May 2017 Summaries

The blog post explores methods to effectively monitor Nginx when deployed on Kubernetes, emphasizing the importance of understanding key metrics and configuring dashboards for optimal performance insights. Nginx, renowned for its speed and resource efficiency, is widely used as a reverse proxy, load balancer, and web cache in cloud environments. Monitoring involves enabling the stub_status module to expose performance metrics, and using tools like Sysdig Monitor, which offers advanced features such as leveraging Kubernetes metadata for detailed segmentation by namespaces, services, and pods. Sysdig provides visibility into Nginx's application layer metrics without code instrumentation, offering insights into HTTP methods, response codes, and request times. The article also touches on Nginx Plus, the commercial version, which offers additional enterprise features and metrics. It concludes by highlighting Nginx's critical role in cloud infrastructure, underscoring its utility in analyzing service responsiveness and predicting backend issues.

The blog post by Gianluca Borello delves into the complexities of container isolation, highlighting how shared kernel resources can cause unexpected interactions between containers. It presents a detailed troubleshooting narrative involving two containers, "worker" and "trasher," where the presence of the latter leads to performance degradation in the former, despite seemingly adequate resource limits. The issue is traced back to the kernel's dentry cache, which stores metadata for file paths, including non-existent ones, leading to a massive accumulation of dentry objects that slow down file operations. The post underscores the importance of monitoring container performance metrics and keeping kernel and distribution versions up to date to avoid such pitfalls. It also notes that newer kernel versions have addressed this specific issue by linking kernel object pool allocations to memory cgroups, preventing excessive dentry object creation.

The blog post provides a comprehensive guide on deploying WordPress in a Kubernetes environment using Helm, a Kubernetes package manager that streamlines the installation and management of applications by packaging them into "Charts." The tutorial outlines the challenges of deploying applications like WordPress, which require various components such as databases and load balancers, and explains how Helm simplifies these processes by managing dependencies and configurations as a single atomic operation. It details the installation of Helm on both OSX and Linux, the initialization of Helm in a Kubernetes cluster, and the deployment of WordPress using Helm Charts, emphasizing the benefits of using Helm for efficient application management. Additionally, the post discusses monitoring practices using Sysdig Monitor, highlighting the importance of setting up alerts for host-level issues, Kubernetes resources, and service-level performance to ensure the continuous operation of applications in a dynamic Kubernetes environment. The author suggests that users can further enhance their deployment strategy by creating custom Helm Charts and integrating user-friendly interfaces like Monocular for simplified Helm management.

Sysdig's Spring 2017 release introduces several enhancements to its monitoring platform, focusing on improved integration, usability, performance, and security features. The update includes support for Docker Swarm with new metrics and metadata labels, Kubernetes 1.6 with multi-cluster support, and automatic detection of metadata and metrics from Mesos platforms. Usability improvements feature a streamlined dashboard creation process, the ability to share dashboards between teams, and enhanced alerting options, including anomaly detection and group outliner alerts. Performance upgrades allow for faster dashboard loading and more efficient metric filtering, while security enhancements include Docker Certified container status for the Sysdig Monitor agent, improved authentication mechanisms, and regular vulnerability scanning. Additionally, new interactive time series charts and the ability to calculate percentiles for common metrics are introduced.

The article explores the deployment and monitoring of Linkerd, an open-source service mesh proxy, within a Kubernetes cluster, addressing the complexities of microservices communication and traceability. Linkerd serves as a scalable solution to manage network logic such as retries, timeouts, and TLS tunneling, thereby simplifying microservices architecture by abstracting service discovery and enabling dynamic request routing. To effectively monitor Linkerd, the article highlights the integration of Sysdig Monitor, which facilitates comprehensive visibility into microservices communication through statsd teleport, capturing real-time metrics like request latency and service health. Linkerd's deployment options, including convoy mode and DaemonSet, are detailed alongside its capability to support distributed tracing systems like Zipkin, aiding in the observation of latency across different microservices. The importance of configuring alerts for service latency, connection failures, and retries budget is emphasized to ensure optimal performance and reliability in microservices environments.

Sysdig offers a comprehensive monitoring solution for Azure Container Service by providing visibility into container and microservices-based applications. It enhances Azure's native tools by enabling in-depth monitoring of container interactions, such as HTTP requests, network connectivity, and resource usage, without needing to install agents inside the containers. Sysdig supports orchestrators like Docker Swarm, Kubernetes, and Mesosphere, and integrates seamlessly with Azure's infrastructure, offering real-time insights and metadata aggregation through its ServiceVision functionality. This allows for a detailed and holistic view of application performance, including service bottlenecks and response times, and supports service-based access control to tailor monitoring to specific teams or applications. Sysdig can be deployed as a SaaS solution or installed on-premises, offering flexibility in how organizations choose to implement it.