Ask Sendai

The Kubernetes implementation of the Container Storage Interface (CSI) has been promoted to GA in the Kubernetes v1.13 release. Support for CSI was introduced as alpha in Kubernetes v1.9 release, and promoted to beta in the Kubernetes v1.10 release. The GA milestone indicates that Kubernetes users may depend on the feature and its API without fear of backwards incompatible changes in future causing regressions. GA features are protected by the Kubernetes deprecation policy.

Uber built Manifold, a model-agnostic visualization tool for ML performance diagnosis and model debugging, to facilitate a more informed and actionable model iteration process. Machine learning (ML) is widely used across the Uber platform to support intelligent decision making and forecasting for features such as ETA prediction and fraud detection. For optimal results, we invest a lot of resources in developing accurate predictive ML models. In fact, it’s typical for practitioners to devote 20 percent of their effort into building initial working models, and 80 percent of their effort improving model performance in what is known as the 20/80 split rule of ML model development.

Declarative configuration management, also known as configuration-as-code, is one of the key strengths of Kubernetes. It allows users to commit the desired state of the cluster, and to keep track of the different versions, improve auditing and automation through CI/CD pipelines. The Apply working-group is working on fixing some of the gaps, and is happy to announce that Kubernetes 1.13 promoted server-side dry-run and kubectl diff to beta. These two features are big improvements for the Kubernetes declarative model.

We are going through an interesting cycle of application decomposition and recomposition. While the microservice paradigm is driving monolithic applications to be broken into separate individual services, the service mesh approach is helping them to be connected back together into well-structured applications. As such, microservices are logically separate but not independent. They are usually closely interdependent and taking them apart introduces many new concerns such as need for mutual authentication between services.

Kubernetes Operators have now become mainstream. An Operator is essentially a Kubernetes Custom Controller managing one or more Custom Resources. The term Operator has become popular as it succinctly captures end result of what Custom Controller+Custom Resource combination is typically defined for, e.g.—declaratively managing a stateful software on Kubernetes (e.g.: databases, off-the-shelf web applications, ML workloads, etc.). In technical terms, Custom Controller+Custom Resource combination is called ‘Custom Resource Definition’ (CRD).

At QCon SF, Suudhan Rangarajan presented ‘Netflix Play API: Why We Built an Evolutionary Architecture’. Key takeaways from the talk included: services that have a single identity/responsibility are easier to maintain and upgrade; engineers should spend time identifying core decisions that need to be made when building a service, and determine whether these are ‘Type 1’ or ‘Type 2’ decisions which require thorough deliberation or rapid experimentation, respectively; and designing an ‘evolutionary architecture’, using tools like fitness functions, provides many benefits.

Test-driven development, or TDD, involves writing tests first then developing the minimal code needed to pass the tests. TDD is an established practice for feature development that can improve code quality and test coverage. What about other, non-functional requirements such as scalability, reliability, observability, and other architectural “-ilities”? How do we ensure operability and resiliency of features when they go to production? How can we encourage teams to build in these architectural standards, just as test-driven development builds in code quality and test coverage?

In OpenTracing the fundamental concept for representing distributed traces is the (time) span: something that starts and then finishes, can be annotated with key-value pairs and can be “causally” related. This representation gained popularity with Google’s Dapper paper and triggered open-source tracing implementations like Zipkin and Jaeger and eventually the OpenTracing specification, but according to the academic literature it is not the only one. In spite its popularity, the span based trace representation has some shortcomings that limit its applicability.

A recently disclosed vulnerability in Kubernetes dashboard (CVE-2018-18264) exposes secrets to unauthenticated users. In this blog post we’ll explore some key takeaways regarding monitoring privilege escalation on Kubernetes. The Kubernetes dashboard is a web based user interface that allows users to manage applications and resources within the cluster. This service has a login functionality starting from Kubernetes 1.7.0. Since then, users are able to authenticate with a kubeconfig file or an access token.

Our security culture is one of the things that sets AWS apart. Security is job zero — it is the foundation for all AWS employees and impacts the work we do every day, across the company. And that’s reflected in our services, which undergo exacting internal and external security reviews before being released. From there, we have historically waited for customer demand to begin the complex process of third-party assessment and validating services under specific compliance programs.