool that provides psychological safety around releases by enabling rapid rollbacks. Francis highlights the importance of avoiding over-processing in automation and shares best practices such as integrating Infrastructure as Code (IaC) in their processes. He advises newcomers to DevOps to start small with automation to achieve quick wins that improve daily operations and foster further improvements. Additionally, Francis recommends key books for those beginning their DevOps journey and reveals his personal interest in woodworking, reflecting his passion for building and creating, both in his professional and personal life.

The 2023 State of CD report, commissioned by the Continuous Delivery Foundation and authored by SlashData, examines the adoption and performance of DevOps and Continuous Delivery practices, highlighting the importance of constructing high-performance DevOps toolchains. Effective toolchains integrate various tools across the Continuous Delivery pipeline, such as work item management, source control, build and test automation, and monitoring, to ensure frequent and reliable deployments. The report indicates that high-performing teams often use a greater number of tools than low performers, with nearly 45% of low performers relying on a single tool and deploying less frequently. To build a successful toolchain, it is recommended to select best-in-class tools, prefer managed solutions over self-hosting, keep workloads uncluttered, and minimize tool dependencies, as these strategies help avoid slowdowns caused by poor tools, self-hosting burdens, and interoperability issues.

Mean Time to Recovery (MTTR) is a key performance metric in software delivery, measuring the time it takes to restore a system after a fault. While MTTR is popularized by the DevOps Research and Assessment (DORA) metrics for its utility in industry research, it can be misleading for teams if used improperly, as it averages out critical incident details. To improve incident management, it is recommended to use detailed metrics and visualizations like scatter plots or box-and-whisker charts to capture trends and outliers. The SPACE framework is suggested as a more holistic approach to incident response, emphasizing satisfaction, performance, activity, communication, and efficiency. By using these diverse metrics, organizations can enhance their incident management processes and system stability more effectively than relying on MTTR alone. Additionally, conducting incident retrospectives and reviews shortly after incidents can help capture learnings and foster continuous improvement by addressing systemic issues rather than focusing solely on individual errors.

In the blog post by Matthew Casperson, methods for tracking CPU and memory usage of Kubernetes pods are explored, focusing on using the metrics server and cgroup interfaces. The metrics server offers a lightweight and scalable solution for monitoring resources in Kubernetes clusters, and while not inherently integrated, it is often easily enabled or bundled with clusters. With the metrics server installed, resource usage can be accessed via `kubectl top pod` and `kubectl top node` commands. In the absence of the metrics server, resource usage can be manually determined through cgroup interface files, though this requires entering an interactive session with a pod and performing calculations to convert raw CPU time into more useful measurements like CPU usage percentage. The text also highlights differences between cgroups v1 and v2 in terms of accessing resource metrics, noting the need for conversion between units of measurement.

Matthew Casperson's post from May 2023 explores a variety of alternative dashboards for monitoring and managing Kubernetes clusters, each offering unique features that go beyond the capabilities of the default Kubernetes Dashboard. The author evaluates several options, such as K8Dash, which provides an automatically updating interface; Konstellate, which focuses on creating and visualizing Kubernetes resources; Kubernator, offering low-level control and a robust YAML editor; Kubernetes Operational View (KOV) for a visual representation of clusters using WebGL; Octant, which provides a web-based dashboard with an intuitive interface; and Weave Scope, known for its process-oriented visualization of nodes, pods, and containers, particularly highlighting inter-pod communication. Casperson notes that these alternatives are easy to install and offer compelling reasons for users to consider switching, showcasing the continuous improvement and innovation within the Kubernetes ecosystem.